A Cryogenic Engineering Conference Publication
Advances in
Cryogenic Engineering
VOLUME 19
K. D. TIMMERHAUS, Editor
Engineering Research Center
University of Colorado
Boulder, Colorado
and
Engineering Division
National Science Foundation
Washington, D.C.
Springer Science+Business Media, LLC 1995
The Library of Congress cataloged the first volume of this title as folIows:
Advanees in cryogenic engineering. v. 1New York, Cryogenic Engineering Conference; distributed
by Plenum Press, 1960T.
Wua., dlagra. 26 on.
Vo'" 1are reprints ot tbe Proceedlnp of tbe Cryogenle En·
glneering Conterenee, 1964Editor: 1960K. D. Tlmmerhaus.
1. Low temperature englneerlng-Congresses.
K. D .. ed. 1I. Cryogenlc Engineering ConterenC9.
TP490.A3
660.29368
I.
Tlmmerhaus,
57-35598
Llbrary ot Congre.s
Proceedings of the 1973 Cryogenic Engineering Conference,
Georgia Institute of Technology, Atlanta, Georgia,
August 8-10, 1973
Library of Congress Catalog Card Number 57-35598
ISBN 978-1-4613-9849-3
ISBN 978-1-4613-9847-9 (eBook)
DOI 10.1007/978-1-4613-9847-9
© 1995 Springer Science+Business Media New York
Originally published by Plenum Press, New York in 1995
Softcover reprint of the hardcover 1st edition 1995
CONTENTS
Foreword. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Russell B. Scott Memorial Award. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1973 Cryogenic Engineering Conference Board. . . . . . . . . . . . . . . . . . . . . . . . .
Awards Committees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
S. C. Collins Award. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xi
XII
xiii
xiii
xiv
xv
Energy Systems
A-I
Cryogenic H2 and National Energy Needs, J. HORD, NBS Institute
A-2
A-3
The Economics of Liquid Hydrogen Supply for Air Transportation,
J. E. JOHNSON, Union Carbide Corporation, Linde Division. . . . . . . . . .
The UCLA Hydrogen Car, A. F. BUSH and W. D. VAN VORST,
A---4
Cryogenic Engineering and Fusion Power, C. E. TAYLOR, Lawrence
for Basic Standards . ........................................ .
12
University of California at Los Angeles. . . . . . . . . . . . . . . . . . . . . . . . . .
23
Livermore Laboratory, University of California. . . . . . . . . . . . . . . . . . .
28
Applied Superconductivity
Machinery and Magnets
B-1
B-2
B-3
B---4
B-5
Superconducting Electrical Generators for Central Power Station
Use, T. M. FLYNN, R. L. POWELL, D. B. CHELTON, and B. W.
BIRMINGHAM, NBS Institutefor Basic Standards. . . . . . . . . . . . . . . . . .
Cryogenic Considerations in the Development and Operation of a
Large Superconducting Synchronous Generator, C. K. JONES and
D. C. LITz, Westinghouse Electric Corporation . .............. , . . . .
Superconducting Alternator Test Results, A. BEJAN, T. A. KEIM,
J. L. KIRTLEY, JR., J. L. SMITH, JR., P. THULLEN, and G. L. WILSON,
Massachusetts Institute of Technology. . . . . . . . . . . . . . . . . . . . . . . . . . .
Alternating Field Losses in the Superconductor for a Large HighSpeed AC Generator, M. S. WALKER, J. H. MURPHY, Y. W. CHANG,
and H. E. HALLER III, Westinghouse Electric Corporation. . . . . ... . .
A Review of Superconducting Magnetic Systems for Generating
Transverse Magnetic Fields, V. V. SYTCHEV, Institute for High Tem-
peratures, Moscow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
44
53
59
67
Applied Superconductivity
Electrical Transmission and Storage
C-l
C-2
European Progress in Cryopower Transmission, G. BOGNER, Siemens
AG, Erlangen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Research on a Laboratory Model of Superconducting Test Cable,
D. V. RAZEVIG, Y. L. BLINKOV, and Y. S. GOLDENBERG, G. M.
Krzhizhanovsky Power Research Institute, Moscow. . . . . . . . . . . . . . . .
v
78
92
vi
C~3
C-4
C~5
Contents
Development of a Rigid AC Superconducting Power Transmission
Line, R. W. MEYERHOFF, Union Carbide Corporation, Linde Division
A Supercritical Helium Facility for Measuring High-Voltage Breakdown, E. B. FORSYTH, R. B. BRITTON, J. DEAN, J. E. JENSEN, and
K. MINATI, Brookhaven National Laboratory. . . . . . . . . . . . . . . . . . . . .
Superconducting Energy Storage, R. W. BooM, H. A. PETERSON, and
W. C. YOUNG, University of Wisconsin, and G. E. McINTOSH,
Cryenco. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
101
109
117
Applied Superconductivity
Transportation
High-Speed Transportation Levitated by Superconducting Magnet,
K. OSHIMA, University of Tokyo, and Y. KYOTANI, Japanese National
Railways. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
D~2 SRI Magnetic Suspension Studies for High-Speed Vehicles, H. T.
COFFEY, Stanford Research Institute. . . . . . . . . . . . . . . . . . . . . . . . . . . .
D~3 AC Losses in Multifilamentary Superconducting Composites for
Levitated Trains under AC and DC Magnetic Fields, T. SATOW,
M . TANAKA, and T. OGAMA, Mitsubishi Electric Corporation. . . . . . .
D-4 Shaped Field Superconductive DC Ship Drive Systems, T. J. DOYLE,
Naval Ship Research and Development Center. . . . . . . . . . . . . . . . . . . .
D~1
127
137
154
162
Applied Superconductivity
Operational Characteristics
E~ 1
E~2
E~ 3
E-4
Alternating Current Losses in Superconducting Conductors for LowField Applications, M. A. JANOCKO, D. W. DEIs, and W. J. CARR, JR.,
Westinghouse Electric Corporation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Application of Loss Models to Superconducting Solenoids,
J. T. BROACH and W. D. LEE, U.S. Army Mobility Equipment Research
and Development Center. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Investigation of the Dynamic Processes Occurring in Superconducting
Windings, V. A. ALTOV, M. G. KREMLEV, V. V. SYTCHEV, and V. B.
ZENKEVITCH, Institute for High Temperatures, Moscow. . . . . . . . . . . .
Analysis of Cryogenic Current Leads with Normal Conductors and
Superconductors in Parallel, B. B. GAMBLE, General Electric Company, and J. L. SMITH, JR. and P. THULLEN, Massachusetts Institute
of Technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
171
181
186
193
Refrigeration
F~l
F~2
F~ 3
A Decade of Involvement with Small Gas-Lubricated Turbines,
M. E. CLARKE, British Oxygen Company Limited.. .. .. .. ... ......
Gas Bearing Cryogenic Expansion Turbines, J.-C. VILLARD and
F. J. MULLER, L'Air Liquide-Centre d'Etudes Cryogeniques. .... . .
Pneumatically Driven Split-Cycle Cryogenic Refrigerator, S. B. HORN,
M. E. LUMPKIN, and B. T. WALTERS, U. S. Army Night Vision
Laboratory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
200
209
216
Contents
F---4
F-5
F-6
F-7
F-8
Theoretical Analysis of Pneumatically Driven Split-Cycle Cryogenic
Refrigerators, S. B. HORN and M. E. LUMPKIN, U. S. Army Night Vision
Laboratory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Refrigeration for the Culham Superconducting Levitron, D. N.
CORNISH and R. E. BRADFORD, Culham Laboratory, and A. J. STEEL,
British Oxygen Company Limited. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Heat Load Due to Ortho-Para Conversion in a Closed-Loop Hydrogen Refrigerator, R. L. PuBENTZ and D. A. VANGUNDY, Argonne
National Laboratory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Prototype Tests on a 200-W Forced Convection Liquid Hydrogen/
Deuterium Target, K. D. WILLIAMSON, JR., J. E. SIMMONS, F. J.
EDESKUTY, J. H. FRETWELL, J. T. MARTIN, and H. FICHT, Los Alamos
SCientific Laboratory, University of California. . . . . . . . . . . . . . . . . . . .
Liquid Hydrogen Pumping for Hydrogen Targets, J. W. MARK,
Stanford Linear Accelerator Center. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
vii
221
231
239
241
248
LNG Storage and Transport
G-l
Effect of Weathering of LNG in Storage Tanks, J. M. SHAH and
J. J. AARTS, Chicago Bridge and Iron Company. . . . . . . . . . . . . . . . . . .
G-2 Design of LNG Receiving Terminals, D. B. CRAWFORD and C. A.
DURR, The M. W. Kellogg Company. . . . . . . . . . . . . . . . . . . . . . . . . . . .
G-3 Some Important Factors in LNG Tanker Design Selection, R. C.
FFOOKS, Conch Methane Services Limited. . . . . . . . . . . . . . . . . . . . . . . .
G---4 Near-Term Trends in LNG Tankship Design, J. L. HOWARD,
G-5
Kvaerner-Moss, Inc. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Status Report on LNG Tanker Designs, A. PASTUHOV, Gazocean USA,
Inc., and M. GONDOUlN, American Technigaz, Inc. . . . . . . . . . . . . . . . . .
253
261
269
276
282
LNG Related Fluid Properties
H-l
Phase Equilibria for Systems Containing Nitrogen, Methane, and
Propane, D. P. L. POON and B. c.- Y. Lu, University of Ottawa. . . . . .
H-2 Liquid-Vapor Equilibria in the Nitrogen-Methane System between
95 and 120 K, W. R. PARRISH and M. J. HIZA, NBS Institutefor Basic
292
Standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
300
Gas-Liquid Equilibria of the CO 2 -CO and CO 2-CH c CO Systems,
L. J. CHRISTIANSEN and A. FREDENSLUND, Instituttet for Kemiteknik,
Denmark, and N. GARDNER, Case Western Reserve University. . . . . . .
H---4 Solubility of Solid Benzene, Toluene, n-Hexane, and n-Heptane in
Liquid Methane, G. P. KEUBLER and C. McKINLEY, Air Products and
309
H-3
H-5
H-6
Chemicals, Inc.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
320
and Iron Company. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Calculation of LNG Excess Volumes by a Modified Hard-Sphere
Model, J. B. RODOSEVICH and R. C. MILLER, The University of
Wyoming. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
327
Phase Behavior ofthe Methane-Carbon Dioxide System in the SolidVapor Region, G. M. AGRAWAL and R. J. LAVERMAN, Chicago Bridge
339
viii
H-7
COBtents
A New Correlation between Heating Values for LNG Custody
Transfer, P. e. JOHNSON, Distrigas Corporation, J. P. LEWIS, Transco
Energy Company, and G. M. WILSON, Brigham Young University. . . .
346
Pure Component Fluid Properties
I-I
1-2
Scaled Parametric Equation of State for Oxygen in the Critical
Region, J. M. H. LEVELT SENGERS, National Bureau of Standards, and
W. L. GREER and J. V. SENGERS, University of Maryland. . . . . . . . . . . .
Superfluid Thermodynamic Transport Limits for Liquid Helium II,
e. LINNET, R. e. AMAR, Y. G. WANG, and T. H. K. FREDERKING,
University of California at Los Angeles. . . . . . . . . . . . . . . . . . . . . . . . . .
358
365
Materials Technology
J-I
A Simple Method for Charpy Impact Testing below 6 K, S. JIN,
W. A. HORWOOD, J. W. MORRIS, JR., and V. F. ZACKAY, University of
J-2
An Iron-Nickel-Titanium Alloy with Outstanding Toughness at
Cryogenic Temperatures, S. JIN, J. W. MORRIS, JR., and V. F. ZACKAY,
J-3
Compressive Load-Deflection Characteristics of Several Foam Materials at Room Temperature, 77 K, and 4.2 K, W. F. STEWART, D. T.
EASH, and W. A. MAY, Los Alamos Scientific Laboratory, University
J---4
J-5
California. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
373
University of California. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
379
of California. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
385
A Structural Plastic Foam Thermal Insulation for Cryogenic Service,
R. B. BENNETT, Amspec Inc.. . . ............ ... . . . . . .... . .......
A Differential Thermal Analysis Apparatus for Use at Cryogenic
Temperatures, E. CATALANO, J. A. RINDE, and J. e. ENGLISH,
Lawrence Livermore Laboratory, University of California. . . . . . . . . .
393
400
Heat and Mass Tramfer
K-I
Forced Convection Heat Transfer to Subcritical Helium I, P. J.
GIARRATANO, R. e. HESS, and M. C. JoNES, NBS Institute for Basic
K-2
Vaporization Onset Heat Flux for Flat Plates in Saturated Liquid
Helium II, D. W. B. MATTHEWS, National Defence Headquarters,
Ottawa, and A. e. LEONARD, Royal Military Col/egeofCanada......
Heat Transfer to Slush Hydrogen, e. F. SINDT, NBS Institutefor Basic
K-3
Standards. . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
404
417
Standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
427
Kinergetics, Inc.. . . . . . . . . . . .. . .. .. . . . . . . . . . . . . . . . . . .. . . . . . . . .
Design and Selection of Cryogenic Heat Exchangers, K. D. TIMMERHAUS and R. J. SCHOENHALS, National Science Foundation. . . . . . . . . . .
437
K ---4 A Countercurrent Heat Exchanger that Compensates Automatically
for Maldistribution of Flow in Parallel Channels, K. W. COWANS,
K-5
K~
Separation of Nitrogen from Helium Using Pressure-Swing Adsorption, G. BIRD, Petro carbon Developments Limited, and W. H. GRANVILLE, University of Bradford. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
/445
463
Contents
ix
Space Technology
L-I
L-2
L-3
L-4
L-5
Cryocontamination of Optical Solar Reflectors and Mirrors, C.-K.
LIU, Lockheed Palo Alto Research Laboratory, and C. L. TIEN,
University of California. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Thermal Performance Characteristics of a Combined External Insulation System under Simulated Space Vehicle Operating Conditions,
F. J. MULLER, L' Air Liquide-Centre d'Etudes Cryogeniques, and P. L.
KLEVATT, McDonnell-Douglas Astronautics Company. . . . . . . . . . . . .
Multistage Radiative Coolers for Spacecraft Sensors, R. P. BYWATERS
and M. C. KEELING, Texas Instruments, Inc. . . . . . . . . . . . . . . . . . . . . .
RF Gauging Efforts with Liquid Hydrogen and Liquid Oxygen As
Applicable to Subcritical Space Vehicle Systems, H. E. THOMPSON,
NASA Marshall Space Flight Center, and W. OTT and N. STANLEY,
The Bendix Corporation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Computed Wicking Rates of Cryogens at Low Gravity for Selected
Screens, R. A. MOSES, D. F. GLUCK, and W. J. HINES, Rockwell
International. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
474
482
490
500
509
Indexes
Author Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Subject Index. ......................................... .............
517
519
FOREWORD
The year 1973 marked the first time that Atlanta, one of the cultural centers of
the South, has hosted the Cryogenic Engineering Conference since its beginning
in 1954. The Cryogenic Engineering Conference gratefully acknowledges the hospitality of the Georgia Institute of Technology and the assistance of W. T. Ziegler and his
staff in making the visit to Atlanta a pleasant and memorable one.
Several significant changes were initiated at the 1973 Cryogenic Engineering
Conference. These included a Conference theme on the subject of "Energy and the
Environment," a new Conference format, and the beginning of a new Conference
frequency of biennial meetings. While retaining the traditional topics of previous
meetings, the 1973 Cryogenic Engineering Conference focused on the role of cryogenic engineering in the generation, distribution, and conversion of energy, and the
related environmental effects. In these areas, much of the current interest stems from
the environmental effects of LNG and liquid hydrogen as compared with other
competing energy forms. These rapidly expanding areas may provide the impetus to
cryogenic engineering in the 1970's that the space program provided in the 1960's.
The Conference format was altered by the use of numerous invited papers
highlighting the theme. These presentations were concentrated in plenary sessions
initiating each day's activities, and in seminars designed to summarize the various
aspects of the theme.
The assistance of the many dedicated workers in the cryogenic field who have
once more contributed their time and talents to the reviewing of the preliminary
papers for the program and the final manuscripts for this volume is again gratefully
acknowledged by the Cryogenic Engineering Conference Board and the editor. The
list of individuals involved in these many important tasks is just as long as in past
years and any attempt to acknowledge individual contributions in the limited space
would be highly inadequate. However, special recognition needs to be given once
again to Mrs. Elva R. Dillman from the University of Colorado for her conscientious
attention to all the details involved in the preparation of the final manuscripts for this
volume.
Over the years it has become a tradition in this series to recognize individuals
who in some way have gone the extra mile in extending the frontiers of cryogenic
engineering or have provided continued assistance to the Cryogenic Engineering
Conference and its associated Advances in Cryogenic Engineering series. In the spirit
of this tradition it is only fitting that we recognize an individual who for the past
twenty-five years has been an active and effective pioneer in integrating basic low
temperature information and principles in various fields to practical solutions of
cryogenic engineering problems. Accordingly, Volume 19 of the Advances in Cryogenic
Engineering is dedicated to the 1973 S. C. Collins Award winner, EDWARD F.
HAMMEL of the Los Alamos Scientific Laboratory.
xi
RUSSELL B. SCO'IT MEMORIAL AWARD
The objectives of this award are to provide increased recognition for the recipients
in the scientific community, and to provide an incentive for higher quality in both oral
and written presentations at future Cryogenic Engineering Conferences.
The winners of the Russell B. Scott Memorial Award for the outstanding papers
presented at the 1972 Cryogenic Engineering Conference, as announced by the
Awards Committee, are as follows: In the cryogenic engineering research category,
R. C. Hendricks, R. J. Simoneau, and R. C. Ehlers of the NASA Lewis Research Center
are recognized for their paper, "Choked Flow of Fluid Nitrogen with Emphasis on the
Thermodynamic Critical Region," while in the application of cryogenic engineering
category, D. E. Daney, P. M. McConnell, and T. R. Strobridge ofthe NBS Cryogenics
Division are acknowledged for their paper, "Low Temperature Nitrogen Ejector
Performance. "
The Cryogenic Engineering Conference extends its congratulations to all of
these award-winning authors.
xii
1973 CRYOGENIC ENGINEERING
CONFERENCE BOARD
R. B. Fleming, Chairman . .............. General Electric Company
B. W. Birmingham .................... National Bureau of Standards
T. H. K. Frederking................... University of California at Los Angeles
L. Garwin ........................... Consulting Engineer
R. C. Hendricks ...................... NASA Lewis Research Center
J. E. Jensen .......................... Brookhaven National Laboratory
A. C. Leonard ........................ Royal Military College of Canada
M. J. Hiza ........................... National Bureau of Standards
W. H. Hogan ........................ Cryogenic Technology, Inc.
W. T. Ziegler " ...................... Georgia Institute of Technology
J. L. Smith, Jr........................ Massachusetts Institute of Technology
M. B. Clapp ......................... Chicago Bridge and Iron Company
J. K. Hulm .......................... Westinghouse Electric Corporation
K. D. Timmerhaus .................... National Science Foundation and University of Colorado, Editor, Advances in
Cryogenic Engineering
L. K. Armstrong...................... National Bureau of Standards
AWARDS COMMITfEES
AWARDS COMMITTEE
J. L. Smith, Jr. Chairman . .............. Massachusetts Institute of Technology
M. J. Hiza........................... National Bureau of Standards
W. H. Hogan ........................ Cryogenic Technology, Inc.
S. C. COLLINS AWARD COMMITTEE
B. W. Birmingham, Chairman .......... National Bureau of Standards
K. D. Timmerhaus.................... National Science Foundation and University of Colorado
R. B. Fleming........................ General Electric Company
S. C. Collins ......................... Naval Research Laboratory
xiii
ACKNOWLEDGMENTS
The Cryogenic Engineering Conference Board is deeply grateful for the support
which the following organizations have given the 1973 conference.
Japanese National Railways
Japan Oxygen Company Ltd.
Linde Aktiengesellschaft
The Arthur D. Little Foundation
Lotepro Corporation
McDonnell Douglas Corporation
Minnesota Valley Engineering
National Bureau of Standards
National Science Foundation
Rivoira S. P. A. (Italy)
Silbrico Corporation
Union Carbide Corporation
University of Colorado
Aerospace Corporation
Air Products and Chemicals, Inc.
British Oxygen Company Ltd.
Chemtron Corporation
Chicago Bridge and Iron Company
Cryogenic Technology, Inc.
Essex Cryogenics Industries, Inc.
General Electric Corporate Research
and Development
Gardner Cryogenics
Garrett Corporation
Georgia Institute of Technology
Grumman Aerospace Corporation
xiv
THE SAMUEL C. COLLINS AWARD
Dr. Edward F. Hammel, associate Q-Division leader for applied technology at
the Los Alamos Scientific Laboratory and a vice president of the International
Institute of Refrigeration since 1963, became the third recipient for the highest award
of the Cryogenic Engineering Conference-The Samuel C. Collins Award for Outstanding Contribution to Cryogenic Technology. This award, named after the
inventor of the first practical helium liquefier, was made at the annual Cryogenic
Engineering Conference held at the Georgia Institute of Technology in Atlanta,
Georgia on August 8, 1973. The award was presented to Dr. Hammel by Dr. S. C.
Collins, first recipient of the award, and Dr. K. D. Timmerhaus, second recipient of
the award and editor of the Advances in Cryogenic Engineering.
Dr. Edward F. Hammel
Third recipient of the
Samuel C. Collins Award
presented by the
Cryogenic Engineering Conference
xv
A-l
CRYOGENIC Hl AND NATIONAL
ENERGY NEEDS*
J. Hord
Cryogenics Division
NBS Institute for Basic Standards
Boulder, Colorado
INTRODUCTION
Hydrogen, as a nonfossil synthetic fuel, is a prime candidate to satisfy many of
our long-term fuel requirements. Specifically, cryogenic hydrogen offers significant
advantages for many applications. The objective of this presentation is to synthesize
the voluminous and sometimes speculative literature, emphasize cryogenic hydrogen
applications, and appraise the prospects for cryogenic hydrogen in the rapidly expanding fuel market.
To suggest that nonfossil hydrogen is a panacea for all of our energy problems
is technically irresponsible and inexcusable. An abundant nonpolluting energy source
is required to mass-produce hydrogen from water. Thus, hydrogen is an energy
carrier (or fuel), not an energy source (see Fig. 1). In this capacity, hydrogen can fill
certain vital needs in future energy conversion requirements. Conservation of energy
_I-•
•
(l1li1_
FIIIiIFIIIs
IFossiIFIIIs-("111l1liI:I1II1- DIcIri:iIy--FIIIs
( s-rc .,.,.. ("'111,~
,'-111--"- ....... 11'.-
'l1li'
-
\ 00ItriciIJ---
--I~(-I­
._.IIc.-
\-.EIodrioI!-
1~
' _ _ 'EIIcIriciIy---
1_---
'(IiI!II_I_:
r
1Y**bi<-
DIcIri:iIy---
\-. .,.,.. (-1'- 00ItriciIJ-
r.
Fig. I. Energy-fuel relationships.
* Invited paper.
lIB
>DIcIrioiIy_
I-FIIIs--
(
S,-FIIIs
[ ::='~l
*.
',,-- . - -----------------------~-~-===
• CIbpriaIy -
......... - - .
•
!OIlIER
2
J. Hord
resources. environmental compatibility. economy. and convenience are essential
criteria of the future-some aspects of the "hydrogen economy" will conform and
others will fail. Since our energy shortage is real, we must also conserve human
resources. Proliferation in the literature of highly speculative engineering is selfdefeating. In-depth technical and economic evaluations are now required to identify
market potentials. The cryogenics industry has the technology and proven capability
to shoulder its share of the burden.
NA TIONAL ENERGY NEEDS AND RESOURCES
Needs and Resources
We refer here to energy needs rather than energy goals because national goals
have not yet been defined. The enormous task of establishing national goals requires
national leadership and will undoubtedly languish in obscurity until a federal energy
management agency [1 J is commissioned. Many authors, particularly nontechnical
ones. consider the current energy shortage as a dilemma rather than a crisis. Irrespective of the terminology. the demand and shortage are real.
Interim solutions to this energy shortage consist in increased importation of
fluid fossil fuels, increased success in discovery of new domestic fossil fuel resources.
coal gasification, and the use of high-grade ore nuclear fission reactors. Forecasts
indicate that our economically recoverable fossil fuel reserves will be depleted prior
to 2100 A.D. High-grade ore fission reactor power reserves are uncertain-one forecast [2J predicts that this energy source will be depleted of low-cost ores by the year
2000. Wind. tide, geothermaL* and other energy sources may help ease the strain on
our fossil fuel reserves but are not expected to be major sources of energy. Long-term
abundant energy sources may be narrowed down to breeder fission reactors. fusion
reactors, and solar power, as illustrated in Fig. 2.
Breeder reactors could supply our country's energy needs for thousands of years
and nuclear fusion rivals solar power as an energy source until the end oflife on earth.
The fast breeder reactor CJ is not expected to be operational for another ten years
and the controlled thermonuclear reactor (fusion) is expected [8J to take at least
twenty-five years. Development of economical solar energy [9~11J is proceeding at
an equally slow pace. All fission and fusion reactors present certain biological and
environmental threats. The handling, disposal, and safeguarding of waste radioactive
Terawatt . Years
ENERGY SOURCE
10'
300
fOSSil fUElS
NUClEAR Iordinary]
UClEAR Ibreeder]
UClEAR'
10 2 . 02 fusion]
SOLAR RAOIA nON
ENERGY DEMAND
11960·2000 A.D.]
J3J0
100
10'
I
300II
10.000
/
10'
10'
WO
10
Reserves at Double Current Cost
300.000 \
1.6110'
'/
/.
12% Depletion of O2 irJ Sea Waler]
1.6110 10
2.8110'
11% land Area, 10% Efficiency, Earth life 10" Years]
1'c'Z:3
140 100
WlI1j R"",,, III 0, - r, ,,"'_ est""'ed " 68111 TW ·ye.,
M.to: II d_. piotto! • ertractoi Jr•• CSt., • MU.iIIert. Ser
c::::::J U.S.
c::::::J World
Im~.
llilll161Sept 19111
Fig. 2. Estimated U.S. and world
energy resources.
* At this writing, the potential of geothermal energy is subject to considerable debate. 3 -
6
Cryogenic Hz aud National Energy Needs
3
materials from breeder reactors is quite a challenge, even for our advanced society.
Tritium containment in fusion power plants appears attainable, but only solar power
emerges as an abundant, nonpolluting energy source with minimal threat to man.
There are currently two major proposals for large-scale harvesting of solar
energy-one proposes land-based solar radiation panels [12] and the other expands
on the old idea ofD'Arsonval to use solar sea power [13-15]. The latter idea uses the
temperature difference of seawater supplied on a continuous basis by solar radiation
to the oceans, to supply our energy needs. Solar energy concepts have always suffered,
as have the electrical utilities, from the inability to effectively store energy. A convenient means of storing energy is in the form of molecular hydrogen.
Relevance of Nonfossil Hydrogen
Man-produced hydrogen, or synthetic fuel, could be mass-produced by thermochemical decomposition or electrolysis of water-the required heat and/or electricity
being supplied by solar or nuclear power. Hydrogen could then be used as a synthetic
fuel to satisfy various energy markets, e.g., household, commercial, industrial, and
transportation. When all fossil fuels are gone, solar or nuclear power can thus be
used to make hydrogen. Hydrogen could be produced continuously or in some cases
during the intermittent off-peak hours for electricity consumption. Whether we use
electricity or hydrogen as the energy carrier depends on production and distribution
costs, availability of energy, environmental insult, and the nature of the energyconsuming device.
It is improbable that all future portable power and energy storage requirements
can be met with either electricity or hydrogen alone. Fossil fuels and other synthetics
will also compete in these markets for at least the next fifty years. In fifty years, the
concept of electric-powered aircraft may be more tenable and storage of electrical
energy in s\lperconducting magnets 6 ] may be common practice; however, the
technology required to produce and use hydrogen exists today. Hydrogen fuel is
appealing because it offers convenient energy storage, portable power, and reduced
air pollution.
The selective substitution of hydrogen for fossil fuels is an attractive partial
solution to our fuel shortage; however, we still need to develop an abundant, nonpolluting energy source to produce the hydrogen. Solar power, using hydrogen as an
energy storage medium is an equally attractive solution to our energy-fuel crisis. Solar
sea power is particularly attractive because there is no need for man-made solar
collectors; therefore, potentially cheaper and more reliable power may be produced
without sacrificing large tracts of land. Minimal adverse environmental impact is
anticipated with solar sea plants. With certain technological advances, off-shore
nuclear power plants [17] are also expected to be environmentally compatible energy
sources for the production of hydrogen (and/or electricity).
The production of nonfossil hydrogen increases the burden on raw energy
sources-the penalty we pay for storing and packaging energy in a convenient form.
From a pure conservation viewpoint, we could argue that nonfossil hydrogen should
be used only when a chemical form of energy is essential. But if we harness vast*
sources of energy and produce hydrogen economically, and if environmental threat is
within acceptable limits, extensive use of hydrogen is technically and morally warranted. In reality, the nonfossil hydrogen market will be limited by the cost of producing it from solar or nuclear power.
e
* Available at anticipated use rates as long as earth will support life.
J. Honi
4
CONSIDER SYNTHETIC FUELS
Although hydrogen is estimated to comprise 90 % of the universe, it does not
occur abundantly as a gas on earth, and so it must be produced synthetically. Further,
we must distinguish between fossil and nonfossil synthetic hydrogen as long as
hydrogen is produced from coal. Detailed comparisons of synthetic fuels are available
in the literature [18-23]; leading candidates are hydrogen, methanol, ethanol, ammonia, hydrazine, and synthetic natural gases (SNG). Fuels produced from farm [24]
and biological wastes are of minor significance [25]. The SNG's are source-limited,
ammonia and hydrazine are toxic, ethanol competes with food crops and is expensive.
Methanol is slightly toxic but emerges as the chieflong-term competitor* to nonfossil
hydrogen. Methanol, natural gas, and SNG can compete with nonfossil hydrogen
until fossil fuel reserves are depleted. Then the cost of methanol, synthesized from
hydrogen and limestone [21] or hydrogen and carbon dioxide from the air [18,19], is
expected to rise markedly. In the last two processes, hydrogen holds a cost advantage
because hydrogen is required to produce methanol. Hydrogen wins either way,
whether it is used in its pure form or is converted into methanol. Fossil hydrogenproduced from coal [27,28], shale, or crude oil-is expected to cost less than nonfossil
hydrogen until fossil fuel prices at least double (or until nonfossil hydrogen production
costs are halved).
Obviously, the application dictates the proper physical form of hydrogen. Roomtemperature hydrogen gas is a candidate fuel in all sectors [18,19,28-30] of the energy
market. It is a potential competitor in all current and future markets served by natural
gas and electricity. Bulk quantities of hydrogen are used by the oil refineries and in
the production of ammonia and methanol. Hydrogen is also used in the electronics,
glass-making, food, pharmaceutical, and metal-working industries. The arguments
for use of gaseous hydrogen in transportation are less convincing than those for
residential, commercial, and industrial uses. In gaseous form, hydrogen must compete
with electricity, natural gas, and SNG. Cryogenic hydrogen is attractive as a transportation fuel [21,22,30--36], has certain potential advantages in hydrogen-electric utility
systems [13,37-39], and is unexcelled in performance as an aerospace fuel. In generating
peak electrical power, the competing fuels are LNG, LSNG,t and methanol. Chief
competitors in transportation are electricity, LNG, LSNG, and methanol. "Thermodynamic thrift" will surely be sacrificed for convenience (and perhaps economy) in
many of these applications. Recent debates [40] accentuate the importance of hydrogen physical form and the competition between the gas and electric industries.
Production
CRYOGENIC HYDROGEN
Recent publications [41,42] indicate that production costs for electrical power
will be about 12 mils/kW -hr within five years. Current (unpublished) estimates indicate
that production costs will soar to 15 mils/kW-hr by 1981. Using the lower unit cost for
energy and extrapolating water electrolysis cost data [18], the projected cost of
electrolytic hydrogen is 0.21 to 0.28 S/lb. This cost could be reduced by lower cost
electricity [13] or by innovative improvements [43.44] in thermochemical decomposition [45] of water. The high-temperature gas-cooled reactor (HTGR), as described by
Quade [44], is frequently suggested as an interim source of process heat. Cost credits
• It appears that methanol will also compete [26] with LNG in the near future.
t Liquefied synthetic natural gas.
Cryogenic Hz and National Energy Needs
s
for byproduct oxygen will be negligibly small unless new, voracious markets are
identified.
Liquefaction
The cost (in 1973 dollars) to liquefy/slush/solidify hydrogen was determined
from the data of Hallett [46] and Strobridge [47] under the following terms: Plant
output of 250 to 2500 tons/day, operation 350 days/year, fixed charge rate of 12 %
on capital investment, plant efficiency @ 40% of Carnot, 12 milsjkW-hr for energy,
integral liquefaction-slush facility, liquid storage capacity of two days output, and
10 K refrigeration for solidification of hydrogen at 13 K. Liquefaction costs range
from 0.072 to 0.098 S/lb of hydrogen. Producing 50 % solid-fraction slush increases
the liquefaction cost by about 24%. Solid hydrogen can be produced from 50% solid
slush for an additional 8 % of liquefaction cost-a total increase of about 32 % over
liquefaction cost. Calculations indicate that solid hydrogen can be produced more
economically from NBP liquid (approximately 27 % increase in liquefaction cost).
From these estimates, the total cost of producing liquid hydrogen is about 0.30
to 0.37 SjIb of hydrogen-slush can be produced for an additional 0.02 SjIb. Using
the appropriate bulk densities and lower heating values for gasoline and hydrogen,
the foregoing liquid prices translate into equivalent costs of 0.63 to 0.78 Sigal for
gasoline (excluding taxes, distribution costs, and assuming identical thermal conversion efficiencies for gasoline and hydrogen). This energy cost may also be expressed
as 5.33 to 6.98 S/10 6 Btu or 18 to 24 mils/kW-hr. Escalating costs of fossil fuels
could soon offset this seemingly exorbitant price of nonfossil hydrogen.
Storage
In the foregoing estimates, storage costs of liquid hydrogen were nearly negligible; however, larger storage capacity will drive hydrogen prices up. Installed costs for
liquid hydrogen storage range from 1.00 to 2.00 Sigal (or perlite-insulated dewars of
less than 106 gal capacity. Hallett [46] estimates that perlite-insulated dewars larger
than 106 gal capacity can be constructed for approximately 0.60 Sigal and urethanefoam-insulated tanks would cost about 0.22 to 0.60 Sigal. Metal hydrides are also under
investigation [29] as hydrogen reservoirs. Most of the advantages [29] and disadvantages [22] of these hydrides have been disclosed. Though frequently touted for containment of hydrogen in the transportation sector, use of these hydrides in stationary
storage applications is more appealing. Liquefaction and storage costs [37.46] for
hydrogen are much higher than those for LNG; thus the price of natural gas can exceed
that of hydrogen gas and LNG will still be competitive with liquid hydrogen. Unfortunately, the life expectancy of natural gas is rather short.
Transmission
The concept of transmitting electrical power and cryogenic hydrogen through the
same transmission line [48] is fascinating and highly futuristic. It is certainly not
obvious at this time that superconducting or cryogen-cooled high-conducting transmission lines will ever be economical. Perhaps the multipurpose transmission line will
permit low-temperature conductors to scale this hurdle. The total cost of transferring
liquid hydrogen through long pipelines is not well known; however, the rule-of-thumb
cost for vacuum-insulated piping is about 15 to $20 per lineal foot for each inch of
inner line diameter. Liquid hydrogen is currently shipped by highway and rail [37.49]
and could easily supply packaged power to remote sites, e.g., hamlets, mountain
cabins, etc.
6
J.Honl
Applications in Utilities
Hydrogen and electricity are not always competItIve-in mixed-utility and
electricity peak-shaving concepts they are complementary. Liquid hydrogen has
been proposed for peak-shaving operations [l9] and in mixed utilities [38] where
hydrogen is the chief energy carrier. Preliminary calculations reflect low overall
energy conversion efficiencies for these systems-unacceptably low if limited energy
sources are used; however, the availability of abundant energy sources makes these
inefficiencies more palatable. In the latter case, our major concern would focus on
environmental compatibility, cost, and convenience. Then, ifnonfossil hydrogen can
be economically produced, the cost ofliquefaction is more readily borne and cryogenic
hydrogen is tenable-the aforementioned utility applications and a variety of others
are possible. Development of solar sea plants should favor the cryogenic form of
hydrogen for transmitting energy from remote ocean sites to population centers.
Similar, though weaker, arguments may be advanced for liquefying hydrogen
produced with offshore nuclear reactors. In any case, cheap production costs add
incentive to liquefy.
Appiications in Transportation
Perhaps the most promising market for cryogenic hydrogen is in the transportation area. About 25 % of the energy consumed in this nation is allocated to the transportation sector and significant opportunities for conservation have been identified
[50]. The most efficient [50-53] means of transportation are by train, bus, plane, and
auto in about that order. * Hydrogen-air turbine-driven trains, trucks, and buses are
distinct possibilities [36]. Such institutional vehicles are not so volume/weightrestricted and cryogenic hydrogen offers significant potential. Electrical propulsion
of these vehicles (and autos) by hydrogen-air fuel cells [54.55] requires further technological development-again onboard cryogenic storage is attractive.
Liquid hydrogen may find its maximum near-future potential in the aircraft
industry. Studies [34.35] indicate that cryogenic hydrogen is essential for aerodynamic
cooling of hypersonic aircraft and highly desirable for supersonic airplanes. Interest in
use ofliquid hydrogen in subsonic [56.57] aircraft has also been renewed [58]. All three
classes of hydrogen-fueled aircraft should show substantial performance gains (range,
engine life, etc.). Impetus from the federal government will be required to initiate this
fueling trend. An evaluation of climatic impact [59] due to hydrogen-fueled aircraft is
needed.
Slush hydrogen also offers some advantages as an aircraft fuel: 50 % solid slush
provides an 18 % increase in heat capacity and a 15 % increase in bulk density when
compared to 20.3 K liquid hydrogen. Handling [60] of slush hydrogen is difficult and
use of helium pressurant is unacceptable in commercial applications. Total helium
resources [61] in the U. S. are about 180 x 109 standard cubic feet. This reserve would
last but a few years if helium were used to service an aircraft system [46] requiring
8000 tons/day of slush hydrogen. Thus, new techniques for handling slush with
hydrogen pressurant may be required, e.g., the innovative use of pumps, hydrogen gas,
honeycomb antislosh baflles, and ullage liquid-var,>r screen separators.
The personal auto is the major consumer [5 ] of energy in the transportation
field. Powered by the highly developed internal combustion engine, it is also a vehicle
* This depends on the definition of "efficiency" and whether freight or passengers are being transported
(bicycles and pipelines are also highly efficient).
7
Cryogenic Hz and National Energy Needs
with stringent volume/weight restrictions. For this reason alone, liquid hydrogen is a
somewhat marginal [22] fuel, as are metal hydrides-dean burning hydrocarbons are
probably better suited for this task (see Fig. 3). The comparative data shown on Fig. 3
are based on the low heat of combustion for each fuel and do not account for the
increased energy conversion efficiency that is possible with hydrogen fuel. Development of suitable hydrogen-air fuel cell propulsion units [55] could drastically reduce
hydrogen weight and volume requirements.
The safety aspects of hydrogen-fueled autos are perplexing. Gasoline is safer
than hydrogen for autos but hydrogen is not too hazardous to use. Safety considerations are different for autos, aircraft, trains, etc. Hydrogen may well be as safe as
gasoline or kerosene with the carefully controlled environment, logistics, and skilled
technicians available to aircraft and institutional vehicles. With autos, we are dealing
with a much larger number of vehicles and unskilled personnel (owners). To debate
the safety pros and cons of hydrogen vs. gasoline is meaningless without specifying
accident criteria [62]. We can draw general comparisons on the basis of fire hazard,
fire damage, explosive hazard, and explosive damage. Gasoline loses in but one
category: fire damage. It is not clear what fraction ofthe auto industry liquid hydrogen
could or should claim. A go-slow policy seems likely in this area since natural and
alternate synthetic fuels are available.
Large ships might also find liquid hydrogen an attractive fuel. Extension of
LNG shipbuilding· and insulation technology to accommodate liquid hydrogen
appears technically feasible but challenging. LNG tankers are currently burning intransit boiloff gas to help propel the ship; exclusive use of crude oil for ship propulsion
and onboard reliquefaction of boil off natural gas are under consideration. Boiloff and
reliquefaction problems are amplified when transporting liquid hydrogen. There are
no adverse environmental effects from oceanic spillage of liquid hydrogen and
recovery of onboard combustion products would provide abundant shipboard
utility water. Shipment ofliquid hydrogen may be the best way to transport energy from
I
c:::::JlfIII'IIII)I_IIIio'
5
4
._-ta*-
~
Ifill • 1111) hIM
btio·
........rustonp·
I
I
.
,
,. ,.
3
,..
'"
"
-
1"'1
'l r
- - 11- .112
1.15
'·II'·
)C,'.)
II)
IEIIIIIl
ICI,.)
LlC
I-Cl~
P)
)2)
1.,1
II)
11111
l1II1.
14)
12)
11) .....,lI1IliII1lJ11is.
12) . . . . ' _ _ SSSlilrial_III ... ).
Fig. 3. Estimated (fuel + tank) volume and weight ratios
for candidate fuels. The ratios are based upon the low
heat of combustion for each fuel and are applicable for
comparison with gasoline tanks of 12.5 to 25 gal capacity.
II) l1li1 SlilrialIIIIs I• •·)·
II) ....._ _ _ I.,I.(21I~_SSSlilrialIllls
~
",,",,"11 %...
1'"_).• ".... -1."
IIoilll ......);_at1lll. ....... o1E.
lI1I:lJss..- .... IIII1,,_ .. _ . _ _ 1Iis ....
).l- ........, . . . . . . . . ., . . . ., . .).IIsI. ..... n ...... IIr .. IIIIIsIIIl."_ .... is ..... IIr ... IIIIIs~
* Transportation of natural gas by airship has also been suggested [63].
8
J. Honl
solar sea plants. Development ofthese plants could initiate a liquid hydrogen shipping
industry.
Small, volume/weight-limited portable power units, e.g., motorbikes, motorboats, lawnmowers, etc., are likely to continue operating on gasoline and eventually
switch to methanol.
Aerospace
Liquid hydrogen is the "standard" fuel and slush [60] and solid [64] hydrogen are
candidates for future deep-space probes. In this limited application, the additional
expense- for slush or solid is immaterial and the frugal use of helium gas is justified.
NASA has scheduled the Space Shuttle for 60 flights/year, carrying 14 x 106 lb/year
of liquid hydrogen into space-a mere dribble compared to potential aircraft consumption rates.· The latter is also about double the 1968 national production [65] of
11 x 109 1b of hydrogen.
RESEARCH EFFORT REQUIRED
Opportunities for technological improvements are abundant. Development of
solar and nuclear energy sources and coal gasification techniques are essential. The
minor energy sources-wind, oceans (tide, waves, currents), geothermal, etc.should not be neglected. An inexpensive and environmentally compatible method for
producing non fossil hydrogen is needed. Application-oriented issue studies are
needed to evaluate alternate fossil and nonfossil synthetic fuels.
Detailed studies are needed to evaluate the benefits of cryogenic hydrogen in
proposed applications-currently concentrated on mixed utilities, peak-shaving, and
transportation. In the well-developed cryogenic field, we can strive for higher efficiencies and lower costs in liquefaction, transmission, and portable and large-scale
stationary storage. Strong, light-weight structural materials and improved insulation
materials for cryogenic service are sought. High-efficiency energy conversion techniques are needed, e.g., the development of hydrogen-air and hydrogen~xygen fuel
cells and gas turbines as well as hydrogen-filled catalytic heaters. Fueling of all
classes of aircraft, institutional vehicles, and autos with liquid hydrogen should be
meticulously analyzed and performance-evaluated where (and when) it is feasible.
Hydrogen-powered ships and transoceanic transport of liquid hydrogen should be
examined. Research efforts on mUltipurpose hydrogen--electric transmission cables
and superconducting energy storage systems can be intensified.
New markets for byproduct oxygen should be identified. Thermophysical
properties of mixtures (hydrogen and methane from coal) are needed and applications
for LSNG should be investigated. Data on hydrogen embrittlement of steels should be
compiled and analyzed and further research performed-particularly as it relates to
transmission of hydrogen and hydrogen-hydrocarbon mixtures through existing
natural gas piping networks. Additional work on metal hydride storage seems worthwhile and compilation and analysis of hydrogen safety data should continue.
Substitution of hydrogen for natural gas in pipelines raises odorization, contaminant detection, and leakage problems. Leakage is also a factor for consideration
in cryogenic hydrogen systems. It has been shown [66] that leakge is inversely proportional to the square root of the density or to the absolute viscosity of the fluid.
• Aircraft
COBSumed
the equivalent of 24 x 109 lb of hydrogen in 1970.
Cryogenic H2 and National Energy Needs
9
Thus, volumetric leakage flow of hydrogen gas will be 1.25 (viscous flow) to three
times as large as methane leakage (at the same temperature and pressure).
Economics, environmental impact, and energy conservation cannot be overemphasized in performance of these studies.
TRANSITION TO SYNTHETIC FUELS
Government leadership and funding are required to accelerate the transition to
synthetics. A national agency, and national goals and commitments are needed.
Increased imports result in higher fuel prices, potential threat to national security,
and staggering economic deficits [67]. With universities and industry cooperating
with government on an intensive national effort, synthetic fuels could soon be implemented. A similar national effort placed man on the moon-a national fuels effort
offers even greater rewards.
Under current circumstances, any timetable for transition to synthetic fuels
must expose personal bias. The time schedule depends upon (1) the resourcefulness of
government and industry leaders, (2) interim solutions, e.g., practiced conservation
and increased success in discovery of domestic fossil fuels, (3) the price paid by the
consumer for enforcement [68J of environmental control standards, i.e., the future
cost of synthetics vs. fossil fuels, and (4) technological developments in the fields of
solar power, nuclear fusion, breeder fission, shale oil, and coal gasification/liquefaction. SNG's from coal [27.69] should be available in quantity by 1985-their earlier use
being limited by competition with domestic natural gas and imported LNG and by
capital investment considerations. Commercial development COJ of shale oil, tar
sands, and coal liquefaction should proceed at about the same pace.
Nonfossil hydrogen will not help alleviate the fuel shortage until inexpensive,
abundant energy sources are developed. Currently, the only prospect is for operational breeder reactors by the mid-1980's. Thus, it appears that a major nonfossil
hydrogen market could not develop prior to 1990. In the interim, fossil hydrogen
from coal could pave the way for future quantity usage of nonfossil hydrogen. With
increased public awareness, appropriate funding and management, frank technical
appraisals, and application of the vast talents and ingenuity of American scientists
and engineers, this time schedule may be shortened. A challenge is issued to the cryogenics industry to develop competitive means of satisfying specific national fuel
requirements with LNG, LSNG, and ultimately liquid hydrogen.
CONCLUDING REMARKS
The scientific and engineering disciplines for large-scale liquefaction, storage,
and handling of hydrogen evolved through our military and space efforts. We are
fortunate to have over twenty years of experience with liquid hydrogen-an appealing
candidate in many future fuel concepts. Modifications of existing technology to tailorfit new specific fuel requirements are well within the capabilities of the cryogenics
industry. So equipped, this industry is in a strong position to meet its responsibilities
and simultaneously has the opportunity to help solve national fuel needs through
cryogenic technology.
REFERENCES
I. D. A. Dreyfus, "Federal Energy Organization," Staff Analysis pursuant to Senate Resolution 45,
Serial No. 93-6 (92-41), Washington, D.C. (1973).
10
J.Hon!
2. M. K. Hubbert, Sci. Am., 224(3):68 (1971).
3. D. W. Brown, M. C. Smith, and R. M. Potter, "A New Method for Extracting Energy from 'Dry'
Geothermal Reservoirs," Los Alamos preprint, LA-DC-72-1157, Los Alamos, New Mexico (1972).
4. A. L. Austin, G. H. Higgins, and J. H. Howard, "The Total Flow Concept for Recovery of Energy
from Geothermal Hot Brine Deposits," Rept. UCRL-51366, Livermore, California (1973).
5. "Summary of Seminar on Geothermal Energy," held at United Nations Headquarters, January 8-10,
1973, New York.
6. W. H. Comtois, "Economics of Electric Power Generation Utilizing Geothermal Energy," presented
at the Am. Power Conference, Chicago, Illinois (1973).
7. A. M. Weinberg, Mech. Engr., 95(6): 14 (1973).
8. R. F. Post, Physics TotkIy, 26:31 (1973).
9. L. P. Gaucher, Mech. Engr., 94(8):9 (1972).
10. F. Daniels, Mech. Engr., 94(9): 16 (1972).
II. H. R. Hay, Mech. Engr., 94(10) 24 (1972).
12. A. B. Meinel and M. P. Meinel, Physics Today, 25:44 (1972).
13. C. Zener, Physics Today, 26:48 (1973).
14. J. D. Isaacs and R. J. Seymour, Intern. J. Environ. Studies, 4:201 (1973).
15. W. E. Heronernus, "Ocean Sited Power Plants," Rept. NSF/RANN/SE/GI-34979/PR/73/1, University
of Massachusetts, Amherst, Massachusetts (1973).
16. W. V. Hassenzahl, J. D. Rogers, and T. E. McDonald, "Magnetic Energy Storage and Its Application
in Electric Power Systems," presented at IEEE 1973 Intern. Conference, New York, Mareh 1973.
17. G. D. Friedlander, IEEE Spectrum, 10(2):44 (1973).
18. J. W. Michel, "Hydrogen and Other Synthetic Fuels," USAEC, TID-26 I 36, September 1972.
19. D. P.Gregory, D. Y. C. Ng, and G. M. Long, in: Electrochemistry of Cleaner Environments (J. O'M.
Bokris, ed.), Plenum Press, New York (1972), p. 226.
20. T. D. Weikel, "Ground Support Equipment: Low Pollutant Fuels," Rept. NAEC-GSED-59 (1972).
21. G. V. Day, Futures 4:331 (1972).
22. A. L. Austin, "A Survey of Hydrogen's Potential as a Vehicular Fuel," Rept. UCRL-5 I 228, Livermore,
California (1972).
23. F. Bacon and T. Fry, New Scientist 55(808):285 (1972).
24. K. Kiang, H. F. Feldmann, and P. M. Yavorsky, "Hydrogasification of Cattle Manure to Pipeline
Gas," presented at the 165th Nat. Meeting, Am. Chern. Soc., Dallas, Texas, April 1973.
25. L. L. Anderson, "Energy Potential From Organic Wastes: A Review of the Quantities and Sources,"
Bur. Mines Infor. Cire. IC-8549 (1972).
26. M. J. Royal and N. M. Nimmo, Oil and Gas J., 71(6):52 (1973).
27. "Clean Energy From Coal-A National Priority," 1973 Annual Report, Office of Coal Res., U.S.
Dept. of Interior, Washington, D.C. (1973).
28. W. E. Winsche, K. C. Hoffman, and F. J. Salzano, in: Proceedings of 7th IECEC,· ACS, Washington,
D.C. (1972), p. 1366.
29. R. H. Wiswall, Jr. and J. J. Reilly, in: Proceedings of 7th IECEC, ACS, Washington, D.C. (1972), p.
1342.
30. K. H. Weil, in: Proceedings of 7th IECEC, ACS, Washington, D.C. (1972), p. 1355.
31. R. J. Schoeppel, Chemtech, 2:476 (1972).
32. W. J. D. Escher, in: Proceedings of 7th IECEC, ACS, Washington, D.C. (1972), p. 1392.
33. L. W. Jones, in: Proceedings of 7th IECEC, ACS, Washington, D.C. (1972), p. 1364.
34. R. D. Witcofski, in: Proceedings of 7th IECEC, ACS, Washington, D.C. (1972), p. 1349.
35. A. A. duPont, in: Advances in Cryogenic Engineering, Vol. 12, Plenum Press, New York (1966), p. I.
36. Washington Science Trends, 30(8):45 (1973).
37. J. R. Bartlit, F. J. Edeskuty, and K. D. Williamson, Jr., in: Proceedings of 7th IECEC, ACS, Washington, D.C. (1972), p. 1312.
38. E. C. Tanner and R. A. Huse, in: Proceedings of 7th IECEC, ACS, Washington, D.C. (1972), p. 1323.
39. D. P. Gregory, Sci. Am., 228(1): 13 (1973).
40. J. H. Chiles III, Sci. Am., 2211(4):8 (1973).
41. "The 1970 National Power Survey," Parts I and IV, Federal Power Commission, Washington, D.C.
(1971).
42. A. M. Weinberg, Science, 177:27 (July 1972).
43. G. De Beni and C. Marchetti, Euro-spectra, 9(2):46 (1970).
44. R. N. Quade, Power Engr., 77(4):50 (1973).
45. J. E. Funk and R. M. Reinstrom,l & EC Process Design and Development, 5(3):366 (1966).
• Intersociety Energy Conversion Engineering Conference.
Cryogenic Hz and Natiooal Energy Needs
11
46. N. C. Hallett, "Study. Cost. and System Analysis of Liquid Hydrogen Production." NASA Rept. CR73226 (1968).
47. T. R. Strobridge. "Cryogenic Refrigerators-An Updated Survey." NBS Tech. Note. to be published.
48. J. R. Bartlit and F. J. Edeskuty. in: Proceedings of 4th Intern. Cryogenic Engineering Conference.
IPC Sci. and Tech. Press. Guildford. England (1972). p. 177.
49. F. A. Martin. in: Proceedings of 7th IECEC. ACS. Washington. D.C. (1972). p. 1335.
50. "The Potential For Energy Conservation." A Staff Study. Office of Emergency Preparedness.
Washington. D.C. (October 1972).
51. C. W. Savery. Traffic Quarterly. 1972 (October). 485.
52. W. P. Goss and J. G. McGowan. Transportation. 1(3):265 (1972).
53. E. Hirst. "Energy Consumption for Transportation in the U. S.... Rept. ORNL-NSF-EP-15. Oak
Ridge. Tennessee (March 1972).
54. Y. Breelle. J. Cheron, and A. Grehier, in: Proceedings of 7th IECEC. ACS, Washington, D.C. (1972).
p. 1.
55. K. V. Kordesch, J. Electrochem. Soc., 118(5):812 (1971).
56. R. C. Mulready, in: Technology and Uses of Liquid Hydrogen (R. B. Scott. W. H. Denton. and C. M.
Nicholls, eds.), Macmillan, New York (1964), p. 149.
57. "Hydrogen for Turbojet and Ramjet Powered Flight." Staff Rept. NACA RM E57D23 (April 1957).
58. R. D. Witcofski, Langley Research Center. NASA. private communication.
59. A. Goldburg, Astro. and Aero., 10(12):56 (1972).
60. C. Sindt, Cryogenics 10:372 (1970).
61. W. M. Deaton and P. V. Mullins, in: Technology of Liquid Helium (R. H. Kropschot, B. W. Birmingham, and D. B. Mann, eds.), U. S. Govt. Printing Office, Washington. D.C. (1968), p. 4.
62. J. Hord, "Explosion Criteria for Liquid Hydrogen Test Facilities," unpublished NBS Report (February 1972).
63. M. H. Sonstegaard, Mech. Engr., 9S (6): 19 (1973).
64. J. Hord, "Solid Hydrogen as a Space Storable Propellant-A Preliminary Study," unpublished NBS
Rept. (March 1972).
65. P. Meadows and J. A. DeCarlo, in: Mineral Facts and Problems, U. S. Bureau of Mines Bulletin 650.
Washington, D.C. (1970), p. 97.
66. J. Hord. "Correlations for Predicting Leakage through Closed Valves," NBS Tech. Note 355 (1967).
67. P. G. Peterson, "Energy Research-The Key to Our Long Range Energy Future," presented to the
55th Anniversary Convention of the Nat. Coal Assoc., Washington, D.C., June 1972.
68. N. de Nevers, Sci. Am., 228(6): 14 (1973).
69. S. A. Bresler and J. D. Ireland, Chem. Engr., 79(23):94 (1972).
70. W. D. Trammell, Chem. Engr .• 80(10):68 (1973).
71. R. D. McCarty and L. A. Weber, "Thermophysical Properties of Parahydrogen from the Freezing
Liquid Line to 5000 R for Pressures to 10,000 Psia," NBS Tech. Note 617 (April 1972).
Pertinent References Not Cited in Text
72. J. G. McLean and W. B. Davis, "Guide to National Petroleum Council Report on United States
Energy Outlook," Nat. Petroleum Council, 1625 K. St. N. W., Washington, D.C. (December 1972).
73. A. L. Austin, B. Rubin, and G. C. Werth, "Energy: Uses, Sources, Issues." Rept. UCRL-51221,
Livermore, California (1972).
74. M. McCormack, "Energy Research and Development," Rept. of the Task Force on Energy, 92nd
Congress, Serial EE, Washington. D.C. (December 1972).
75. "National Gas Supply and Demand 1971-1990," FPC S-218, Staff Rept. No.2, Bur. of Nat. Gas,
Federal Power Commission, Washington, D.C. (February 1972).
76. V. D. Arp, A. F. Clark, and T. M. Flynn, "Some Applications of Cryogenics to High Speed Ground
Transportation," NBS Tech. Note 635 (February 1973).
77. W. Berry, R. Calleson, J. Epsil, C. Quartero, and E. Swanson, "A Fuel Conservation Study for
Transport Aircraft Utilizing Advanced Technology and Hydrogen Fuel," NASA CR-112204 (November 1972).
78. B. M. Abrahim and F. Schreiner, Science, 180(4089):959 (1973).
79. "An Assessment of Solar Energy as a National Energy Resource," prepared by the NSF/NASA
Solar Energy Panel, Dept. of Mech. Engineering, University of Maryland, College Park, Maryland
(December 1972).
80. S. Weiss, "The Use of Hydrogen for Aircraft Propulsion in View of the Fuel Crisis," NASA TMX68242 (March 1973).
81. R. S. Lewis and B. I Spinrad (eds.), The Energy Crisis, Science and Public Affairs, Bulletin of the
Atomic Scientists, Chicago, Illinois (1972).
A-2
THE ECONOMICS OF LIQUID HYDROGEN
SUPPLY FOR AIR TRANSPORTATION
J. E. Johnson
Union Carbide Corporation, Linde Division
New York, New York
INTRODUCTION
The growing energy crisis within the United States is causing energy-consuming
industries to focus on the problems of supplying their specific energy needs. Hydrogen
has been suggested [1] as a particularly suitable fuel to power our future society, which
demands, in addition to efficient energy supply, an avoidance of environmental
degradation. As with the production of any synthetic fuel, the conversion of energy
from some other less desirable form is required to make hydrogen's clean energy
available. Hence hydrogen should be viewed as an energy carrier (as is synthetic
natural gas and electricity). Significant fuel reductions have been reported [2] when
energy conversion devices such as internal combustion engines, fuel cells, and
catalytic heaters have been operated on hydrogen. Previous studies [3] have shown
that large-scale projects can produce liquid hydrogen at prices that could make this
fuel competitive if appropriate load factors and low-cost energy sources are available.
The purpose ofthis presentation is to identify the prospects and analyze the economics
for an early application of liquid hydrogen which could substantially contribute toward easing the fuel shortage.
The near-term problem in supplying sufficient energy to meet our society's needs
is an increasing dependence on foreign oil imports (see Fig. 1). The problem is also
being exacerbated by insistence on stemming environmental degradation which has
been caused largely by the extravagant use of energy, particularly in the transportation
area. The U. S. has considerable reserves of coal and nuclear energy (see Fig. 2) but
these are unsatisfactory fuels to directly power transportation systems. In order to
make this energy available to serve transportation, it must be converted to a more
suitable form, by converting coal or nuclear energy to a synthetic fuel.
Although hydrogen has been recognized for its potential as a universal fuel, its
near-term benefits are likely greatest in aircraft operation because there is a significant
opportunity to increase the amount of transportation that can be obtained per unit of
energy consumption, principally because of the low density of hydrogen. The air
transport industry is ideally suited to introduce the benefits of the efficient and
nonpolluting energy conversion potential of hydrogen. The industry has highly
qualified personnel, is extremely safety conscious, and conducts its operations in large
discrete segments, which will reduce logistics problems of introducing a new fuel.
There have been fifteen years of excellent experience in the production, distribution, and utilization of liquid hydrogen to draw upon [4] which has proven that
liquid hydrogen can be handled safely. Over a billion gallons have been transported
12
The Economics of Liquid Hydrogen Supply for Air Transportation
13
TRILLION BTU
_
~
~
1970
1975
1980
1985
Fig. l. Projected relation between imported
oil and domestic production.
Proved
Possible
Speculative
Relative Energy Reserves
total U.S. oil possible
1
total U.S. coal reserves
80
$10/1b Uranium
23
$50/1b Uranium
400
Fig. 2. Relative energy reserves of the U.
s.
with an excellent safety record. In examining the overall safety aspects, one finds in
many respects that liquid hydrogen is safer than other, less volatile hydrocarbon
fuels. Today, liquid hydrogen is commonly transported anywhere in the U. S. by railcar and tractor trailer to customers requiring service. The accomplishments of the
Apollo program are an adequate demonstration that liquid hydrogen can be handled
safely.
In order to evaluate the economics of liquid hydrogen supply, two systems were
analyzed-one originating with coal, the other from nuclear energy. To assess their
competitiveness, they are scaled from similar energy conversion systems [5] that are
currently being developed to provide additional conventional energy resources.
CONVERSION OF COAL TO HYDROGEN
The abundant U. S. domestic coal reserves that must be developed to fuel our
economy for the coming century are well distributed geographically. Although
there have been environmental objections to processing these reserves, they can be met
economically by adequately engineered systems.
Present studies indicate that the preferred system, as shown in Fig. 3, for conversion of coal to gaseous hydrogen would be accomplished near the mine. Polluting
materials, eliminated in solid form., if of no value could be returned to the mine. The
process of producing hydrogen improves the economics of pollutant removal because
the separation can be accomplished before dilution ofthe undesirable elements in the
combustion products and nitrogen. There would be no smoke and only carbon dioxide
would be vented. Gaseous hydrogen would then be piped to the airport for liquefaction. The conversion and storage of the liquid hydrogen would be accomplished in a
facility adjacent to the airport. In the case analyzed, the liquefaction power is obtained
by reacting part of the hydrogen with oxygen to produce steam directly for driving a
J.E.Joo..-
14
co,
GAS
PIPELINE
I
I
I
I
I
I
~~~~~~
________________ JI
AIRPoR'T---------- --- ----------------- --,
I
I
TO
~~~
LIQUID HYDROGEN
H1
I
I
I
I
STEAM
Fig. 3. Liquid hydrogen from coal.
turbine. The development of power in this manner, to be competitive with conventional drives, is dependent on exploitation of the high efficiency potential that has been
projected [6] for direct hydrogen-oxygen combustion systems.
CONVERSION OF FISSION ENERGY TO HYDROGEN
With the imminent development of breeder reactors, the U. S. will considerably
enlarge the fissionable reserves and energy production capability of its uranium
resources. The process to convert nuclear energy to liquid hydrogen is shown in Fig. 4.
In this case, it is assumed that hydrogen would be produced from nuclear energy by
electrolyzing desalted sea water and liquefying at a nearby "Nuplex" located offshore
for environmental considerations. The hydrogen is shipped by barge to the airport.
Storage facilities would be maintained at the airport to receive and distribute the
hydrogen.
0,
NUCLEAR
PLANT
A'R"ORT-----l
I
I
TO
PLANES
Fig. 4. Liquid hydrogen from
nuclear energy.
The Economics of Liquid Hydrogen Supply for Air Transportation
15
LIQUID HYDROGEN COST
Analysis was based on a 2500 tons/day capability in order to develop economics
on a scale of operations comparable to contemporary fueling systems. The specifications and investment for these two systems are summarized in Table I, and the projected liquid hydrogen costs are developed in Table II as functions of various fuel costs
and financing plans. The liquid hydrogen costs developed in Table II are on standalone economics based on current and near-term technology without benefit of any
credits that might be available due to:
1. Synergistic system benefits (power complexes, load sharing, off-peak energy
utilization, multimode transmission systems).
2. Supplemental hydrogen productions from "waste fuels" (garbage, oil refinery
residuals, char residues from coal conversion facilities).
3. Uses of obsoleting energy distribution systems (natural gas pipelines, urban
fossil-fired power stations).
4. Tax or other incentives that might become available to encourage investment
in the development of new domestic fuel capability.
These systems benefits, plus presently unforeseen technological developments,
will undoubtedly reduce the costs projected and accelerate closing of the hydrogen/
hydrocarbon cost ratio as the application is developed.
Table I. Investment for 2500-TonjDay Liquid Hydrogen Supply System
Investment,
million dollars
Coal conversion process
Coal requirements
Coal conversion plant
Pipeline, H2 gas
Compressor stations
Refrigeration/air compression
Oxygen generators
H 2 liq uefiers
H2 storage tanks
Distribution area
31.000 tons/day (12.000 Btu/lb). 31.000 MM Btufhr
Four 400 million cfd H2 gas generators, elf. = 0.58
One 300 mile, 36 in. diameter, 900 psi
Two 12,000 Hp ea. gas drive
Ten 125,000 kW, steam drive, elf. = 0.65
Ten 1200 ton/day cold boxes
Ten 250-ton/day cold boxes
Ten 12,500,OOO-gal flat bottom tanks
Five filling stations per tank
Total investment
Nuplex---electrolysis process
Nuclear energy requirements
Nuclear plants
Desalination plant
Electrolysis units
Refrigeration compressors
H 2 liquefiers
Barges, H2 liquid
H 2 storage tanks
Distribution area, docks
6200 MW, 64,000 MM Btu/hr
Six 1033-MW reactors, elf. = 0.33
Six one million gal/day units
Six 175 MM cfd H 2 , 820,OOOkW ea. train, I atm
pressure
Ten 125,000 kW, electric drive
Ten 250-ton/day cold boxes
Eight I,IOO,OOO-gal barges, one day turnaround
Ten 12,SOO,OOO-gal flat bottom tanks
Total investment
$500
147
3
250
50
100
63
5
$1118
$2188
5
220
125
100
20
63
IS
S2736
Total'
(cumulative energy cost)
Liquid hydrogen delivery cost
Fixed charges
Operating cost
Total 6
(cumulative energy cost)
Hydrogen liquefaction cost
Fixed charges
Energy cost
Total'
(cumulative energy cost)
Hydrogen pipeline cost
Fixed charges
Energy cost
2.02
1J29
0.60
0.52
0.15 9
0.90
0.14
0.01
1.29
0.60
0.69
0.15
0.14
0.01
2.48
1.19
1.62
0.91
0.71
0.23
0.22
0.01
0.83
0.71
1.049
50.71
0.12
2.85
1.23
1.009
1.79
0.91
0.88
0.23
0.22
0.01
0.95
3.30
1.51
1.28
0.33
--
50.71
0.24
0.33
0.17
0.17
SO.33
SO.l7
S8/ton
Refinery/chemical
14/ ton
10.47
0.24
Hydrogen gas production cost
10.47
Fixed charges
Energy cost
0.12
Total'
0.59
(cumulative energy cost)
0.75 9
0.33
0.17
0.33
10.33
18/ton
10.17
14/ton
Utility
0.17
TotaP
(cumulative energy cost)
Energy production cost
Fixed charges
Fuel costs
Financial rules 2
Fuel cost
Coal energy
0.32
0.05
0.27
1.05
3.21
4.58
4.26
--
0.32
0.73 9
3.21
0.38
0.05
0.33
1.25
0.32
0.93
4.03
5.66
5.28
4.03
--
10.32
3.71
0.88
(O.9¢/kWe)
0.69
(0.7¢/kWe)
10.32
2.89
10.59
10.29
3mils/kWe
4.70
0.45
0.07
0.38
1.55
6.70
6.25
0.49 9
1.06
4.70
10.49
4.21
1.00
(1.O¢/kWe)
10.90
10.10
0.50
0.07
0.43
1.76
0.49
1.27
5.52
10.49
5.03
7.78
7.28
5.52
1.19
(1.2¢/kWe)
10.90
10.29
3mils/kWe
Refinery/chemical
1 mil/kWe
Nuclear energy
10.59
10.10
1 mil/kWe
Utility
Table II. Liquid Hydrogen Cost Analysisl
2.24
11.7¢/lb
$1.76
0.48
2.24
$0.11
0.11
0.229
2.72
14.1¢/lb
$1.76
0.96
2.72
$0.11
0.13
-0.24
Stream time
Hourly fixed
charge
calculation
8%yr
3%yr
2%yr
20%yr
7%yr
0.0026 % Inv./hr
=
0.0017 % Inv./hr
15yr
320 days/yr
4%yr
3%yr
2%yr
14%yr
= 5%yr
Refining
chemical
345 days/yr
Depreciation
20yr
Average interest
and return
Maint. and labor
Taxes, insurance
Total
Utility
Notes:
1. Unit costs in $/\06 Btu (net heat value basis).
2. Fixed charges calculated as percent of investment
from Table I as follows:
¢/Ib
Total hydrogen system cost 8
Fixed charges
Fuel cost
Total S/MM Btu
Liquid hydrogen distribution
cost
Fixed charges
Losses
Total"
(cumulative energy cost)
3.17
3.65
19.0¢/lb
$2.69
0.96
3.65
$0.16
0.18
0.34
4.94
25.6¢/lb
$4.31
0.63
4.94
$0.11
0.25
0.36
3. Nuclear heat not competitive irrespective of
finance plan.
4. Coal gasification plant is a secure investment since it is convertible to other services
-some utility financing and S4 + coal.
hydrogen generation cost-budget:
SO.90/MM Btu
5. Gas pipelining cheaper than bargingliquefier should be at airport; utility
financing possible for pipeline.
hydrogen transportation cost-budget:
SO.15/MM Btu
6. Liquefier is a single purpose facility-utility
financing possibly not available. If public
utility nuclear electric power available, use
electric drive.
hydrogen liquefaction cost-budget:
SI.20/MM Btu
16.5¢/lb
$2.69
0.48
3.17
$0.16
0.16
0.329
6.08
7.23
37.6¢/lb
$6.59
0.64
7.23
$0.16
0.37
0.53
8.37
43.5¢/lb
$6.59
1.78
._8.37
$0.16
0.43
-0.59
7. Distribution cost-part utility (airport),
part .industrial airline financing.
hydrogen distribution cost-budget:
SO.25/MM Btu
8. Overall systems analysis-note hydrogen
cost sensitivity greatest with investment
and finance plan-fairly insensitive to fuel
cost variance. Conversion of nuclear heat
to hydrogen must be accomplished with
somewhat more than double the efficiency
of electrolysis to become competitive-thermal splitting techniques [8] may offer
this potential.
9. Best estimated hydrogen cost-budget (sum
of items in notes 4-7): S2.50/MM Btu.
31.6¢/lb
$4.31
1.77
-6.08
$0.11
0.31
0.42
18
J.E.Johnson
ECONOMIC ANALYSIS
The conversion of commercial aviation to kerosene fueling in the late 1950's was
economically fortuitous. Kerosene has been extensively available and has generally
been part of the petroleum products not easily marketable to other customers. However, the demand for clean fuels to meet the environmentally restricted energy requirements is now causing supply shortages. The automotive, home heating, peak power
generation, and air transport requirements of the country are forcing increased
competition to obtain the available fuel supplies. Certainly, those consumers whose
fuel dollar contains a large segment of distribution and storage cost (particularly
automotive fuels and peak power generation requirements) will pay premiums to
purchase the more easily distributed hydrocarbon fuels. Prices for turbine fuel must
rise significantly, then, as present pressures act to eliminate these formerly excess
hydrocarbon fuels from the bulk energy markets, and divert them to the more highly
distributed markets.
Hydrogen, on the other hand, can be produced from the lowest cost energy
sources available. From the data presented in Table II, coal or abundant carboncontaining residues (shale oil, garbage, byproduct chars) are likely to be the source of
hydrogen fuel. As recoverable coal resources are economically depleted, nuclear
energy will provide the hydrogen, and then, as other, more distantly projected resources are developed (geothermal, fusion, solar, etc.), they, in economic turn, will be
available to produce hydrogen fuels. In this manner, then, the cost of a constant-quality
hydrogen fuel will be pegged to the lowest cost energy resources and hence will not be
subject to the price elasticity and quality variation that will exist for hydrocarbon fuels.
Alternative fuel routes must be examined which might better satisfy air transportation requirements than liquid hydrogen. Clearly, without developing a source of
carbon there is no other readily available energy carrier other than hydrogen which
would be available in exploiting nuclear energy reserves. Ammonia, besides being a
toxic pollutant, is too heavy to be satisfactory. Coal, in addition to generation of
hydrogen, could be converted to methane, methanol, or a synthetic hydrocarbon fuel.
Production of any of these commodities generally requires conversion of the coal or
part ofthe coal first to synthesis gas by the reaction
C
+ H 20
+ heat
ICO
+ H2
The synthesis gas may then subsequently be reacted to produce the desired fuels by
one of the following routes:
CO
+ 3H 2 ~CH4 + H 20
CO
+ 2H2 ~CH30H (methanol)
(methanation)
There are other possible routes to convert the carbon in coal to more suitable
fuels (partial carbon-hydrogen reactions, partial distillation of coal) but it is a requirement that hydrogen be generated in order to get the total carbon resource to
market in an environmentally acceptable form. Since hydrogen must be produced as
the initial step, conversion to carbon carrier "hydrides" only contributes to further
The Economics of Liquid Hydrogen Supply for Air Transportation
19
inefficiency and resource consumption. Interestingly, the differences in bulk distribution costs of these respective fuels are minor
and not significant in fuel selection,
particularly in a high-throughput, steady-load requirement that air transportation
offers.
The cost of liquefying the hydrogen is the factor that prevents immediate
rationalization of hydrogen as the superior cost and resource efficient energy carrier
for application to aviation fuel. Methane, too, must be liquefied for aviation use but it
has specific weight and combustion characteristics that are not too much different
from kerosene. Methanol is already a partially oxidized fuel and because of its weight,
will not likely become a superior aviation fuel. Synthetic liquid hydrocarbon fuel
(which is feedstock restrictive and technically difficult to produce efficiently) becomes
too valuable in highly distributed applications to clearly settle on this option for the
bulk fuel air transportation market.
Although hydrogen must ultimately be the most economical fuel (because our
fossil carbon resources are finite) it is not necessary for hydrogen to achieve cost
parity with other fuels in order to gain acceptance. Consumers that can convert
hydrogen energy to work more efficiently than hydrocarbon energy will opt for this
route preferentially to minimize total fuel purchase, not fuel unit price. The reported
aircraft benefits that are uniquely derived from hydrogen fuel make this application a
very likely candidate for initial adoption on its economic merit alone. The opportunities that liquid hydrogen fuel provide to improve aircraft operation by the elimination
of design constraints imposed by kerosene fuel open up possibilities with regard to the
following factors:
n
1. Lighter weight structure.
Less fuel weight and fewer engines to support.
Lighter landing gear, braking systems.
2. Improved engine performance.
Utilization of the refrigeration energy of a cryogenic fuel.
Improved combustion.
Greater design flexibility for emission reduction.
Higher efficiency for any prescribed NO x specification.
Lower maintenance, less erosion, no coking, sulfidation.
3. Improved operation
Greater range or increased payload per unit air frame weight.
Shorter runway.
Faster takeoff, noise suppression tradeoffs.
Higher economical altitude.
Sonic boom reduction.
In accelerating the timing for liquid hydrogen applications, the very complex
technical, economic, and environmental arguments for hydrogen fuel adoption must
be properly coordinated. Acceptance of liquid hydrogen will be benefited by four
factors:
1. The extraordinary inflation in cost of liquid hydrocarbon fuels, which will
reduce the present fuel cost disparity.
2. Technological improvement in the production and distribution of liquid
hydrogen, which will accelerate and broaden the opportunity for application.
3. Identification of ancillary aircraft benefits (reduced engine maintenance, less
pollution), which would encourage consumers to pay a premium for hydrogen
J.E.Johnson
20
energy and cause the acceptance at even less than overall fuel cost parity.
4. Optimization of the relative efficiency improvements possible with liquid
hydrogen fuels, which will reduce overall fuel costs.
It is the demonstration of reduced aircraft fuel consumption which remains the
most meaningful technical contribution in developing a persuasive argument for
early acceptance ofliquid hydrogen fuel on economic merit alone. This potential and
the pressure of decreasing hydrocarbon turbine fuel availability are the factors that
could vector an acceptance campaign over the shortest course to success.
A red uction of the relative hydrogen fuel and aircraft specific energy consumption
in the range of 40 %, plus the imminent reduction of the hydrogenjhydrocarbon fuel
cost ratio to less than two would make the prospects for hydrogen fuel economically
viable in the near-term. This relationship is portrayed more clearly in Fig. 5. Demonstration of the combined potential oflarge-scale, low-cost liquid hydrogen availability
and reduced aircraft energy consumption (and other operational savings) could also
provide substantial leverage in containing and controlling overdependence on imported fuels, with its implied and real threats to both the economy and security of
the nation.
IMPLEMENTATION
The essential demonstration to prove the viability of application of liquid hydrogen fuel for air transportation is that the efficiency and operating improvements
projected for its use are safely obtainable. The introduction ofliquid hydrogen aviation
fuel as a successful economic venture requires attention in the following four areas:
1. Technical demonstration of the benefits of hydrogen fuel usage.
2. Evaluation of efficiency and performance improvements in various engine and
air frame concepts.
3. Identification of economical route structures for commercial introduction of
hydrogen fuel.
4. Advance planning to accommodate liquid hydrogen facilities at airports.
Technology to convert and fly existing aircraft with liquid hydrogen is available
(it was done in 1957). The significant technology advances achieved in the Apollo
program make this now more feasible and will provide the essential components
(lightweight tanks, fuel transfer systems, engine conversion techniques, principles of
.4
r-------------------------------,
.5
c: c:
o
0
"g,.~
.6
E E
""
88
~
~
.7
a:;"ii
"
" .8
U.N~
J:J:
.9
1.0
~
_____
1.5
2
Hydrogen Fuel Cost/Hydrocarbon Fuel Cost
Fig. 5. Hydrogen fuel cost/hydrocarbon
fuel cost.
The Economics of Liquid Hydrogen Supply for Air Transportation
21
safe design). This should permit early initiation of the testing required to prove component airworthiness and develop additional know-how.
The marketability of a hydrogen-fueled aircraft should occur first in those
application concepts that offer the greatest economic improvement to meet existing
or developing air transport market needs. It would seem that the low weight and
cryogenic properties of hydrogen would benefit long-haul or high fuel-weight/payload
missions the most. The development and adoption of a liquid hydrogen aviation fuel
capability which is based on a unique and exportable domestic technological resource
will also permit continued dominance of the world aircraft markets and provide a
credit rather than a fuel debit in meeting critical balance of payments obligations.
Introduction of hydrogen fuel also requires identification of economical route
requirements. Obviously, one flight per day from every airport in the country poses a
serious distribution problem for introducing a new fuel. Certainly a high-density
route structure between several city pairs will permit initiation of larger hydrogen
production and distribution facilities, which will gain an economy of scale and
reduce start-up costs. Although hydrogen costs were developed for 2500 tons/day,
introduction at 110 the scale of operation analyzed (250 tons/day) should not adversely
alter the economic analysis.
The refueling operations and energy conversion operations to produce and distribute liquid hydrogen at airports are going to require considerable real estate and
environmental planning to achieve the efficient support system that is a requirement
for safe and economical distribution of a cryogenic fluid. In addition, provision for
contingency support plans needs to be developed (such as fuel distribution by tanker
planes or ground vehicles for recovery of aircraft forced down at nonstocked airports).
Inadequate airport planning and development could easily foreclose the possibility
for achieving the economic and environmental benefits obtained through orderly
conversion of air transportation to liquid hydrogen.
CONCLUSIONS
Liquid hydrogen can be produced and safely distributed to the air transportation
industry by conversion of domestic coal reserves in an environmentally compatible
manner for approximately $2.50/10 6 Btu. Depending on the degree of projected
aircraft performance improvement that would result from a switchover to liquid
hydrogen, conversion of domestic coal reserves to hydrogen aviation fuel could provide
significant near-term relief in meeting the growing energy requirement. It would
appear that a liquid hydrogen aviation fuel capability offers the best domestic alternate
fuel strategy to counter overpricing and overdependence on imported hydrocarbon
liquid fuel for air transportation while gaining credits through continuing technical
dominance of the export aircraft market. Careful overall system planning is required
to permit early, economical introduction of liquid hydrogen fuel and provision for
later accommodation of the cryogenic equipment to serve the requirements of a
switchover to hydrogen fuel by the air transportation industry.
REFERENCES
1. D. Gregory, Sci. Am., 228(1): 13 (1973); L. Lessing, Fortune 86(11): 138 (1972); c. Marchetti, Euro
Spectra, 10(4): 117 (1971).
2. "Hydrogen and Other Synthetic Fuels," Rept. TID 26136, U. S. Government Printing Office, Washington, D.C. (September 1972).
22
J. E. Johnsoo
3. N. C. Hallet, "Study, Cost, and System Analysis of Liquid Hydrogen," NASA Rept. CR 73, 226, Ames
Research Center, Moffet Field, California, (June 1968).
4. F. A. Martin, in: Proceedings 7th IECEC, ACS, Washington, D.C. (1972), p. 1335.
5. W. E. Winsche, K. C. Hoffman, and F. J. Salzano, in: Proceedings 7th IECEC, ACS, Washington, D.C.
(1972), p. 1366; W. E. Winsche, K. C. Hoffman, and F. J. Salzano, Science, 180(4093): 1325 (1973).
6. W. Hausz, G. Leeth, and C. Meyer, in: Proceedings 7th IECEC, ACS, Washington, D.C. (1972), p.
1316.
7. J. E. Johnson, "The Storage and Transportation of Synthetic Fuels" ORNL Rept. TM 4307 (July 1973).
8. C. Marchetti, Chern. Economy & Eng. Rev. 5(1):7 (1973).
A-3
THE UCLA HYDROGEN CAR
A. F. Bush and W. D. Van Vorst
University of California at Los Angeles
Los Angeles, California
INTRODUCTION
The use of hydrogen as a fuel for internal combustion engine was suggested*
as early as World War I, in connection with the flight of dirigibles. It did not prove
feasible with the engines of the day-largely because of problems of knock and
backfire, although the latter was apparently solved in principle.
Interest in hydrogen was revived during World War II due to difficulties in
supplying motor fuel to Australia. Studies were encouraging and prototype vehicles
were successfully operated [2] before the end of the war removed the stimulus for
development. Interest has continued, largely at the academic level, and solutions
found for the problems of knock [3], preignition [4], and backfire. Research has
become more intensified comparatively recently as the magnitude of the automobile's
contribution to the pollution of the atmosphere has become recognized [1,4-6]. The
need to minimize the concentrations of hydrocarbons, carbon monoxide, and oxides
of nitrogen in engine exhaust has become generally acknowledged and accorded top
priority. Hydrogen offers the complete elimination of the first two, and considerable
reduction of the last; it is, therefore, worthy of the most thorough investigation.
HYDROGEN-FUELED INTERNAL COMBUSTION ENGINE
Considerable experimental work has been done by way of adapting a variety of
engines to run satisfactorily with hydrogen as the fuel. It seems neither appropriate
nor necessary to detail this effort here, but the characteristics of the final engine
developed might be noted briefly.
The engine was a Ford V-8, "Boss," with a 351 in. 3 displacement. The compression ratio was lowered from 11.7 to 8.9: 1, and sodium-filled exhaust valves were used
to maximize heat transmission from the valve heads. Spark plugs gapped at 0.9 mm
(the usual) were found to be satisfactory.
One of the techniques employed to eliminate backfiring is that of reducing the
overlap period during which both intake and exhaust valves are open simultaneously.
This was accomplished by replacing the camshaft supplied with the engine by one of
shorter duration (270° vs. 324°), less overlap (46° vs. 92°), and higher lift (1.38 cm vs.
1.25 cm).
The most novel modification-and there is reason to think it the most effectiveis that of recirculating a fraction ofthe exhaust gas (EGR). This has the effect ofreducing the combustion chamber temperature and thereby eliminating backfiring due to
* An interesting history of hydrogen-fueled engines is given by Billings and Lynch [I].
23
24
A. F. Bush and W. D. Van Vorst
preignition, curtailing knocking, and reducing NO x formations. Earlier tests indicated
recycle of at least 25 % of the exhaust gas is feasible. This suggested a remarkably
simple and effective control system: the exhaust from two of the eight cylinders was
manifolded separately, cooled, and ducted back to the intake manifold. This, of course,
does not give exactly 25 % recycle, due to the higher temperature of the exhaust, but
does give a satisfactory mix. Efforts to define the optimum condition are continuing.
The EGR is mixed with fresh air prior to carburation. An IMPCO LPG, four-barrel,
air-valve carburetor proved highly satisfactory (Fig. 1).
Ignition timing is sensitive to the air-fuel ratio-as in the case of gasoline and
other fuels-and also to the EGR rate. Excessive spark advance must be avoided so
as to minimize preignition and the buildup of local hot spots to which it leads. Overretarding ofthe spark must also be avoided; this permits the development of hot spots
in the chamber, resulting in preignition and subsequent backfire through the (still
open) intake valve. Satisfactory performance was obtained with a mechanical advance,
varying the ignition timing from 17° btdc (before top dead center) at idle to 20° btdc at
6550 rpm.
While the combustion of hydrogen per se produces no unburned hydrocarbons,
pyrolysis of lubricating oil leaking by the piston rings has been shown to be a source of
preignition as well as hydrocarbon in the exhaust. In order to minimize this, certain
modifications of the ring-and-groove arrangement were made. The top and second
ring grooves were widened from 2.0 to 3.2 mm. Two 1.6-mm molybdenum-faced,
ductile iron compression rings were installed in the top groove; two 1.6-mm cast iron
torsional oil scraper rings in the second. A standard size (2.0 mm) oil control ring was
used in the third groove, but was modified to have radiused rail faces to increase the
tension for better sealing.
The fuel system is shown schematically in Fig. 2. The regulators reduce the
hydrogen pressure from the storage condition (initially 136 atm) to a vacuum of 3 cm
water. The solenoid valve shuts automatically when the engine is not running.
With these modifications, the engine has operated most satisfactorily with fuelair ratios from 30 to 65 % of the stoichiometric amount. Work is continuing on this
Fig. I. Mixing of EGR, air, and
hydrogen.
25
The UCLA Hydrogen Car
STAINLESS STEEL TUBING
ENGINE
HYDROGEN
CYLINDERS
CHECK
VALVE
SECONDARY
REGULATORS
Fig. 2. Hydrogen fuel system.
and other engines, and conclusions may seem premature at this point. There is
considerable evidence, however, to support the following statements.
1. Under comparable conditions, hydrogen has yielded a higher brake thermal
efficiency than gasoline; this is not an original finding, but a verification of
results reported by others
2. Operation at a fuel-air ratio of 45 % of stoichiometric amount seems optimal;
the thermal efficiency is maximum and the NO x emissions minimal at this
condition.
3. Exhaust emission products are well below those allowable under the 1976
Federal Standards. Table I gives a comparison; the operating values were
determined at the California State Air Pollution Board facility in Los Angeles,
according to the 1972 standard test procedure [8].
4. The fuel economy is good; under normal driving conditions, 0.035 km/g
(10 miles/lb) was obtained-or 1.37 km/10 6 cal (200 miles/l0 6 Btu).
5. Safety aspects of hydrogen as a fuel are no more stringent than for gasoline,
methane, or propane. There is a greater possibility of leakage from faulty
connections, but tube joints in this laboratory were routinely tested up to
pressures of 4000 psi and temperatures up to 1400°F without any occurrence
of leakage. If a leak did occur, it is noteworthy that the consequence of a
hydrogen leak is potentially far less serious than with gasoline, because of
hydrogen's low heat of combustion on a volume basis and its high rate of
diffusion in air.
6. Hydrogen embrittlement has not proven to be a problem to date.
n.
Table I. Exhaust Emissions
Exhaust component
UCLA hydrogen car
1976 Federal Standards
CO,glmile
HC, glmile
NO" glmile
0.0
0.0
0.205
0.41
3.4
0.4
26
A. F. Bush and W. D. Van Vorst
In summary, work to date-both in this laboratory and in others-has demonstrated the basic feasibility of using hydrogen as an internal combustion engine fuel.
However, much work remains to be done.
FUEL STORAGE AND DELIVERY PROBLEMS
As indicated above, there seem to be no major unsolved problems with the use of
hydrogen as an automotive fuel from the combustion standpoint. There are, of course,
others. The crux of the matter is the problem of storage. The UCLA car simply uses
two bottles of hydrogen gas, charged at 6000 psia. * About 6lb may be stored, giving a
cruising range of 60 miles. For a more reasonable range, say 250 miles, clearly gaseous
storage is impractical.
Neither is the storage of liquid hydrogen altogether attractive. Aside from the
low holding temperature required-and the attendant problems-a volume of over
40 gal is indicated for a cruising range of 250 miles.
The large volume requirement should not of itself be used to rule out the possibility of using liquid hydrogen. Development of an appropriate cryogenic system would
appear to simplify the control problem as well as the overall distribution problem
from production to consumption. However, the distribution of liquid hydrogen on a
massive scale remains a debatable issue.
The great hope would appear to be in hydrides. Densities of8 and 9 lb of hydrogen
per cubic foot of hydride have been reported [9]. This is for the metallic hydride per se,
however; assuming the reservoir would be a packed bed of granules of the hydride,
this figure must be reduced to allow for the nonsolid volume ofthe bed. Thus a figure
of five seems realistic, implying a storage volume of 30 gal or more for a cruising range
of250 miles. While this is not an unreasonable volume, the additional weight may be a
problem. Using hydrides of lanthanum nickel type, a suitable vessel would weigh in
excess of a ton; even a magnesium-base hydride would weigh over 400 lb. Let us
assume, however, that weight per se is not a problem of dominant seriousness. A much
more intriguing problem is that of control of the hydrogen release and feed to the
engine. Energy must be supplied to effect the dissociation, the amount required being
dependent upon the nature of the hydride, as does the temperature-pressure relation of
the dissociation. Thus the fuel tank must also serve as a heat exchanger and one of
novel design at that. If the heat of dissociation and the temperature of dissociation are
low enough, the exhaust gases may serve as the source of the heat required. Indeed, if a
hydride with a very low dissociation temperature were developed, the atmosphere
might well serve as the heat source. An auxiliary heater or source of hydrogen might be
required for start up, though the bed itself might have sufficient heat capacity to
initiate flow if the heat of dissociation were low enough.
Sketchy and superficial as this summary of the stage problem is, it does serve
to point up two critical problems remaining to be solved: the development of a
suitable hydride for storage as such, and the design of an effective heat transmission
system for the release of hydrogen. This question of storage and delivery is of overwhelming importance, and in a very real sense the potential development of the
hydrogen car awaits its answer.
There remain the broader questions. How will the hydrogen be produced, what
will it cost, and how will it be delivered to the consumer? The first two are receiving
* Maximum pressure; 2000 psia is the usual charge pressure.
The UCLA Hydrogen Car
27
considerable attention from an increasing number of studies directed toward the
"hydrogen economy," most notably those of Gregory [10]. Recently, Zener [11] has
suggested the use of solar sea power to produce hydrogen and oxygen. Exclusive of
distribution costs, the base cost of hydrogen per vehicle mile appears to compare
favorably with that of gasoline. The overall system problem of supplying hydrogen to
the consumer is perhaps the final problem to be attacked. Whether a spent hydride
bed would be replaced with a full one at a suitable service station, or charged with
fresh hydrogen (cooling of the bed would have to be provided), whether the service
stations would be supplied by tankers hauling liquid hydrogen or by pipeline, or
whether residences will be supplied by piped hydrogen and cars could be fueled at the
residence-these and a host of ancilliary questions await serious investigation.
In summary, it is not the engine aspect that impedes the development of the
hydrogen car. It is the handling of hydrogen.
ACKNOWLEDGMENTS
Donations from industry provided over 80 % of the support for the hydrogen car described here. The
authors are most grateful for this support, and regret that space does not permit a specific listing of the 40plus contributors. The authors also take great pleasure in further acknowledging the dedication of the
students working on the project-No R. Baker, D. Bell, 1. G. Finegold, F. E. Lynch, J. Lu, and R. Takashi.
REFERENCES
I. R. E. Billings and F. E. Lynch, "History of Hydrogen-Fueled Internal Combustion Engines," Energy
Research Publication 73001, Brigham Young University, Provo, Utah (1973).
2. J. S. Just, Gas and Oil Power, Ann. Rev., 38: 326 (1944).
3. R. O. King, S. V. Hayes, A. B. Allan, R. W. P. Anderson, and E. J. Walker, Trans. ETC, 2: 143 (1958).
4. R. G. Murray and R. J. Schoeppel, SAE Rept. No. 700608 (1970).
5. R. G. Murray and R. J. Schoeppel, SAE Rept. No. 719009 (1970).
6. M. R. Swain and R. R. Adt, Jr., SAE paper no. 729217, in: Proceedings of 7th Intersociety Energy
Conversion Engineering Conference, American Chemical Society, Washington, D.C. (1972), p. 1382.
7. R. O. King, W. A. Wallace, and B. Mahaptra, Can. J. Res., 26:264 (1948).
8. R. J. Wiswall, Jr. and J. J. Reilly, in: Proceedings of 7th Intersociety Energy Conversion Engineering
Conference, American Chemical Society, Washington, D.C. (1972), p. 1342.
9. Environmental Protective Agency, "New Motor Vehicles and New Motor Vehicle Engines, Control
of Air Pollution," Federal Register, 37(221, Part II): 24,250 (November 15, 1972).
10. D. P. Gregory and J. Wurm, "Production and Distribution of Hydrogen as a Universal Fuel," in:
Proceedings of 7th Intersociety Energy Conversion Engineering Conference, American Chemical
Society, Washington, D.C. (1972), p. 1329.
II. C. Zener, Physics Today, 26(1):48 (1973).
Note added in proof: The authors wish to report that their continued work with hydrogen-fueled
engines indicates that a cryogenic liquid system is more competitive than might be inferred from this
paper as presented.
A-4
CRYOGENIC ENGINEERING AND FUSION
POWER *
c. E. Taylor
Lawrence Livermore Laboratory
University of California
Livermore, California
INTRODUCTION
The goal of controlled fusion research is to eventually produce electric power;
after some twenty years of worldwide research effort. there has been much progress
toward this goal. We can now outline a reasonable timetable for the U. S. fusion
program which leads to the possibility of commercial power in about thirty years [1];
this assumes that a scientific feasibility demonstration will be successful within the
next ten years. and serious design studies are now underway on several scientific
feasibility demonstration experiments.
There has been increasing interest in making speculative design studies of
commercial-scale fusion power plants, mainly to delineate engineering problems and
to estimate costs. Some of these designs are spectacular in scale and present interesting
cryogenic engineering applications. The outstanding central feature of fusion systems,
other than laser pellet schemes. is that the reacting plasma must be confined by
magnetic fields. If such magnets were to be conventional. such as water-cooled copper.
they would consume all or most of the reactor power output. Therefore. it has long
been realized that fusion reactors must use either cryogenically cooled or superconducting coils. It is the purpose of this presentation to delineate some of the
cryogenic aspects of these reactor designs.
TOKAMAKS
The most complete design study yet made of a superconducting coil system for
fusion reactors has been made by the ORNL group [2] based on a relatively low-field
Tokamak. Choice of 79-kG maximum field and Nb-Ti conductor was dictated by
the ability to extrapolate present experience to such large sizes with confidence.
This reactor design, shown in Fig. 1. has a coil with forty-eight circular sections,
each of 11.2 m ID and about 0.5 m square in cross section. arranged to produce a
toroidal field with a 21-m major centerline diameter. Each coil weighs 78,000 kg and
total stored energy is 4 x 10 10 J. It is envisioned that the entire reactor could be
housed in an evacuated chamber 165 ft in diameter and 90 ft high built of steel-lined
reenforced concrete. Heat leak could be as low as 1.7 kW, including joints and coil
leads. but neutron radiation heating would be about 5 kW, assuming the coil shielding
* Work performed under the auspices of the U. S. Atomic Energy Commission.
28
Cryogenic Engineering and Fusion Power
Fig. I. ONRL reactor design.
29
30
C. E. Taylor
transmits only about 10- 6 of the neutron energy. However, if refrigerative power of
this magnitude is installed, cooldown time would be about two months. It is probable
that a cost tradeoff between refrigerator size, shielding thickness, and cooldown time
might well give a much larger refrigerator for an optimized system. Although this
ORNL design is conservative, it gives reasonable costs and leads one to conclude
that present-day techniques for multilayer thermal insulation, load support using
reenforced epoxy members, steel structural materials, and compound conductor
designs can be used for Tokamak coils.
Most reactor coil designs, however, assume significant technological advance
in the next two decades and therefore utilize higher field designs. It is likely that
reactor fields will ultimately be as high as mechanical stresses will allow, and in a
twenty to thirty year time frame, it is reasonable to expect significant developments
of practical high-field materials; therefore 150 to 200 kG fields have been assumed
for several speCUlative designs.
A Tokamak coil as visualized by the PPPL group [3J assumes fields greater than
150 kG, which demands the use of Nb 3 Sn-type, mechanically brittle superconductors.
They have proposed a coil shape, shown in Fig. 2, which practically eliminates bending
forces and therefore results in a lighter-weight structure [4]. For the design shown in
Fig. 2, the coil is about 16 m high and 0.5 m wide, and has a toroidal diameter of
about 18 m. This figure illustrates the complex problem of placing a diverter on a
reactor. Particles which are continuously diffusing across field lines out of the plasma
will eventually strike the vacuum wall and generate a large amount of gaseous impurities. The diverter consists in shaping the magnetic field to allow these particles
to strike the wall in a place which is "around the corner" from the main plasma, thus
allowing the impurities to be pumped away with only a small chance of returning
to the plasma region.
Tokamak experiments to date have used conventional copper coils. However,
for the next generation of these devices, because of the large size and cost, the alter-
Fig. 2. Cross section of the Princeton Tokamak
reactor design.
Cryogenic Engineering and Fusion Power
31
native of using superconductors is being considered, both in the U. S. and in Europe.
Unlike a future power reactor, physics experiments require a field duration only
long enough to observe relevant plasma processes, which can be a fraction of a
second; therefore, superconductors are not essential and the choice is one of economics, reliability, and predictability. To oversimplify, the choice is largely between
investing in superconductors or in conventional power supplies. If one of these huge
devices is superconducting, the resulting development will advance the state of art
much closer to reactor requirements; on the other hand, there is a natural desire to
minimize all nonessential risks when building large experiments with costs in the
range of tens of millions of dollars.
One interesting alternative which is being studied is to use liquid-nitrogen-cooled
coils in these pulsed systems. The resulting decrease in power supply capital investment must be balanced against the refrigeration costs and added complexity of
thermal insulation.
Another approach to CTR is the LASL theta pinch, an inherently pulsed device
[5]. The plasma is contained in a small-diameter (about 60 cm) tube bent into a 112-mdiameter torus. The power pulse consists of two stages; first, a fast toroidal field is
applied in about 0.1 Jl.sec which creates a shock-heated, dense plasma; then a much
slower, 100 to 200 kG compression field is applied on a several millisecond time scale
which squeezes the plasma to a smaller diameter, heating by adiabatic compression,
whereupon thermonuclear "burning" occurs on a millisecond time scale. The resulting
intense neutron burst is absorbed in a blanket located outside the coils and the
increase in plasma energy caused by radiation and charged particle production causes
an expansion of the plasma.
The compression coil must supply about 6500 MJ to a 114-m-diameter reactor
and a low-cost energy storage technique is required. Although capacitors are certainly
technically feasible and have been almost universally used for fast pulsed fusion
experiments, the cost would be impossibly high for a reactor. Therefore, superconducting inductive storage has been proposed. The storage coil need not have a high field
but must be efficiently coupled to the compression coil because of the enormous
amount of power oscillating back and forth between the coils. This coupling problem
is similar to that of a superconducting synchrotron and a solution has been proposed
utilizing a time-varying coupling inductance system [6].
It appears possible, in principle, for the coupling system losses, combined with
resistive losses in transmission lines and compression coil, to be about equal to the
energy added to the coil during each pulse by the expanding plasma which pushes
the field back into the coil. This represents a direct conversion into electricity of about
7 % of the fusion energy. Therefore, the compression coil system may need very little
make-up energy.
Such a coupling system may utilize high-field superconductors in a mechanically
rotating system. Figure 3 shows a cross section of the theta pinch reactor.
A major problem, of course, is the design of a conductor with low enough loss
when pulsed to remain in the superconducting state. This appears to be within the
present state of the art. Although a reactor application will probably use a nearly
lossless coupling system, more immediate plasma physics experiments can use the
much simpler transfer method of connecting the coil to the load and then opening
a shorting switch which is across the storage coil C]. This method can transfer a
maximum of only one-fourth of the stored energy. The switch may be a superconductor
which is driven normal. Of course, all energy deposited in the switch must be removed
32
C. E. Taylor
Fig. 3. Schematic of a theta pinch fusion
reactor (cross section of a torus).
by a refrigeration system and maximum efficiency is achieved if the switch rises to a
high temperature; then for the cooldown cycle, heat is removed at successively
decreasing temperatures, which poses an interesting refrigeration problem. An extensive development program is underway at LASL on all aspects of magnetic energy
storage and transfer. Small (30 kJ) systems have been built and a 300-kJ module is
under construction for testing in anticipation of a requirement for an 850-MJ fusion
feasibility experiment. Here, as in the case of the next generation Tokamak experiments, a decision between superconducting and conventional approaches depends
on reliability, predictability, and economics.
Another major approach to fusion power involves the open-ended magnet system
on mirror machines. In this type offield, plasma particles can escape by being scattered
along the magnetic field lines, whereupon they are lost from the main confinement
region; this is in contrast to the toroidal systems in which field lines essentially close
upon themselves within the plasma and particles are lost by the much slower diffusion
process across the magnetic field. In fact, an energy balance on a mirror reactor
shows that the energy carried by the particles escaping along field lines must be
recovered at high efficiency if net output power is to be obtained [8]. Therefore, a
mirror reactor can be visualized as a steady-state system with an intense stream of
particles being accelerated to reaction energies by external sources; this particle
stream impinges on the reacting plasma, and is thereby ionized, randomized by
collisions, and trapped, supplying both fresh fuel and energy to sustain the plasma
in spite of end losses. Plasma escaping out the ends is led along field lines to a low-field
region. Here the electrons are separated when the field lines are turned abruptly and
the ions which are carrying nearly all of the escaping energy in the form of directed
kinetic energy are slowed down by electrostatic fields and ultimately deposit their
charge on a high-voltage electrode. High recovery efficiencies of the order of 85 to
90 % have been demonstrated in laboratory models and are expected in reactors [9].
Thus, a mirror reactor consists of a large minimum-B superconducting coil,
probably of the "yin-yang" shape, and a direct recovery system which includes a
very large low-field expansion coil. Figure 4 shows the coil for a 1000-MWe reactor.
The most striking feature is that, unlike toroidal machines, the coil is not easily
subdividable into many smaller identical sections and the conductors must be wound
in three dimensions. In this example for a maximum field at the conductor of 200 kG,
the coil weight is 2.8 x 106 kg and, because of the very large forces which are directed
toward bending the coils into a single circular loop, a large spherical structural shell
is provided. Structural weight is 55 x 106 kg; stored energy is 4 x 1011 J. Such a
Cryogenic Engineering and Fusion Power
33
huge mass would require at least a month for cooldown using a 30-kW refrigerator.
In this case, it is attractive to consider using a warm structure and low-cost minimumconductivity compression material which would also serve to minimize heat leak.
This reduces cooldown time and structure cost at the expense of more refrigeration
power.
There have been two large dc superconducting mirror machines constructed,
the IMP experiment at ORNL [10] and BB II at LLL 1]. The 2.4-MJ IMP coil
system used Nb-Ti circular coils combined with a quadrupole field component
produced by Nb 3 Sn noncircular coils. BB II used conventional Nb-Ti. However,
as in the case for Tokamaks, the next generation mirror experiment can in principle
utilize conventional pulsed coils because the plasma parameters of interest have
characteristic times of a fraction of a second. Superconducting coils for these large,
multidimensional minimum-B magnets, using high-field materials, will require considerable conductor development and have been underway at LLL for several years.
Long before a power-producing fusion reactor is built, regardless of which
reactor confinement scheme is used, there will be a need for a fusion device for
materials testing purposes. Such a device need not produce net power but it must
produce a fusion power density as intense as possible to provide realistic testing of
many possible materials.
For example, one of the most pressing problems is to assess radiation damage
caused by 14-MeV neutrons at fluxes of about 10 14 neutrons/cm 2 -sec. Such a device
will require a maximum intensity dc field having a size of about 2 to 3 m in diameter.
This application, which must have dc coils, may represent one of the next large-scale
fusion applications for large high-field coil systems.
e
100 ,..
t-tlAAOR ReACTOR
III·AGNET
COIL S
SECTION
20
-A - A·
~
Fig. 4. Plan and cross-section views of the Livermore mirror reactor with direct converter.
34
C. E. Taylor
The confinement coil systems mentioned here have stored energies between
15 and 400 GJ. The largest existing superconducting coil, at the CERN Bubble
Chamber, has 0.8 GJ. Discharge of these coils must be controlled to prevent overheating, sparking, excessive helium coolant pressure, etc. In case of an accident, such
as loss of refrigeration, the coil must be discharged rapidly enough to prevent overheating, usually within several minutes, depending on current density and heat
capacity. To prevent excessive voltages, this will require subdivision of the coil into
electrically isolated sections capable of handling a discharge of a few gigajoules each.
When such a discharge begins, the coolant will be heated rapidly and provision
must be made for rapid venting to keep pressures at reasonable levels. For example,
a large volume venting of liquid may be desired. Hollow, internally cooled conductors
appear attractive because of the minimum amount of helium within the winding
and ability to contain this helium safely at high pressures.
The most difficult cryogenic engineering problems of fusion reactors are mainly
those caused by the large size of the superconducting magnets. It seems reasonable
to count on extrapolating existing techniques of low heat-leak supports, superinsulation, refrigeration, and structural design. Refrigeration power should be less
than 30 kW for a l000-MWe plant and is not a very significant fraction of total cost.
There will be a need for more practical, rugged, high-field superconductors and, in
the fusion time scale, their development can reasonably be expected.
When we look ahead fifty years and assume success with fusion, it is conceivable
that helium requirements will exceed our expected reserves. The need to extract
atmospheric helium, presumably much more expensive than for present supplies,
may be an important cost factor in reactor design.
REFERENCES
I. "Fusion Power," Division of Controlled Thermonuclear Research, U. S. Atomic Energy Commission,
WASH-1239 (February 1973).
2. M. S. Lubell, W. F. Gauster, K. R. Efferson, A. P. Fraas, H. M. Long, J. N. Luton, C. E. Parker,
D. Steiner, and W. C. T. Stoddart, in: Proceedings 4th Conference on Plasma Physics and Controlled
Nuclear Fusion Research, Vol. III, IAEA, Vienna, Austria (1971), p. 433.
3. J. File, R. G. Mills, and G. V. Sheffield, IEEE Trans. on Nuclear Science, NS-18:277 (1971).
4. J. File and G. V. Sheffield, in: Proceedings 4th Intern. Conference on Magnet Technology, Brookhaven
National Laboratory, CONF 720908 (1972), p. 240.
5. S. C. Burnett, W. R. Ellis, T. Oliphant, and F. L. Ribe, "Pulsed High Beta Fusion Reactor Based
on the Theta Pinch." Los Alamos Scientific Laboratory Rept. No. LA-DC-720702 (July 1972).
6. K. I. Thomassen, "Reversible Magnetic Energy Transfer and Storage Systems," Los Alamos
Scientific Laboratory Rept. No. LA-DC-72-1325 (November 1972).
7. H. L. Laquer, J. D. G. Lindsay, E. M. Little, and D. M. Weldon, "Superconducting Magnetic Energy
Storage and Transfer," Los Alamos Scientific Laboratory Rept. No. LA-DC-72-470 (May 1972).
8. R. W. Werner, G. A. Carlson, J. D. Lee, R. W. Moir, R. F. Post, and C. E. Taylor, in: Proceedings
4th Conference on Plasma Physics and Controlled Nuclear Fusion Research, Vol. III, IAEA, Vienna,
Austria (1971), p. 329.
9. R. W. Moir, W. L. Barr, R. P. Freis, and R. F. Post, in: Proceedings 4th Conference on Plasma Physics
and Controlled Nuclear Fusion Research, Vol. III, IAEA, Vienna, Austria (1971), p. 315.
10. K. R. Efferson,D. L. Coffey, R. L. Brown,J. L. Dunlap, W. F. Gauster,J. N. Luton,andJ. E. Simpkins,
IEEE Trans. on Nuclear Science, NS-18:265 (1971).
II. C. D. Henning, R. L. Nelson, A. K. Chargin, B. S. Denhoy, and F. Harshbarger, IEEE Trans. on
Nuclear Science, NS-18:290 (1971).
B-1
SUPERCONDUCTING ELECTRICAL GENERATORS
FOR CENTRAL POWER STATION USE
T. M. Flynn, R. L. Powell, D. B. Chelton, and B. W. Birmingham
Cryogenics Division
NBS Institutefor Basic Standards
Boulder, Colorado
INTRODUCTION
The electrical industry is faced with a need for dramatically larger generators to
optimize the utilization of large new power sources such as gigawatt-sized fast
breeder reactors. Earlier increases in generator capacity were achieved by impro¥ing
the cooling of the heat-producing components ofthe generator. For the first time in the
evolution of power generators, a unique technology-superconductivity- is available
to the electric machine designer. Superconductivity may offer a way of achieving
higher capacities than otherwise possible.
In the fall of 1971, this division of NBS began a program in cooperation with
industry to accelerate the application of superconductivity to large-scale generators.
The purpose of this program was to develop a more effective way to generate and use
electricity, utilizing superconductivity. The goal was to help solve the national energy
crisis through better use of natural resources. It would also possibly develop entirely
new products and exports. Unfortunately, the program stopped short of actual
experimental construction since sufficient funding did not become available, but did
endure sufficiently to demonstrate outstanding industry-government cooperation in
this area of national need. It is felt, however, that the initial investment provided a
sound foundation for the application of superconductivity to energy systems.
Specifically, the program was successful in placing three key contracts with
industry to document commercial, user's, and manufacturer's needs. They were:
1. Commercial: A market appraisal and economic analysis to assess commercial
potential and economic impact, including the beneficial effect on import/
export balance of trade.
2. User: Applications of superconducting machinery to determine future machinery needs of the electric power industry.
3. Manufacturer: Technical problem identification to determine the critical
components, materials, and systems and suggest an appropriate Rand D
program.
MARKET APPRAISAL
The total market for turbine-generator sets is expected to grow from about S350
million in 1970 to approximately SI725 million in the year 2000, based on 1970
dollars. The generator portion of the set will account for an increasing percentage of
35
T. M. Flynn, R. L. Powell, D. B. Cbelton, aod B. W. BirmiDgbam
36
the total, rising from 66 %, or $230 million, to 76 %, or S1310 million over this time
period.
The average-size turbine-generator installed annually will increase to about
1()()() MW in 1985 and about 1600 MW in 1995. Based on these averages, as many as
60 new steam turbine generator units will be needed annually from 1985 to 1995 in
order to supply the rapidly growing capacity requirements called for in the Federal
Power Commission forecast.
Technological leadership in superconducting machinery in the United States
could help to reduce generator imports by up to 50 %. In recent years, imports have
increased to over 10% of the U. S. market. The technology can be exported profitably
by U. S. companies, but because tariffs and nontariff trade barriers will continue, it
would be exported by means of licenses and joint ventures. For generator imports,
the absolute amount of imports may not be as good an indicator as the trend. For
instance, from 1955 to 1969, the U. S. share of the world market has continually
declined from 32 to 20 %, while West Germany's share increased from 18 to 24 %, and
Japan's share grew from 1.3 to 8.5 %.
Environmental considerations are not expected to have a direct effect on the
development of superconducting machinery. However, indirect effects such as limited
plant locations, increased plant sizes, and new processes consuming electricity will
influence the future needs ofthe market. In these instances, superconducting machinery
gives new options for larger units if they are environmentally and economically
desirable.
The major obstacles to increasing the level of private development are: (1) inadequate return on investment based on still uncertain economic advantages, and
the long-term nature of the technical and market development programs; and (2) the
high risk which must be borne by the utilities which first purchase large (over 1000
MW) plants with superconducting generators.
The development of fast breeder reactors will make possible larger reactors
than can be built with light-water reactor technology. Superconducting generators
have the potential to be built in larger sizes than are possible with existing technology.
Thus, it may be desirable to accelerate the development of large superconducting
generators in order to take full advantage of the economies of scale which are expected
from large breeder reactors.
The last finding is perhaps the most formidable, namely that superconducting
generators may be essential to realizing the full advantage of the large, fast breeder
reactors on which the country is depending to resolve part of the energy crisis.
APPLICATIONS OF SUPERCONDUCTING MACHINERY
This study addressed a complex set of technical and economic factors which
would determine how, if at all, superconducting generators might playa significant
role in the future generation of large quantities of economical electrical energy from
nuclear and fossil-fueled plants.
U. S. Market
Table I shows the estimated steam turbine-generator additions in the United
States broken down into estimates of the numbers of fossil and nuclear generators
based on assumed average generator sizes for each time period. The assumptions used
regarding the derivation of the estimates of the number of generators, as well as the
anticipated maximum ratings, are shown in the table.
37
Supereooducting Electrical Generators for Central Power Statioo Use
Table I. Steam Turbine-Generator Additions Based on Average of Low and
High Projections·
Nuclear
Fossil
Time periods
1986--1990
1991-1995
1996--2000
Number of
l000-MW
additions
55
64
80
Generator
ratingt
max/avg,
MVA
Number
of
generators
Number of
l000-MW
additions
Generator
ratingt
max/avg,
MVA
Number
of
generators
1680/950
1800/1050
1920/1125
69
73
85
219
275
363
2160/1530
2370/1710
2460/1825
172
193
239
* Assumptions: 1. Significant successful extension of tandem-compound (single-shaft) steam turbine-
generator technology will be achieved. 2. The utilities' need for continued emphasis on availability,
reliability, and efficiency will set the need for significant service experience feedback when considering
size increases. 3. Steam supplies will consist of fossil boilers, breeder water reactors, pressurized water
reactors, high-temperature gas-cooled reactors, liquid-metal fast-breeder reactors, and gas-coal fired
reactors, with only nominal increases in turbine inlet steam conditions. If steam supply sizes grow faster
than turbine-generator sizes, multiple turbine generators for each steam supply will be used.
t MV A = 1.2 MW, based upon an average power factor of about 0.8.
It is felt that conventional (nonsuperconducting) design concepts can continue to
provide generators of high reliability and efficiency in the rating ranges required for
the next two decades. This is because one can reasonably expect the more effective
utilization of materials, improvement in rotor cooling, and toleration of more field
site erection, all of which will permit larger ratings for generators.
Turbine--Generator Equipment Cost Curves
Figure 1 gives estimated turbine-generator equipment cost to the user for steam
turbine-generator sets utilizing fully loaded turbines and either anticipated extensions
of present generator design technology or superconductive generators. The figure
gives the costs for the combined turbine and generator set.
Prices for sets with conventionally designed generators were taken from published prices applicable to 1971, which were then converted to the basis of 1975 dollars,
for shipment in 1975. Transportation costs to deliver the turbine-generator to the rail
siding nearest to the purchaser's site, consistent with current practices, are included
in the price.
II
II
Fig. 1. Estimated turbine-generator equipment cost to user
measured in 1975 dollars per kVA.
"
.....-·I~
I~~~._"~~F=~.
0 .....11
T. M. Flynn, R. L. Powell, D. B. CbeltOll, aod B. W. Birmingham
38
Other assumptions underlying the data in Fig. 1 include the following considerations:
1. Costs are mature values, assuming high market penetration of each respective
type of generator.
2. Extra costs to amortize the development expenses of superconductive generators are not included in the price.
3. The same turbine costs are used for both types of generator, although overspeed control will be more difficult for the superconductive machine.
4. Superconductive generator costs were based on present concepts and projected proportions and materials. These costs might increase or decrease,
although the latter seems more likely as additional progress and understanding
are gained.
Figure 1 shows both four pole, 1800 rpm (upper family of curves) and two pole, 3600
rpm (lower family of curves).
A range of expected values, or prices, for steam turbine-generator sets utilizing
superconductive generators is shown by the upper shaded bands for each speed. The
range of prices shown expresses the uncertainty of exact design details and of labor,
material, and overhead cost involved in producing a superconductive product line.
A range of expected values, or prices, is also shown for conventionally designed
generators. The increasing band width merely expresses the increasing uncertainty
in predicting the cost of conventional generators over longer time scales. The band
becomes wider to account for such possibilities as inability to continue to increase
effectiveness in use of materials and to improve gas cooling system capability, or
failure to control increased electromagnetic and mechanical force levels. These contingencies would interrupt the historical trend of increasing generator power density.
The relatively higher prices for turbine-generator units with superconductive
generators are a result of the following factors, which tend to offset the smaller
physical size and weight of superconducting generators:
1. New manufacturing and assembly procedures for the stator and the rotor
2.
3.
4.
5.
are needed; possibly clean-room-type manufacturing areas could be required,
thereby resulting in somewhat higher manufacturing and engineering costs.
There is the possibility of high materials cost in such areas as the superconducting winding, the long cylindrical fabrication for the rotor, and the airgap stator winding.
More difficult processes are required in many new areas involving joining,
control of outgassing, etc.
The hermetic nature of the rotor assembly makes it costly to correct errors.
The severe operating duty on the armature winding requires the utmost in
control of component and assembly tolerances.
Estimates were made to 3000 MW, although there appears to be no technological
barrier to building machines up to 5000 MW, at 3600 rpm.
Efficiency Improvement
Offsetting the possibly higher purchase cost of superconducting generators are
the potential improvements in generator efficiency, ranging from 0.3 to 0.5 % for
1800-rpm machines and from 0.5 to 0.8 %for 3600-rpm machines. Increased generator
efficiency is obtained for several reasons. The principal causes of higher efficiency are:
Superconducting Electrical Generators for Central Power Station Use
39
1. Essentially zero field winding or excitation system losses.
2. Low friction and windage losses due to a smooth rotor, and stator, small
rotor, and low bearing loads.
3. Low stator iron losses due to a small iron core and elimination of the stator
teeth.
4. Small stator winding volume.
These reduced losses are only slightly offset by the refrigeration system power requirement, which is approximately 0.02 % of the machine rating.
The increased efficiency can amount to almost 0.8 %for large 3600-rpm machines,
which is equivalent to reducing the machine losses by one-half. The efficiency gains
for large 1800-rpm machines are a little lower. An increased efficiency of 0.8 % for a
lOOO-MV A generator will result in a benefit of more than $2,000,000 per machine in
capitalized losses.
In addition, these improvements in efficiency have a high evaluated worth to the
electric utility user and hence to the general public because of the fuel savings and the
incremental plant capacity which they provide. The fraction of the benefit due to fuel
vs. incremental capacity will vary depending on the type of steam supply, fuel, and
other factors.
The estimated benefits vs. rating due to improved effciency for superconducting
generators are shown in Fig. 2. Equal reliability levels are assumed, and extra development costs to evolve superconductive generators to this mature state are not included.
Reliability
Reliability has an extremely high worth to the electric utilities and hence to
the public, so much so that an additional two days or so a year of outage time for a
superconductive generator can offset all of the potential economic benefits. The R
and D challenge, therefore, is to bring about a superconductive generator technology
capable of providing economically attractive generator designs and reliability levels
equal to those achieved today.
Fortunately, superconducting generators have a number of features which indicate a potential for excellent reliability and very long life. These features include:
1. The active parts of the rotor are maintained at one temperature for long
periods. This results in very few thermal cycles in the lifetime of the machine.
Once the generator is cold, startup time can be very rapid.
2. The stator winding may be constructed as an integrated structure and be
supported so that very little relative motion between the conductors, insulation, and structure due to thermal changes will take place.
ISlI'S TlIlL IYAlUTIi
ClST. 1115 $'s/Ill
Fig. 2. User's total evaluated cost: superconductive vs.
conventional generators.
40
T. M. Flyma, R. L. po"en, D. B. CbeItoa, I11III B. W. BirmiogIIam
3. The stator winding may be constructed in the form of a high-strength barrel
and be designed to withstand the steady-state and transient torques and
forces due to load changes and short circuits with a high degree of confidence.
4. Critical fault clearing times are equal or superior to conventional generators
without the need for field forcing during the fault period. This results in very
simple low power excitation and regulating systems which could be very
reliable.
These potential reliability advantages tend to be partially offset by the need for a
helium refrigeration plant to cool the superconducting winding. However, reliable
refrigeration plants have been built and operated for a number of years, and this
technology should be transferable to electric utility applications.
TECHNICAL PROBLEMS
Because the development of superconducting generators is at an early stage,
the number of problems and technical opportunities which could legitimately command attention is, for all practical purposes, very large. Therefore, the section below
will highlight only a few of the problems that are peculiar to the introduction of
cryogenics into this established industry of electrical machinery. Problems primarily
related to the stable operation of superconducting generators on power systems will
not be considered in the discussion. Thus, the list of problems is limited to essentially
cryogenic disciplines.
Refrigeration System
The refrigeration system should be closely coordinated with the generator
performance requirements with the goal of optimizing heat transfer, flow characteristics, and reliability of the generator-refrigerator system as a whole. At present,
no large-capacity refrigerator has yet demonstrated the long-term, highly reliable,
maintenance-free operation that is required for electric utility systems.
Helium Transfer System
The helium transfer system is conceptually simple, but problems associated with
rotating seals for such systems are of critical importance to long life and high reliability.
The rotating seals will necessarily operate in a dry helium atmosphere inducive to
high wear rates. The wear rate may be significantly reduced by oil lubrication, at the
risk of potential contamination of the field winding due to diffusion of the lubricant
to the low-temperature region.
Helium availability itself is not considered to be a significant problem. The
volume of helium contained in the generator-refrigerator closed cycle is very small,
and loss due to leakage and maintenance operations will be minimal. Even with a
large number of generator units in operation, the helium requirements will be relatively small compared to those of other helium users.
Heat Tnmder and Flnid Dynamics
Since heat transfer and fluid dynamics are strongly interrelated, a thorough
understanding of the fluid dynamic aspects is essential. The important consideration
is that all flow irreversibilities ultimately show up in the form of heat, thus reducing
the capacity of the helium to absorb other losses.
Irreversibilities peculiar to superconducting generators will arise from instabilities
in parallel flow paths, centrifugal losses, and mechanical vibrations. For instance, if
Superconducting Electrical GeDerators for Central Power StatiOD Use
41
the losses in one of the two flow paths are higher than expected, the fluid will be at a
higher temperature and therefore less dense. For a given mass flow, this will increase
the pressure drop, tending to reduce the mass flow, and compounding the problem.
In a rotating superconducting device, as the fluid is moved from a given radius
to a larger radius, work is done on the fluid, causing an increase in temperature and
pressure. This change reduces the capacity of the fluid for absorbing other losses and
is most severe in large-diameter and/or high-speed machines. Heat transfer equations
and coefficients must be better understood for the complex flow channels and unusual
force fields present in high-speed rotating machinery.
Machine-caused mechanical vibrations may induce hydraulic oscillations and
flow instabilities in the helium, a significant problem encountered with helium. The
end result is to reduce the thermal capacity of the refrigerant.
Structural Materials and Metal Joining
The development of composite materials in recent years has provided a new
form of structure that has potential in superconducting machines. The materials
have promise because of their high strength and light weight and because they are
electrically nonconducting. These materials have definite potential uses in the air-gap
armature winding, but little information is presently available on the fracture toughness properties of composite materials. Every candidate material, including conventional structural materials, must be fully characterized down to liquid helium temperatures for mechanical behavior, including yield and tensile strengths, fatigue resistance,
fracture toughness, fatigue crack growth rate, fracture mode, and structural stability.
The most important requirement is to evaluate the materials for very long life in the
cryogenic environment. Design data must also be obtained for other thermal and
electrical properties such as electrical resistance, thermal expansion, and thermal
conductivity, including their variation in high magnetic fields.
Superconductor Technology
The heart of a superconducting generator is its superconductor field winding.
The superconductors are faced with two limitations: ac losses and stability. Alternating current losses are caused by the magnetization loss of the superconductors, by
the eddy current loss of the normal metal matrix in the superconductor-normal metal
composite conductors, and by the self-field loss of the transport current variation of
the conductors. The stability against quenching of the superconductors into their
normal states depends on the heat capacities and the heat transfer of the conductors
and their liquid helium coolant, and on the mechanical stability of the conductors.
The number of superconducting materials available for use in the construction of
high-field devices (50 kG and above) is not large when compared to the number of
materials known to exhibit superconductivity. This is primarily due to the limitations
imposed by the requirements of fabricability and achievable current-carrying
capacity in high fields. Fabrication is the most important requirement since presentday conductors are in general required to have a rather complex multifilament
configuration in which the conductor consists of many filaments embedded in a copper
or copper alloy matrix and twisted about the axis of this composite. Many of these
conductors may then be braided into a larger cablelike conductor. These requirements
plus cost limitations have reduced the materials used in commercial production to the
ductile alloys NbZr, NbTi, ternary alloys based on NbTi, and the compound superconductors Nb 3Sn and V 3Ga. The early development of filamentary Nb 3Sn would be
exceedingly useful to generator development. Table II lists the properties of the
T. M. Flynn, R. L. Powell, D. B. Chelton, and B. W. Birmingham
42
Table II. Superconducting Conductor Materials
Material
Alloys
Nb-Ti
Nb-Zr
7;,K
8.7
9.5
Compounds
Nb 3Sn
V3Ga
18.05
16.5
Possible new materials
V 2Hfo.sZro.s
V 3Si
Nb3Alo.7sGeo.2s
Nb 3AI
10
17.1
21
18
He2 at 4.2K,
kG
145
115
220
>220
230
235(OK)*
410
295
Jeat 4.2 K,
A/cm 2
lOS at 50 kG
3 x Hf at 50 kG
lOS at 150 kG
lOS at 120 kG
lOS at 130 kG
lOS at 70 kG
104 at 120 kG and 13 K
* Extrapolated.
present commercial materials and possible new materials. At the present time, there
are ten companies in the world producing superconducting conductors suitable for
use in electrical machinery, and accordingly a reliable source of supply is not considered a problem.
CONCLUSION
The superconducting electrical generator is a major step in meeting future civilian
power needs. This new approach to central power station generation will require
considerable transfer of technology from existing rotating machinery, supercond ucting
static magnets, cryogenics, materials, and related fields.
However, there appear to be no significant technical barriers to such a development. On the contrary, it is in the national interest to accelerate this application of
cryogenic technology to realize many potential benefits. Among these are: increased
generator rating to take full advantage of the large, fast breeder reactors; improved
efficiency, reducing present losses by half; possible improved reliability; reduced
transportation and site erection problems; reduced cost resulting from economies of
scale and better materials use. Early development would help stem imports and
possibly develop new industrial products for export.
The economic crossover between superconducting and normal generators
appears to be around 1000 MV A, although clearly the assumptions involved in this
estimate could shift this point a few hundred MVA either way, most probably in the
lower direction. In any case, one can reasonably expect that once superconducting
technology is developed for the large machines, these techniques will cascade down
to much smaller machines, as has always been the case with any innovation in this
industry.
The program described stopped short of actual experimental construction
because of budgetary restrictions late in 1972, but did endure sufficiently to demonstrate outstanding industry-government cooperation in this area of national need.
We are proud to have been a partner with industry at this exciting time in the evolution
of cryogenic technology and regret that the program did not develop fully.
Superconductiag Electrical Generators for Central Power Station Use
43
ACKNOWI,EDGMENTS
This presentation does not have a conventional scientific bibliography, for it is based almost entirely
on reports submitted for this program. A limited number of these original reports are available from the
authors.
In particular, this paper draws upon the work of three key contracts and at times quotes verbatim
from them, although the synthesis and conclusions are those of the authors of this presentation. Accordingly,
the authors wish to give special acknowledgment to: J. Bishop, Jr., Arthur D. Little, Inc., Cambridge,
Massachusetts; G. R. Fox, Corporate Research and Development, General Electric Company, Schenectady,
New York; and C. J. Mole, and C. C. Sterrett, Superconducting Electrical Machinery Systems, ElectroMechanical Division, Westinghouse Electric Corporation, Pittsburgh, Pennsylvania.
DISCUSSION
Question by J. K. Hulm, Westinghouse Electric Corporation: You exhibited curves of total user
cost for superconducting vs. normal generators as a function of power rating in which the superconducting
machines become more economically attractive than normal machines above about 1200 MVA. This is
a very crucial number as far as industrial decision-making is concerned. Could you give us some indication
as to how these cost curves were derived in the superconducting case, and what detailed assumptions were
made?
Answer by authors: The economic crossover point is indeed very crucial to industrial planning.
Estimates of that point vary from 500 to over 2000 MVA. The value given in this paper is one based
upon calculations from the General Electric Laboratories, primarily from Section 4 of their report prepared for the Cryogenics Division of NBS. They assumed a superconductive field-winding rotor but a
normal temperature, iron-free armature stator. The rotor incorporated an electromagnetic shield and
the stator was surrounded by an iron yoke for shielding. They assumed a design criterion for short-circuit
transients to exceed steady-state condition by a factor of five. The detailed design assumptions were
covered in ten pages of their report.
However, other equally competent, experienced designers have calculated much lower crossovers.
There are significant differences in estimates of the costs of constructing large, high-technology machines.
Another factor difficult to assess at this time is the proper prorated amount to charge for development.
If the U. S. Government or EPRI were to support most of the development of superconducting machines,
the ultimate costs for production models would be much less.
Comment by H. H. Woodson, University of Texas: I question the cost curves for superconducting
and conventional generators. I do not believe that conventional generators can be built up to a size of
3000 MW with two poles; 1500 to 2000 MW is more realistic. I fail to see how superconducting generators
can use one-third as much material and cost twice as much as conventional machines.
Answer by authors: The actual rating of the largest conventional machines that are possible is also
in considerable dispute. At present, generators rated near 1500 MVA are having operational difficulties.
Larger generators would undoubtedly have water-cooled rotors. Many designers feel that such rotors
would be more difficult to operate reliably than helium-cooled ones. The General Electric estimate of
3000 MVA for the largest possible conventional generator is an extrapolation from present designs and
is based upon many assumptions, including those concerning transportation, field erection, and solution
of the water-cooled rotor problem.
An accurate resolution of the various estimates of costs for superconducting machines can only be
achieved after further research and development of the next generation of machines.
B-2
CRYOGENIC CONSIDERATIONS IN THE
DEVELOPMENT AND OPERATION OF A LARGE
SUPERCONDUCTING SYNCHRONOUS
GENERATOR
c. K. Jones and D. c. Litz
Westinghouse Electric Corporatioh
Pittsburgh, Pennsylvania
INTRODUCTION
The idea of using superconducting windings in high-power electrical machinery
can be traced back almost to the discovery of superconductivity by Kamerlingh
Onnes in 1911. It was only as recently as 1960, however, when materials were found
that are capable of remaining in a resistanceless condition at high transport current
densities in large de magnetic fields [1], that some of these applications began to
appear feasible. The use of superconducting windings in high-amplitude alternating
magnetic fields is still not in general attractive at power frequencies, primarily due to
the excessively large eddy current and hysteresis losses in any known material,
thereby excluding from serious consideration, for the present, such applications as
the stator windings of an ac machine. However, the reductions in size and weight to
be derived from their employment in the field windings, where only steady fields are
normally experienced, appear to be substantial. The size and weight reductions make
superconducting machines attractive for electric-drive ship propulsion systems and
for consideration for generation of power on electric utility systems [2). A number of
small model machines of this type have already been built and successfully operated,
demonstrating the feasibility of the idea on a scientific basis [3-5). The purpose of the
experiment described in this presentation was to extend this work into the realm of
practical engineering, and to determine more clearly the potential technological and
economic promise of this innovation. For this investigation to be meaningful, it was
necessary to design and construct a machine which was large enough to define design
approaches for future machines. A rating of 5 MVA was established as being close to
the minimum value, to demonstrate potential capabilities for ship propulsion and
power generation.
DESIGN PHILOSOPHY
A basic decision was necessary at an early stage in the program to select the
machine configuration. Although the choice of a stationary field machine together
with a rotating armature would have been a considerably simpler development,
particularly with respect to the cryogenic requirements, it would not have represented
the optimum design for scaling to much larger ratings. Power transfer from the rotat44
Cryogenic Considerations in a Large Superconducting Synchronous Generator
45
ing armature winding by slip rings at hign ratings is not attractive. Also, preliminary
calculations indicated that a much lower power density machine would result.
Consequently, it was decided to direct the approach to the long-term solution
by the use of the present-day synchronous generator configuration of a rotating field
inside a stationary armature, accepting the complications associated with cooling the
superconducting field winding to liquid helium temperatures while subjected to large
rotational forces.
Selection of an intrinsically stable multifilamentary superconductor was demanded by the high winding current density required. Limiting the peak magnetic
field in the winding to the vicinity of 50 kG (5 T) simplified the mechanical and
thermal design of the rotor and enabled predominantly conventional manufacturing
techniques and materials to be employed, while still achieving a power density
considerably greater than that of a conventional machine.
The field winding design must provide localized cooling of the conductors to
ensure a near-isothermal winding while providing sufficient support to minimize
losses due to conductor motion. The field winding must also be provided with
adequate thermal insulation to minimize convection, radiation, and conduction
heat flow into the low-temperature region. Convective heat transfer was eliminated
by means of vacuum insulation around the low-temperature region. In order to avoid
the use of large-diameter vacuum shaft seals, a permanently sealed rotating dewar
design was used, requiring, however, that provision be made for the differential
thermal contraction of the field winding support structure relative to the outer shell.
Radi~tive heat input was intercepted at an intermediate temperature by vapor-cooled
radiation shields. Support of the field winding and transmittal of machine torque
requires the use of low-heat-leak structures. The thermal design provided for staged
vapor cooling of these structures, as well as the field excitation leads, to reduce
conductive heat transfer to a low level. The structural design of the rotor assembly
made operation below the first critical speed possible.
Because of the selection of a rotating field configuration, a helium transfer
system with rotating seals was required. Since commercial applications would make a
closed-circuit refrigerator essential, the transfer system was accordingly designed to
handle both input and exhaust helium streams at cryogenic temperatures, although in
the initial phases of the test program, no provision was made for recovery of the
exhaust helium.
The stator design philosophy adopted required minimum innovation. However,
because of the high flux sweeping the armature winding, the losses in ferromagnetic
teeth normally used in ac machines would be prohibitive. Therefore, the teeth were
eliminated and an "air gap" stator winding design was used.
MACHINE DESIGN AND CONSTRUCTION
The above design requirements have resulted in a machine configuration shown
conceptually in Fig. 1. The superconductor used in the present machine is shown in
cross section in Fig. 2. It consists of approximately 2000 fine filaments of Nb-Ti alloy
approximately 40 ,um in diameter embedded in a copper matrix and twisted about the
composite axis at approximately one twist per inch. The entire composite is insulated
with a thin layer of enamel insulation. The conductor is approximately 1/8 in. x
1/16 in. with a copper to superconductor ratio of two to one with a measured short
sample critical current of 2300 A in an applied transverse field of 50 kG at 4.2 K.
46
C. K. Jones and D. C. Litz
Fig. 1. 5· MY A superconducting generator.
Fig. 2. Superconductor cross section.
Possible excitation rates up to 200 A/min were established. The design point along the
field-current load line of the superconductor was conservatively selected to be only
60 %of the value of the short sample critical current, obtained by extrapolation of the
load line to the short sample critical current-field curve at the operation temperature
of approximately 5 K. The rotor design is shown conceptually in Fig. 3.
The superconductor was wound in slots on a stainless steel pole piece and then
wedged to minimize losses due to conductor motion. Cooling was provided by
channels through the field winding. These channels directed the flow from one end
turn area through the active length to the opposite end turn area, where it was collected.
Adequate heat transfer area was available assuming a heat transfer coefficient of only
0.06 W/cm 2 -K, well below the nucleate boiling limit. In addition, further areas for
helium coolant flow between adjacent turns was available due to the rounded corners
of the superconductors. This area alone is estimated to be sufficient to remove the ac
losses anticipated under transient conditions. The porosity of the windings was
Cryogenic Considerations in a Large Superconducting Synchronous Generator
Rctor Structure and Dewar Wall
47
Vacuum
Torque Tube
Field
I
Excitation ---1
leads
ljlm~~~~~~~:rl""""Radiation
Cooling Shield
Tube
Radiation Shield
Support Spokes
Axial Radiation
Shield
Fig. 3. 5-MVA rotor cross section.
confirmed by flow impedance measurements at room temperature. Highly polished
gold-plated copper radiation shields, maintained between 36 and 90 K by exhaust
helium gas, are provided to intercept radial and axial radiation from the warm rotor
structure. The rotor structure and dewar wall provide a barrier for the permanently
sealed vacuum enclosure surrounding the field winding. The field winding is supported by a thin-walled torque tube on one end and a radial-field support structure
on the other end. The machine torque transmitting structure was conservatively
designed to withstand three times rated torque at operating speed. The torque tube
is vapor-cooled to reduce the cold end heat leak. The field support structure has axial
flexibility, to absorb the thermal contraction of the cold region, and sufficient radial
rigidity to allow operation below the first critical speed.
The rotor was designed to minimize welds and brazes in the low-temperature
area. Seal welds. wherever used, were supported mechanically to reduce weld stresses.
All welds and brazes were nondestructively tested and where practical were subjected
to a low-temperature helium leak test to confirm vacuum integrity.
The material properties of many structural metals were surveyed for each application within the rotor. Minimum ductility requirements used for larger turbine generator rotors were the major criteria for acceptance of a structural material. A design
stress of two-thirds of material yield strength was used with the structural design
criteria specified above. The materials survey showed that austenitic stainless steel,
Inconel alloys, and aluminum had sufficient ductility for use in the cold regions of the
rotor. Type 310S stainless steel was chosen as the structural material for the lowtemperature assemblies and for the torque tube. This material has high mechanical
strength and good ductility from 300 down to 4.2 K. The room-temperature structural
parts were fabricated from type 304 stainless steel. The radiation shield was fabricated
from OFHC copper because of its high thermal and electrical conductivity. The steadystate stresses in the shield are sufficiently low to prevent large creep deformation.
The field excitation leads were fabricated from stranded copper wire, and vaporcooled internally by exhaust helium gas from the field winding. The leads were
conservatively designed to ensure that a field winding quench or loss of coolant would
not cause a burnout and were conduction-dominated to ensure no current dependence
of cold end heat leak.
The helium flow system within the rotor was designed to pass all of the coolant
through the field winding channels to provide maximum cooling capability. After
48
C. K. Jones ud D. C. Litz
Clamp~
Vacuum Vacuum Insulated
rayon«
\,
?
/
0- Ring
~~
-Gcuum
/
/
Vacuum Insulated
Tube
(a)
stationary Vacuum
Insulated Bayonet
Fig. 5. Helium supply system.
Fig. 4. (a) Static joint for transferring liquid helium.
(b) Rotating helium transfer system.
leaving the field winding, the flow was divided into three paths, one of which was used
for torque tube and radiation shield cooling, and another for excitation lead cooling.
The helium used for support structure and lead cooling was exhausted through flow
control orifices at or near ambient temperature. The remaining path provided a
recirculation path to allow independent adjustment of field cooling flow. The helium
in this path was exhausted from the machine at a temperature below 10 K. The total
heat leak into the field winding structure was estimated conservatively to be approximately 7 W, including transient winding losses during excitation.
In order to maintain flow of liquid helium to the rotor, a rotating seal assembly
had to be developed. To accomplish this, a rotating version of a static bayonet joint,
shown in Fig. 4a, was used. The rotating version shown in Fig. 4b substitutes a rotating
face seal for the static O-ring seal. The face seal operates near room temperature since
it is isolated from the cold zone by the vacuum insulation and the thermal distance
length of the bayonet. The helium transfer system used for the machine was designed
for two cold channels. One of the channels was used to inject the helium coolant into
the rotor, while the remaining channel was used to exhaust the recirculation flow from
the field winding.
The liquid helium was furnished from a two-dewar system shown in Fig. 5. The
two-dewar system was required to provide continuous flow during the test program.
A liquid helium-bath-cooled heat exchanger was provided to minimize transfer losses
and to permit operation of the rotor over a wide range of input temperature and
pressure conditions.
The air gap stator design is shown in Fig. 6. The high flux sweeping the armature
winding required that the armature windings be fabricated from finely stranded LITZ
wire and that the stator teeth be nonconducting to minimize eddy current losses. The
machine torque has to be transmitted from the windings to the machine frame without
the benefit of electrically conducting structural stator teeth to prevent excessive losses.
The coils were wound on an epoxy-glass tube between phenolic laminate teeth. The
teeth were bonded to the tube and banded with fiber glass on the outside surface to
provide a strong, nonconducting conductor support. The iron shield, which is
Cryogenic Considerations in a Large Superconducting Synchronous Generator
49
Fig. 6. 5-MV A stator winding cross section.
operated below magnetic saturation, was fabricated from normal electrical iron
laminations stacked and keyed to the stator frame. The stator structure was conservatively designed for three per unit machine torque. Cooling of the stator structure was
accomplished by passing oil on the outside of the cond uctor insula tion through cooling
ducts located outside of the coil insulation.
The rotor is supported in the stator with spherical seat, pivoted pad oil film journal bearings.
The machine is provided with instrumentation in the cryogenic and normal
temperature regions to monitor important temperatures during the test program.
Individually calibrated carbon resistors were used in the cryogenic region, and ironconstantan thermocouples were used in the stator. The field winding inlet and outlet
temperatures as well as radiation shield and torque tube temperatures could be
continuously monitored, readout being via instrumentation slip rings mounted on an
extension of the drive shaft. Stator winding and iron shield temperatures were also
continuously monitored during the test program.
In addition to the measurement instrumentation for temperature, accelerometers
were installed on the bearing brackets to monitor bearing vibration.
OPERATION AND TESTING
The testing and evaluation ofthe machine is being conducted through a series of
operations which began in June 1972. Initially, the mechanical integrity of the rotor
was established and a preliminary estimate of its cryogenic performance made prior
to the complete assembly of the generator. The cooling of the rotor was accomplished
in two stages. During the entire cooldown period, the rotor was continually rotated at
300 rpm to ensure an even distribution of the coolant and to prevent bowing of the
shaft. The cooling rate was limited to 20oK/hr to minimize mechanical stresses arising
from differential cooling.
After an initial purging of the rotor for 1 hr with dry nitrogen gas at room
temperature, cooling commenced with liquid nitrogen. Continuous monitoring of
temperature by the field winding resistance and the thermometers showed that
thermal equilibrium was reached throughout the rotor, with the winding structure at
77 K, after 12 hr with a consumption of 300 liters ofliquid nitrogen. A back flow of dry
helium gas was then established and maintained until the winding temperature
exceeded 90 K. The liquid nitrogen containers were removed and the helium supply
50
C. K. Jones and D. C. Litz
5 MVA. 0.3 Pf. SC R 3.0.
1.0 r
41~
fIOTVOI
;...;I;.:,".;.:
30;:...fIO
r--H'T.--+-----ri"'T-,
.8
~
0>
1. 0
-
..,~ ~:;
.~
- .5 ~u
~
.2
Fig. 7. 5-MVA generator on test stand.
.4
Per Unit Field Cu rrenl
Fig. 8. Saturation curves of the
5-MVA generator.
system was attached to the machine with the rotor stationary. The flow of liquid
helium was initiated and rotation resumed, with the winding temperature of 4.2 K
being achieved in 8 hr using approximately 500 liters of liquid helium. The cooldown
times to nitrogen and helium temperatures are in good agreement with calculated
values. Operation at speeds up to 3850 rpm confirmed the rotor mechanical stability.
The field winding was then excited to a low level with a current of 50 A to check for
electrical short circuits and to permit a plot of the radial component of the magnetic
field. Assembly of the machine was then completed and the machine was installed on
the test stand as shown in Fig. 7.
The cooldown sequence for the machine closely followed previous experience
with the rotor alone and thermal equilibrium was established in approximately the
same time. In order to determine the electrical characteristics of the generator, a
series of basic tests was then carried out, and confirmed design calculations. The
design performance of the machine was achieved with respect to both open-circuit
voltage of 4160 V and short-circuit current of695 A. The generator was also subjected
to five sudden three-phase-to-ground short-circuit tests from voltages varying from
2.5 to 10 % of rated voltage in order to determine the principal machine reactances.
The results of the electrical tests are summarized in Fig. 8 and Table I. The superconducting winding was maintained at approximately 5 K throughout the electrical
test proced ure.
Table I. Test and Calculated Reactances and Time Constants
Parameter
Test
value
Calculated
value
Direct axis synchronous reactance, X d' %
Direct axis transient reactance Xd" %
Direct axis subtransient reactance, X/, %
Direct axis short-circuit· transient time constant rd" sec
Direct axis short·circuit subtransient time constant r/" sec
33.0
21.0
13.9
2.50
0.73
30.1
20.6
13.9
4.0
0.15
• With O.lO-ohm resistance in series with the field winding.
Cryogenic Considerations in a Large Superconducting Synchronous Generator
51
The ultimate realization of the full potential advantages of superconducting
rotating machinery depends critically upon the characteristics of the superconducting
field windings. Accordingly, the operational flexibility of the field winding under a
wide range of cooling conditions has been investigated in some detail, with the following extremely encouraging results.
1. Excitation to full field current at rates up to 200 A/min could be achieved
without quenching at speeds up to 3600 rpm.
2. Deliberate quenches of the field winding current were initiated at current
levels up to 1060 A (at a reduced rotor speed of 1500 rpm to avoid machine
overvoltage) without apparent damage. A maximum rise in winding temperature of only 50 K was observed after a quench and recovery to 5 K as achieved
in a few minutes.
3. Variation of winding temperature by control of coolant flow over the range
4.3 to 6.5 K resulted in field winding currents between 85 and 90 % of the
short sample critical value on the current-field load line for the temperature
concerned, independent of the rotational speed.
4. Changing the cooling mode from supercritical to boiling (i.e., two phase)
helium by variation of the rotor input pressure from 20 to 42 psia had no
significant effect on the winding performance.
No major problems have been encountered during the test program so far.
Occasional plugging of the liquid helium supply system with solid air was experienced,
due to contamination when helium supply dewars were changed inexpertly, a problem
which would be avoided with a closed-loop refrigerator. Subsequent inspection of the
helium transfer system revealed substantial wear rates on the seal faces.
CONCLUSIONS
The successful development and operation of a 5-MVA superconducting
synchronous generator has confirmed the feasibility of this technological innovation.
It is evident that no cryogenic problems of a fundamental nature exist which would
necessitate dramatic changes in machine configuration to permit scaling to larger
ratings for applications in ship propulsion systems or for power generation. In
general, the performance ofthis machine has been found to be quite satisfactory and in
agreement with the theoretical predictions. No difficulties were encountered in finding
engineering solutions to the peculiarly cryogenic problems offield winding mechanical
stability under rotation at low temperatures, adequate cooling of the superconductor
at high current densities, and the efficient transfer of liquid helium into a rotating
structure. In addition, it has been shown that heat leaks to the field winding support
structure can be maintained at acceptably low levels by judicious mechanical and
cryogenic design. The production of superconducting electrical machinery of this
kind is clearly feasible with modern manufacturing techniques employing readily
available materials.
Considerable research and development efforts will be required, however, in
order to fully exploit this important step forward in the evolution of electrical rotating
machinery. Optimization of the superconducting field winding will require much
better understanding of the behavior of superconductors under transient conditions
and different modes of cooling. Long-term reliability of the helium transfer systems
makes mandatory the development offar superior rotating seals. Any real commercial
or military application will require the availability of compact, economical refrigera-
52
c. K. J _ ... D. C. Litz
tion systems of high reliability. None of these requirements appears to present
insurmountable difficulties.
ACKNOWLEDGMENTS
The authors wish to express their indebtedness to the Westinghouse Superconducting Electrical
Machinery Team, without whose efforts this generator could not have been designed, built, and tested. The
comments and criticism ofT. J. Fagan, H. E. Haller III, C. J. Mole. J. H. Parker, Jr., C. C. Sterrett, and M. S.
Walker during the preparation of this paper are particularly appreciated.
REFERENCES
I. J. E. Kunzler, E. Buehler, F. S. L. Hsu, and J. H. Wernick, Phys. Rev. Letters, 6:89 (1961).
2. C. J. Mole, W. C. Brenner, and H. E. Haller, III, Proc. IEEE, 61 :95 (1973).
3. Z. J. J. Stekley, H. H. Woodson, A. M. Hatch, L. O. Hoppie, and E. Helas, Trans. IEEE, PAS85:274
(1966).
4. P. Thullen, J. C. Dudley, D. L. Greene, J. L. Smith, Jr., and H. H. Woodson, Trans. IEEE, PAS90:611
(1971).
5. D. Eckert, F. Lange, M. Endig, G. MUlier, and W. Seidel, in: Proceedings 1972 Applied Superconductivity
Co'!ference, IEEE Publ. No. 72CH0682-5-TABSC (1972), p. 128.
DISCUSSION
Question by W. F. Gauster, Office of Naval Research Laboratory: What was the magnetic field
strength in the air gap?
Answer by author: The average magnetic field in the air gap was 20 kG.
Question by H. H. Woodson, University of Texas: From what open-circuit voltages were short
circuits applied?
Answer by author: The machine was subjected to a sudden short circuit at 2.5, 5, 7.5, and 10% of
rated voltage. Higher voltages were not possible since the drive stand rating is 250 Hp.
Question by Z. Posedel, Brown Boveri & Company, Switzerland: How many liters of helium were
needed during steady-state operation of the system?
Answer by author: The steady-state helium consumption varied from 30 to 60 Iiters/hr during the
test program.
B-3
SUPER CONDUCTING ALTERNATOR TEST
RESULTS
A. Bejan, T. A. Keirn, J. L. Kirtley, Jr., J. L. Smith, Jr.,
P. Dullen, and G. L. Wilson
Massachusetts Institute of Technology
Cambridge, Massachusetts
INTRODUCTION
The first industrial application of superconducting materials appears to be their
use as the field windings of rotating electrical machines. Research is being carried
on throughout the world on this application, and many reports have appeared in the
literature on the testing of ac and dc machines. A research group at the Massachusetts
Institute of Technology, supported by the Edison Electric Institute, has been studying
the application of superconductors to the rotating field windings of alternators since
early in 1967. Two machines have been constructed by this group. The first machine,
which is rated at 45 kV A, has been discussed in the literature [1-3]. Details of the
design and construction of the second machine are also in the literature [4,5). Initial
tests have been performed on the second superconducting alternator constructed.
This presentation describes the operation of the machine during these tests, and
outlines their results.
DESCRIPTION OF ALTERNATOR
The second superconducting alternator has a superconducting, rotating field
winding, and a stationary, oil-cooled, copper armature winding. The nominal rating
of the machine is 2000 k VA. A cross section through the machine is shown in Fig. 1,
and the completed machine is shown in Fig. 2.
The field winding is wound of stabilized niobium-titanium superconductor. The
outside dimensions of the rectangular conductor are 0.125 x 0.050 in. There are
twenty-four strands of superconductor (Supercon T-48B) of approximately 0.010 in.
diameter which are twisted at four turns per foot. The copper to superconductor
ratio is 2.6: 1. At the operating flux density of 25 kG, the short sample current is
2300 A. The stability limit based on nucleate boiling and maximum heat transfer
area is 1500 A. Operating current is 800 A. The wire is insulated with copper oxide.
There are 4660 ft of wire in the winding, with no internal joints, and the winding
inductance is 0.18 H.
The field winding is attached to the 6-in.-diameter stainless steel rotor mandrel
by fiberglass bands in an epoxy matrix, and is covered by an 8-in.-diameter stainless
steel cover tube. The field winding region is surrounded by a copper shell, which
serves as a damper winding, electrical shield, and a thermal radiation shield.
53
54
A. Bejan, T. A. Keirn, J. L. Kirtley, Jr., J. L. Smith, Jr., P. Thullen, and G. L. Wilson
Fig. I. Cross-section view of 2-MVA alternator.
The entire rotor runs in an evacuated micarta tube which is sealed at the shaft
penetrations by oil-buffered carbon face seals. The vacuum space is continuously
pumped by two oil-filled diffusion pumps.
The rotor is mounted in self-aligning, spherical-race ball bearings. The bearing
pillow blocks are attached to the bearing support pedestals with tuned rubber mounts.
A rubber coupling is used to connect the rotor to a 25-Hp direct current, controlledspeed drive system.
Some operating experience has been gained during tests of this machine, and
this experience is described in subsequent sections.
COOLDOWN
The field winding has been cooled to operating temperature on five occasions
between December 4,1972, and May 25,1973, using a cooldown procedure developed
with the first machine. During all phases of cooldown, the rotor speed is set at 90 rpm
to ensure uniform refrigerant distribution.
Liquid nitrogen is transferred to the rotor through the' helium transfer system to
begin the cooldown procedure. Winding resistance is monitored to determine the
Fig. 2. The 2-MVA alternator
ready for test.
55
Superconducting Alternator Test Results
MAIN RETURN GAS TEMPERATURE
••0
•
200
'"2
%
0
"
...
::>
.....
0:
•
"0
~
2
100
Z
...'"
~o
Z
i
0 .'
LN,
0
0:
'"z
<5
WINDING TEMPERATURE
I
Fig. 3. Winding resistance and coolant
outlet temperature during cooldown.
.
~
;;;
0:
~
..u;
'. ---- - - -. -.
0
0 .'
0 .1
O,Ql
TIME . HOURS
rotor temperature. A nitrogen flow rate of 16liters/hr is achieved with a storage dewar
pressure of 17 psig. This phase of cooldown requires about 3 hr, and is concluded when
the field winding temperature is about 100 K. Any liquid nitrogen which remains in
the rotor is removed by heating the field winding. Absence of liquid nitrogen is
indicated by an increase offield winding resistance with time. A typical cooldown curve
is shown in Fig. 3.
Liquid helium is transferred to the rotor during the second phase of cooldown.
A liquid flow rate of 24liters/hr is used, and about 1 hr is required to reach the
superconductor's transition temperature.
Rapid cooldown of the rotor appears to limit overall thermal distortion of the
rotor. Slower cooling rates have produced nonplastic rotor warping oflarge amplitude
(±0.025 in. at the shaft ends.) The cooldown procedure described represents a
compromise between the problem of internal structural stresses due to the low
thermal diffusivity of stainless steel and the problem of overall deformation due to
large-scale nonuniform cooling. The entire rotor contracts 0.145 in. in length during
cooldown. This contraction is taken up by slip in the transfer-end vacuum-shaft-seal
and bearing. The driven end of the rotor is fixed in axial position.
During tests, the helium flow rate is maintained at 15 to 20liters/hr and the
storage dewar pressure required is about 1.5 psig. An operating time of 2 to 3 hr is
achieved with a 50-liter storage dewar. The temperature of the helium gas at the
main return vent is typically 60 K.
TRANSFER COUPLING OPERATION
The helium transfer coupling, which is used to transfer helium refrigerant to
and from the rotor, requires careful design and maintenance for its successful operation. Detailed consideration has been given to transfer couplings in the literature [3].
Several modifications have been made on the transfer coupling for the second alternator to obtain successful operation at 3600 rpm.
The helium transfer coupling, shown in Fig. 4, uses three vacuum-insulated
passages to transfer helium to and from the rotor. Each flow passage is isolated from
each of the others by a narrow annulus terminated by a carbon face seal, to prevent
56
A. Bejan, T. A. Keim, J. L. Kirtley, Jr., J. L. Smith, Jr., P. ThuIlen, and G. L. Wilson
LEII[15 CDOL " MT OU'TL..tJ It,,}
I,t; .i Cl$ (OOU NT IlIffl
~
.. .. 1" III c tUIIN" A" O'""LIT 11'1£1
Fig. 4. Cross-section view of transfer coupling.
thermal communication and direct flow mixing. Maintenance of the relative motion
annulus, and the carbon face seals, are problems requiring careful attention.
Mechanical alignment of the rotating portion of the transfer tube is maintained
by two ball bearings, as shown in Fig. 4. The central portion of the rotating tube has
been straightened in place and the inside of the stationary portion has been lapped to
ensure a near uniform, noncontacting clearance gap. The average gap thickness is
0.010 in.
Proper operation of the carbon face seals has been obtained after the addition
of cooling and lubrication systems. Cooling of the largest-diameter seal is provided
by a flow of room-temperature nitrogen gas as indicated in Fig. 4. The remaining seals
are conduction-cooled by water jackets.
Operation of the carbon face seals without lubricant in the dry helium gas
environment found in the transfer coupling causes high seal temperatures and
rapid wear. To prevent this, seal lubrication systems have been installed.
Each seal has a separate lubrication system. The largest seal is lubricated by an
oil mist carried in the nitrogen seal-cooling gas. The middle seal has a drip oiler.
Excess oil from this seal leaves the transfer system in the helium flow used to cool the
electrical leads. The smallest seal is lubricated by an oiled felt located on the liquid
helium side of the carbon ring. No oil contamination has been encountered.
During a test run at 3600 rpm with the rotor cooled to operating temperature,
the temperature of the nitrogen seal coolant was 340 K, and the lead coolant was
365 K. These temperatures appear low enough for successful steady-state operation.
MECHANICAL OPERATION
The rotor of the machine, shown in Figs. 1 and 5, is a slender, hollow structure,
which has many mechanical components and which is supported on spherical-race
ball bearings. Operation of the rotor near or above 3000 rpm is not possible with the
pillow blocks rigidly attached to the bearing pedestals, due to the presence of a
critical speed at 3000 rpm. Transfer coupling damage was sustained while attempting
to operate above this speed. Operation at 3600 rpm has been made possible by proper
tuning of the bearing mounts.
Superconducting Alternator Test Results
57
.... ltrII HE"UUlI III(TUIllIffi"&
Fig. 5. Cross-section view of rotor.
Tuning of the bearing mounts was accomplished by suspending the pillow
blocks on rubber pads [6]. These rubber pads were cut from a vibration isolation mat
which is compounded to incorporate both elasticity and damping. The system comprising the rotor and its mounts is tuned to place the natural frequencies of the mount
vibration modes weB below 3600 rpm. In these modes, the rotor vibrates on its
mounts as a rigid body. The natural frequency of the rotor acting as an elastic beam is
consequently moved above 3000 rpm. In practice, the mount criticals were clustered
around 1500 rpm, and the rotor critical speed appears to be about 5000 rpm, based
on data taken with a shaker. A band of relatively low vibration exists around the
operating speed of 3600 rpm. The use of rubber mounts has introduced some uncertainty about rotor position during operation, but this problem does not appear to
be serious.
Dynamic balancing was performed with the rotor at room temperature. No
change in balance is noted when the rotor is cooled to operating temperature. Successful operation to 3600 rpm has been achieved with the rotor at operating temperature.
An overs peed test to 4000 rpm has been performed at room temperature with no
change in performance from operation at 3600 rpm.
LEAD FAILURE
Maximum anticipated operating current of the superconducting field winding is
800 A. This current is carried to the winding from the room-temperature copper slip
rings by means of a pair of helium-gas-cooled copper braids. Each braid consists of
336 strands of number 30 wire. The braid is 23 in. long and is terminated in the
junction box shown in Fig. 5. At this point, the braid is soldered to the superconductor.
The superconductor passes through a Teflon-lined, i-in. diameter stainless steel tube
to the winding. Liquid helium flows from the winding space over the superconductor,
through the junction box, over the braid, and out the vent. The gas flow rate and exit
temperature are monitored.
During apparent normal operation at 900 rpm and 800 A field current, one lead
failed as shown in Fig. 6. This failure was not indicated by the monitoring instruments.
Failure was caused by overheating and burnout of a 2-in. length of superconductor.
Signs of overheating were also evident in the second, unfailed lead.
Portions of the field winding and the outer stainless steel cover were also damaged
during this failure. This was a result of the discharge of the coil. The inductive field
winding maintained an arc between the ends ofthe failed lead. When the voltage across
the arc grew too high, new arcs were formed to complete the circuit between the ends of
58
A. Bejan, T. A. Keirn, J. L. Kirtley, Jr., J. L. Smith, Jr., P. Thullen, and G. L. Wilson
JUNCTION lOX
HEUUtoi INLET
. -~ .--.---.--
--.-----.--
Fig. 6. Detail of lead failure.
the winding through the stainless steel rotor structure, principally through the thin
(0.030 in.) cover tube outside the winding (Figs. 1 and 5). The arcs burned holes
through the cover tube opposite the lead tubes. Components of the arc and helium
exiting through these holes entered the vacuum space and resulted in an overpressure.
There is no evidence that loss of superconductivity occurred in the winding proper
prior to the lead failure. Helium flow at the time of failure was 16.6liters/hr through
the main return and 2 liters/hr along the leads.
The exact location and nature of the failure is shown in Fig. 6, which is a drawing
of the stainless steel lead tube. The failed lead has been placed on the drawing to
illustrate the location of the failure. Undamaged superconductor is evident on both
sides of the failure zone.
FUTURE OPERATION
Following redesign and repair of the current leads, the test program will be
resumed. The projected tests include: operation at 3600 rpm to full field current
(armature open-circuited), operation at 3600 rpm to full armature current (armature
short-circuited), and operation as a synchronous condenser on a 13.8-kV bus of the
Cambridge Electric Company. Completion of the tests is anticipated by mid-1974.
ACKNOWLEDGMENlS
The authors wish to thank the Edison Electric Institute and the Electric Power Research Institute for
their continuing support for this research project.
REFERENCES
I. P. Thullen and J. L. Smith, Jr., in : Advances in Cryogenic Engineering, Vol. 15, Plenum Press, New York
(1970), p. 132.
2. P. Thullen, J. C. Dudley, D. L. Greene, J. L. Smith, Jr., and H. H. Woodson, IEEE Trans . Power
Apparatus and Systems, PAS-9O(6) :611 (1971).
3. P. Thullen, J. L. Smith, Jr., and W. D.Lee, in : Progress in Refrigeration Science and Technology , Vol. I,
AVI Publishing Co., Westport (1973), p. 535.
4. P. Thullen, A. Bejan, B. Gamble, J. L. Kirtley, Jr., and J. L. Smith, Jr., in : Advances in Cryogenic
Engineering, Vol. 18, Plenum Press, New York (1973), p. 372.
5. J. L. Kirtley, Jr., J. L. Smith, Jr., and P. Thullen, "MIT-EEl Program on Large Superconducting
Machines," IEEE Transactions Paper T 73137-7, presented at the IEEE PES Winter Meeting, New
York, January 28-February 2, 1973.
6. A.Tondl, Some Problems of Rotor Dynamics, Publishing House, Czechoslovak Academy of Sciences,
Prague, Czechoslovakia (1965), p. 225.
B-4
ALTERNATING FIELD LOSSES IN THE
SUPERCONDUCTOR FOR A LARGE
HIGH-SPEED AC GENERATOR
M. S. Walker, J. H. Murphy, Y. W. Chang·
Westinghouse Research Laboratories
Pittsburgh, Pennsylvania
and
H. E. Haller III
Westinghouse Electromechanical Division
Cheswick, Pennsylvania
INTRODUCTION
During the past several years, great interest has developed in the use of superconductors in rotating electrical machines. The possible applications range from lowspeed propulsion motors to high-speed airborne generators. Utilization of superconductors frequently offers machines which are of smaller size, lower weight, or
higher efficiency. The requirements on the superconducting windings for rotating
machinery may be more stringent than for other applications, however, because of
the high current densities which are used, the alternating and/or transient electromagnetic fields in which the windings may operate, and the rates of excitation which
field windings must achieve.
The interest in superconducting machinery has been spurred by the development
of the multifilamentary composite superconductor, which has made it possible to
build stable field windings with relatively high current densities. Alternating or transient magnetic fields produce eddy current and hysteresis losses in this composite,
however, and the losses must be removed to maintain the low temperature of the
superconducting state, adding to the refrigeration requirements for the machine.
An understanding of the loss behavior for the conditions which exist in the machine
is clearly desirable. The theory of alternating field losses and experimental verifications of the various loss behaviors is reported throughout the literature [1-8], but
the situation of low-amplitude, high-frequency alternating fields with large fixed bias
fields which is encountered in the field winding of a synchronous machine has not
been examined.
This presentation reports an examination of the loss characteristics of a superconductor which has been used in the first phase of development of the rotating field
winding for a 12,OOO-rpm, 4OO-Hz generator [9]. Calorimetric measurements were
performed on a small sample of the conductor at 4.2 K in environments simulating
* Present address:
Hughes, Torrance Division, Torrance, California.
S9
60
M. S. Walker, J. H. Murphy, Y. W. Chang, and H. E. Haller III
those which might be encountered in the machine. The results are analyzed in
light of recent theoretical work by Carr [1.2J which has included higher frequency
effects.
EXPERIMENTAL APPARATUS
Sample losses were measured using the apparatus shown in Fig. 1.* The sample
is held firmly in a micarta can which contains a heater coil and rests inside concentric
superconducting solenoids as shown. The can and the solenoids are immersed in
liquid helium at 1 atm, and the can may be opened to the surrounding bath for filling
through the valve shown.
The solenoids are used to simulate (on the sample) the superimposed bias and
alternating magnetic fields B = Bo + Bm sin rot which may be encountered in the
field windings of the generator. Both solenoids are wound of multifilamentary wire.
The outer, 2i-in.-bore solenoid is used to provide a fixed bias field of up to 5 T.
The inner, 2-in.-bore solenoid is driven with a signal generator and bipolar dc
amplifier to provide alternating fields at frequencies of 10 to 1200 Hz and amplitudes
of up to 0.06 T. A search coil is wound around the can to sense the alternating field.
r.'Ielerlngvllv!'
V ;t.~
Conlrol Roc!
V.cuum JiC ttl
HeU urn 'Surlk'
rill VIIY.
___ runntl
Blufieid
So~noid
AlterniUng
fl.~ So~noi.
---i---Y-1lln
Fig. I. Calorimeter schematic.
Fig. 2. Photomicrograph of a cross section of the
sample. NbTi filaments are embedded in a
copper matrix.
• A more complete description of the experimental apparatus by Y. W. Chang and M. S. Walker will be
published elsewhere.
Alternating Field Losses in the Superconductor for a Large High-Speed AC Generator
61
Losses generated in the sample result in helium boiloff gas which is channeled
through an insulated tube to room temperature,'" then through a copper tube,
which serves as a heat exchanger, and a metering valve. The pressure drop across the
valve is measured with an MKS Baratron capacitance bridge and # 77H-l00 transducer and is calibrated with known power inputs to the can through the heater.
The pressure drop varies linearly with heater power over at least a decade in power
for each valve setting, so that adjustment of the valve provides a sensitive measurement of the power from 1 to 30 mW. A sensitivity of about 10 J1.W has been achieved,
but long-term drift in the background limits the resolution oflosses to about 200 J1.W.
The calorimeter was checked without a sample, and no losses were observed over the
entire range of field conditions. Depending upon the valve setting and power level,
the time required to establish and ensure equilibrium for each calibration or measurement point requires as long as 1 hr.
A cross section of the superconductor, which was purchased from Supercon,
is shown in Fig. 2. It is a monolithic composite consisting of 294 filaments of Nb50 wt %Ti, 47 J1.m in diameter embedded in a copper matrix and twisted about the
composite axis at 100 twists/m. The conductor has a rectangular cross section of
10- m by 1.4 x 10- 3 m with 2.4 x 10- 4 m radii at the corners, and is coated with
5 x 10- 5 m of polybondex insulation. A 3.46-m length of this material was wound
into a small, tight, open-ended coil 2.54 x 10- 2 m in diameter and 2.54 x 10- 2 m
long. The coil was oriented coaxially in the center of the solenoids as shown in Fig. 1,
so that the magnetic fields were parallel to the wide side of the conductor and nearly
perpendicular to the conductor axis.
EXPERIMENTAL RESULTS
Figures 3 and 4 illustrate the dependence of the losses per unit volume upon the
amplitude of the applied alternating magnetic field. The losses, shown in Fig. 3 for
zero bias field, display a nearly field-squared dependence. The losses shown in Fig. 4,
for the 5-T bias field, display a field-squared dependence at the high frequencies and
slightly greater than a field-squared dependence at the low frequencies. The losses
can thus be conveniently scaled with field squared, as shown in Fig. 5, to illustrate
104
103
r =4.? K
Fig. 3. Field amplitude dependence of
alternating field losses for a zero bias
field .
B =0 resl.
o
.. Although it was originally intended that the tube be insulated by vacuum, it was found that the performance of the probe was improved by allowing air to freeze in the vacuum section.
62
M. S. Walker, J. H. Murphy, Y. W. Chang, and H. E. Haller III
104
'j)
1
;;-
~
~
Icr
' ~ 4 .l
K
80 = 5 lesla
4 ~_-+~_+-----,~.j--JL,.---+--+-----;t--t--'.
10-
Fig. 4. Field amplitude dependence of
field losses for a 5-T bias
field.
8 10- 1 alternating
8m . ,,,I.
frequency dependences and the effect of the 5-T bias field . The solid lines are best
fits through the data. The dependence of the loss on bias field at fixed field amplitude
was examined for the special cases of 100 and 200 Hz, which are frequencies of
particular interest in the analysis which follows. The fractional change of the loss
with bias field is identical for these two cases, as is shown in Fig. 6.
The results are estimated to be accurate to ± 10 %, with an additional uncertainty
due to a possible calibration drift of about 400 W1m3 , which may account for as
much as an additional ± 10 %for the lowest loss levels presented. Scatter in the data
of Fig. 5 is indicative of the accuracy of the scaling as well as experimental error. The
data are sufficiently accurate to clearly demonstrate the essential features of the loss
behavior.
loS
T = 4. 1 K
III
Ze ro Bias fIeld Data
. 5 Tesla Bias Fiek1 Data
6
8 100
f . h'
6
8 11m
Fig. 5.Frequency dependence of scaled
alternating field losses.
Alternating Field Losses in the Superconductor for a Large High-Speed AC Generator
63
o.6r--,---,---,---,-------,-----,
• 8m = 8.5 x 10--3 Tesla, f =100 hz
0..0
;;:0
0.4
o 8 m = 5.6 x 10-3 Tesla, f = 200 hz
I
~ 0.2
Theorehcal (.urve lor
RRR = 60
°O~~I~.O~-+I.~O-~l.~O-~4.~O-~5~.O-~6.0
Fig. 6. Bias field dependence of alternating field losses.
80 , Tesla
COMPARISON WITH THEORY
All of the losses, excepting those in large bias fields at frequencies below about
100 Hz, which are believed to be mainly hysteretic, result predominantly from an
eddy current loss mechanism. Carr [1,2] has developed an excellent method for
calculating these losses, using an anisotropic continuum model which assumes that
the composite can be characterized by an average conductivity O"II(Bo) parallel to
the filaments and a conductivity 0"1.(Bo) transverse to the filaments. The model has
been extended here to include an average isotropic relative permeability of the continuum I1r' Because of the effective diamagnetism produced by the shielding of the
field within the filaments by persistent currents, I1r can be less than unity, even for a
nonmagnetic matrix.
The observed losses are in good quantitative agreement with the theory, although
several approximations must be made regarding the geometry of the sample. First,
the analysis applies to loss behavior dominated by the core region of the composite
which contains the filaments, and neglects the contribution of the copper outer sheath
to the losses. The effect of the outer sheath will be discussed further in this section.
It is further assumed that the close stacking of turns of the sample does not seriously
affect the loss results, and that because of the small aspect ratio of the sample it can
be described in terms of a cylindrical composite with core and total cross-sectional
areas equivalent to those in the sample and characterized by Ro = 5.4 X 10- 4 m
and R t = 6.5 X 10- 4 m, respectively.
J2
The low-frequency results,
Ro « fJ and L « 2nfJ, for zero bias field approach
anf2 dependence, as shown in Fig. 5, and are believed to be described by
P/~ =
Bm
(Ro)2 I1/O"1. R 0 2W2 [1
Rt
8
+ 4(~)2J =
2nR o
1.17 x 104f2
W/~3
T
e·
(1)
for uniform field penetration of the wire 4 •5 ]. As frequency is increased to about
150 Hz, the skin effect (f/O" 1.)1/2 limit, fl Ro » fJ, described by
P/~ = !...-(RO) 2I1rW[l + (~)2J-I = 3.09
Bm
R o R,
110
2nR o
X106fl/2
W/rr;-3
T
(2)
is reached [I]. Eventually, at sufficiently high frequencies, the losses in the copper
sheath dominate and an increase in slope is observed.
Solving (1) and (2) for the conductivity and permeability, one finds that I1r =
0.306 and 0"1.(0) = 3.53 x
mho/m. The observed achievement of skin effect
behavior at 150 Hz is consistent with fJ = 1.64
Ro) calculated using the above
values for I1r and 0"1.(0).
109
(J2
M. S. Walker, J. H. Murphy, Y. W. Chang, and H. E. Haller III
64
Since for zero bias field the alternating field amplitudes employed in the present
experiment are always small compared to the penetration field [6]. nearly complete
shielding of the filaments results. The permeability should therefore be less than or
equal to 1 - A., depending upon the filament arrangement. because part of the matrix
can also be effectively shielded. For this sample . .Ie is equal to 0.560. Therefore 1 - .Ie
is equal to 0.440. which is greater than 0.306. indicating that a substantial amount of
shielding of the matrix is provided by the persistent currents in the filaments.
The average conductivity transverse to the superconducting filaments can be
compared with a measured value of 10 10 mho/m for the matrix resistivity at 4.2 K
determined independently for this sample by extrapolation of the results of axial
current conduction measurements in liquid hydrogen. For this measurement. the
sheath was etched away. but the copper matrix in the core region containing the
filaments was retained. The lower conductivity indicated by the loss measurements
is consistent with observations by Critchlow and Zeitlin [8] that an alloy diffusion
layer may be formed around the filaments which lowers the transverse conductivity.
It can easily be shown that if the lower conductivity is due to a diffusion layer.
then a magnetoresistive decrease in (J.1. should result. that is the same as would be
expected for copper of conductivity (J(O) = (J .1.(0). Consistent with a magnetoresistive
effect for copper CO] having a conductivity equal to the transverse conductivity
determined above. f l 2 losses should increase with bias field by 41 ~/o at 5 T. The
shift in losses shown in Fig. 6 agrees with this analysis.
F or those measurements in the 5-T bias field that are at sufficiently low amplitudes
for nearly complete shielding ofthe alternating field from the interior of the filaments.
11, can be assumed to be 0.306. From the skin effect losses in the 5-T field. (J.1.(5) was
determined to be (J jO)/2.1 L using
P/~
Bm
=
(Ro)2 ~fl'W[1
R, Ro flo
+ (~_)2J-1
2nR o
= 4.49
X
106fl/2
W/~3
T
(3)
The observed approach to a skin effect behavior at 200 Hz is consistent with the
above values for II, and (J.1.' which give a skin depth of 1.69(j2 Ro) at that frequency.
The hysteresis losses can now be separated from the eddy current losses at the
low frequencies using these values of fl, and (J .1.(5). The eddy current power loss at
low frequencies is calculated to be
P/V = (Ro)2fl/(J.1.R02W[1
Bm
2
R
8
'
+
4
(~)2J =
2nR 0
5.54 x 103j2
W/m3
T2
(4)
and is subtracted point by point from the total power losses to give the hysteresis
losses. For the caSe of partial penetration of the filaments. the hysteresis losses are
given by
(5)
The resulting values for ;c calculated from (5) are consistent within a factor of two
with M5T) = 0.525 x 10 9 A/m2 in the filaments. Critical current densities for practical
multifilament composites of the Nb-50 wt ~;;, Ti alloy are generally of the order of
109 A/m2 [II] at 5 T. somewhat higher than the result obtained. indicating that the
approximation of very small penetration of the alternating field into the filament~
may not apply for the 5-T bias field in the present experiment.
Alternating Field Losses in the Superconductor for a Large High-Speed AC Generator
65
Preliminary theoretical calculations of the effects of inclusion of the sheath in
the analysis of/2 andf1/2 loss regions indicate that a 33 % decrease in the calculated
J-Lr and corresponding decrease in U.L may result. These calculations are a part offurther
work which is being done to determine a general description of losses for a composite
with a multifilament core and a metal or alloy sheath.
The frequency-independent loss given by [1,5]
~=
J2 Ro
L 2nb
V
(Ro)
8Bm2n2
J-Lo 2 uL2 ~
2
W
m3
(6)
for
« band »
was investigated as a possible explanation of the loss
behavior in what has been described as thef 1/2 region. The magnetic field dependence
of the loss in this region is inconsistent with the conductivity that results from the
magnitude of the loss, however, and it is therefore concluded that the measured loss
cannot be explained by a frequency-independent loss equation.
SUMMARY
Extension of the theory of alternating field losses in a cylindrical, multifilamentary, superconducting composite using the anisotropic continuum model has
been made for low field amplitudes and high frequencies by inclusion of an average
isotropic relative permeability for the continuum. Loss measurements have been
made on a rectangular conductor over a frequency range of 10 to 1200 Hz and
analyzed using this model, and the usefulness of these new analytical expressions
has been demonstrated.
The experimental results over a portion of the frequency range covered show
a loss which is proportional to (f/U.L)1/2. This loss is attributed to the skin effect in
the composite core due to normal currents, and its dependence on frequency is
consistent with the equations developed using the anisotropic continuum model.
These results would indicate that conductors that operate in the skin effect region at
low alternating field amplitudes should have large diameters with long twist pitches
and high matrix conductivity material in order to minimize the eddy current power
loss per unit volume. The loss for small-amplitude alternating fields at 800 Hz is
approximately 108 W1m 3 - T [2] and is nearly independent of bias fields in the 0 to 5 T
range.
As higher frequency alternating fields are applied, the losses in the metallic
sheath of the composite become dominant. Further investigation, both theoretical
and experimental, of the effect of the sheath is planned.
ACKNOWLEDGMENTS
The a uthors wish to thank J. H. Uphoff and A. L. Foley for their help with the measurements and the
design and construction of the apparatus for this experiment. The authors also wish to thank W. J. Carr, Jr.
for his advice on the theory of alternating field loss and D. W. Deis for his measurements of the conductor
resistivities. This work has been supported in part by the U. S. Air Force, Aeronautical Systems Division,
Aero Propulsion Laboratory, E. L. Boyer, Project Engineer, under Contract F33615-71-C-1591.
NOTATION
B
B..
Bo
d
= magnetic flux density
= amplitude of applied alternating magnetic field
= magnitude of bias magnetic field
= diameter of filaments
M. S. Walker, J. H. Murphy, Y. W. CIumg, aDd H. E. HaUer III
fi6
f
j,
= frequency of applied field
=
=
=
L
P
Po
Ro
=
=
R,
=
RRR =
=
t
V
=
critical current density of filaments
filament twist pitch
power loss at Bo
power loss at Bo = 0 T
average radius of composite core
average outside radius of wire
(1.1.(4.2 K, Bo = 0)/(1.l(300 K, Bo = 0)
time
volume of sample
Greek symbols
;.
8
Jl
Jlo
Jl,
(111
(1.l
w
= fraction of composite core that is superconductor
= skin depth in composite core, (nJlf(1.l)1/2
average permeability of core, Jl,Jlo
permeability of free space
isotropic relative permeability of composite core
= average conductivity parallel to superconducting filaments
= average conductivity transverse to superconducting filaments
= angular frequency of alternating field, 2nf
=
=
=
REFERENCES
I. W. J. Carr, Jr., "AC Loss in a Twisted Filamentary Superconducting Wire," to be published in
J. Appl. Phys.
2. W. J. Carr, Jr., "AC Loss in a Twisted Filamentary Superconducting Wire II," to be published in
J. Appl. Phys.
3. M. N. Wilson, G. R. Walters, J. D. Lewin, and P. F. Smith, J. Physics D: Appl. Phys., 3: 1517 (1970).
4. G. H. Morgan, J. Appl. Phys., 41 : 3963 (1970).
5. G. Ries and H. Brechna, Publ. KFK 1372, Institute fUr Experimentalle Kernphysik, Karlsruhe,
West Germany.
6. R. Hancox, Proc.IEE, 113: 1221 (1966).
7. M. J. Chant, M. R. Halse, and H. O. Lorch, Proc. lEE, 117: 1441 (1970).
8. P. R. Critchlow and B. Zeitlin, J. Appl. Phys., 41: 4860 (1970).
9. R. D. Blaugher, J. L. McCabria, and T. J. Fagan, Jr., "Cryogenic and Mechanical Design ofa 12,000
RPM, 4 Pole Superconducting Rotor," paper presented at 1973 Cryogenic Engineering Conference,
Atlanta, Georgia, August 8-10, 1973.
10. F. R. Fickett, Annual Report-Incra Project No. 186 (to be published).
II. J. Wong, private communication.
B-5
A REVIEW OF SUPERCONDUCTING MAGNETIC
SYSTEMS FOR GENERATING TRANSVERSE
MAGNETIC FIELDS*
V. V. Sytchev
Institute/or High Temperatures
Moscow, USSR
INTRODUCTION
The discovery of type II nonideal superconductors featuring high values of
critical current density in strong magnetic fields during the early 1960's has provided
the impetus for the recent dramatic growth in the technical utilization of superconductivity. Superconducting devices now find their application in experimental
physics, in power engineering and electrical technology, in charged particle transport,
in radio electronics, in magnetic devices, etc. In all of these fields the use of superconducting devices provides improved performance when compared with conventional devices and, in a number of cases, presents the only possible solution.
The most advanced programs among the various research studies presently
being carried out in the field of superconducting devices are those involving the
development of superconducting magnet systems (SCMS) for various applications.
Studies have shown that these superconducting magnet systems are decidedly
superior to conventional electromagnets with respect to cost and operating characteristics and, in many cases, provide the only practicable solution.
Depending upon their function, superconducting magnet systems can be divided
into three groups, namely: (1) solenoid-type systems for generating longitudinal
magnetic fields inside a solenoid; (2) systems designed to generate transverse magnetic
fields in a channel; and (3) systems designed to generate magnetic fields of complex
configuration (these are mainly systems for thermonuclear fusion research).
This review will deal with superconducting magnet systems of the second group,
i.e., systems designed to generate transverse magnetic fields in a channel. Such systems
are essential for MHO generators, for use as magnetic dipoles in elementary particle
accelerators, for use in transport systems for charged particle beams, for some types
of electric motors and generators, etc.
The development of such systems is a complex task involving the solution of
an extensive set of diverse problems in physics and engineering. Such problems
include:
1. Metallurgical and metal physics problems related to the manufacture of
superconducting alloys and intermetallic compounds featuring high critical
properties.
* Invited paper.
67
68
V. V. Sytcbev
2. Determination of the optimum configuration of the superconducting magnet
system (minimum weight at maximum field intensity and given degree of
field uniformity).
3. Compensation for the various forces acting upon the windings.
4. Provision of a reliable operation for the superconducting magnet systems.
5. Development of a cryostat and a means of cooling the system.
6. Development of a suitable power supply for the superconducting magnet
system.
Let us now discuss specific examples of solutions of these problems.
BENDING MAGNETS
In experimental work with a modern accelerator a large number of bending
(dipole) magnets are used. The latter serve to transport a charged particle beam to a
detector or target. In a modern high-energy physics research laboratory up to 150
to 200 such magnets may be used. In conventional magnets, where the field is generated
with the aid of a water-cooled copper winding and an iron yoke, the field intensity
is usually 2 T. With an increase in the energy of the charged particles, the magnet
length and effective cross section have to be increased in order to maintain a given
bending angle. For large accelerators presently being designed, the length of a
conventional bending magnet can reach 4 m with a cross-sectional diameter of 40 cm.
The use of superconducting dipoles permits a shorter magnetic system with a higher
field intensity. The study of superconducting dipoles developed up to this time shows
the possibility of attaining field intensity values on the order of 4 to 5 T. This would
double the parameters of the conventional systems. Prototypes of bending magnets
have been developed over the last few years by both laboratories and manufacturing
companies.
It should be noted that technologically it is much more difficult to develop a
winding of a complex configuration than an axisymmetric winding. In order to
obtain a uniform field, one should distribute the conductor appropriately in the
winding cross section depending upon the azimuth angle qJ. An ideal dipole is obtained
if the current density varies as cos qJ over the perimeter of a circle or an ellipse, as
well as in the case when the windings are in the form of two intersecting ellipses
(channel in the intersection zone). The actual winding is some approximation to the
ideal case. Special attention needs to be paid to the provision of a suitable structure
for the turns affected by magnetic forces whose direction usually rules out the
possibility of utilizing the structural properties of the conductor itself. One should
also bear in mind that in windings of a complex configuration the current degradation
effect is always greater than in simple axisymmetric windings. The utilization of the
principle of full cryostatic stabilization in such windings always encounters considerable difficulties for a variety of reasons (among these are the difficulties in
fastening the turns in a "loose" winding, the need for ensuring a relatively high current
density due to stringent requirements placed upon the field topology, the compactness
requirements, etc.). In this connection, the use in such windings of intrinsically
stabilized conductors appears more expedient.
Superconducting magnets for beam transport may be constructed with or
without an iron yoke. The provision of a magnetic circuit brings about some increase
in the magnetic field for the same dimensions of the winding (20 to 30 % at fields of
4 to 5 T and accordingly more at low fields). The magnetic circuit is more effective
Inner diameter of the winding,
cm
Length, cm
Magnetic field, T
Stored energy, kJ
Date of testing
10
90
(3.5)
(80)
1973
10
90
2.7
50
1972
Argonne
National
Laboratory
8.5
(Cold)
50
4.6
64
1972
8.5
(Cold)
183
(4.0)
(175)
1973
Brookhaven
National
Laboratory
13.2
(Warm)
195
3.4
370
1972
European
Center of
Nuclear
Research
(Geneva)
35
3.5
25
1969
10
29
(20 Warm)
83
4.0
655
1971
Lawrence
Radiation
Laboratory
5x13 20x60
7.6 x 12.7 7.6 x 12.7
10
(Warm) (Warm) (Warm)
(Cold)
(Cold)
300
180
300
300
300
2.5
2.3
(3.0)
(4.5)
1.8
112
600
120
1972
1972
1971
1973
1973
National Accelerator Laboratory
Table I. Design Parameters for Recently Developed Superconducting Magnet Systems
Used in Accelerator Studies
70
v. v. Sytchev
when it is brought closer to the winding and is inside the cryostat. In this case, however,
the cooling of a relatively large mass of metal is required. The systems developed up
to now include magnets with both cold and warm magnetic circuits. It should be
borne in mind that the saturated iron of the magnetic circuit causes some distortions
in the magnetic field configuration.
Table I presents basic parameters of magnet systems for charged particle beam
transport as designed by various research organizations [1]. At present, the efforts
of a number of research laboratories are principally devoted to the problem of developing pulsed magnets for accelerators. Accelerator magnets differ only slightly, in a
structural sense, from magnets designed for beam transport and therefore the experience gained through the use of pulsed magnets leads almost automatically toward
the development of reliable and efficient dc magnets.
PULSED DIPOLES
A modern charged particle accelerator on the order of several hundred GeV,
such as a synchrotron, consists of a large number of pulsed magnets arranged in a
circle with a diameter of the order of several kilometers. The use of superconducting
magnets in place of conventional magnets helps reduce the accelerator diameter by
a factor ofthree to four for a given particle energy, or alternatively permits an increase
of the particle energy for the same accelerator diameter.
Several research programs are now being carried out which involve the development of a superconducting accelerator. A number of research organizations in
Western Europe are collaborating in a study of a European superconducting synchrotron (GESSS). This group includes researchers from the Rutherford Laboratory
(RHEL), the Institute of Experimental Nuclear Physics in Karlsruhe (IEKP), and
the Center of Nuclear Studies at Saclay (CENS). The group's goal is to increase the
charged particle energy of the synchrotron being developed at CERN from an energy
level of 300 to 400 GeV to an energy level of tOOO GeV by the replacement of conventional magnets having a magnetic field of 1.8 T with superconducting magnets
having a field of 4.5 T (there is also the possibility of developing a lOOO-GeV superconducting accelerator using a 3OO-GeV accelerator as injector [2].
At the Brookhaven National Laboratory (BNL) in the United States, work is
progressing on the development of a superconducting accelerator with intersecting
beams (intersecting storage accelerator) for a capability of 200 GeV [3]. The existing
accelerator with a variable gradient (AGS) will presumably be used as the injector.
In the Soviet Union, studies are underway at the Radiotechnical Institute to
determine the possibilities of developing superconducting synchrotrons with energies
of up to 4000 to 5000 GeV (with a diameter of up to 5435 m and a field strength of
6 to 8 T) [4].
In all of these efforts, the researchers are heavily involved with the problem of
developing pulsed dipoles, since the latter present the weakest link in the accelerator
designs.
As already pointed out, structurally a pulsed dipole does not differ greatly from
a dc dipole. However, new problems emerge, and some ofthe existing problems appear
to acquire a different coloring. Not only does the iron yoke present in pulsed magnets
bring about an increase in the field intensity and serve as a shield, but it also causes
a considerable reduction in the size and rating of the power supply due to the reduction
in energy stored in the magnetic field. Practically, all of the presently designed pulse
Superconducting Magnetic Systems for Generating Transverse Magnetic Fields
71
dipoles involve the use of a "cold" magnetic circuit. At the same time, the problem
of field distortions due to a saturated iron yoke becomes more difficult in the case
of pulsed magnets because of the necessity of allowing for such distortions both at
the maximum and all the intermediate field values.
In the case of dc magnets, the need for cooling the conductor is related only to
the problem of cryostatic stabilization. However, in pulsed windings, irrespective
of their stability, the inner layers ofthe winding have to be cooled to remove the heat
generated by field variations. The cooling is accomplished either with the aid of
channels for liquid helium or by thermal conduction in a monolith winding.
In a pulsed winding, the conductor displacement causes both a degradation of
current and a marked increase in the losses. Therefore, a full or partial impregnation
of the winding is preferred.
The most complicated problem underlying the ability of pulsed dipoles to
compete successfully in this application is that of reducing the losses in the structural
elements immersed in the helium bath. All of the pulsed magnets now in existence
or in the design stage employ NbTi-based multiple-core compositions. The duration
of a magnetic field cycle in superconducting accelerators of the future is estimated
to be from several seconds to 1 to 2 min. Accordingly, the rate of field change may
be essentially different in these magnets and different requirements are placed upon
the superconducting materials. Quite a few measurements have already been made
investigating the various losses occurring in pulsed dipoles.
The accumulated experimental knowledge and existing calculation techniques
make it possible to estimate to a first approximation (the error is presumably within
a factor of two) the heat that will be released in the helium bath from the magnets
of future superconducting synchrotrons. For example, in the case of typical dipoles
with an induction of 5 T, an inner diameter of 100 mm and a superconducting core
diameter of 10 ]lm, the calculated losses for 1 m of dipole vary from 10 to 40 W
(depending upon the cycle duration).
Table II provides results from pulsed dipoles which have already passed the
test stage; Table III contains design parameters of a number of dipoles now in the
development stage [2,5-9].
Estimates show that the cost of a 1-GeV superconducting accelerator is onethird to one-fourth as much as a conventional accelerator. This estimate includes
the cost of the electric power for 10 yr of operation.
SUPERCONDUCTING MAGNETIC SYSTEMS OF MHD GENERATORS
The magnet systems of MHD generators are structurally similar to superconducting dipoles designed for application in accelerator technology. In most cases,
however, large magnets are required for MHD generators, while the requirements
placed upon the uniformity and reproducibility of the field are considerably lower
than in the case of the superconducting dipoles for particle transport. As a rule,
MHD magnets already in existence and those being designed have no iron yoke;
otherwise they would be too bulky. The most difficult problem connected with large
magnets is that of providing for the magnetic forces. The magnets ofMHD generators
have either some variety of saddle winding or a winding in the form of spaced oval
coils. The saddle winding is generally preferable for an MHD generator since such
a configuration provides for the maximum length of a uniform transverse magnetic
field from a given amount of superconducting material. However, it is much more
difficult to develop a large superconducting magnet system with a saddle winding
v. V. Sytc:hel'
72
Table ll. Design Parameters of Pulsed Dipoles Developed and Tested by Various
Research Laboratories
Lawrence
Radiation
Laboratory
Designation of dipole
Inner diameter or
cross section. mm
Length. mm
Central magnetic field
(dc). T
Magnetic field/current
rise. time. T/sec
Method of cooling
Magnet iron
Type of conductor
Number of strands
Strand diameter. mm
Number of filaments
per strand
Filament diameter. JIm
Cross-section ratio
(Nb-Ti):Cu:(Cu-Ni)
Table
Institute of
Experimental
Nuclear Physics
at Karlsruhe
Brookhaven
National
Laboratory
Rutherford
High Energy
Laboratory
No.8
63.5
2F
50
2G
50
DT
80 x 40
AC-3
100
AC-4
90
400
3.9
350
3.4
350
4
400
4.5
400
3.9
720
4.5
3.5/0.5
-
-
3.2/5
3.5/1
4.3/2
Channels
-
-
Channels
Wire.
laminated
Cable
133
0.2
211
Laminated Laminated
Laminated
Heat
Channels
conduction
None
Laminate d
Braid
132
0.2
210
Braid
132
0.2
210
Cable
10
0.5
1045
10
I: 1.25:0
10
1:2:0
7
1:2:0
10
I: 1.5 :0.2
Cable
0.4
1045
Cable
25
0.85
2035
8
I: 1.3 :0.2
12
I: 1.3 :0.5
90
m. Design Parameters of Pulsed Dipoles under Development by Various
Research Laboratories
Institute of Experimental
Nuclear Physics at
Karlsruhe
Designation of dipole
Year of completion
Inside diameter. mm
Length. mm
Nominal magnetic field. T
Current rise time. sec
Laminated magnetic iron
Method of cooling
D2a
1972
80
1400
4.5
10--20
Yes
Chanllels
Type of conductor
Number of strands
Strand diameter. mm
Number of filaments per
strand
Filament diameter. pm
Cable or braid impregnator
Cross-section ratio
(Nb-Ti):Cu:(Cu-Ni)
Nominal current. A
Cable
12
0.54
1000
12
In-Sn
I: 1:0
1500
D3
1973
80
2800
4.5
3-10
Yes
Channels
8-10
Rutherford High
Energy Physics
Laboratory
Center of Nuclear
Studies at Saclay
AC-5
1973
90
MOBY
1972
100
1~00
500
6.0
4.5
3
5
Yes
Yes
Channels
Heat
conduction
Cable
Braid
24
5-15
1.1-0.7
0.44
10.000--20.000 1000
ALEC
1973
110
1500
5.5
5-15
Yes
Heat
conduction
1000
5
Epoxy resin
I: 1.3 :0.2
10
Epoxy resin
1:1:0
I: 1.6:0
-4000
1500
2500
SupercoodactiDg Magnetic Systems for Geaeratiog Tnmsverse Magnetic Fields
73
than with conventional Helmholtz coils or a couple of oval coils.
First, the intricate configuration of the winding seriously complicates its manufacture; second, the distribution of magnetic forces in the winding is complicated,
resulting in the problem of compensating for the latter forces with the aid of support
structure. It is because of these difficulties that only a few superconducting systems
have been developed up to now for use with MHD generators.
One of the first attempts to develop a practical use for supercond ucting magnets
was made by the Westinghouse laboratories when two spaced, round superconducting
coils were used to study energy conversion in an MHO system [10].
A small superconducting magnetic system with oval coils was developed at the
Institute of High Temperatures (Soviet Union) in 1964--1965 and later used in an
explosive MHO generator with a power of over 1 MW. The channel section (at room
temperature) of this system was 35 mm x 70 mm, the length of the uniform field
portion was 150 mm, and the field in the center was 1.5 T [11.12].
In 1966, a considerably larger magnetic system was tested in the Avco-Everett
laboratories. The channel diameter (cold) was 305 mm, the length of the uniform
field portion was about 1200 mm, the field in the center of the coil was 3.7 T, and
the stored energy was about 4 MJ. This magnetic system was one of the first fully
stabilized systems [13].
In 1968, Hitachi laboratories (Japan) tested several model MHD-type magnets.
The largest of these generated a field of 1.6 T in a channel with a cold diameter of
200 mm [14]. Somewhat later, the Electrotechnical Laboratories (Japan) tested a
saddle magnet generating field of 2.4 T in a channel with a cold diameter of 290 mm.
The magnet winding together with the support structure weighed over a ton and
had an external diameter of 760 mm and a length of 910 mm. The magnet operated
in combination with the channel of an MHO generator rated at 1 kW [15.16].
Table IV presents the parameters and results of the magnets associated with
MHO generators which have been described in the literature since 1970. These
magnets were tested by the Institute of High Temperatures, USSR (IVTAN) [17],
Hitach~ Japan [18], Magnetic Corporation of America (MCA) [19], Mitsubishi,
Japan [20], Institute of Technical Physics at Jiilich, Federal Republic of Germany
(ITP) [21], and Electrotechnical Laboratories, Japan (ETL) [22]. Some additional
data on these magnetic systems will be supplied below.
The IVTAN saddle magnetic system tested in 1970 has an external diameter of
285 mm and overall length of 500 mm. It is manufactured from a stranded, indiumimpregnated cable (21 superconducting and 28 copper strands) with an additional
copper braiding. Interlayer stainless steel wire banding has been used in the winding.
Thus the winding has no outside support structure. Small gaps (0.3 mm) are provided
between the winding layers. Helium penetrating these gaps serves as a heat sink and
provides enthalpy stabilization for the winding. The system is supplied with a superconducting shunt which provides the operation under "frozen flux" conditions.
A somewhat larger superconducting magnet system has been developed and was
successfully tested in 1973 in the Physico-Technical Institute of Low Temperatures
of the Ukrainian Academy of Sciences, Soviet Union [23]. The system provided for
a uniform magnetic field with an intensity of up to 3.5 T in a warm channel 168 mm in
diameter. The saddle winding is made of a round, superconducting cable 0.8 mm in
diameter. The winding frame is a cylinder with slots 14mm wide and 28 mm deep
milled in the external surface of the pipe. The inner diameter of the frame is 249 mm,
the outer diameter is 326 mm, and the length is 600 mm.
Saddle
Oval coils
Oval coils
Mitsubishi
Institute of Technical Physics at Jiilich
Electrotechnical Laboratory (Japan)
• With cryostat.
t Two sections.
Saddle
Saddle
Saddle
Institute of High Temperature (USSR)
Hitachi
Magnetic Corporation of America
Organization
Type of
winding
390 x 1300 (Hot)
140 (Cold)
380 (Cold)
180 (Hot)
214 (Cold)
214 (Hot)
278 (Cold)
225 x 295 (Hot)
Diameter or
crosssection of
duct, mm
1200
800
250
200
600
900
Length of
uniform
field,
mm
48,000
9,500*
395
-100
-7,000
423
Weight of
winding
with force
sustaining
elements,
kg
1000
Nb-Ti-Zr
Nb-Ti-Zr
Nb-Ti
Nb-Ti-Ta
Nb-Ti-Zr
5
3
1.86
3.1
4.5
5
Central
field,
T
1275,1570t
-300
998
-500
485
Superconductor
Nb-Ti
Current,
A
Table IV. Design Parameters and Test Results of Magnetic Systems Associated with
MHO Generators at Various Research Laboratories
70
5
0.155
0.92
-0.12
MJ
Stored
energy,
1...
[IJ
;<
,<
it!
Superconducting Magnetic Systems for Generating Transverse Magnetic Fields
7S
The experience gained during 1970 and 1971 in the successful development and
operation of the model saddle-type superconducting magnet system IVTAN, generating a transverse magnetic field of 3 T at a current of 975 A in a channel 140 mm in
diameter provided the basis for work begun in 1972 aimed at building a large superconducting magnet system with a saddle winding for one of the MHD generators
developed at the Institute of High Temperatures [24].
The Hitachi magnet system is constructed from a copper strip with a section,
1.6 mm x 7 mm, containing ten superconducting cores. It is designed for conditions
of full cryostatic stabilization. The overall length of the winding is 1800 mm, the
outer diameter being about 1450 mm (880 mm over the winding). The magnet has a
superconducting shunt. In the course of testing, the maximum field in the center was
4.7T.
The Magnetic Corporation of America saddle magnet is constructed from a more
up-to-date composite superconductor: 180 superconducting cores in a copper
matrix with a 1.63 mm x 1.63 mm section, with a twist pitch of 25 mm and a ratio of
copper to superconductor of 1.8. In developing this magnet, special attention was
paid to the red uction of the overall weight ofthe system (561 kg including the cryostat).
No channels for liquid helium are provided in the system-the designers rely fully
upon the principle of intrinsic stabilization. The rated current density in the winding is
15,000 A/cm 2 • First tests ofthis system have been carried out in which the rated field
value of 5 T could not be attained; the field value obtained in the center was of the
order of 4 T. As a result, some improvements in the design are envisaged.
A distinguishing feature of the Mitsubishi magnet system is that the support
structure is placed outside the cryostat and forces from the winding are transmitted
through the layers of superinsulation. Conditions for full cryostatic stabilization are
provided. The superconductor consists of ten superconducting cores soldered in the
slots of a copper strip with a 1.2 mm x 10 mm section.
The Institute of Technical Physics at Jtilich magnet system consists of two oval
coils spaced in such a way as to provide a gap for accommodating the channel of an
MHD generator. The winding is designed for conditions of full cryostatic stabilization. It is constructed from a composite conductor with a 2.78 mm x 14.5 mm section
(21 superconducting cores in a copper matrix, copper to superconductor ratio of 2.33,
with a twist pitch of 75 mm). The rated value of the current density is 6500 A/cm 2 •
The rated field in the center of 4.2 T could not be reached in this system. At field values
exceeding 3 T, the occurrence of the normal zone was observed. This phenomenon
was of a statistical nature. Presumably, the inadequate operation was caused by the
presence of a faulty portion in the conductor, as well as by the use in the structure of
shunting strips of normal metal for each layer of the winding.
The Electrotechnical Laboratory magnet system, built with the aid ofthe Hitachi
Company, is the largest ofthe nonaxisymmetric magnet systems. With respect to the
energy stored by the magnetic field, it is only exceeded by the magnets of the National
Accelerator Laboratory and the European Center of Nuclear Research (CERN)
bubble chambers. The results of testing the Electrotechnical Laboratory magnet have
not yet been published. The magnet consists of two oval coils separated from each
other by a distance of 700 mm. The rated field value of 5 T is attained at a current
density in the windings of (2.5-3 x 103 A/cm 2 ; the maximum field on the winding
is 7.5 T. The winding is constructed from a composite conductor with a copper
matrix section, 3.5 mm x 8 mm. The copper to superconductor section ratio varies
from twenty-eight for the outer sections to six for the inner sections. Full cryostatic
stabilization has been provided for. Magnetic forces are mainly transmitted to a
76
v. v. SytcbeY
thick-walled tank of liquid helium. It is planned to use this magnet system with a
MHD generator having an electric power output on the order of 1 MW.
CONCLUSIONS
From this review it is evident that some progress has already been achieved in the
development of superconducting magnet systems generating transverse magnetic
fields in a channel. The experience which has been gained over the past few years in the
various laboratories around the world gives reason to hope that these activities will be
expanded in the near future.
REFERENCES
I. W. Gilbert, "Summary of International Progress on Superconducting Magnets," paper presented at
Particle Accelerator Conference, San Francisco, California, March 5-7, 1973.
2. "Towards a European Superconducting Synchrotron," GESSS-I, May 1972.
3. J. P. Blewett, in: Proceedings 8th Intern. Conference on High Energy Accelerators, CERN (1971),
p. 501.
4. A. L. Mints, A. A. Vasiljev, and E. L. Burshtein, in: Proceedings 7th Intern. Conference on High Energy
Accelerators, Vol. 1, Erevan (1969), p. 60.
5. W. S. Gilbert, R. B. Meuser, and F. L. Toby, in: Proceedings 4th Intern. Conference on Magnet Technology, Brookhaven National Laboratory, New York, September 19-22, 1972, p. 324.
6. J. H. Coupland and D. E. Bangham, in: Proceedings 4th Intern. Conference on Magnet Technology,
Brookhaven National Laboratory, New York, September 19-22, 1972, p. 737.
7. H. Brechna and M. Green, in: Proceedings Applied Superconductivity Conference, IEEE Pub!. No. 72
CH0682 5-TABSC (1972), p. 226.
8. G. Bronca, P. Genevey, F. Kircher, J. Perot, J. P. Pouillange, and G. Prast, in: Proceedings 4th Intern.
Conference on Magnet Technology, Brookhaven National Laboratory, New York, September 19-22,
1972, p. 203.
9. A. Berruyer, R. Blondet, G. Bronca, P. Genevey, F. Kircher, J. Perot, and J. P. Pouillange, in:
Proceedings 4th Intern. Conference on Magnet Technology, Brookhaven National Laboratory, New
York, September 19-22,1972, p. 316.
10. Chern. Eng., February 18 (1963), p. 96.
II. V. A. Kirillin, A. E. Sheindiin, E. I. Asinovsky, V. V. Sytchev, V. B. Zenkevitch, A. M. Maksimov,
and V. A. Altov, Doklody Akad. Nauk SSSR, 177:77 (1967).
12. V. A. Kirillin, V. A. Altov, E. I. Asinovsky, A. N. Dremin, F. I. Dubovitsky, V. B. Zenkevitch,
Yu. A. Kusnetzov, E. F. Lebedev, S. D. Savrov, V. V. Sytchev, and A. E. Sheindlin, Doklady Akad.
Nauk SSSR, 185:316 (1969).
13. Z. J. J. Stekly, in: Les champs magnctiques intenses, Colloques internationaux du CNRS, Grenoble,
Septembre 12-14, 1966, Editions du CNRS (1967), p. 237.
14. T. Doi, H. Kimura, S. Sato, K. Kuroda, H. Ogata, and U. Kudo, Cryogenics, 8:290 (1968).
IS. I. Todoriki, K. Koyama, A. Matsumura, Y. Matsuda, H. Hitotsuyanagi, K. Tada, and T. Murayama,
Sumitomo Electr. Tech. Rev., N-Il:65 (1969).
16. K. Koyama, K. Ushio, I. Todoriki, S. Shimamoto, T. Onishi, K. Kaiho, K. Agatsuma, H. Nomura,
Y. Kawasaki, M. Obata, K. Miura, K. Komuro, and S. Tamura, in: Proceedings ICEC-J, Iliffe Sci.
and Tech. Pub!., London (1971), p. 351.
17. V. V. Sytchev, V. B. Zenkevitch, and V. A. Altov, Pribory i Tekhn. Eksperimenta, N-2: 185 (1972).
18. H. Kimura, T. Doi, S. Sato, T. Kasahara, and T. Iizuka, in: Proceedings ICEC-J, Iliffe Sci. and
Tech. Pub!., London (1971), p. 364.
19. Z. J. J. Stekly, R. J. Thome, E. J. Lucas, and R. F. Cooper, in: Proceedings ICEC-4, Iliffe Sci. and
Tech. Pub!., London (1972), p. 218.
20. K. Fushimi, K. Akashi, O. Ogino, T. Moriguchi, M. Iwamoto, T. Satow, M. Tanaka, I. Hirata,
K. Ishihara, and K. Fujiwara, in: Proceedings 5th Intern. Conference on MHD Electrical Power
Generation, Vol. I, Munich, Germany, April 19-23, 1971, IAEA, p. 459.
21. T. Bohn and P. Komarek, in: Proceedings 4th Intern. Conference on Magnet Technology, Brookhaven
National Laboratory, New York, September 19-22, 1972, p. 265.
22. Y. Aiyama, K. Fushimi, K. YasukOchi, T. Kashara, R. SaitO, and H. Kimura, in: Proceedings ICEC-4,
I1iffe Sci. and Tech. Pub!., London (1972), p. 227.
Superconducting Magnedc Systems for Generating Transyerse Magnetic Fields
77
23. B. I. Verkin, A. V. Pogorelov, E. A. Amelin, V. M. Boytshuk, B. E. Zaporozhchenko, E. M. Medvedev,
V. N. Pavlov, and A. E. Janov, Cryoelectricity and Power Engineering, Naukova dumka Publishing
House, Kiev (1973), p. 6.
24. Institute of High Temperatures of the USSR Academy of Sciences, 1972 Annual Report, Nauka
Publishing House, Moscow (1973), p. 68.
C-l
EUROPEAN PROGRESS IN CRYOPOWER
TRANSMISSION*
G. Bogner
Siemens, Aktiengesellschaft
Eriangen, Germany
INTRODUCTION
European research and development activities in cryopower transmission were
initiated more than ten years ago. In fact, as early as 1963 the English BICC (British
Insulated Callenders Cables Ltd.) decided to design and build a laboratory-scale
superconducting link in order to test the feasibility of superconducting ac transmission. This link was designed for a high-current, low-voltage transmission. Its length
was about 3 m and it consisted basically of a single-phase conductor system in the
form of a coaxial pair of tubular niobium conductors. By the end of 1967, the work
had achieved a superconducting ac transmission of 2080 A e]. Results of another
early study on cryogenic cables were presented by Klaudy [2] of Graz, Austria at the
1965 Cryogenic Engineering Conference.
Research on cryoresistive cables was carried out and completed during this
period. In this connection, a dc low-temperature aluminum cable project should be
mentioned which was performed jointly by two French companies, CGE (Compagnie
Generale D'Electricite) and Air Liquide. This project successfully tested a 20-m-Iong
dc cable (cooled to 25 K by pressurized helium gas) up to 3500 A and 40 kV [3].
Since those early studies, European efforts in developing cryogenic cables have
increased considerably. Today several laboratories are engaged in this research,
working exclusively on superconducting cables. The following European laboratories
are now performing work on superconducting cables: Central Electricity Research
Laboratories (CERL) in Leatherhead, England, Centre de Recherches de la CGE and
Air Liquide France, Electricite de France (with extensive theoretical studies), Laboratoire Central des Industries Electrique (LCIE) in France, Anstalt fUr Tieftemperaturforschung (ATF) in Graz, Austria, the study group of AEG-Kabelmetal-Linde and
the research laboratories of Siemens AG in Germany. This presentation will outline
the state of the art, present a brief description of the different research projects by
these laboratories, and, where possible, provide information about future plans.
PROJECT DESCRIPTION AND CABLE TYPES
UNDER DEVEWPMENT
CERL (Leatherhead), ATF (Graz), and Siemens have essentially investigated ac
cables from the very beginning; however, they have not excluded theoretical studies
* Invited paper.
78
European Progress in Cryopower Transmission
79
on dc cables. CGE has done practical work on both types but prefers ac cables. AEGKabelmetal-Linde have concentrated their earlier efforts on dc cables, but intend to
emphasize ac cables in the future. As far as rigid cables are concerned CGE proposed a
± 100 kV, 4 GW dc cable which consists offour helium-tight tubular conductors with
an electrical insulation of evacuated multilayered Mylar. The four conductors are
transposed helically to take care of both thermal contractions and magnetic forces [3].
Siemens and CERL in an earlier stage of their programs also developed rigid ac
cables. These developments were successfully concluded several years ago. Since that
time, both laboratories have favored a cable type which is designated as semiflexible [4-6]. The work of CGE-Air Liquide has also focused on this type of cable
which, in the ac case, consists of a rigid thermal insulation jacket and three (or
multiple) flexible coaxial conductor pairs contained within the insulation jacket. Figure
1 shows a cross section of this type of cable as developed by CERL. The design is also
representative of the systems studied by CGE-Air Liquide and Siemens, with the
exception that the designs of the latter two provide a separate thermal insulated tube
for the helium return flow. The preliminary design specifications for the ac cables on
which the research and development of the three laboratories are based are the following:
CGE-Air Liquide: 125 kV, 3000 MW
120 kV, 2500 MW
Siemens:
CERL:
132 kV, 1400 MW and subsequently
275 kV, 4000 MW
Figure 2 shows a section of a coaxial conductor fabricated by Siemens. The actual
phase conductor consists of Nb-Al wires helically wound on a flexible former. It is
separated from the return or screen conductor (also of Nb-Al wires) by a wrapped
multilayer insulation of polyethylene. Figure 3 shows the corresponding conductor
Fig. I. Cross section of a semiflexible superconducting ac cable.
(Designed by CERL, but also representative for CGE and Siemens.)
O I ELECTR~
80
G. Bogner
Fig. 2. Flexible ac cable conductor
(developed by Siemens).
fabricated by CGE. Instead of Nb-Al wires, this conductor uses Nb-Al strips-the
only striking difference from the Siemens conductor. These conductors are suitable
for winding on transportable drums in lengths up to 500 m and for inserting them into
the rigid, thermally insulated envelope in the field. They are also able to accommodate
thermal contraction.
The work of ATF (Graz) and AEG-Kabelmetal-Linde has been concentrated on
totally flexible cable types where not only the conductor but also the cryogenic jacket
are flexible [,.8]. Figure 4 shows a model of a totally flexible dc cable (one pole) which
was fabricated by AEG-Kabelmetal-Linde as part of their program to develop a
±200 kV, 5 GW dc cable. Replacing the hollow conductor (shown in this figure),
which is subdivided into helical strips, by a coaxial pair of corrugated tube conductors
would lead to the ac cable design of ATF (Graz). These complete cables are suitable
for winding on a drum and for transportation in lengths up to 200 m. However, at
Fig. 3. Flexible ac cable conductor (developed by
CGE).
81
European Progress in Cryopower Transmission
- S<worI rings
_ Su~sq,s
_I---:-~-
- &:-.
iL---~~ -_ Saeen
m.Gtion
_ imennosl c:orrugaed lbe
_
~
and spaceos
1>1ho vacwn
- 2nd conugaIod tube
- Dud tcr liquid niOOgen,
metaHic spac:as
_ Superinsulalion and spac:as
ln1ho_
Fig. 4. Construction of a single-phase flexible
dc cable (AEG-Kabelmetal-Linde).
higher power ratings, they have to be fabricated and installed as single-conductor
systems, resulting in higher thermal losses and wider trenches.
THERMAL INSULA nON
Rigid as well as flexible thermal insulating envelopes are provided within the
European cable designs. Every concept employs a thermal shield, between the
ambient temperature steel pipe and the helium tube, which is kept at an intermediate
temperature of about 80 K by liquid nitrogen. In general, superinsulation in the form
of multilayered aluminum-coated Mylar foil is placed between the outer steel pipe
and the thermal shield. This kind of superinsulation is preferred because of its good
mechanical behavior and handling. CGE-Air Liquide, however, recommend the use
of alumina powder insulation because their design permits the rigid cryogenic jacket
to be built in the field, rather than being prefabricated in the factory in 20-m-Iong
sections and transported to the site. They claim that it should be easier to handle the
powder on the site since it requires no winding. Figure 5 shows an 18-m, full-scale
model of such a cryogenic jacket, successfully tested at the Research Center of Air
Liquide at Sassenage [3].
In the case of a rigid thermal insulation, the interior tubes and thermal shields are
suspended or supported by thin steel or plastic cables or by fibreglass-reinforced epoxy
resin studs. For the flexible cryogenic jacket, the concentrically placed corrugated
tubes are separated by helical spacers fabricated from low thermal conductivity
plastics. To minimize the thermal contraction problems for the rigid tube insulation, a
material with a low thermal contraction (such as Invar) is used for the helium pipe
and the nitrogen shield. With a flexible cryogenic envelope of corrugated tubes,
evacuation of the space between the helium pipe and the nitrogen shield may be
G. Bogner
82
Fig. 5. Full-scale 18-m cryogenic
envelope model suitable for a superconducting 3000- MW ac cable
(eGE).
troublesome, since the latter consists of two concentric tubes forming a tight annular
channel. This problem can be solved by purging this space with carbon dioxide until
the air is removed, and then evacuating the carbon dioxide to a low pressure. When the
shield is cooled with liquid nitrogen, the residual carbon dioxide is condensed as a
solid on the cold surface
The thermal losses of the different types of insulation have been determined in a
series of experiments. The rigid type with multilayer superinsulation has the lowest
values, namely ~2 W/m 2 at 80 K and 0.1 W/m2 at 4.2 K. The flexible type has the
highest values, i.e., 4 W/m2 at 80 K and 0.4 W/m2 at 4.2 K. (The area dimension in
both types refers to the surface area of the liquid nitrogen shield at 80 K or that of the
helium tube at 4.2 K). The thermal conductivity of the powder insulation is about a
factor of three to four higher than that of the multilayered superinsulation.
n.
CONDUCTORS
For dc cables, the CGE group has produced tubular and strip conductors
consisting of commercial NbTi multicore conductors embedded in high-purity
aluminum by an extrusion process. The AEG-Kabelmetal-Linde group has provided commercial Nb 3 Sn ribbons, stabilized by thick copper layers soldered to the
ribbons, in support of their dc cable project.
For ac cables, niobium in the form of thin layers on substrates of copper or
aluminum is the preferred conductor material. CERL (Leather head), in collaboration
with IMI and Siemens, together with Vacuumschmelze Hanau, have developed coextrusion processes by which rigid tubular copper niobium composite conductors can
be fabricated in lengths up to 20 m and diameters of around 10 cm [5.9]. The thickness
of the niobium layers varies from 25 to 50 jlm. For flexible conductors which are
subdivided into helical strips or wires, both CGE and Siemens have developed highpurity aluminum wires and ribbons (resistivity ratio ~ 2000) which are covered on all
surfaces with thin layers of niobium. The latter is to avoid eddy current heating in the
normal metal. The bond between the aluminum and niobium was found to be as
good as in the case of copper and niobium. Thus, these kinds of conductors should
provide reasonable fault current capability. But the fault-current behavior still needs
thorough experimental verification. Appropriate experiments are under preparation.
83
European Progress in Cryopower Transmission
Another solution of the fault current problem has been proposed by CERL, namely,
the use of a triple structure of Cu-NbTi-Nb, where the overcurrent could flow in the
NbTi, exhibiting lower losses than would occur in the aluminum under fault current
conditions [10]. ATF (Graz) in collaboration with Kabelmetal (Hannover) are engaged in the development of corrugated tube conductors consisting of a Nb-Cu or
Nb-AI combination. These conductors would have the advantages of flexibility, long
length fabricability, and balance of axial thermal contraction.
Detailed theoretical analysis of the origin of the Nb surface losses has been
carried out at CERL [11 J and many laboratories have made extended loss measurements on niobium conductors stabilized with normal metals and fabricated under
real manufacturing conditions. Figure 6 shows the 50-Hz losses of Nb-AI wires vs. the
peak flux density on the conductor surface at temperatures from 1.6 to 6 K. These
results, obtained by Siemens, demonstrate that this type of niobium conductor can be
operated at surface peak flux densities up to 0.1 T and temperatures up to 6 K without
exceeding, on average, the permissible dissipation of 0.1 W/m2 [12]. Similar results
have been obtained by CGE with a 5.1-m-Iong flexible coaxial cable consisting of two
layers of fifty-four rectangular Nb-AI conductors [3J. CERL, on the other hand,
measured the hysteresis losses of a 5.5-m-Iong coaxial Cu-Nb rigid tube pair (63 and
100 mm diameter) and found the losses to be lower than 0.1 W/m2 at 5 K and a surface
current density of 40 kA/m [6]. Figure 7 shows the CERL test facility in which the loss
measurements were made. The equipment is also suitable for fault current simulation
(maximum current of70kA for 1 sec) and for separate dielectric experiments with lineto-neutral voltages up to 100 kV and impulse voltages up to 300 kV.
With regards to possibly a better fault current capability, Nb 3 Sn was also
considered by Siemens as an ac conductor material. Alternating current loss measurements on laboratory-fabricated, vapor-deposited material, however, indicated that the
50-Hz dissipation of the material was about a factor of two to three too high to be
competitive with Nb, even when taking into account the possible higher operating
temperatures. Continuation of this work is planned when better material is available.
ELECTRICAL INSULA nON
As in the case of conventional cables, the quality of the electrical insulation is of
crucial importance to the performance of a superconducting cable. Therefore,
10"
Fig. 6. Losses of niobium at various temperatures at 50 Hz.
6o--..J.ao----'100
--1.L..2o~14L...O-1..L60-1...L80-2....L.OO--'-m....J
T - 2.J....J
10.3 '-----'60
~
Stlrf.ac~ flux d@f1sltV
84
G. Bogner
Fig. 7. View ~fthe test facility at CERL
during final assembly (photo by courtesy of DOC).
relatively early laboratory-scale measurements on the dielectric strength arid losses
of the various interesting electrical insulations have been initiated by all the European
laboratories. The following types of electrical insulation have been considered to date:
liquid helium under various conditions (temperature, pressure), vacuum, and wrapped
multilayer insulation of normal paper or synthetic tapes in vacuum or impregnated
with helium. Test facilities for investigating the dielectric strength of the different
types of insulation under various conditions with dc and ac voltages up to 200 kV
and impulse voltages up to 300 kV are available. Figure 8 shows the 200-kV ac test
facility of Siemens.
Helium, with temperatures of 4.2 to 300 K and pressures of 1 to 10 bar, has been
investigated with electrode spacings up to 70 mm by CGE, LCIE, ATF (Graz), CERL
and Siemens C,13-15]. A significant degradation of the dielectric strength with
Fig. 8. 200-kV ac breakdown test
facility (Siemens).
8S
European Progress iD Cryopower Transmission
e
increasing temperature was observed (for instance, the dielectric strength at 4.2 K is
20 kV/mm, while at 10 K, it is only 6.5 kV/mm 3 ]. Application of elevated pressure
leads to an improvement ofthe dielectric strength (for example, at 10 bar, the dielectric
strength is 70 % greater than at 1 bar) [13]. An enlargement of the electrode gap also
leads to a remarkable diminution of the dielectric strength (1 mm, 26 kV p/mm;
10 mm, 16 kV p/mm; with subscript p indicating peak value).
Investigations of cryovacuum as an insulator have been conducted by CGE and
LCIE. At smaller gaps, vacuum is superior to helium; at larger gaps (> 10 mm), the
reverse seems to be true. Considerable attention has also been focused on the magnitude of prebreakdown currents in vacuum. Experimental results show that the currents
are significantly reduced (by factor of 102 to 104) when lowering the temperature
from 300 to 4.2 K. With covered electrodes, the dissipation due to these currents can be
kept within permissible limits at reasonable field strengths. Problems associated with
solid spacers in vacuum were also explored. Tests demonstrated that with selected
materials and shapes of the spacers, the dielectric strength reaches about 80 % of the
bare vacuum value [3].
In view of the promising flexible cable constructions, a series of ac and dc breakdown measurements was carried out on solids in the form of single foils or wrapped
multilayer configurations immersed in helium or vacuum. The materials investigated
to date are: conventional insulating paper, Mylar, polyethylene, and Tyvek. Generally,
the dielectric strength of the combination of solids and helium or vacuum is higher
than that of the pure immersion medium, with one exception, namely Tyvek. Within
the applied gap widths, the breakdown voltages of this material are less or equal to
those of helium.
Figure 9 shows a compilation ofthe results of 50-Hz breakdown voltage measurements obtained by European laboratories. From this figure, one notes that, in general,
the dc dielectric strengths are higher than the ac dielectric strengths and that vacuum
is superior to helium as an immersion medium. Based on the results of Fig. 9, one
anticipates that in real cable insulations (thickness 2:: 10 mm), dielectric strengths of
about 10 kV/mm (rms) for ac and up to 20 kV/mm for dc can be realized with the
insulating media considered above. With multilayer insulations wrapped in paper or
synthetic foils, not only is the dielectric strength of great importance, but so is the
mechanical behavior (with regard to bending, cooling, etc.). For this reason, extended
200~--~-----r---.r---'-----'----'
LHe '='liQuid Helium
mm
Vat =Vacuum
PE
100
•
....
50 ••••
• Mylar-Vac-D C
• ••••••••••
y.~~........
••••••••••• ,LHe
LHe
••••~ ••
~~
20
Supercritical He
----.:~;;.~.~
Tyoek·LHI
Fig. 9. Summary of electrical breakdown
measurement results obtained by European
laboratories.
DC
• PE·Supercr'Hcal He-O C
r
~
~Polyethylene
C.GE •••••••••••••• AC
Slemens---
- - . : : : . : ••• ~
10L---~----~--~----~----~---J.
0.1
0.2
0.5
2.0
thickness
86
G. Bogner
experiments are being conducted to find the best compromise between electrical and
mechanical properties.
In addition to the dielectric strength, the dissipation factor of solids at various
temperatures has also been determined. Figure 10 shows that nonpolar dielectrics,
such as polyethylene, have dissipation factor values, in the neighborhood of 10 - 5 at
temperatures of 4.2 K. Measurements of the loss angle were conducted at ATF (Graz)
and CGE by both electrical bridge and calorimetric methods.
CABLE COOLING
The cooling concepts of the European cable projects have all been based on a dual
fluid cooling by single-phase helium at 4 to 8 K and a single-phase nitrogen at about
80 K. The proposed solutions for the helium return flow, however, differ between
projects; some use it for cooling one conductor (CERL), while others form a heat
sink between the nitrogen shield and the actual helium pipe (Siemens). The investigation of the possibility of a single-fluid helium cooling cycle is presently being planned.
The various transmission line designs provide ranges of operating temperature and
pressure that are characterized by the following values: 77 to 84 K and about 5 bar for
nitrogen, 4.4 to 6 K and 3 to 4 bar for niobium cables, and 6 to 8 K and about 6 bar
for Nb 3 Sn cables. In the case of helium, the temperature and pressure ranges have been
selectively specified to take into account the maximum heat capacity of the coolant
and thus provide minimum mass flow rate. Care has also been exercised to achieve on
the average a positive Joule-Thomson coefficient (dp/dT > 0) as the pressure of the
helium decreases during its flow along the transmission line C].
In a series of experimental closed cooling cycles with single-phase helium, it was
verified that the design values are realistic. Cooldown and warmup procedures, as
well as steady-state operations, were performed near the transposed critical line
without any disturbing oscillations. To avoid the latter, especially during cooldown, it
is necessary to impose a long cooldown time. Evaluations ofthe cool-down procedures
for cable sections several kilometers long yield cooldown times of several weeks. In
contrast to the other European laboratories, where closed cooling cycles with
refrigerators were used for these experiments, the Siemens laboratories employed a
1O-"OL-.---10-'-O-----200..L-----K--'300
Temperature T
Fig. 10. Temperature dependence of dissipation factor of solid dielectrics; f = 50 Hz
(ATF, Graz).
87
European Progress in Cryopower Transmission
I/hr
600
500
>
I
~
400
300
~
Q
200
100
a
Fig. 11. Pumping capacity ora liquid Hepump.
0
60
80
100 120
150
180 200
240r p m
NumbEr of rotations
helium pump which was specially developed for this purpose. This pump, which is
capable of mass flow rates up to several thousand liters per hour, distinguishes itself
by an unusually high efficiency (> 90 %). It is also suitable for cooling a real cable line.
Typical pumping characteristics which are representative only for this special type of
pump (500 liters/hr) are given in Fig. 11.
Considerable effort has also been put into the design of proper connections for
the coolant helium to flexible cables. This connection can be accomplished either
within the terminals or, for longer cable lines, by individual intermediate connections.
In every case, part of the helium has to be fed into the conductor core, passing through
the wrapped electrical insulation across which the line-to-neutral voltage is applied.
It is a common feature of all the proposals made to date that the helium passes through
an axial channel in the electrical insulation. By either using materials of different
permittivity for the electrical insulation or interleaving metallic layers, the electrical
field strength in the channel can be linearized.
TERMINALS AND JOINTS
With respect to the design of cable terminals, it is assumed that superconducting
cables are used as links in an otherwise normally conducting system. Thus the problem
arises of bringing heavy currents at high potential from ambient to low temperatures
with losses as low as possible. Terminals for dc and ac voltages up to 200 kV and
currents up to 12.5 kA have already been designed and important components have
already been constructed and tested. In general, gas-cooled current leads for reducing
the heat flow from the ambient temperature to the low-temperature end are preferred,
since considerable experience with current leads used in large superconducting
magnets is available in most laboratories. The gas-cooled lead where the conductor
acts as a heat exchanger has the same losses as one cooled at discrete temperature
levels by heat exchangers connected with a refrigerator, but is easier to construct
than the latter. The problems of transmitting the high voltage from ambient to low
temperatures were solved to a large extent during the construction of reliable highvoltage terminals for measuring the dielectric strength at low temperatures. Both the
experience with gas-cooled current leads and that with high-voltage terminals have
88
G. Bogner
provided guides for terminal designs which, not surprisingly, are in many respects
similar to conventional terminals.
Figure 12 shows the design for a 200 kV, 12.5 kA dc terminal [8]. The upper lefthand side of the figure shows the conductor, constructed as a heat exchanger (about 1
m high), in which the transition from helium to ambient temperature takes place. The
superconductors (strips in the present case) are connected to the base of this heat
exchanger. For a good helium flow to the cable core and the exchanger, the cable
insulation is perforated. The insulation in this part and around the heat exchanger
is a fiberglass-reinforced epoxy resin tube. The gradual transition from the smalldiameter cable insulation to the larger tube diameter is performed by a paper cone. A
suitable duct in this cone allows the helium to enter the heat exchanger and conductor
core. On the right-hand side of Fig. 12, the upper part of the terminal is shown. This
part of the terminal consists of a conventional porcelain insulator, again using a paper
cone to provide the transition from the smaller diameter ofthe epoxy resin tube to the
larger diameter where the electrode ends. The hollow copper bus bar through which the
warm helium gas returns to the refrigerator has a cross section of 20,000 mm 2 in order
to limit the Joule heating at 12.5 kA to 350 W/m.
The following problem is encountered with helium gas-cooled leads: The warm
helium gas, which has an extremely low dielectric strength (O.l5kV/mm), has to be
brought from a high to a low potential. Two elegant methods for an easy solution of
this problem were suggested by Siemens [16.17]. It was experimentally verified that
the addition of a small volume of electronegative gases to the helium gas or the use of
finely subdivided tubes with large electron trapping surfaces considerably improves the
dielectric strength of ambient temperature helium gas. In the first case, the added gas
can easily be condensed from the mixture at ground potential.
Detailed experiments have also been carried out to prepare for the successful
realization of high-performance cable joints. In these experiments, it was substantiated
that niobium--<:opper composite conductors in the form of rigid tubes can be welded
without substantially increasing the ac losses at the welded seam. For this purpose,
the ends of the composite tubes were provided with pure Nb rings during fabrication
!~r~~~~s bar ---t~
290 K
Heet exchanger
~~e~~brr pOJeelain,_---.:l'"r-N
FibrB'llass·
reinforced
cast resin
R::Itential areas
metalli,ed
Field control cone
--'I~_- Helium admission
F;o,ld control cone _~.
Cryogenic jock.,
Fig. 12. 200-kV terminal for a superconducting
dc cable (constructed by AEG).
European Progress in Cryopower Transmission
89
which permitted a pure niobium welded seam. In the case of flexible cables, the
connection of conductors with aluminum cores or NbTi conductors with aluminum
claddings can be accomplished by welding just the aluminum. Nb 3 Sn ribbons with
copper strips as stabilizing material may be overlapped and soldered. The resistance
of these normal conducting contacts is low enough and their spatial extension so
small that the heat generated within them can be neglected. The thickness of the
wrapped multilayer electrical insulation is enlarged near the joint and is fabricated
by self-welding polyethylene tape. Additional bellows to compensate for thermal
contraction of the cold pipes are also installed near the joint. It is a general feature of
these joints that the interior parts are fastened to the outer ambient temperature tube
to prevent axial movements.
PRESENT AND FUTURE STUDIES
Besides the larger test facilities already mentioned in the foregoing sections (Figs.
5 and 7), some other tests with large model cables have been carried out or are planned.
ATF in Graz, Austria in collaboration with AEG-Telefunken, Kabelmetal, and
Linde AG in Germany have built a flexible cable of 50 m length consisting of four
corrugated tubes, the innermost one containing a few copper wires coated with lead.
The electrical insulation was provided by the liquid helium coolant and polyethylene
spacers. The cable section was designed for a line-to-line voltage of 20 kV and a current
of 5 kA. The test section, wound in a large helix, was used to study the thermal behavior
of a cable under actual operating conditions. Figure 13 shows the test arrangement.
Currently, a 15-m-long piece of flexible cable with Nb 3 Sn conductors and
wrapped paper insulation (construction as in Fig. 4) is being tested in an oval loop
at the AEG laboratories to provide dc current data and information on the thermomechanical behavior of the cable. The cable contains (in one of its straight lengths) a
Fig. 13. 50-m-flexible ac model cable (ATF, Graz).
90
G. Bogner
joint in which the Nb 3 Sn conductor loop is short-circuited. The cable current (several
tens of kiloamperes) is induced into the loop by four transformerlike inductors, so
that no terminals are required in the test. A high dc voltage test (200 kV to ground)
with a 20-m-Iong flexible cable section and two terminals is also in preparation at this
company. The tests will be carried out toward the end of 1974, concluding the first
phase of the development program of the AEG-Kabelmetal-Linde group.
At the Siemens laboratories in Erlangen, a 30-m-Iong, single-phase ac test cable
is under construction. The cryogenic envelope for this cable is rigid. The conductor is
flexible and consists of Nb-AI wires helically wound on a flexible hollow former.
Actual conductor and shield are separated by a wrapped multilayer insulation. The
test length is equipped with terminals on both ends adequate for an operational
voltage of 120 kV and a rated current of 10 kA. The program provides synthetic
current and voltage tests, which will start during the first half of 1975.
With these tests, the first part of the Siemens cable program will be completed
at the end of 1976. In the second phase, it is planned to study both operational safety
and control in more detail by constructing a test line with a length of around 100 m
and operating it under real conditions in parallel with an existing line. A similar
program is planned by the AEG-Kabelmetal-Linde group but with striking differences in the concept, particularly with regards to the conductor material, the thermal
insulation, and the mechanical construction of the cable; thus, both programs will
provide a valuable contribution. (It should be noted, however, that the funding for
these programs has not yet been appropriated). The ATF (Graz) also plans to install a
60 kV superconducting experimental link of around 50 m in parallel with an
existing overhead line. It is the hope of AFT (Graz) that the project can be completed
early in 1975.
The funded programs of CERL (England) and CGE (France) extend to the end of
1975. The aim of the research at CERL within this time frame is to determine whether
superconducting cables are technically and economically feasible for bulk transmission of electrical power in and around cities. The installation of longer test cables
is apparently not planned within this time period. The same seems to be true for
CGE, which intends to continue the measurements on dielectrics and conductors.
SUMMARY
A number of ac and dc superconducting cable projects presently exist in Europe.
Special attention is being focused on semiflexible cable types, i.e., with rigid cryogenic
envelopes in which flexible conductors are contained. In addition, totally flexible
cables are being developed which enclose flexible conductors in flexible cryogenic
jackets constructed with the aid of corrugated tubes. Intensive investigations of the
properties of thermal insulations, stabilized superconductors, and electrical insulations have been carried out and are being continued. Special manufacturing processes
for niobium-copper or niobium-aluminum ac composite conductors have been
developed which exhibit 50-Hz losses at 4 to 6 K and at surface peak flux densities of
0.1 T and which are well below the design limits of ac cables. Measurements of the
breakdown voltages and dissipation factors at about 4 K indicate that wrapped
multilayer insulation of polyethylene tapes will be a suitable electrical insulation.
Different types of dc and ac joints and terminals have been designed or have already
been constructed and tested. A number of larger laboratory model tests with cable
lengths of 10 to 50 m have been carried out or are planned for the near future with the
aim of studying the interaction of the different components and of demonstrating the
European Progress in Cryopower Transmission
91
feasibility of superconducting cables. Tests with longer sections under real operating
conditions parallel to existing conventional lines will follow in a few years.
ACKNOWLEDGMENTS
The author wishes to thank the colleagues who supported him in the preparation of this paper. He
particularly acknowledges the assistance ofP. A. Klaudy, ATF (Graz, Austria), J. A. Baylis, CERL (Leatherhead, England), P. Dubois, Centre de Recherches de la CGE (France), and last, but not least, H. Heumann
and H. Voigt of AEG-Telefunken (Germany). The work on superconducting cables in Germany is supported by the Bundesministerium fiir Forschung und Technologie.
REFERENCES
I. D. R. Edwards, Electrical and Electronics Tech. Eng., 1968 (March): I.
2. P. A. Klaudy, in: Advances in Cryogenic Engineering, Vol. 11, Plenum Press, New York (1966), p. 684.
3. P. Dubois, I. Eyraud, and E. Carbonell, in: Proceedings Applied Superconductivity Conference, IEEE
Publ. No. 72 CH0682 S-TABSC (1972), p. 173.
4. G. Bogner and F. Schmidt, Elektro-Technische Zeitschrift, 92:740 (1971).
S. G. Bogner, "Supraleitendes Hochleistungs-Drehstromkabel," presented at the Herbstschule iiber
Anwendung der Supraleitung in Elektrotechnik und Hochenergiephysik, Oktober 1972, Titisee,
Gesellschaft flir Kemforschung, Karlsruhe, Germany (1973), p. VI.
6. J. A. Baylis, Electrical Times, 1973 (Issue 4232, May 24):9.
7. P. A. Klaudy, Elektrotechnik und Maschinenbau, 89:93 (1972).
8. H. Heumann, "Kabeltechnische Probleme bei Projektierung und Bau von Obertragungsstrecken mit
Supraleitem," Mitteilungen der Kabelwerke der AEG-TeIefunken-Gruppe (April 1972), pp. 1-7.
9. C. G. Barber and B. J. Maddock, in: Proceedings Applied Superconductivity Conference, IEEE Publ.
No. 72 CH0682-S-TABSC (1972), p. 211.
10. M. T. Taylor, in: Proceedings Conference on Low Temperatures and Electric Power, IIR, London
(1969), p. 61.
11. P. H. Melville, J. Phys. C, 4:2833 (1971).
12. P. Penczynski, "Messung der Temperaturabhiingigkeit der Wechselstromverluste von supraleitendem
Niob," presented at DPG Spring Meeting, Miinster, Germany (1973).
13. J. Thoris, B. Leon, A. Dubois, and J. C. Bobo, Cryogenics, 10: 147 (1970).
14. R. J. Meats, Proc. IEEE, 119(6):760 (1972).
IS. Siemens Research Laboratories, unpublished results.
16. F. Schmidt and P. Massek, "Stromzufiihrung fiir elektrische Einrichtungen mit auf Tieftemperatur
gekiihlten Leitem," Deutsche Patent Auslegeschrift No. 2,164,706 (1973).
17. F. Schmidt, G. Matthiius, and P. Massek, "Stromzuflihrung flir elektrische Einrichtungen mit auf
Tieftemperatur gekiihlten Leitem," Deutsche Patent Auslegeschrift No. 2,163,270 (1973).
C-2
RESEARCH ON A LABORATORY MODEL OF
SUPERCONDUCTING TEST CABLE
D. V. Razevig, Y. L. Blinkov, and Y. S. Goldenberg
G. M. Krzhizhanovsky Power Research Institute
Moscow, USSR
INTRODUCTION
Superconductors make it possible to increase the current density in currentcarrying components by over two orders of magnitude and to bring the capacity of a
single line up to 10,000-30,000 MW. At the same time, the sizes and the total energy
losses are substantially reduced during transmission.
The development of superconducting power transmission lines includes a large
number of closely related tasks; however, since different branches of science and
technology are involved, the approach to these tasks is as independent as the disciplines
themselves. The distinguishing feature of superconducting lines is the presence of
two kinds of energy transport: electrical, which determines the electromagnetic state
of the superconducting cable, and thermal, which determines the thermal state. The
relationship between these two states is very close and preconditions the stability of
the superconductor when it is in the superconducting state.
The variety of scientific and technical tasks to be solved requires that the research,
development, and technological work be pursued on a broad basis and following a
unified plan and program involving several consecutive phases.
RESEARCH PROGRAM
Phase One. The first phase of the program covers a study of physical and mathematical models to describe the processes and obtain the necessary initial data for
substantiating the choice of materials and the key engineering solutions for ac and dc
superconducting cables and their components. At this stage, design methods are also
to be developed for the electrodynamic, thermophysical, and hydrodynamic processes
occurring in superconducting cables. It is only through a fundamental understanding
of these principles that the characteristics and possible parameter ranges of the
superconducting cable can be determined. The· final choice of these parameters
requires a thorough technical and economic study, including questions of defining
expedient fields for employing superconducting cables within power systems, their
operational stability and reliability, their behavior as components of a power system,
and a rational control of the operating conditions.
Phase Two. This phase is limited to the development of a commercial prototype
of the superconducting cable including the manufacturing technology and technical
solutions for manufacturing the superconducting components of the test line.
91
Research on a Laboratory Model of Superconducting Test Cable
93
Phase Three. This portion ofthe program involves a comprehensive investigation
and testing of the individual components of the line to determine the operating,
critical characteristics and parameters, compare various construction versions of
superconducting cables, and also examine the constructional and technological
solutions.
Phase Four. The final phase of the program covers building and testing of an
experimental line on the test facilities but within the power system. The specific
operational peculiarities of a long-distance transmission line are to be studied here,
since they cannot be determined on short cable sections. This includes study on such
questions as thermal stability, distribution of the pulsating waves associated with the
flow of coolant along the line, the shifting rate of thermal fluctuations along the line,
the impact and interrelationship of the electromagnetic process upon the thermal
one, the distribution of vacuum by cryopumping, and the effect of vacuum waves.
The G. M. Krzhizhanovsky Power Research Institute is engaged at present in
carrying out the first phase of the program. In 1971, researchers at the Institute
constructed the first laboratory model of a single-phase superconducting cable which
operated in a semiopen circuit. Components of the system included a gas holder,
refrigerator, and superconducting cable with niobium superconductor. The cable was
energized by both ac and dc. The laboratory model verified the possibility of attaining
the design current density in the superconductor with forced coolant circulation,
and provided information on the efficiency and stability of the model's components.
An improved model of a three-phase superconducting cable was then designed and
built on the basis of the data obtained from the first model.
The main purpose of developing this modernized model was to investigate the
principal technical, technological, and constructive solutions to be embodied in a
cable of this kind and to obtain the interrelation between individual unit elements
and systems under heavy current loadings and forced coolant circulation. Some
matters of concern, such as the choice of the coolant state for commercial lines, the
electrical insulation material, etc., were also to be investigated experimentally.
SUPERCONDUCTING CABLE
Figure 1 shows a schematic of the experimental system including a three-phase
superconducting cable (I), an electrical power source (II), a cryogenic-vacuum system
(III), and a metering system (IV). The three-phase superconducting cable includes
all the components of a possible commercial line. The test cable is made up of two
end couplings (A), two segments of cable (B), and an intermediate coupling (C). The
cable segments consist of a cryogenic jacket and conducting system and are the main
object of study.
Figure 2 depicts the cross section of the cable. The cryogenic jacket consists of
an outer protective shield (1) and a combined liquid nitrogen-cooled thermal shield
(2,3,4). The conducting system includes an electrostatic shield (5) and three conducting cores (8). Tubes, with a Nb 3 Sn superconductor deposited on the inside
surface, serve as core foundation. This is the case of a reverse composition, i.e., the
normal metal is in no way affected by the electromagnetic field. The space inside the
tubes forms a path for the flow of inlet coolant. The electrical insulation of the cores
also serves as insulation between the inlet and exit flows of coolant. The coolant
exits in the space between the electrostatic shield and the cores.
The 3-m-Iong cables are joined by an intermediate coupling. The end couplings
are mounted on the opposite terminals of the cables, and the coolant and electrical
D. V. Razevig, Y. L. Blinkov, and Y. S. Goldenberg
94
,- -
.- - - -
- -
--
-
-
- - - - - - - ----,
!~I ~r~~
,-
,mrnh
---------,
I
I
I
: )""1
I
I
I
I
L __ _____ _________ -1
I
I
r-
_J
I
:re,
I
I
-t.~~
...
I
' cNii
~~
I
I1_~ _..i"_
__
_
-
-
_
_
_
. 1
Research on a Laboratory Model of Superconducting Test Cable
95
Fig. 2. Cross section of superconducting cable and protective sheath.
(1) Outer protective sheath; (2)
vacuum-multilayer insulation; (3)
intermediate thermal shield; (4)
nitrogen shield; (5) electrostatic
shield; (6) inner centering and
protective sheath; (7) electrical insulation; (8) superconducting core;
(9) inlet line for coolant; and (10)
exit line for coolant.
power enter and leave from the experimental system through the couplings. The
galvanic current leads consist of a rigid part which connects the external busbars of
the supply source to the cold system and a superconducting, flexible part which connects the rigid current lead to the current core. Such a current structure of the leads
provides a linear compensation for temperature deformation of the cores and a
simultaneous removal of the heat inleak from the current leads because the flexible
bond is cooled directly by the main coolant.
The cable is cooled with helium (at temperatures of 4.5 to 12 K and pressures
up to 20 atm). The cooling process is carried out in the following manner: The cold
helium flows from the refrigerator to the regulator valve, which is mounted directly
at the entry to one of the end couplings. The coolant is then distributed in the insulation collector to the three current-carrying lines. After this, the coolant flows onto
the opposite collector, where a portion of it is removed to cool the current leads.
The remaining portion of the coolant flows back to the distributor through the interphase space.
The cryogenic system for the cable (Fig. 1) consists of a refrigerator-liquefier (2),
helium conduit (3), refrigerator control system (4), valve system for the flow regulation
(5), parameter control system (6), and preliminary cooling assembly (7).
Maintenance of the temperature and pressure parameters in the cable and simultaneous removal of some of the coolant to minimize heat inleaks from the end facilities
results in the complex operation of the refrigeration unit. The large coolant consumption by the current leads causes an imbalance in the lower heat exchangers of
the refrigerator and calls for special measures to eliminate this imbalance.
The cryogenic system is designed to fully cool the cable with a single refrigerator
without resorting to intermediate coolants (nitrogen, hydrogen, etc.).
The cryogenic system provides four basic stages of operation:
1. Refrigeration of the cable to 100 K by means of the preliminary cooling
assembly.
96
D. V. Razevig, Y. L. Blinkov, and Y. S. Goldenberg
2. Refrigeration of the cable to 25 K by means of a compressed gas motor.
3. Bringing of the cable to a steady state, with the working temperature between
4.4 and 12 K, by compressed-gas and throttle cycles.
4. Operation of the model under steady and unsteady conditions, with a refrigerant removal for cooling the current leads.
The electrical power source permits experimental tests using either dc or threephase ac of industrial frequency. The control system ensures a smooth current rise
from 0 to 10 kA. An automatic interlocker disconnects the source when one of the
lines in the superconducting cable goes normal.
The main design parameters and characteristics of the superconducting cable
are given in Table I. Some of the experimental results obtained while investigating
the three-phase superconducting cable are given in the remaining presentation.
COOLING OF CABLE
The investigation of the cable's behavior during refrigeration over a temperature
range from 300 to 5 K and the determination of the least possible refrigeration time
in relation to the thermal state of the current cores and helium vessel was carried
out in three stages.
Stage I-Reducing the temperature from 300 to 100 K. The helium coolant was
admitted to the preliminary cooling assembly and circulated through the cable by
means of the inlet and exit lines. As a result of the experimental investigation, the
optimum flow rate for the coolant was found to be 12 kg/hr at a flow velocity of
1.5 m/sec. The average rate of temperature decrease amounted to 25 KO/hr, while
the temperature in the nitrogen shield was maintained at 85 K.
Stage II-Cooling the conducting system from 100 to 25 K. In this case the
flow rate of coolant through the current leads dropped to 0.05 g/sec. This value was
determined on the basis of minimum heat in leak to the helium zone from the current
leads. The gas consumption in the refrigerator-cable-refrigerator scheme was 10 kg/hr
and the maximum rate of temperature decrease was 15 KO/hr.
Table I. Significant Design Parameters of the Superconducting Cable
Electromagnetic
Nominal current J.
at 4.5 K. kA
10
Test voltage. kV
35
Inductive phase reactance. H
at 4.5 K
2.4 x 10- 7
at 20 K
2.5 X 10- 7
at 300 K
2.6 X 10- 7
Interphase capacitance
at 4.5 K. F/m
46 x 10- 12
Losses in cable at
J. = 0.5. W/m
at 5 K
0.14
at 7.5 K
0.22
at 10 K
1.5
at 12 K
2.8
Thermophysical
20
Heat inleak to nitrogen
shield. W
0.39
Heat inleak to cable
system. W
Heat inleak through current
leads for a flow rate of
0.1 g/sec. W
0.064
at current of I kA
0.585
at current of 3 kA
1.676
at current of 5 kA
4.69
at current of 8 kA
Cryovacuum
Refrigerator capacity. W
at 4.5 K
at5.5K
at 7.5 K
dlOK
dl2K
108
134
147
I~
I~
Refrigerator capacity with gas
removal for cooling current
leads for a flow rate of
0.1 g/sec. W
at 5.5 K
117
at 7.5 K
139
at 10K
171
Vacuum in the cable.
10- 6
mmHg
Research on a Laboratory Model of Superconducting Test Cable
97
Stage III-Reducing temperature from 25 K to that required by the working
conditions ofthe experiment. This was done by adjusting the refrigerator. The amount
of coolant consumed by the current leads increased to 0.1 g/sec and the average gas
consumption in the exit flow was 3 kg/hr. The average rate of temperature decrease
in the conducting system was 4 KO /hr. Figure 3 shows the temperature change as a
function of time for the external piping system of the conducting system (curve 2)
and for the cores (curve 1). The operating conditions for the unit were smoothed out
by adjusting the cryogenic configuration which provided the maximum refrigeration
capacity for a given state. Thus, at 10 K, the coolant consumption in the inlet flow
was 3.5 g/sec for a refrigerator capacity of 180 W.
CRYORESISTIVE STATE
The investigation of the cryoresistive operating conditions was carried out for
the purpose of determining possible current loads under transition conditions and
heat release, temperature field distribution, time characteristics, thermal change in
current cores, and helium configurations. The experiments were carried out using
helium gas at 19 K as the inlet coolant to the cores. At this stage measurements were
made of the current, voltage drop in the cores and in the separate sections, voltage
T,K
l00_t_- -+ - --tr- - - - - -t -- - - - -_t
f~t-----~_++------_t_-----_t
Fig. 3. Temperature dependence as
a function of cooling time for the
superconducting cable.
o
-±_____--:;!,'l" h~5
L.._ _ _ _ _-I!:"-_ _ _ _ _
to
I()
40 '
98
D. V. Razevig, Y. L. Blinkov, and Y. S. Goldenberg
drop in the flexible bond and the current lead itself, and temperature values in the
middle of the core, at the exit from the core, and entry to the current lead, as well
as at several locations in the exit coolant piping system. Potential probes for metering
the voltage drop were soldered inside the tube wall and led out through the inner
channel of the core through the inlet path of the coolant and then through the
current lead to the metering instruments. The temperature sensors were mounted
on a special frame installed in the core. The sensors recorded the temperature of the
flow, the core wall, and the coolant through the current core.
Alternating current was supplied to all three phases. Negative direct current
was supplied to phase C, while positive dc was supplied interchangeably to either
phase A or phase B. Before energizing the superconducting cable, the temperature
gradient along the length of the core was 5 degrees. This gradient is preconditioned
by the existing heat inleaks at the given gas flow rate in the cable and the refrigerant
capabilities of the inlet and exit helium streams.
Figure 4 shows the voltage-current characteristics (curve 1) of phase 'A' with
the following conditions: inlet flow rate of 3.5 g/sec and flow through the current
leads of 0.3 g/sec. (The refrigerator was operated at such a capacity as to provide
a temperature of 19 K for the gas at the entry to the cores). The temperature of the
gas flow at the exit varied with the current rise. Curve 2 characterizes the temperature
change at the exit from the core of phase 'A' when the current is increased. Thus, for a
current value of 160 A, the temperature at the exit of the core was 25 K, while at
1600 A, the corresponding temperature increased to 32 K. The presence of such a
temperature gradient in the cores and its linear dependence upon the current is
probably due to the influence of the exit flow. It was found that lowering of the temperature gradient along the core length was possible by reducing the amount of the
exit flow, i.e., by directing most of the gas through the current leads to the gas holder.
T~ K;JJ
JS
m8
,!
~
Jl
3.l
II
JO
ff
Z9
2,
11
IQ
Z6
u
.f
~t
Fig. 4. Voltage-current characteristics of the superconducting cable under phase 'A' conditions. (See text
for complete operating conditions.)
99
Research on a Laboratory Model of Superconducting Test Cable
This operation, however, led to a gas imbalance in the heat exchangers of the refrigerator and eventually to a change in the operation of the entire plant.
As a result of this investigation, an important requirement for the construction
of superconducting cables has been determined. It implies the necessity of separating
the inlet coolant flow from the exit coolant flow by appropriate thermal insulation
in order to essentially eliminate the temperature gradient between them.
The voltage-current characteristics obtained for ac and dc operation were the
same under identical thermal conditions and methods of installing the potential
probes. The value of the pure resistance of the core turned out to be practically equal
to the Ohmic resistance and was about 10- 5 ohm at the mean gas temperature of
24K.
SUPERCONDUCTING STATE
The superconducting operating conditions were also investigated in the presence
of a high temperature gradient in the cores. Figure 4 shows the voltage-current
characteristics of phase 'A' for dc (curves 3,4) and ac (curves 5, 6) operation. Curve 3
was obtained for an inlet gas temperature of 12 K and an exit gas temperature of
17.5 K. Superconductivity was observed along the length of the core, in the flexible
bond of the current lead, and in the part of the current lead situated in the collector.
The flow rate of the coolant was 3.35 g/sec and the flow through the current leads
was 0.55 g/sec. Voltage drop in the core began to appear at a current of 200 A. At
600 A, the temperature sensors in the vessel registered 18 K and the flexible bonds
entered the normal state. When the current was disconnected, superconductivity was
still observed in all the sections.
The current was increased in such a fashion so as not to disrupt the superconductivity in the phase because, at high current values, heat release occurs which
changes the thermal state of the entire plant. It is impossible to maintain the thermal
state without changing the flow rate of the coolant.
Curve 4 was obtained for an inlet gas temperature to the core of 10 K and an
exit gas temperature of 16.5 K. The coolant flow rate was maintained at the same
level as for the previous case. The nature of curve 4 is analogous to that of curve 3.
In this case, however, a voltage drop first appeared at 740 A. In the former case, the
capacity in phase 'A' at a current value of 1000 A was 1 W, while in this case, the
capacity was 0.2 W (Table 11). Warming of the exit gas in phase 'A' began to appear
for a current value of 1260 A. Superconductivity in the flexible bond failed at 1800 A
while the gas temperature in the collector rose to 17 K.
Table II. Power in Watts Liberated in Core of Phase 'A' under Different Operating
Conditions*
Power, W
Curve
Mode of
operation
'T;nl.JT..itt
3
4
5
6
dc
dc
ac
ac
12/17.5
10/16.5
8/15
6/14
at 1000 A at 2000 A at 3000 A at 4000 A at 5000 A
* Temperatures in K.
1
0.2
t Current conditions refer to current through phase 'A'.
t Calculated for cryoresistive working conditions.
22
9
0.4
98t
59.4
28
2.96
170t
170t
88t
32
260t
260t
260t
140t
100
D. V. Razerig, Y. L. Bliakoy, IIIld Y. S. Goldeaberg
The voltage---<:urrent characteristics depicted by curves 5 and 6 were taken with
ac operation. The gas temperature was 8 and 6 K at the inlet conditions and 15 and
14 K at the exit conditions for curves 5 and 6, respectively. The gas requirement for
the inlet flow was 3.3 g/sec and 0.55 g/sec for the current leads. The voltage---<:urrent
characteristics under ac operation are slightly different from those under dc operation.
Under ac operation, the temperature rise with increasing current is considerably less
in the region where the magnetic field penetrates the superconductor. In addition,
starting from the moment that the coil is energized, a voltage drop occurs which
remains practically unchanged up to 2000 A for curve 5 and up to 2800 A for curve 6.
The value of the power which is liberated in the core of phase 'A' is given in
Table II. This value was calculated only for the region where the voltage---<:urrent
characteristic is changing appreciably. No calculations were made for the ac power
value where the voltage-current characteristics remain essentially unchanged.
CONCLUSIONS
The experimental research on the three-phase superconducting test cable which
operated with a forced coolant circulation scheme from a single refrigerator (with
gas removal for cooling the current leads) revealed that:
1. Uneven cooling of the cores along the cable length occurs under a low coolant
consumption by the current leads. The temperature of the exit flow is strongly
influenced by the flowrate of the inlet stream.
2. The gas removal in excess of 10% from the inlet flow for cooling the current
leads abruptly decreases the capacity of the refrigerator and causes gas
imbalance of the heat exchangers.
3. The presence of a large mass of metal in the cold zone (cores, helium piping,
couplings) has no adverse effect upon the temperature stabilization of the
superconducting cable during heat release.
4. Cooldown from room temperature to 100 K required 8 hr, from 100 to 25 K
required 6 hr, and from 25 to 10 K required an additional 4 hr.
5. The velocity control of the coolant flow through the superconducting cable
can be accomplished by means of an outboard valve which provides a
smooth regulation.
A series of electrical characteristics has been obtained recently. However, an
analysis at this time would be premature, since there were no systematic observations
for the various experiments and the metering system behavior.
C-3
DEVELOPMENT OF A RIGID AC
SUPERCONDUCTING POWER TRANSMISSION
LINE *
R. W. Meyerhoff
Union Carbide Corporation, Linde Division
Tarrytown, New York
INTRODUCTION
The availability of less costly and more efficient means for the underground
transmission of large blocks of power may be a significant factor in determining the
ability of the electric power industry to meet future demands for electric power. It
is generally agreed, however, that the present technology of underground power
transmission leaves much to be desired when one looks ahead to the anticipated
requirements. At the present time, underground superconducting power transmission
lines appear to be an attractive alternative to present technology and for this reason
have, in recent years, been the subject of research and development programs in
several countries. One such program has as its objective the demonstration of both
the technical and economic feasibility of ac superconducting power transmission.
This three and one-half year program, initiated in late 1971, will conclude in April
1975 with the testing of a laboratory model of a rigid coaxial line having a power
rating of approximately 3400 MV A.
This presentation summarizes recent work on the development of the conductor
and dielectric systems for the proposed rigid coaxial system illustrated schematically
in Fig. 1. Since most of the design parameters for this system have been discussed in
7 ], only a brief summary will be given here. The proposed ac
detail elsewhere
superconducting power transmission line consists of three-phase conductors each
mounted by means of dielectric spacers coaxially within its own superconducting
shield. These three conductor and shield assemblies are placed along with a helium
return line within a single 18- to 25-m-Iong factory-evacuated and -sealed multishielded cryogenic envelope. The primary dielectric between the conductors and
shields will be pressurized helium, which will be pumped through the spaces between
the conductors and shields. Field installation of the preassembled sections will
consist of joining in the following order: the conductor, shields, helium return line,
and the cryogenic enclosure.
The following two sections will describe recent work on the development of
the dielectric design and of the conductors and shields for the proposed system.
Details of the cryogenic enclosure and refrigeration system are described elsewhere
by Morihara et al. C].
e-
* Study
sponsored by the Electric Power Research Institute with additional support from the U. S.
Department of the Interior.
101
102
R. W. Meyerhoff
SHIELD BACKING
N IOSIUM SHIELD
DIE I.ECT R Ie
HELIUM suPpLy
-\-1. . - -
HEL IUM RETURN
~-H---
NIOBIUM COND UCTOR
~'1I!H~--
CONDUCTOR BACKING
CRYOGENIC ENVELOPe
Fig. 1. Schematic (not to scale) of the cross
section of the ac superconducting power
transmission line considered in this study.
DIELECTRIC SYSTEM DEVELOPMENT
One of the most frequently raised questions concerning the technical feasibility
of an ac superconducting power transmission line of the type described above and
illustrated schematically in Fig. 1 relates to the ability of pressurized helium to serve
as the primary electrical insulation between the conductors and shields. While there
have been a number of studies [8-12] of the breakdown behavior of helium at low
temperatures, all of these earlier investigations employ maximum voltages and gap
lengths much smaller than those required for the proposed design. Hence, one of
the first objectives of this program was to determine the breakdown properties of
pressurized helium and dielectric spacers at gap lengths and voltage levels similar
to those expected to be employed in such an ac superconducting power transmission
line. For this reason, a high-voltage testing laboratory designed specifically for the
investigation of the performance of solid dielectrics at liquid helium temperatures
and of pressurized liquid helium was set up, containing the following four principal
pieces of equipment: a ± 200-kV dc supply, a 200-kV ac supply, a ±750-kV impulse
generator designed to produce both a 1.5 x 40 J1.sec lightning surge and a 200 x 2000
J1.sec switching surge, and a liquid helium dewar approximately 1 m in diameter and
2 m in height designed to operate with internal pressures up to 6 atm. For sphereplane breakdown measurements, the sphere is rigidly mounted to the dewar cover.
The ground plane can be adjusted by means of a gear train from outside the dewar
to give gaps between 0 and 7 cm with an accuracy of ±0.002 cm.
The first goal was to determine the intrinsic breakdown behavior ofliquid helium
at gap lengths and total voltages similar to those envisioned in an ac superconducting
power transmission line. The results of measurements made with boiling helium at
4.2 K using a polished 25-cm brass sphere and a polished brass ground plane are
summarized in Fig. 2. Also included in Fig. 2 are the results of five other published
studies [8- 12J of breakdown voltage for boiling helium vs. gap. These results show
that helium is a well-behaved dielectric with the breakdown voltage VBO increasing
nearly linearly with gap up to 200 kV. The ac and dc results are consistent with extrapolations of the previously published data of the other studies included in Fig. 2.
with the exception of the work by Fallou et al. [IIJ, which shows a considerable
departure from a linear relationship between VBO and gap. Fallou and associates,
however, used small-diameter spheres at gaps greater than the sphere diameter. If
Fallou's results are recalculated on the basis of peak electric stress at breakdown vs.
gap, the data are then consistent. The impulse results also illustrated in Fig. 2 show
that boiling helium has an impulse strength greater by a factor of approximately
two than the ac or dc breakdown strength.
The results presented above show that boiling helium has a breakdown strength
for ac or dc voltages of approximately 16 kVjmm and an impulse (1.5 x 40) strength
103
Development of a Rigid AC Superconducting Power Transmission Line
200
•
185
•
•
160
.
•
140
C
.
120
•
•
C
100
60
•
c
t
0
•
60
40
Fig. 2. Experimental values of the
breakdown voltage of boiling liquid
helium at 4.2 K and 1 atm as a
function of gap length.
20
10
12
GAP, MM
of 27 kV/mm. When the helium is pressurized (1 to 2 psig), there is only a slight
rise in the impulse strength. The ac and dc breakdown stress increases from 16 k V/mm
for boiling helium to approximately 22 kV/mm for pressurized (1 to 2 psig) nonboiling
helium.
Breakdown measurements were also made with a 25-cm-diameter copper sphere
and a 30 x 30 cm copper plate which were both electroplated with niobium. Sphereplane ac, switching surge dc, and lightning impulse voltage breakdown measurements
in liquid helium were made using the as-plated niobium surfaces (the plated sphere
and ground plane were washed with water to remove salts from the plating bath,
but no polishing or other surface treatments were used). These measurements, made
with helium at temperatures between 4.2 and 5.0 K at pressures between 0 and 30 psig,
gave breakdown voltages essentially identical to the results for measurements made
with the well-polished brass sphere and ground plane. In addition, values of the
breakdown voltage for a 200 x 2000 Jlsec switching surge were determined with 0
and 2 psig helium between the electroplated niobium sphere and ground plane.
These results are summarized in part in the first two columns of Table I. The values
of the electric stress at breakdown EBD are given for a 5-mm gap for a 1 x 40 Jlsec
lightning impulse, 200 x 2000 Jlsec switching surge, and dc and ac voltages for
helium at 0 and 2 psig. In the case of an electroplated sphere-plane, as reported
above for the brass sphere-plane, the suppression of boiling in the helium by applying
a small, positive pressure ('" 2 psig) raised the dc and ac breakdown voltages and,
hence, EBD by approximately 34 %but had little or no effect on the lightning impulse
breakdown stress.
Switching surges had not previously been studied with the brass sphere-plane
combination. The 200 x 2000 Jlsec results shown in Table I for the electroplated
104
R. W. Meyemoft'
Table I. Electroplated Niobium Sphere and Ground Plane Test Results
EBD • kV/ern (5 mm gap)
Test
1 x 40 psec
200 x 2000 IAsec
de
ac (peak)
Boiling
helium
Opsig
clean
Nonboiling
helium
2psig
clean
Boiling
helium
Opsig
Si0 2 + Cu
Nonboiling
helium
2psig
Si0 2 + Cu
272
180
154
165
272
238
210
228
188
128
244
123
178
niobium sphere-plane combination appear to be consistent with what one would
expect from the ac, dc, and impulse results, namely, results quite close to but slightly
higher than the dc or ac peak stress at breakdown and somewhat lower than the
results obtained for the impulse wave.
Three series of measurements of the effect of impurities in helium on EBD have
been made. In the first test, air was blown into the dewar, producing particles of
frozen air in the helium. This produced no observable effect on the measured values
of EBD • For the second test, - 325 mesh Si0 2 was added to the liquid helium. A
motor-driven fan submerged in the helium was used to keep the Si02 in suspension.
As in the case of the frozen air, no observable effect on EBD was noted. Copper powder
( - 325 mesh) was then added to the helium. Measurements of EBD for the helium
containing both Si0 2 and copper are presented in the last two columns of Table I.
Comparison of these two columns with the first two columns of Table I shows that
the addition of the copper powder produced a measurable decrease in the values of
EBD • Nevertheless, even with the copper and Si0 2 powder addition, the helium still
gave values of EBD higher than the 150 kV/cm required for withstanding lightning
impulse for the designs being considered. Additional measurements of the effect of
larger metal and dielectric contaminants will be carried out in the near future.
The very encouraging performance observed for helium in the sphere-plane
experiments have thus far confirmed beliefs that pressurized helium is a suitable
material for use as the primary electrical insulation in the ac line being developed.
For this reason, a major portion of the dielectric development program is now directed
toward the development of the cable spacers used to maintain the coaxial arrangement of the conductors and shields. A number of materials potentially suitable for
use in spacer fabrication have been evaluated on the basis of properties such as loss
tangent, breakdown strength, surface arc resistance, thermal expansion, mechanical
properties, and cost. While this portion of the program is not yet completed, the
fabrication and testing of spacers using the more promising materials has commenced.
Although to date the spacers have been only one-half the length required for the
cable designs, they more than meet the required 60-Hz breakdown strength and very
nearly meet the required lightning impUlse and switching surge requirements for a
138-kV cable. The testing of full-size spacers will commence in the near future using
those materials and spacer shapes that appear best suited on the basis of the one-half
scale spacer tests.
Development of • Rigid AC SupercoaductiDg Power Transmissioo Line
105
CONDUCTOR-SHIELD DEVEWPMENT
Two additional frequently raised questions of the technical feasibility of a rigid
niobium conductor concern the ability of this design to accommodate fault currents
and the differential thermal expansion between the conductor-shield assembly and
the outer wall of the cryogenic envelope. The theoretical and experimental work
described below indicates that both the electrical and mechanical requirements can
be met by employing an Invar inner wall in the cryogenic envelope and using a
niobium-copper-Invar composite for both the conductors and shields.
In order for a superconducting line to be acceptable to an electric utility, it
must be able to continue to deliver its rated power immediately following a current
overload without any service interruption. This can be accomplished in two ways.
On the one hand, the cable can be designed so that even the largest current
overload expected does not raise the line current to a value greater than the critical
current of the superconductor. This can be accomplished in three ways. (1) Use highspeed current-limiting circuit breakers which would limit the maximum overload
currents to values not more than 1.5-2 times the rated current. (2) Use a hard superconductor such as Nb 3 Sn or Nb-Ti at a current well below its critical current (this
is not practical if niobium is used, but is possible if the hard superconductors are
employed since the larger ac losses in these materials force them to be used at current
densities well below their critical currents). (3) Use a composite of niobium and a
hard superconductor so that at rated current operation, one has the advantage of
the low-loss niobium with the underlying layer of hard superconductor having the
capability of remaining superconducting when the overload current exceeds the
critical current of the niobium.
On the other hand, the cable can be designed so that even if the overload current
drives the superconductor normal and raises the conductor temperature above the
critical temperature of the superconductor, the conductor will cool down and return
superconducting when the overload is removed and the current returns to the rated
value.
We believe that the latter method is preferred and that it is possible to design a
niobium line that will meet the system requirements. This requires that the line design
meet the following two criteria
(1)
and
(2)
The first criterion, as expressed by (1), is that the total energy dissipated in the
conductor both during the current overload and during the time following the overload while the conductor is cooling back to its critical temperature is less than
Qmax = Cv!J.T
(3)
where C v is the heat capacity of the line and !J.Tis the maximum allowed temperature
rise.
The second criterion, as expressed by (2), is that following the current overload,
the ]2 R dissipation in the line must be sufficiently small to allow the line to cool back
down to the critical temperature of the superconductor at the rated current level.
For the designs being considered in this program, current overloads up to ten
times the rated current for times up to six cycles will produce a total ]2 R dissipation
R. W. Meyerbotr
106
that will raise the system temperature, at equilibrium, by less than 0.1 K. (For example,
a 138-kV transmission line with an 8-cm-diameter conductor has a rated current of
14,000 A and a capacity of 3400 MVA. The radius of the cryogenic enclosure for
this cable is 20 cm. Thus the total heat capacity of the contained helium is 5 x 10 7
J/mile/K. The electrical resistance of the copper substrate is 10- 3 a/mile. Thus, a
single-phase fault of 140,000 A for 0.1 sec will produce 2 x 106 J. Hence, the equilibrium temperature rise would be 0.04 K.) Thus, there appears to be no problem
meeting the first criterion. The ability of the proposed designs to meet the second
criterion depends on the value of the heat transfer coefficient between the conductor
and the pressurized helium. For a tubular conductor of radius r, backed by a copper
thickness greater than the skin depth b, the electrical resistance is given by
R = pL/2nrb
(4)
Since the area is given by 2nr L,
(5)
and
(6)
Equation (2) can be written for copper, with a resistivity ratio of 100, as
h>
KH/
- 9.2[1 - (H p /HJJ1!2 - To
(7)
where K = 63.7 X 10- 9 W/cm 2 -G 2 •
Values of h from (7) are plotted in Fig. 3 vs. H p for various values of He and To.
Also shown in Fig. 3 are steady-state values of h which have been reported in the
literature for various conditions of helium flow, temperature, and pressure. If the
values of h under transient heating and cooling conditions do not differ markedly
from the steady-state values of h reported in the literature, there should be no difficulty
in choosing a value of H p close to 1000 G which will ensure acceptable behavior
following a fault. Measurements are presently underway to demonstrate that the
proposed designs will indeed recover and return to the superconducting state following current overloads in the manner predicted by the analysis summarized above.
Various methods of dealing with the thermal contraction of the conductors
have been discussed in detail by Baylis [13J. There are three basic categories in
which the methods of accommodating the thermal expansion can be grouped:
1. Expandable sections of conductor such as bellows, omega, and dog-leg
bends
2. Sliding joints
3. Restraint
The bellows method is an acceptable solution but creates other problems that
add to the complexity of the system. One is the lack of rigidity. This could result in
mechanical oscillations induced by the electromagnetic forces which act during
fault currents. These oscillations could, in tum, increase the conductor losses and
precipitate mechanical fatigue and eventual failure of the bellows. Second, the
intensification of magnetic and electrical fields at the apex of the convolution aggravates the ac loss and electric field problems.
107
Development of a Rigid AC Superconducting Power Transmission Line
, .•
.------r---r--.----.---.---~--....,
.5
.1
.05
,
><
N
'~"
..•...
.01
.II .005
h :: 0 .0&5. / cm 2 - ~
NO F LOW
.00\
R. ;;; loS
h ~ 0 .2 3
• I eme! - K
R. _106
h = 1.1111
W / c",,2_ 1<
.0005
Fig. 3. Coefficients of heat transfer
calculated from equation (7) as a
function of peak surface magnetic
field for various material parameters and coolant temperatures.
.
0.0001 '---_---'-_ _--'-_ _- ' -_
.2
•6
_
.l........._
.8
Hp, KI LOGA US S
_
...
" - -_ - - - ' -_ _-1
Sliding joints, while attractive in principle, have not been considered due to the
formidable problems associated with making high-current sliding superconducting
contacts.
The most favorable solution at the present time and seemingly the simplest is
to use an Invar substrate and dewar wall and then physically attach the conductors
to the dewar. The conductor would be a composite of Invar, copper, and superconductor. The amount of copper is reduced to a thickness of the order of the skin
depth for fault protection. The thickness of Invar selected is that sufficient to maintain
mechanical stability. The stress in the copper is now uniformly distributed over the
Invar surface and buckling will not occur if the system is warmed back to room
temperature as it would if a niobium-copper composite were used. If a niobiumcopper conductor were restrained, it would elongate plastically when cooled from
room temperature to 4.2 K and hence would buckle if warmed to room temperature.
The thermal contraction of Invar between room temperature at 4.2 K is less than
the elastic limit for Invar; hence, no buckling will occur since the composite can be
designed to keep the total strain within the elastic range for the composite. This
approach is not free of all problems. Namely, it requires the transfer of an axial
restraining force across the dielectric spacer design. A niobium-copper-Invar composite conductor has been fabricated and is being used to study the behavior of this
system on thermal cycling.
108
R. W. Meyerboft'
ACKNOWLEDGMENTS
The author wishes to acknowledge the contribution of the many members of the research department
of the Linde Division of the Union Carbide Corporation who contributed to the work described in this
paper. The author is particularly indebted to W. T. Beall, L. H. Day, and D. A. Haid.
NOTATION
surface area, cm l
C.
heat capacity of superconducting line, J/sec-K
h
heat transfer coefficient, W/eml-K
He
peak value of magnetic field at conductor surface when transport current equals critical current, G
H,
peak value of magnetic field at conductor surface at the rated current, G
1r
rated current, A, rms
1,
fault current, A, rms
K
= constant, 63.7 x 10- 9 W/cml_Gl
L
= length of superconducting line, em
r
= radius of tubular conductor, cm
Q..... = maximum allowable energy dissipation in line, j
R
= electrical resistance, ohms
7;(1 R) = critical temperature at rated current, K
To
= normal operating temperature, K
tl
= fault duration, sec
tl
= time following fault for line to cool and return superconducting, sec
lJ
= skin depth
A
=
=
=
=
=
=
=
REFERENCES
I. H. M. Long, W. T. Beall, L. K. Eigenbrod, R. W. Meyerhoft', and J. Notaro, "Superconducting Cable
System," Final Rept. Edison Electric Institute Project RP78-7 (October 1969).
2. H. M. Long and J. Notaro, J. Appl. Phys. 42: ISS (1971).
3. L. K. Eigenbrod, H. M. Long, and J. Notaro, "Conceptual Design and Economic Analyses of a
Superconducting AC Power Cable System," Paper No. 70TPI68-PWR, presented at IEEE Winter
Power Meeting, New York (1969).
4. R. W. Meyerhoft', in: The Science and Technology ofSuperconductivity (W. D. Gregory, W. N. Mathews,
Jr., and E. H. Edelsack, eds.), Vol. 2, Plenum Press, New York (1973), p. 433.
S. R. W. Meyerhoft', "Superconducting Power Transmission," ASME Publ. No. 72-PWR-7, presented
at IEEE-ASME Joint Power Generation Conference, Boston, Massachusetts (1972).
6. R. W. Meyerhoft', "Cost Analysis of Dielectric Systems for AC Superconducting Power Transmission
Line," IEEE Publ. No. 71638EL, p. 216, presented at 10th Electrical Insulation Conference, Chicago,
Illinois (1971).
7. H. Morihara, D. J. Webster, and K. C. Kather, "Design and Construction of a Rigid Cryogenic
Envelope with Low Heat Loss for a Superconducting Power Transmission Line," paper presented
at 1973 Cryogenic Engineering Conference, Atlanta, Georgia (August 8-10, 1973).
8. J. Galand, Compt. Rend., 266: 1302 (1968).
9. K. N. Mathes,IEEE Trans., EI-2:24 (1967).
10. J. P. Lehmann, "Measures dielectriques dans les ftuides cryogeniquesjusqu'a 200 kV-SO Hz," Revue
Gen. Elec., 79: IS (1970).
11. B. Fallou, J. Galand, J. Bobo, and A. Dubois, in: Low Temperature and Electric Power, Intern. Inst.
Refrigeration, London (March 1969), p. 201.
12. R. J. Meats, Proc.IEEE, 119(6):760 (1972).
13. J. A. Baylis, in: Proceedings 1972 Applied Superconductivity Conference, IEEE Pub\. No. 72CHQ682S-TABSC (1972), p. 182.
C-4
A SUPERCRITICAL HELIUM FACILITY FOR
MEASURING HIGH-VOLTAGE BREAKDOWN*
E. B. Forsyth, R. B. Britton, J. Dean, J. E. Jensen, and K. Minati
Brookhaven National Laboratoryt
Upton, New York
INTRODUCTION
Several groups working in the field of power transmission using superconducting
cables have proposed the use of flexible cables 2 ]. In these cables, the insulation
between the inner and outer conductors is a dielectric tape; at the operating temperature, the interstices between the tapes become impregnated by helium coolant.
Although some work has been carried out on nitrogen-impregnated plastic [3]
at low temperatures, there has been almost no attention paid to helium impregnation,
particularly in the supercritical region. More extensive work on helium breakdown [4]
is not applicable to gas-impregnated solids since the gap spacing is much greater
for transmission systems insulated only by helium. Tests on solid and tape polyethylene in helium at 4 x 105 N/m2 have been reported from France [2]. The apparatus
described below was built in order to determine the high-voltage characteristics for
a variety of materials, tape sizes, and lay-up techniques under various helium conditions. The helium itself will be in the supercritical region, in the temperature range
from 6 to 9 K and 5 to 16 X 105 N/m2 [5]. Thus, the apparatus required consists of
a winding machine, which can place the sample material in a suitable configuration
on a mandrel, and a high-voltage feed through bushing and cryostat, which permits
a room-temperature, high-voltage connection while maintaining the sample under
the desired helium conditions.
e·
DESIGN OF APPARATUS
Cryostat
The test chamber is mounted within a commercial helium storage dewar which
can hold about 100 liters. The dewar is mounted in a pit below floor level. The general
arrangement of the cryostat is shown in Fig. 1. The test chamber is formed by the
pressure tube, which is insulated from the liquid helium by a vacuum jacket. The
pressure tube is designed for 16 x 105 N/m2. To provide a uniform test temperature,
the lower portion of the pressure tube is made of heavy-walled copper, while the
upper portion is made of thin-walled stainless steel to provide a low heat leak. The
cryostat and mandrel are shown in Fig. 2. The pressure seal is made at the upper
• Work supported by the National Science Foundation.
t Operated by Associated Universities, Inc., under contract to the U. S. Atomic Energy Commission.
109
110
E. B. Forsyth, R. B. Britton, J. Dean, J. E. Jensen, and K. Minati
TEMPERATURE GRADI EN T
ALONG TlilS l [ NG TM
9- OIA. 1I0.
HELIUM OEWAR
CIRCULAR
TES T MANDREL
CONTROLLED
-++---t!IrtI'1I 1
TEMPERATURE
ALONG THIS
LENGTH
L.IQUID HELIUM
IIHIIIII8ft--f-t- PREs5UAE
HllI-- !+ VACUUM
TUBE
JACKET
HEAT
E.C"ANGEA -+i----1~~
Fig. I. Dielectric test cryostat.
Fig. 2. Cryostat and mandrel, top of dewar.
PRESSURE
SEAL ARE'
T
TEMPERATURE
GRADIENT
l , U. ) INrI
t
STRESS
.00 _
rEST
12 .' _
OIA.
SECTION
TueE
STRESS
CONE
----L
Fig. 3. View of high-voltage bushing extending
from the dewar.
A Supercritical Helium Facility for Measuring High-Voltage Breakdown
111
flange of the pressure tube. A standard O-ring seal is used between the flange and the
test mandrel, since it is at room temperature.
A cylindrical weldment extends above the cryostat and supports the 150-kV
bushing* for a high-voltage connection. The mandrel passes through the center of
the bushing and is connected at the top corona ring. The bushing and weldment are
pressure-sealed so that a dielectric fluid may be used inside the bushing to prevent
internal corona discharge at high voltages. The bushing is shown in Fig. 3.
The temperature of the supercritical helium and lower portion of the mandrel
in the test chamber is controlled by balancing the heat input to the pressure tube
against the available refrigeration. Heat leak down the stainless steel support tube
holds the test chamber at an equilibrium temperature. A small tube is wrapped
around the outside of the test chamber and passes through the vacuum jacket into
the liquid helium reservoir. The other end of this tube passes out of the cryostat. A
needle valve and variable-area flow meter are attached. The refrigeration available
to the pressure tube is increased by opening the needle valve and transferring liquid
out of the reservoir. This refrigeration cools the test chamber to the desired temperature. When a test chamber temperature above the equilibrium temperature is required,
the helium flow is stopped and heat is applied to the pressure tube by means of a
nichrome wire electrical heater. The temperature is measured by means of both
copper vs. constantan and gold (0.07 % at. % iron) vs. chromel thermocouples. Two
differential thermocouples (of both types) referenced to the liquid bath are located at
the top, middle, and bottom of the pressure tube.
Test Mandrel and Winding Machine
The test mandrel consists of a center conductor with stress cones which permit a
smooth increase in electric field intensity to a maximum that can be maintained along
a 60-cm test section. Samples of tape material can be wound over this section to
simulate the butt-spaced dielectric of a wrapped coaxial cable.
The mandrel is essentially a 2.7-m-Iong, 0.25-mm-wall tube of stainless steel
which is clad with epoxy resin and glass fiber tape to provide electrical insulation.
The design is shown in Fig. 4. There is enough cross-linking of the glass fibers in the
composite so that the combination can withstand the thermal stress of being cooled
rapidly to 4 K and back to room temperature. If kept free of moisture, the plastic is
also a good electrical insulator. A test sample of glass-epoxy laminate was capable of
withstanding 100 kV dc over a 20-cm length in air. In position, the mandrel is attached
at the top to the ceramic bushing which is filled with room-temperature sulfur hexafluoride (SF6 ) gas at 2.06 x 10 5 N/m2, while the lower, cold, end is exposed to helium
gas under the same pressure and density as the dielectric sample. Both ends are
vacuum-cleansed prior to pressurizing. For low-voltage tests, the upper end of the
mandrel is left exposed to the atmosphere and capped with an anticorona ring.
Precise uniform taping requires accurate taping angle adjustment, uniform
tape tension and mandrel rotation. For uniform butt-spacing of the tape, the taping
angle must be adjusted after every layer of tape, since the angle is a function of a
winding diameter and tape width. A general view of the winding machine is shown in
Fig. 5.
Tape tension is provided by a mass suspended from the end of the tape. A mass of
1.1 kg was used with 1.9-cm-wide tape for a tension of 567 N/cm of width. The mass
* Manufactured by Ceramaseal Inc.
112
E. B. Forsyth, R. B. Britton, J. Dean, J. E. Jensen, and K. Minati
Fig. 5. General view of tape winding machine.
must be centered on the tape width, since it acts as a plumb bob in setting the taping
angle.
The taping angle is adjustable between 14 and 32° for both left- and right-hand
positions. The angle is changed by turning a screw crank, and is read with a revolution
counter. The adjusting screw passes through a nut fastened to the lower side of the
bearing channel. Rotating the screw will move the nut and channel along the axis of
the screw. Since the nut is 91 cm from the pivot point of the channel and the pitch of
the screw is 0.23 cm, the angular sensitivity is 8 min of arc per revolution or 2 min of arc
per count.
PERFORMANCE
Cryogenic
Heat leak down the stainless steel tube supporting the test chamber and heat
leak down the mandrel warm the test chamber and sample to approximately 14 Kat
thermal equilibrium when no helium flow exists in the heat exchanger. Transferring
liquid out of the reservoir through the heat exchanger at the rate of 0.03 g/sec will
reduce the sample to a temperature below 5 K. Intermediate temperatures can be
obtained by adjusting the flow rate to a lower value. The flow rate required to achieve
a given temperature varies slightly with the liquid level, mandrel construction, and the
sample insulating vacuum. However, sufficiently constant (± 0.25 K) temperatures
can be obtained by establishing a constant flow rate. This is achieved by regulating the
back pressure on the liquid helium reservoir. The vent gas of the dewar is bubbled
through a 25-cm head of low-vapor-pressure oil.
The liquid helium level was found to affect the temperature uniformity of the
sample, causing a temperature gradient of 2 or 3 K along its length. This condition
was corrected by adding a liquid helium heat exchanger (not shown in Fig. 1) to the
test chamber support tube at a location a few centimeters above the test chamber. As
with the test section heat exchanger, a small liquid helium flow is controlled by an
externally mounted needle valve. Regulating this flow brings the sample temperature
A Supercritical Helium Facility for Measuring High-Voltage Breakdown
113
gradient to well below 1 K. A temperature uniformity along the 6oo-mm sample
length to within the accuracy of the thermocouple (± 0.25 K) is obtainable.
The time required to achieve thermal equilibrium is mainly dependent on the
thermal capacitance of the helium used to pressurize the sample. Since the thermal
capacitance is a strong function of the pressure and temperature near the critical point,
the system time constant varies considerably over the operating range. One hour is
sufficient time to establish a constant temperature for most operating conditions.
High temperatures are achieved by supplying a small amount of electrical heat
to the sample holder. Although the upper temperature of interest in the current
experiment is 15 K, the test chamber can be raised to a much higher temperature.
The sample can be warmed up and removed from the cryostat, a fresh mandrel
installed, and cooled down again without refilling the loo-liter liquid helium reservoir.
Changing a mandrel vaporizes about ten liters; thus, several mandrels can be tested
before the dewar must be refilled. The static heat leak of the cryostat without the
sample installed is 0.75 W.
Electrical
Measurements have been performed on a variety of plastic samples using both ac
and dc voltages. Breakdown stress is significantly less using ac voltages. For example,
a 0.762-mm test thickness of Teflon ruptured at 125 MV/m on dc and 36 MV/m
average stress at 60 Hz. Teflon, however, shows a tendency to split badly during the
tests due to electromechanical stress. Synthetic papers using spun-bonded polymers,
such as Tyvek, have superior mechanical properties and do not tend to crack or split.
However, tests to date have shown that paperlike materials have lower breakdown
stresses than films. Calendering improves the breakdown to some extent, as can be
seen when comparing tests 15 and 16 with tests 5 through 7 in Table I.
Test 19 et seq. were performed with a partial discharge detector connected to the
test unit. By detecting the point of corona inception, it will be possible to learn more
about the mechanisms of breakdown in helium impregnated tapes.
The planned program for insulation breakdown testing is as follows:
1. Determination of best material for inner and outer electrode surfaces.
2. Relationship between ac, dc, and impulse breakdown for specific samples.
3. Effect of butt-spacing, tape width, and winding tension on breakdown.
4. Effect of coolant temperature and pressure on breakdown level.
5. Testing of various materials, such as Nomex, Mylar, and Teflon.
6. Quantitative measurement of corona levels.
CONCLUSION
The cryogenic performance and range of the apparatus are adequate to simulate
all normal and abnormal conditions in a superconducting cable. The tests carried out
so far have shown that some materials can provide an acceptable level of breakdown
stress. However, a dielectric design compatible with all the electrical and mechanical
constraints of a cryogenic cable has not yet been found. In addition, the problems of
corona inception and long-term insulation degradation have only just begun to be
attacked. The facility has proven to be reliable and versatile and promises to be a
very useful tool in the investigation of dielectric behavior at cryogenic temperatures.
The design of the test apparatus solves some of the problems of a complete pothead termination. The control of electric field intensity in the presence of a temperature gradient and pressure differential has been accomplished successfully.
2.03
0.76
10
10
2 x o.08mm}
10 x 0.20mm
2 x O.OSmm
15
15
Tefton
Tyvek
Tefton
Nomex-cd
Nomex-cd
10
12
11
7.92
7.92
7.92
7.92
14.19
2.03
10
Tyvek
Same as No.5
Tyvek
Same as No.7
Tefton
5
6
7
S
9
0.76
0.76
2.32
7.92
2.03
10
Tyvek
4
14.12
14.19
14.19
4.47
7.92
2.03
10
Tyvek
4.47
3
2.03
10
2.06
Helium
pressure,
105 N/m 2
Tyvek
Bare mandrel
Dielectric·
Radial
thickness,
mm
2
Test No.
Number of
layers
12.9
7.0
7.0
6.6
-7
9.0
7.4
S.2
-S
6.6
5.S
-200
SO
Helium
temperature,
K
23
51 dc
65-70dc
30.1
66.9
29.9
1I.S
7.6
9.S
5.9
125
11.2
23-24dc
24dc
15.6
20--22 dc
12
95dc
7.S
11.2
7.3
16-1S dc
23 dc
-15dc
16dc
Breakdown
level,t
kV
Average
stress at
breakdown,
MV/m
Remarkst
Velostat wrapped overall; mandrel
C = 1090pF
USI conductive polyethylene wrapped
overall
Velostat conductive polyethylene
wrapped overall
Surface leakage over wet-ice coateddielectric
Sample spoiled by condensation;
mandrel C = 620 pF
BD strength improved with time;
conductive polyethylene plus copper
tape outer wrap
Stainless steel ribbon wrapped over
dielectric
Dielectric wrapped with No. 14
manganin wire
Conductive polyethylene overwrapped
with stainless steel ribbon
BD at end of sample at coincidence of
seven butt spaces
Table I. Summary of Breakdown Experiments at Cryogenic Temperatures
ct.
f
~
JI
!"l
~
j'
~
J
pi
PI'
j
pi
t"'l
...
......
4.06
3.98
54
14 x 0.07mm}
l2 x 0.25mm
Mylar
Mylar
22
13.3
Mandrel C = 445 pF ; 23 min BO;
punctured at termination
25 min BO; punctured at termination
and through sample
Bare SS ID; AI foil 00; BO
t rms ac, otherwise dc; ac voltages applied in 4-kV increments for 4 min each to 49 kV, and for 15 to 30 min each above 40 kV.
breakdown.
55
14.1
17.0
t BO =
7.5
6.9
7.0
7.0
= spun-bonded polyethylene. Nomex = spun-bonded polyamide (nylon). Mylar = solid polyester tape. cd = a calendered tape.
14.12
14.12
14.47
14.12
* Tyvek
2.03
1 min BO; AI sheath
3 min BO; Al sheath
Al foil 10 and Al sheath
21
8
17.4
18.1
22.3
Mylar
36.8
46.0
5 (corona start)
36.0
3 (corona start)
35
16 (corona start)
75.5
20
7.0
7.2
6.8
17
18
19
14.12
14.12
14.12
17.6
32.2
6.9
14.12
1.83
24
2.03
2.05
2.03
15.5
27.6
13.6
14.12
1.78
14
Tyvek-cd
(10790)
Tyvek-cd
(1058B)
Mylar
Mylar
Mylar
15
8
27
8
4 min BO; Al sheath
36.2
27.5
6.7
14.19
0.76
10
Teflon
14
16
2 min BO; USI-CP sheath; Teflon
split badly
20 sec BO; Alum-tape sheath; Teflon
split badly.
4 min BO; Al sheath
31.4
24
6.9
14.12
0.76
10
Teflon
13
..
(II
......
I
i
i
110
I·
~
S'
~
i
i
;:
~
fa:
>
116
E. B. Forsyth. R. B. Brittoa, J. Deu, J. E. J _ , ... K. MiDati
REFERENCES
I. E. B. Forsyth, J. P. Blewett, R. B. Britton, M. Garber, D. H. Gurinsky, J. M. Hendrie, J. E. Jensen,
G. H. Morgan, and J. R. Powell, Trans. IEEE on Power Apparatus and Systems, PAS92(2):44 (1973).
2. P. Dubois, I. Eyraud, and E. Carbonelle, in: Proceedings 1972 Applied Superconductivity Conference,
Annapolis, IEEE Publ. No. 72-CH0682-5-TABSC (1972), p. 173.
3. M. J. Jefferies, S. H. Minnich, and B. C. Belanger, Trans. IEEE on Power Apparatus and Systems,
PAS 92(2):514 (1973).
4. R. J. Meats, Proc.IEE, 119(6):760 (1972).
5. J. E. Jensen, "Large Helium Systems for Superconducting Power Applications and High Energy
Physics," presented at the Cryogenic Society of America, Chicago, Illinois (October 3-5, 1972); also
available as Brookhaven National Laboratory Rept. 17228, Brookhaven, New York (1972).
DISCUSSION
Question by M. S. Lubell, Oak Ridge National Laboratory: Have you had a chance to measure the
dielectric properties of the Dupont material Tefzel?
Answer by author: Not yet; we have looked at only a few common materials. Due to their high cost,
we do not regard Teflon or Tefzel as likely candidates for cable insulation until many ofthe less expensive
plastics have been investigated. Teflon was measured for reasons of general interest and to determine the
performance of the apparatus with a high-breakdown material.
Question by D. L. Atherton, Queen's College: At room temperature, the dc dielectric strength of
Teflon degrades seriously with the deviation of the impressed dc electric field. For how long were your
dc voltages applied at low temperature?
Answer by author: The voltage was increased at a steady rate until breakdown occurred. We have
not had the opportunity to study the time effect, but intend to do so. The voltage is applied in 4-kV increments for 4 min each up to 40 k V, and for 15 to 30 min each above 40 k V.
C-5
SUPERCONDUCTING ENERGY STORAGE
R.
w. Boom, G. E. Mcintosh: H. A. Peterson, and W. C. Young
University of Wisconsin
Madison, Wisconsin
INTRODUCTION
Energy storage with large superconducting magnets is one of the possible new
components in a power system. Serious feasibility studies are under way in the United
States at the University of Wisconsin and at the Los Alamos Scientific Laboratory.
The preliminary opinion by both groups is that such units should be technically
feasible. As of the date of this writing (summer 1973), there is less certainty that such
units can be designed, constructed, and operated economically.
In 1972, total electric energy sales amounted to 1574 x 106 MWhr in the United
States, including Alaska and Hawaii. The average generating capability, which
increases at about 30,000 MWfyear, was approximately 380,000 MW during 1972.
If an allowance is made for scheduled maintenance on generating units and reserve
capability requirements, about 300,000 MW could have been generated continuously
for 8760 hr producing a total of approximately 2600 x 106 MWhr during the year.
Even making an additional allowance for system losses, generating capability in
terms of MWhr continues to be nearly twice the actual MWhr produced.
Continuing difficulties in siting new power plants make it attractive to double
electric energy production without adding new generating capacity. The very real
need for some means of load leveling (or peak shaving) to achieve a more constant
system load with reduced daily cyclic variations is greater every year. Ideally, one
would like to achieve a condition such that peak load and average load would be
essentially the same for an entire system. In that case, peak loading of the transmission
system need not coincide with the peak load on the system. The need for some means
ofload leveling and the enormous payoff to be realized is clearly evident in the ind ustry
annual statistical reports. The industry plans are to double pumped hydrostorage
during 1973 to achieve a total capacity of 8000 MW with approximately 25,000 MW
total installed within ten years [1].
REVIEW
Inductive storage concepts have developed over the past fifteen years. Walker
and Early [2] and Carruthers [3] considered inductors as sources of pulsed energy.
Stekly [4] suggested superconducting inductors for energy storage, discussed toroidal
* Visiting Professor from CTi/Cryenco, Denver.
117
118
R. W. Boom, G. E. Mclatosb, H. A. Petenoo, and W. C. Young
configurations, and emphasized structural problems. Sole [5J reported on various
magnet configurations for pulsed use. Ferrier [6J estimated the optimum shape of
magnets and calculated operating losses for large superconducting inductors useful in
a power system. Irie and Yamafuji
computed hysteresis losses for layer wound
magnets carried through charge-discharge cycles. Brechna et al. [8J considered different magnet structures and concluded that superconducting magnets are best used
for milIisec to second discharge times. Powell and Bezler [9J suggest that the superconducting magnet structural costs may be reduced by using warm reinforcement
structure and suggest the use of in situ bedrock as inexpensive structure.
Boom and Peterson [10J have shown that large magnets are more efficient than
small magnets. that high pinning strength IX = Hj is the important superconductor
parameter, and that low-field units are as satisfactory as high-field units if cryogenic
losses do not become excessive. They also present a basic six-pulse Graetz bridge to
connect a dc magnet to a three-phase line and discuss how controlled power reversibility is achieved by varying the delay angle of firing thyristors in the bridge. Operating
losses from the converter bridge and the refrigeration required to balance electrical,
magnetic, mechanical, and thermal losses into the cryogenic enclosure were estimated
to be as low as 10% total for a 10,OOO-MWhr system. A preliminary cost graph
indicated possible satisfactory economies for large units. It was shown that system
electromechanical oscillations can be damped effectively within one or two cycles
which is typically about 1 sec.
Hassenzahl et al. [11J also discussed the Graetz bridge method for controlled
power reversibility and present a good case for the utility of small. medium, and large
units. Present work 2 J at Los Alamos includes some preliminary cost estimates for
various storage systems. As an example, by assuming 2500 hr/year delivered power, the
cost in mils/kWhr for pumped hydrostorage would be 21 to 40, for batteries 23 to 31
eSJ, for a 28,OOO-MWhr toroid 30 to 34, for a 10,OOO-MWhr solenoid 32 to 36, and
for a 280-MWhr solenoid 75 to 83 mils/kWhr. The low cost in each case represents
charging with cheap off-peak power at 2.5 mils/kWhr, while the high cost represents
15 mils/kWhr prime-time charging. The large toroid and solenoid designs utilize
warm reinforcement, while the smaller unit could use either warm or cold reinforcement.
Although extensive optimization studies have not been carried out, short, largediameter, low-field solenoids are presently recommended. One such example at
10,000 MWhr is included in this presentation.
n
e
AC/DC REVERSIBLE POWER CONVERTER
The power converter interface between the three-phase power system and the
superconductive energy storage inductor (SESI), often referred to as a Graetz bridge,
is shown in Fig. 1. The six valves of Fig. 1 are SCR's or thyristors which can be turned
on or "fired" at any angle IX. If IX = 0, the valves operate as diodes. The bridge output
voltage shown by the heavy solid line is a unidirectional series of pulses as shown in
Fig. 2(a). There are six "pulses" of voltage, corresponding to the six "commutations"
of current among the six valves for each cycle of the 6O-Hz power system source. For
all output voltages of Fig. 2, it is assumed that Id = 0 and hence there is no commutation overlap interval or angle.
The average bridge output voltage decreases to zero at IX = 90° and becomes
negative thereafter, so that we may write
Ed = Eav = EdO cos IX
and
Pd = Eid = Ed01d cos IX
(1)
119
Sapen:onductiDg Eaergy Storage
(oj
a-o
WI
a
~30·
WI
GRID CONTROLLED REVERSIBLE POWER
At/DC BRIDGE CONVERTER
a-w
(dJ
a-
90·
WI
\0 THRE~E-':P~HASt;E~~d=d
60 Hz AC SYSTEM
Fig. 1. Basic six-pulse bridge
ac/dc grid-controlled reversible
power converter and inductor.
Fig. 2. Instantaneous output voltage of six-pulse
bridge converter with ignition delay angle a but
no overlap. Average output voltage also shown.
Since the bridge current 1d is not reversible, it follows that the bridge output power Pd
is uniquely a function of cos IX, which can be positive or negative.
Viewed from the power system side of the converter, varying IX changes the power
factor, the angle by which the ac voltage of one phase of the power system leads the
corresponding fundamental frequency component of ac line current (see Fig. 3). Since
there are three phases. the power from the power system is Pac = 3Ea1al cos fJ, where
Ea is the effective line-to-neutral ac system voltage, 1al is the effective value of the
fundamental frequency component of the ac system line current, and fJ is the angle by
which the voltage Ea leads the current 1al in time or phase position. If losses in the
converter are negligible. then Pac = Pd and the unique matching capability of the
converter (including reversibility) as an interface between the power system and the
SESI is clearly established.
The effect of a commutating reactance Xc corresponding to the transformer
between a power system and the bridge is to introduce a time delay interval or overlap
angle in the commutation process. Equation (1) is modified to become
Ed
= Eav = EdO(cos IX)
-
1dXc
(2)
Normally the product term 1dXc will be less than 20% of EdO ' However. it generally
cannot be neglected.
Converter bridges can be connected in series to obtain a higher voltage Ed or in
parallel to obtain a higher current 1d • For best efficiency, bridges should be operated
at or near maximum rating during the total charge--discharge cycle. which requires
switching from the series.low-1d connection to the parallel, high-1d connection during
use.
110
R. W. Boom, G. E. Mc:IJdoIIa, H. A. P-.-, ad W. C. Y-a
~
;",
(.1 a.o.8
1.-.
._
"....,.~V .
~
....."..~
(.,a'ed"8~
t::
,,10,
t:·~I",
,. ..".m~-
(fla"50"8~
... .,....~
Fig. 3. Alternating current system terminal phase dis·
placement angle Bas determined by ignition delay angle rl.
Results of system analyses reveal excellent dynamic performance of such power
system components. Control appears to be straightforward with no particularly
difficult problems to solve. Certain harmonic voltages can be eliminated by proper
design using well-established techniques for either series or parallel bridges. For
example, a twelve-pulse bridge automatically eliminates the sixth harmonic and
tuned filters can eliminate other undesirable voltages.
MAGNET AND STRUCTURE DESIGN
In an earlier paper [10], solenoids, toroids, and Brook's coils were compared and
it was concluded that the superconductor material optimization favors solenoids
or Brook's coils. It was shown that superconductor materials are best used at midfield
values; for example, 4 T for NbTi at 4.2 K and 5 T at 2.5 K are favored. In another
paper [13], methods were considered to design toroids with cold support so that
strains in the composite conductor are limited. The scheme proposed was to use a
constant-tension toroidal shape made up of forged wafers having slots in which precompressed composite conductors were imbedded. In this way, copper could be
carried from compression to tension without yielding while the steel developed higher
stresses.
The five major system design problems are: structure, cryogenic thermal load,
composite conductor, component fabrication, and system construction. To keep
these problems in perspective, note that all efficiencies favor large magnets; the
example included here is football stadium size and contemplates cryogenic components much larger than ever before constructed. The 5-T solenoidal system discussed
below (see Fig. 4) uses bedrock support, low-weight, low-thermal-loss epoxy-fiberglass structure to transmit forces to the bedrock, a single-turn, unreinforced NbTi-AI
composite conductor, and a 1.8 K supedluid helium cooling system. One feature of the
Superconducting Energy Storage
121
SUB-COOLER
SUPERflU'D
P I P'lN(j
IN
THE
CONDUCTOR
Fig. 4. Ripple conductor, structural thermal insulation,
cooled shields, and vacuum can for lO,OOO-MWhr superconductive energy storage inductor.
Fig. 5. 1.8 K superftuid helium cooling system. The helical tube represents 50 tubes in parallel. Only one
of many circuits is shown.
system is the use of "ripple" conductors and "ripple" outer dewar walls which can
adjust to variable dimensional requirements without excessive material strain during
construction, cooldown, and use.
Structure
Bedrock was chosen because it is free except for excavation costs. Cold support by
totally internal structure such as stainless steel would require at least 100 kg/kWhr
independent of size or magnet shape according to the virial theorem [10]. A solenoid
with external bands stressed in only one dimension requires 200 kg/kWhr; the
optimum utility of 100 could be achieved with internal core members stressed in
biaxial tension. If aluminum alloy 2219, which can also support 3.45 x 108 N/m2
(50,000 psi), replaces the stainless steel, then one might expect the optimum to become
30 kg/kWhr, which is only marginally acceptable at SI or S2/kg fabricated. Other
candidates for cryogenic use, such as 9 % Ni steel and titanium alloys, are even more
expensive.
Warm structure could utilize less expensive material. High carbon steel might
require 200 kg/kWhr at room temperature. Even if it could be purchased at the
attractive price of SO.20/kg, the brittle fracture problems associated with large masses
of welded ferrous material plus the cost of the cryogenic thermal insulation and its
operating cost make such construction questionable.
Thermal Load
The room-temperature power required to refrigerate the magnet system is the
major operating loss in these energy storage systems. Therefore the thermal loss is
the overriding design factor. In the case of warm support, the loss is almost totally
the thermal conduction loss through the epoxy-fiberglass structure which transmits
forces from low temperature to room temperature. In order to minimize the conduction loss at the surfaces, the magnet configuration should possess a large
122
R. W. Boom, G. E. MclDtosb, H. A. Petel'SOll, and W. C. YOUDg
volume-to-surface ratio since the thermal loss is proportional to the surface pressure
multiplied by the surface area, while the energy stored is the same pressure multiplied
by the volume. Such reasoning led to the choice of a relatively short solenoid with
a large diameter; the 10,000-MWhr example has the aspect ratio P = length/diameter = t. By comparison, a P = 3.14 solenoid storing 10,000 MWhr requires four
times more epoxy structure with four times the heat load.
The axial forces of compression are about the same magnitude as the radial forces
for p = t solenoids. The axial load component can be carried by massive internal cold
structure or by external warm structure via the epoxy-fiberglass support bridges. The
latter approach was chosen in order to save material but with the resulting additional
penalty of the associated thermal loss through the additional epoxy structure.
A 1.8 K, superftuid helium system was selected over a 4.2 K system in order to
utilize the much higher current density of NbTi at 2.5 K; in fact, the total refrigeration
system can be purchased with this saving in superconductor cost. Each conductor is
cooled via internal tubes filled with superftuid helium which transmits heat to a
reservoir without pumping, all of which requires a relatively small helium inventory
(see Fig. 5). The 1.8 K refrigeration is especially useful for a p = t solenoid since the
end turns are in such high radial fields that NbTi at 4.2 K could not be used, unless
the end turns are staggered radially.
Conductor
A composite conductor of 1000 resistance ratio aluminum 5 x 50 cm in cross
section enclosing 50 cm 2 of cooling tubes and braided, small, filamentary NbTi is
planned for the single-layer solenoid. Conductor fabrication might be completed
on site by the coextrusion process whereby the superconductor braid around an
aluminum core and an array of tubes for internal cooling are passed through an
extrusion die for the addition of the high-purity aluminum matrix. Several simple
aluminum coextrusions have been completed in the laboratory [14], using a 60-ton
press, and the conclusion is that coextrusion could be used here. The aluminum need
not be zone-refined and might cost less than $2/kg [is].
Extrapolating from the work of Linnet et al. [16], it is estimated that 50 cm 2 of
l-atm superftuid helium could conduct 25 W with insignificant temperature rise and
possibly up to 100 W before exceeding the A-point. This means that 4O-cm normal
sections could be continuously cooled by the helium. Providing cooling for a singlelayer, high-current solenoid is a formidable task; in this case, 314,000 A generates
5 W/cm. It is expected that more careful study might indicate the use oflarger helium
passages, more helium surface for heat transfer, or a subdivision of the simple single
turn into smaller conductors.
System
The 10,000-MWhr system as visualized might be constructed in a 16-m-wide
trench which is 100 m deep and 100 m in radius. The rock walls will be slightly stairstepped so that axial loads will not accumulate to the median plane but will be subdivided and carried separately by several rock ledges. In this way, no internal structure
other than the aluminum conductor and the epoxy struts are needed. At S35/m 3 ,
that cost would be approximately $40 x 106 plus $30 x 106 to cast in place the stepped
cylindrical liners, for a rock structure cost of approximately $7.0/kW-hr.
The virial theorem states that magnetic fields must be held in place by tensile
members. In the case of rock, in situ stresses of a few hundred psi in compression
should exist. Therefore, although preexisting radial cracks might negate tensile
loads in the circumferential layer adjacent to the hole, compression of rock wedges
Supereoaductillg Euergy Storage
123
will carry the load out to a much larger radius where the tension can unload the in situ
compression to carry the magnetic load.
In order to accommodate rock compaction strains as well as cooldown and
magnetic strains, the "ripple" conductor and "ripple" outer dewar wall can change
length to fit the rock wall without excessive material strain. Since the rippled conductor
can accommodate large fields because of the small local radius of curvature, the
conductor can be sized for electrical and thermal reasons and shaped to provide the
structure as well. The single conductor in a single-layer solenoid also minimizes
mechanical hysteresis losses and eliminates most of the interturn frictional losses. The
design procedure is summarized in Table I (see Fig. 4).
For the 10,000-MWhrsolenoid with 103.8-mradius using a 5 x 50 cmconductor.
the following are obtained: '-,4 = 25.6 cm, '-B = 3.6 cm, RA = 103.35 m, ()A = 0.30°,
(}A = 44°, 4JB = 43.7°, 1= 314,000 A, and eTal = 3.45 X 10 7 N/m2. For the 2.5 m
length of the. epoxy-fiberglass support structure stressed to 3.45 x 108 N/m2 and
assuming in situ compression in the rock of 1.38 x 106 N/m2 (200 psi) and a modulus
of 3.45 x 10 10 N/m2 (5 x 106 psi), the radially outward motion of the conductor is
17 cm. In addition to this, the cooldown to 1.8 K imposes a 0.4 % reduction in conductor length. Thus, to assure that the necessary radii are present in the conductor
when cold and under load, an additional 1 % must be added to the length of the
conductor to achieve the cold location illustrated in Fig. 4.
The total radial load is 5.32 x 1011 N and the total axial load is 6.16 x 1011 N.
The greatest axial loads are on the end turns. Groups of conductors will be supported
by epoxy-fiberglass struts to appropriately inclined ledges on the stepped concrete
Table I. Design Procedure for lO,OOO-MWhr Solenoid
Parameter
1. Maximum magnetic field
2. Proportions and shape of magnet
3. Choice of stabilizer material in the conductor
4. Locations and sizes for superconductor
strands and cooling passages in the
stabilizer
5. Conductor size
6. Radius rA.
7. Number of epoxy-fiberglass supports
(determined by 0,1.)
8. Width of epoxy-fiberglass supports
9·I/JA.,I/JB,andrB
10. RA.
11. Length of epoxy-fiberglass supports
12. Design details of support
13. Radial expansion of the rock walls
14. Added length of conductor necessary to
accommodate the radial expansion of the
rock walls, compression of the epoxyfiberglass supports, and cooldown strains
Basis for evaluation
Optimization of cooling and superconductor costs
Minimization of cooling costs
Cost, strength, and electrical and thermal
properties of the conductor materials
Electrical, thermal, and magnetic properties of
the stabilizer and superconductor and the field
strength and temperature distributions in the
cross section
Field strength, current density, and cooling
Field strength and conductor strength
Acceptable conductor length (cost of conductor)
Strength of material used
Width of supports, geometry and static
equilibrium of conductor
Required area inside midthickness of conductor
to produce necessary energy storage
Permissible heat loss by conduction
Structural stability considerations
Mechanical properties ofthe rock and the
tectonic residual stresses in the rock
Material properties of epoxy-fiberglass supports,
radial expansion of rock walls, and expansion
coefficients of conductor and supports
114
R. W. Boom, G. E. MeiJduIII, H. A. P*'-, .... W. C. YOIIIII
shell (Fig. 4). The epoxy cross section required at 1.8 K is therefore 3329 m 2 total,
which accounts for 14.4-kW refrigerator load at 1.8 K.
CRYOGENICS
The cryogenic system for the 5-T, 1O,OOO-MWhr storage magnet imposes several
unusual requirements:
1.
2.
3.
4.
Low operating loss relative to the amount of stored energy.
High reliability compatible with electric utility service.
High radial and axial forces supported by warm structure.
1.8 K operating temperature.
5. Necessity to minimize helium volume in consideration ofthe domestic supply.
6. Sheer size considerations in a solenoid 105 m long and 210 m in diameter.
With the low-temperature end established at 1.8 K for preliminary design purposes, it
is possible to make a cryogenic conceptual design. As shown in Fig. 5, this consists of
the superconductor stack with internal cooling and a superfluid piping system, epoxyfiberglass support structure, and cooled shields at 77 and 20 K with superinsulation.
This design represents a compromise between the greater thermal efficiency of
adding a 4.5 K shield and the lesser cost and complexity of only two thermal barriers
with the added advantage that, at least, the 77 K shield may be cooled with external
liquid nitrogen in case of refrigerator failure. It may also be possible to exchange heat
with external liquid hydrogen to temporarily hold the second shield at 20 K, but the
safety and logistic aspects of this operation have not been evaluated. The values
shown in Table II were obtained for the design shown in Fig. 4 using data from
Table III.
Table II
Insulation/radiation
heat leak,
kW
Temperature,
K
Support structure
heat leak,
kW
Total heat
load,
kW
14.4
75.8
683
79.6
758
0.5
1.8
20
3.8
77
75
Power required,
MW
20·
16.6
4.5
5.5
26.6
• 5.1 kW is allocated for internal heat generation, lead losses, hysteresis, and reserve to quench local
instabilities.
Table m. Low-Temperature Refrigerator Performance
Temperature,
K
Carnot power,
W, powerjW, refrig.
77
20
4.5
1.8
2.9
14
66
166
Actual power,
W, powerjW, refrig.
7.25
56
265
828
%ofCarnot
40
25
25
20
Supercoaducting Energy Storage
125
Unpublished [17] refrigeration data indicate that good cryogenic design with
special emphasis on the number of heat removal (expander) devices will result in
reasonably efficient machines for the required temperature levels. Table III summarizes results that may be expected from large machines which are designed for
optimum performance, not for minimum capital cost.
COST ANALYSIS
A very preliminary cost analysis for a 10,OOO-MWhr, 1000-MW reversible
system shows that the cost might be about $430 x 106 , or $430/kW reversible. The
cost is divided as follows: 45 %for the composite conductor, 30 %for rock preparation
and structural insulation, 12 %for on-site assembly, 8 %for ac/dc conversion, and 5 %
for refrigerators.
Further optimization could be expected to reduce costs as follows:
1. Larger helium volume would decrease the amount of aluminum for the same
hot spot stabilization reliability.
2. Addition of high-strength aluminum to the conductor could eliminate the
axial thermal load and would simplify and reduce the cost of the radial epoxyfiberglass support structure.
3. Determination of rock characteristics from several sites under consideration
may indicate that less conservative assumptions may be made for conductor
and vacuum wall design.
4. Refrigeration uses 638 MWhr/day or 6.38 % of W,OOO MWhr. This could be
reduced by 25 % by adding a 4.5 K shield. If the one-way converter loss is
assumed to be 1 %, then the maximum daily converter loss is 200 MWhr.
5. The economics presented above might be assigned to 20,OOO-MWhr storage
since two units of opposite polarity probably could be located close together
to reduce the net external field. Alternatively, a larger-diameter guard ring
coil might be used. Otherwise, the external field from one coil equals the
earth's field at about 5 km.
ACKNOWLEDGMENTS
The authors wish to thank project staff members who contributed substantially to this work, especially
M. Kuchnir, who developed the 1.8 K system, R. W. Moses for magnet shape optimization, J. K. Ballou
for magnet design, and D. P. Hartman for electrical engineering. Contributing faculty include F. J. Worzala
on materials, B. C. Haimson on rock mechanics, and Visiting Prof. R S. Claassen on materials.
This project is supported by the National Science Foundation, the Wisconsin Alumni Research
Foundation, The Wisconsin Utilities Research Foundation, and the General Electric Company.
REFERENCES
I. Electrical World, 179:51 (1973).
2. R. C. Walker and H. C. Early, Rev. Sci. Instr., 29(11): 1020 (1958).
3. R. Carruthers, in: High Magnetic Fields (H. Kolm, B. Lax, F. Bitter, and R. Mills, eds.), MIT Press,
Cambridge, Massachusetts (1962), p. 307.
4. Z. J. J. Stekly, "Magnetic Energy Storage Using Superconducting Coils," IDA/HQ 63-1412, Proceedings Pulse-Power Conference (February 1963), p. 53.
5. J. Sole, "Stockage D'Energie Possibilites Supraconducteurs En Vue Des Decharges De Grandes
Puissances," Rapport CEA-R3243, Comm. Energie Atomique, Limeil, France (June 1967).
6. M. Ferrier, in: Low Temperatures and Electric Power, Intern. Inst. Refrigeration, London (March
1969), p. 425.
7. F. Irie and K. Yamafuji, "Some Fundamental Problems with Superconducting Energy Storage," in:
Low Temperatures and Electric Power, Intern. Inst. Refrigeration, London (March 1969), p. 411.
8. H. Brechna, F. Arendt, and W. Heinz, in: Proceedings of 4th Intern. Con/. on Magnet Technology,
Brookhaven, New York (1972), p. 29.
126
R. W. Boom, G. E. MelDtosh, H. A. Pet-, and W. C. Youog
9. J. R. Powell and P. Bezler, "Warm Reinforcement and Cold Reinforcement Magnet Systems for
Tokamak Fusion Power Reactors: A Comparison," BNL-17434, Brookhaven National Laboratory,
New York (November 1972).
10. R. W. Boom and H. A. Peterson, IEEE Trans. Magnetics, MAG-8(3):701 (1972).
II. W. V. Hassenzahl, J. D. Rogers, and T. E. McDonald, "Magnetic Energy Storage and its Application
in Electric Power Systems," presented at IEEE Intern. Conv. and Exposition (March 26--30, 1973),
New York, New York.
12. W. V. Hassenzahl, W. E. Keller, and B. L. Baker, "The Economics of Superconducting Magnetic
Energy Storage Systems for Load Leveling-A Comparison with Other Systems," LASL-LAMS
Report, in press.
13. W. C. Young and R. W. Boom, in: Proceedings of 4th Intern. Magnet Conference, Brookhaven, New
York (1972), p. 244.
14. R. W. Boom, P. Brown, J. C. Laurence, and F. J. Worzala, Bull. Am. Phys. Soc., 17(12): 1195 (1972).
15. C. N. Whetstone, ALCOA, private communication.
16. C. Linnet, V. Purdy, Y. W. Chang, and T. H. K. Frederking, "Unsaturated Helium Cooling Limits,"
Rept. DAAK02-68-C-0064, UCLA (October 1970).
17. G. P. Coombs, CTi, Waltham, Massachusetts, private communication.
18. M. L. Kyle, E. J. Cairns, and D. S. Webster, "Lithium/Sulfur Batteries for Off-Peak Energy Storage:
A Preliminary Comparison of Energy Storage and Peak Power Generation Systems," ANL-7958,
Argonne National Laboratory (March 1973).
DISCUSSION
Question by D. C. Litz, Westinghouse Electric Corporation: Is there any provision for shielding the
energy storage coils?
Answer by authors: We have considered a larger diameter guard ring coil which could cancel out
the dipole moment or, alternatively, the construction of solenoids in pairs of opposite polarity which
could partially short-circuit the flux from the other solenoid. All solutions to the problem of external
magnetic fields, the two mentioned above or a toroidal structure, are expensive.
Question by J. K. Hulm, Westinghouse Electric Corporation: Could you give some rough economic
comparisons between this proposed energy storage system and pumped hydrostorage now in use or
lead-acid batteries under consideration?
Answer by authors: We have not completed our cost analysis of the superconducting storage system
but know that capital costs must be competitive with generator costs and with pumped storage costs
before the superconducting system can be considered. Our target range is therefore 200-500 $/kW.
Battery storage is under development (see references 12 and 18) with somewhat uncertain costs at present.
Question by D. Atherton, Queen's University: Would you comment on the shear forces in the rock
caused by the axial forces?
Answer by authors: As described in the paper axial forces could be conducted from a tum or group
of turns to support ledges attached to the rock face. The ledges probably should not be of rock; the shear
problem in the rock behind the ledge structure can be minimized by joint grouting and rock surface
pinning with long bolts.
D-l
HIGH-SPEED TRANSPORTATION LEVITATED BY
SUPERCONDUCTING MAGNET
K. Oshima
University of Tokyo
Tokyo, Japan
and
Y. Kyotani
Japanese National Railways
Tokyo, Japan
INTRODUCTION
Operation of the world's fastest commercial train (maximum operating speed
of 210 km/hr) was initiated in 1964 by the Japanese National Railways (JNR) with
the completion of the Tokaido Shinkansen (New Tokaido Line) between Tokyo and
Osaka (515 km). The line has now been extended to the San-yo district, a distance of
676 km from Tokyo to Okayama. Present plans call for further expansion of the line
to a Nationwide Shinkansen Network and a maximum operating speed of 250 km/hr.
The high-speed transportation system presently carries about 250,000 passengers
per day compared with 60,000 passengers per day at the start. Approximately 230
trains are used for the daily operation. However, the demand for additional transportation facilities is still on the upswing. Even though another line connecting Tokyo and
Osaka through the northern district is planned, predictions are that the capacity of the
Tokaido Shinkansen will be saturated by 1983 to 1984. Thus, by 1985, it will be
necessary to construct additional rail facilities in the Tokaido district (Pacific coast
district between Tokyo and Osaka).
JNR has made an extensive study of the economic and technical aspects of the
transportation system to meet such a situation and has come to the conclusion that
the most desirable solution is to run a new high-speed ground transportation system
with a train having an average speed of over 300 km/hr and preferably 450 km/hr. The
latter would translate into a maximum speed of 550 km/hr. In 1970, JNR and the
Ministry of Transportation made a decision to start research and development
efforts for such a high-speed railway system.
It has been concluded, after considering various factors such as geographical
conditions, safety, noise, pollution, transportation capacity, and technical feasibility,
that a linear propulsion system whose active side is on the ground coupled with a
superconducting magnet which provides levitation for the suspension and guidance
system will be the best system, provided it is technically successful in the available time
period. Construction is aimed at around 1985, with development of the system
expected to be completed by 1977.
127
128
K. 0sIIima .... Y. Kyotui
BASIC CHARACTERISTICS OF mE SYSTEM
The basic system components for the new train are the driving system and the
supporting and guiding system. In the conventional system, the train is driven by the
wheels and relies upon the frictional force between the rail and the wheel for traction.
The supporting and guiding system is also the wheels and the rail. With an increase
in the speed of the train, the resistive force increases, while the adhesive force by
friction decreases. Data from the Shinkansen show that the upper limit in speed of
the conventional system is around 300 to 350 km/hr. Here is where the resistive force
becomes larger than the adhesive force. Thus, beyond this limit it is not only necessary
to adopt a nonadhesive driving system but also to reevaluate the supporting and
guiding system.
Several alternative driving systems investigated by JNR include jet or propeller
engines, linear turbines, linear motors, and tube systems. As for the supporting and
guiding system, such alternatives as mechanical systems of wheels or a gliding sleigh,
aeronautical systems with wings, airflow systems with air-cushions, and magnetic
levitation systems have been considered. The basic concept finally chosen for the
new train is a linear motor drive with the primary on the ground and utilizing magnetic
suspension with the superconducting magnet installed on the vehicle and loops or
sheets of normal conductor on the track. It is expected that a combined system of
linear synchronous motors utilizing the superconducting magnets of the guiding
system will be developed for the driving system.
In order to give a general idea of the system, some of the design values proposed
as guidelines for evaluation and planning of the research and development program
are listed in Table I.
MAGNETIC LEVITATION SYSTEM
The use of a permanent magnet or normal conducting magnet was originally
considered for the magnetic levitation system. However, a system utilizing a superconducting magnet eventually proved to be a more attractive one. There are many
aspects which have to be investigated or developed before the final system can be
constructed. Some of these items are listed in Table II.
Table I. Design Criteria for High-Speed Levitated Train
Maximum number of coaches/train
Maximum operation speed
Maximum acceleration and deceleration:
Acceleration
Deceleration, normal brake
Deceleration, emergency brake
Starting speed of levitation
Effective levitation height (between coil centers)
Accuracy of the track
Hours of operation
Period of operation without maintenance service
Number of superconducting magnets
Levitation
Guiding and drive
Carriage
Weight
Dimensions
Propulsion
16
S50km/hr
3km/hr/sec
Skm/hr/sec
lO km/hr/sec
100km/hr
250mm
±10mm/lOm
From 6 AM to 12 PM at 1S-min intervals
18hr
4 x 2 rows/coach
4 x 2 rows/coach
30 tons
2Sm x 3.4m x 3.4m
Linear synchronous motor
High-Speed Transportation Levitated by Superconducting Magnet
129
Table II. Magnetic Levitation System
Aspects Requiring Further Study
Levitation system
Loops vs. sheets
Repulsive vs. differential mechanism
Lift force
Magnetic drag force
Dynamic behavior
Interactions of mechanical and magnetic system
Superconducting magnet system
Stability
Interaction of ac magnetic fields and shielding
Leads and magnetization modes
Weight reduction
Superconducting materials
Cryogenic system
Refrigeration and cooling system
On-board system vs. on-ground system
Refrigeration cycle
Lightweight refrigerator and compressor
Lightweight cryostat
Insulation and load supporting mechanism
Safety
Materials of construction
Control and instrumentation
With respect to the cryogenic system, an evaluation of several alternative
systems has already been made. As for the cooling system, the present intent is to
develop an on-board refrigeration system with a separate unit for each train coach.
Such factors as weight, input power, volume, reliability, control, and safety have to be
given serious {;onsideration. Comparative studies are presently underway to investigate several different schemes, including such systems as one refrigerator for each
coach and one refrigerator for each magnet cryostat.
[
10--0
r1
CarrlGQ'
Guldl.t mat·o'
1:t
, .....- - - - SolpI.llo. mot.I'
1 lfi"]!
L
6000
!l!,
~
6000
-----.I...-
l~
6000--.1
Fig. I. Superconducting magnet arrangements.
130
K. Oshima and Y. Kyotani
The heat load for each cryostat is estimated to be 3 W at 4.5 K and 28 W at 77 K.
Schematic arrangements of the cryostats are shown in Fig. 1. The superconducting
magnets being used in the evaluation have length-to-width dimensions of 4000 mm
and 500 mm, respectively, pitches of 6000 mm, and maximum ampere-turns of
500,000 A-turns. The design target for the total weight of the cooling system has been
set at two tons.
TESTS FOR SUPERCONDUCTING MAGNET LEVITATION
Most of the present research effort in industry and at JNR is involved with
determining the characteristics and performance of the superconducting magnet
levitation system. As a start, JNR has constructed a basic test facility to provide a
direct comparison between the theoretical calculations and the actual lift and drag
forces of the system. The facility provides a capability for carrying out dynamic tests
on the levitation system with superconducting magnets on the coach and loops of
normal conductor in the track (up to a speed equivalent to 100 km/hr). A round
cryostat containing two semicircular superconducting magnets is suspended above a
rotating disk of 1420 mm diameter with loops of normal conductors. The time
constants of the normal conductor loop can be varied by connecting inductors, while
the relative pitches of the superconducting magnets and loops can be varied by changing the number of normal conductor loop units. The upper part of the facility can be
replaced with a movable cryostat containing the superconducting magnets and can be
levitated by the repulsive force to investigate the dynamic characteristics. In this case,
the superconducting magnet is magnetized by a persistent current mode. Figure 2
shows the test facility.
As an extension to the rotating disk experiment, a test facility with linear drive
was constructed. The levitated vehicle can thus be tested on a 220-m track. This
permits an evaluation ofthe dynamic characteristics of magnetic levitation in combination with a linear synchronous motor drive. Aseparate line, parallel to the running
LI f t
fOr
ell
Su,penl i on fram .... - - --<Dr Cl9 for c.
Mealur l no po r t
I
Spacer
Con cr . .. bed
Fig. 2. Levitation test facility.
High-Speed Transportation Levitated by Superconducting Magnet
131
Table III. General Features of Test Vehicle
Vehicle dimension (I x w x h)
Weight
Maximum testing speed
Acceleration and deceleration
Drive force
Distance of levitation
Superconducting magnet
length
width
pitch
Normal conducting loop
length
width
pitch
Persistent current switch
Low-speed support
4m x 1.6m x O.7 5m
2 tons
50km/ hr
>O.2g
500 kg
100m
1200mm
400mm
1490mm
300mm
280mm
468mm
Mechanical
Wheel
path of the vehicle, is also available to permit testing of the magnetic levitation by
itself. Table III presents the general features of the vehicle, while Fig. 3 shows a
schematic drawing of the facility.
The world's first known demonstration of a test vehicle run on an actual straIght
track and levitated by a superconducting magnet was performed in July 1972. The
experimental results from these tests were in good agreement with earlier theoretical
predictions. Vehicle levitation was 40 mm over a distance of 100 m at a speed of
about 50 km/hr.
AU:IIi i i i 0 r y d r i v i n 9 de'l i c:.
t l in eor
SuperconductinQ mOQnet cor
Induction motor)
Oraw bar
No rmal
con d .c !
i "9
l oop
Su p. rcond.c!in Q lo op
Fig. 3. Linear synchronous motor test facility.
K. Oshima and Y. Kyotani
132
oor---------~--------,_--------_r--------_,
°8~~~~~~9~~~~-+-IO~------~,~1--------J,·2
TINE., •• e
Fig. 4. Levitation characteristics for LSM vehicles.
Some of the experimental results are shown in Figs. 4and 5. Figure 4 shows the
mean height of levitation and the difference in levitation height between the front
and back of the vehicle due to vehicle movement. The up and down frequencies and
the pitching oscillations approximate the characteristic frequencies of the magnetic
suspension system (1. 7 Hz) and the pitching caused by the magnetic moment (1.4 Hz),
respectively. Figure 5 shows the levitation characteristics as a function of the running
speed.
In another study, two prototypes of lightweight cryostats were designed and
constructed to determine how much the heat influx and weight of such cryostats
could be reduced. Type A is a flat cryostat with one superconducting magnet, while
type B is shaped like a dumb-bell with a superconducting magnet in each end. The
cross-sectional diagrams of each prototype are shown in Fig. 6,while Table IV lists
the features and dimensions of these units.
300.-----,------.------.-----~----__,
...
...•
cp
RO l l
lurfac e
for wh • • l
c
"
0
200 r-----~----~------+-----_+----~
,
I
I
I
I
/
I
I
-2 . 0 i
~
u
o O, I •• n b
.!:
LSIoI
'0
•
o eoad in
'io
c
c
I without fa il
I
100r-t---i-----~--~~+-~~~~~__1 -I. 0
...o·
...
iifa
40
•~
0
...
so
0
0
.;
Fig. 5. Mean levitation height H,
drag force FD , and angle of
inclination 8, vs. speed of LSM
vehicle.
High-Speed Transportation Levitated by Superconducting Magnet
133
co o l i n V p i p.
Rad i a t i on
Ih i . l d
Fig. 6. Cross section of two
prototype cryostats.
Typ .
B
Table IV. Design Features of Prototype Cryostats
Type A
Shape
Superconducting magnet (dummy)
Total weight, kg
Maximum acceleration tolerance, g
Liquid helium capacity, liters
Liquid helium loss, liters/hr
Material
Flat
300 x 1500 mm
Two units
100 kg (for two)
550
10
250
4
Tialloy
TypeB
Dumb-bell
800 x 4000mm
One unit
200 kg
700
10
250
6
Tialloy
The heat leak into the cryostats was reduced to four to six liters/hr with the use
of special support and neck tube devices.
DEMONSTRATION VEHICLE
As a part of the development program and in commemoration of the Japanese
railway centenary, a demonstration vehicle capable of carrying four passengers was
constructed utilizing both superconducting magnet levitation and linear induction
motor drive. Figures 7 and 8 show a schematic drawing and photograph, respectively,
ofthe demonstration vehicle. The general features ofthe vehicle, designated as ML 100
(Magnetic Levitation-looth Railway Anniversary Commemoration), are presented
in Table V.
Some ofthe experimental results are shown in Figs. 9 and 10. Figure 9 shows the
levitation characteristics at a maximum speed of 60 km/hr. The up and down oscillation frequency of the vehicle is around 1.8 Hz, which is the characteristic frequency
of the magnetic suspension system. Figure 10 shows the levitation characteristics
as a function of the running speed. Extrapolation of these results indicates that at a
speed of 500 km/hr, the lift-to-drag ratio will be about 80 and the levitation height
will be about 70 mm.
K. Oshima and Y. Kyotani
134
2500
o
'"
'"
F l oor
Roocl l on plole
s. c:O ndary conductor
(on.oh i clo)
leve l
Fig. 7. Schematic of the ML lOOdemon·
stration vehicle. Dimensions are in
millimeters.
Fig. 8. Photograph of the ML 100 demonstration vehicle.
High-Speed Transportation Levitated by Superconducting Magnet
135
Table V. Design Features of Demonstration Vehicle
General
Running speed
Acceleration and deceleration
Levitation height
Lift force
Load
Drive force (maximum)
Vehicle
Dimension (I x w x h)
Weight (including load)
Superconducting magnet
Number of cryostats
Weight (including two magnets)
Dimension (I x w x h)
Superconducting coil
Ampere-turns
Coil dimension (I x w)
Coil pitch
Persistent current switch
Liquid helium capacity
Normal conductor coil
Coil dimension (I x w)
Coil pitch
Track length
Number of coils
Material
:i
E
6Okm/hr
3.5 km/hr/sec
100mm
3500 kg
250kg
950 kg
7000 mm x 2500 mm x 1930 mm
3500 kg
Two
1000 kg each
4250 mm x 600 mm x 400 mm
Two/cryostat
250,000 A-turns/coil
1550mm x 300mm
1800mm
Mechanical
200 liters
480 x 330mm
675mm
240m
355 x two rows
Aluminum
100
·• .
.. . .
• E
~
-
0:
~ -;, EE
v
---
~
.., ... :z:
~
!
H
50
I
o
17
~
.....
~
~'\
18
~
V
19
V4
20
T I lot E . . . ~
Fig. 9. Levitation characteristics of the ML 100 demonstration vehicle.
136
K. Oshima and Y. Kyotani
300
r------y- - ---r---,--- --.---------y-- --, -
1 .~
200 r-----+----;'''--+----t----t----+---~ - 1. 0
E
E
...
·."
••
... 100
..
•
c
-0.5
u
»
·
....
·
·"
(;
...
"
2
U
"
·...
.
"
~
·
...
...
"
o
<>
0~---1~0~--L~--~~~0~--4~0~--~~~0--~ 0
60
S PEE D .
km/ h
Fig. 10. Mean levitation height H. drag force FD • and angle of inclination () vs.
speed of ML loo demonstration vehicle.
CONCLUSION
Experimental studies on superconducting magnet levitation have just begun in
Japan. This review has only described the experimental studies performed to date
by the Japanese National Railways. Additional experimental and theoretical work
is being carried out simultaneously in both industry and universities.
Plans have now been formulated to perform a field test of a levitated superconducting magnet vehicle at the speed of 500 km/hr on a 7-km track in 1975. Studies
for designing the test vehicle are presently under way and it is expected that this test
will open the way to final development and design of the actual train which has to
be completed by around 1977 in order to be operational in 1985.
DISCUSSION
Question by K. D. Timmerhaus, University of Colorado: Were the performances of the cryostats
during the vehicle runs in agreement with the design and static tests?
Answer by the author: The cryostat was constructed as a prototype for the test vehicle to be tested
in 1975 on 7-km test track. Therefore no running tests have been performed as of this time. The cryostat
for the demonstration vehicle ML 100 involved only minor considerations of the liquid helium loss. The
evaporation rate was about 13 Iiters/hr per cryostat and did not change appreciably for either the static
or running condition.
Question by K. D. Timmerhaus, University of Colorado: What was the energy required to maintain
the planned high speed? What was the quality of the roadbed necessary for these high speeds and what
are some of the economics of the system?
Answer by the author: The energy required for a train of 16 cars is estimated to be about 90 MW.
The accuracy of the roadbed needed will be ± 5 mm/ 10m. The economics of the system are estimated
as follows: Capital investment required for the construction of the system will be less than 2000 billion yen
(seven billion dollars). The proposed fare for travelers between Tokyo and Osaka is estimated to be about
5000 to 6000 yen compared to 4000 yen for the present Tokaido Shinkansen.
D-2
SRI MAGNETIC SUSPENSION STUDIES FOR
HIGH-SPEED VEHICLES·
H. T. Coffey
Stanford Research Institute
Menlo Park, California
INTRODUCTION
The fact that vehicles can be levitated magnetically has been known for quite a
long time [1], but practical means of constructing them are only now emerging. The
effort being applied to the development of these vehicles and their associated transportation systems is considerable, both in the United States and elsewhere.
There are two basic types of magnetically levitated (MAGLEV) vehicles; those
levitated by attractive and those levitated by repulsive magnetic interactions. In the
attractive case, conventional copper-wound electromagnets attached to the vehicle
are attracted upward to a ferromagnetic rail. Since this type of interaction is unstable
in that a displacement tends to increase the destabilizing force, servo control of the
magnet current is required to stably levitate the vehicle 1 to 2 cm below the rail.
The repulsive system uses superconducting magnets on board the vehicle to generate
eddy currents in a passive conducting guideway, resulting in the levitation force
that suspends the vehicle above the guideway. By shaping the guideway properly,
the same type of interaction produces the forces required to guide the vehicle. Since
the levitation and guidance forces increase as the vehicle approaches the guideway
and decrease as it moves away, the motion is inherently stable.
As might be expected at this state of development, many variations have been
proposed in the levitation and propulsion systems to be used in each type of vehicle.
For example, a continuous sheet of aluminum [2-18], a slotted sheet of aluminum
[19,20], an aluminum "ladder" [21-24], and various arrangements of discrete passive
coils [25-30] have been proposed for the guideway in the repulsive system. The attractive systems have been built with guidance provided by mUltiple coils beneath the
rail and with coils reacting against vertical rails. The propulsion system can be
independent of the levitation system in either type ofvehicle. The propulsion system
most often mentioned is the linear induction motor, either on board the vehicle or
in the guideway.- It is also possible to incorporate on-board propulsion with the
levitation system in attractive-type vehicles by applying alternating current to the
coils to make a linear induction motor [31]. This is not feasible in the repulsive
system since the power losses in the superconducting windings are prohibitive. The
repulsive system, however, is ideally suited to linear synchronous propulsion [32).
• Work supported in part by the Federal Railroad Administration. The opinions expressed are those of
the author. Invited paper.
137
138
H. T. Coffey
This is accomplished by electrifying the guideway so that the guideway is analogous
to the stator and the vehicle to the rotor of a synchronous motor.
There are inherent advantages and disadvantages to each of the two basic systems.
The attractive system operates from a standstill to rather high velocities. At very high
velocities, however, this system suffers from its small clearance (1 to 2 cm) and associated high control frequency requirements as well as loss in lift caused by eddy
currents in the ferromagnetic rails. The repulsive system operates well at very high
speeds and provides large clearances (20 to 30 cm) but requires wheels at low speeds.
It also requires superconducting magnets and their associated cryogenics. The
attractive system, because of its small clearance, will require a secondary suspension,
while it is possible the repulsive system will not.
The list of differences between the two types of MAGLEV systems and their
comparative merits is extensive, but it is not the purpose of this presentation to
compare them. The primary objective is to discuss the design and performance of
the magnetically levitated test vehicle under evaluation at Stanford Research Institute
(SRI).
mE NEED FOR HIGH-SPEED GROUND TRANSPORTATION
Since the concept of magnetically levitated vehicles is so old, it is reasonable to
ask why it was so long in reaching fruition. There appear to be two reasons for this.
First, the technology required to implement a practical system of this type has only
recently been developed. This is especially true of the repulsive system, where superconducting magnets are necessary to keep the suspension power requirements within
reasonable bounds. The attractive system awaited the development of practical
magnet control devices. Second, the need for a high-speed ground transportation
system has only recently become sufficient to warrant the development of an entirely
new system. Since the MAGLEV concept arose in the early 1900's, there has been a
rapid increase in travel by automobile and airplane. These modes of transportation
225.------,------,-------,
TOTAL U.S.
POPULATION
200
- 20
:f
E
1400 . - - - - - - , - - - - - , - - - - - - - ,
TOTAL INTERCITY
TRAVEL
15
AUTO
z
o
~
TOTAL INTERCITY
PASSENGER MILES
>cr
~z
109
>-
to
~
PER CAPITA
INTERCITY TRAVEL
5 ;;:
>-
0:
u
"
25
cr
~
oL----~---~
1966
1965
____~o
1975
400
~
1985
Fig. 1. Growth in U. S. population, total intercity
trave~ and per capita passenger demand (source:
Transportation Facts and Trends; July 1972 and
January 1973 supplement).
200
TOTAL INTERCITY TRAVEL
BY PUBLIC CARRIER
~L55---~,~-5----J,97-5--~,~5
Fig. 2. Comparison of total intercity
travel by public carrier and intercity
travel by automobile (source: Transportation Facts and Trends; July 1972
and January 1973 supplement).
139
SRI Magnetic Suspension Studies for High-Speed Vehicles
have served the nation well, as did the train. As we look to the future, however, we
must anticipate the saturation of airline transportation in at least some corridors
and develop an alternative mode of ground transit.
In the last ten years, we have experienced a sharp increase in intercity travel
caused by a growing population and an increase in per capita travel demand (Fig. 1).
The ways in which we have satisfied this need for travel are shown in Fig. 2. The fact
that we rely on the private automobile for most all intercity travel surprises most
people. Even when traveling distances of 500 to 1000 miles, 62 % of this travel is by
automobile (in 1967, the latest year for which statistics could be found). In addition,
only 16 % of the travel is for business reasons. This extent of reliance on the automobile will not change rapidly, and we can expect the automobile to dominate as a
form of transportation for the foreseeable future.
The place where high-speed ground transportation might be expected to have
an impact is in the public transportation of passengers. The changing emphasis on
modes of public transportation is clearly shown in Fig. 3. Obviously, air transportation is preferred to other modes by a considerable and growing margin (the decrease
in the last two years is caused, at least in part, by a change in the method of compiling
the statistics). Bus transportation has remained relatively constant and rail transportation has continually declined. Although air transportation is obviously preferred
to any other form of public transportation, it still represents only 9 % of the total
intercity travel. Much has been done and more will be done to expand the capacity
of the air transportation system, but there is a question as to how far this system can
be expanded to meet future transportation needs. Unless we are to expand the use
of automobiles with their associated problems in direct relation to the demand for
transportation, another form of passenger transportation will have to emerge. This
is the need that can be fulfilled by high-speed ground transportation.
Vehicles traveling at speeds of 300 mph on the ground between the centers of
urban areas can compete with the travel time of airplanes over distances of 400 to
500 miles since the terminals can be located in the centers of cities. MAGLEV vehicles
can be made to be quiet enough for operation in the cities, relatively pollution free,
and as attractive in appearance and service as modern jetliners. Further, their
operation will not be affected by fog and rain, although deep snow might cause some
problems, especially with the attractive system. Consequently, scheduling should be
very reliable. The technology exists and the need is apparent for a new high-speed
TOTAL
140
AIR
Fig. 3. Comparison between modes of public transportation.
(Data for 1971 and 1972 reflect a change in means of compiling
statistics. Source: Transportation Facts and Trends; July 1972
and January 1973 supplement).
~~oo<"--- BUS
RAIL
O~~~~====~W~AT~ER~__~
1955
1965
1975
1985
140
H. T.Coft'ey
ground transportation system. If the costs are not prohibitive, these systems should
be available in the next decade or so.
EFFORTS IN MAGLEV DEVELOPMENT
The United States' efforts in developing the repulsive type of magnetic levitation
system under Federal funding have been centered in Stanford Research Institute,
Ford Scientific Research Laboratories, Massachusetts Institute of Technology, and
Brookhaven National Laboratories. Earlier work was performed at Atomics International and the Sandia Corporation, and private work is underway at the General
Motors Research Center. Several universities are also participating.
The efforts in these organizations have concentrated on developing the suspension systems, although work is being performed on integrating the suspension with a
linear synchronous motor using the magnets as the rotor as described earlier [33.34].
Most notable in this respect is the effort at MIT, although most of the organizations
listed have considered this possibility to some extent.
A very significant and dedicated effort is underway in Japan to develop the
repulsive type of MAGLEV under the sponsorship ofthe Japanese National Railways
(JNR). The major emphasis in Japan has been on the repulsive system with coil or
"ladder" type guideways as opposed to the continuous sheet guideways most often
considered in the United States. Major companies engaged in this work in Japan are
Mitsubishi Electric Corp., Sumitomo Electric Industries, Hitachi Electric Corp.,
Kobe Steel, Ltd., Fuji Electric Co., and Tokyo Shibaura Electric Co.
. The major effort in Germany is on the attractive MAGLEV system. This work
is being performed at Krauss-Maffei and Messerschmidt-Bolkow-Blohm. The
repulsive system is under development at Siemens AG in cooperation with AEGTelefunken and Brown-Boveri et Cie [35].
Canada's work on magnetically levitated vehicles is centered in Queen's University's Institute for Guided Land Transport under the sponsorship of the Canadian
Ministry of Transport [34]. Although the work here is concentrated on the repulsive
type of MAGLEV system, Canada will be the first country to install a revenueproducing, low-speed, attractive MAGLEV transportation system. This low speed
system will be installed in Toronto by Krauss-Maffei.
Studies are being conducted in England on the repulsive system at the University
of Warwick and on the attractive system at the University of Sussex. It has recently
been announced that a test vehicle will be constructed and tested at Warwick.
SRI-DOT MAGLEV TEST VEHICLE
The first study of magnetic levitation at this facility was performed cooperatively
with Atomics International in 1967 for the Sandia Corporation [2-5]. The objective
was to assess the feasibility of magnetically levitating rocket sleds at speeds up to
Mach 10 ( -7000 mph). It was concluded that such a vehicle was technically feasible.
At the suggestion ofD. Williams of Sandia, this facility began to consider the possibility
of applying this same technology to high-speed ground transportation. A study was
performed in 1971-72 for the Federal Railroad Administration to experimentally
and theoretically evaluate the technical feasibility of magnetically levitating highspeed ground transportation vehicles [11]. The basic theory of the magnetic interaction
forces was developed and a small, nonlevitated vehicle carrying a single 0.19 x 0.25 m
SRI Magnetic Suspension Studies for Higb-Speed Vebicles
141
superconducting magnet was constructed and used to confirm the theoretical calculations. These tests were performed on a magnet held at a fixed position in aU-shaped
guideway 500 ft in length.
A fully levitated test vehicle was constructed in 1972 and tested on a modified
guideway [12]. This vehicle consisted of a 4.3 x 1 m chassis, four 0.27 x 0.32 m
superconducting magnets, a battery for power, and a telemetry system for data
acquisition. The vehicle weighs 300 kG, but almost 400 kG was levitated on one
occasion. The overall guideway is 500 ft long, 400 ft of which is lined with aluminum
as shown in Fig. 4. Each horizontal plate is 0.46 m wide and the vertical plates are
0.15 m high. The plates are 1.9 em thick. This rectangular geometry largely decouples
the vertical and horizontal motions of the vehicle.
The vehicle was completely levitated and guided by the magnetic forces. Propulsion and braking were provided by an endless cable powered by a glider towing
winch visible at the end of the guideway in Fig. 4. The cable was attached to a wood
skid guided by an aluminum channel in the center of the guideway. Contact to the
vehicle was made through a vertical pin attached to the skid. The pin was mounted
using two linear bearings to permit the vehicle to move laterally in the guideway.
A third linear bearing was placed around the pin to permit the vehicle to move in
the vertical direction. This bearing was in turn placed inside a ball bushing which
permitted the vehicle to roll, pitch, or yaw through 15° of motion without restriction.
The only constraint was in the direction of motion.
Each of the 0.32-m long and 0.27-m wide magnets (Fig. 5) contains 580 turns
of 0.76-mm-diameter twisted niobium-titanium composite conductor with a
Cu: superconductor ratio of 5: 1. Insulation was provided by copper oxide and a
20 % coverage of 0.25-mm-diameter nylon spiral wrapping. The conductor was
designed to conduct a cryogenically stabilized current of 100 A and a partially stable
Fig. 4. SRI-DOT magnetically levitated test vehicle.
142
H. T. Coffey
Fig. 5. One of four superconducting magnets
used to levitate the vehicle (shown before contact
blocks and switch were installed).
current of 160 A. Since it was not intended to operate the vehicle at partially stable
currents, and since the possibility of damage existed, the magnets were not tested to
their maximum current. A current of 140 A was unintentionally applied to the
magnets without quenching, however. The current in the magnets was controlled
by simple carbon pile rheostats, and the mutual inductance between the magnets
and the guideway caused the current to increase when the vehicle was in motion.
The in-flight current was not recorded, but appears to have been as high as 115 A
on one occasion.
The helium dewars were liquid nitrogen-shielded and had a capacity of about
6.3 liters of nitrogen and 10 liters of helium. Compressed, evacuated fiberglass was
used for insulation. It provides good insulation under a compressive load and is
relatively insensitive to the pressure in the vacuum space. To install the insulation,
the helium container and its insulation were enclosed in a plastic bag which was
evacuated until the desired compression was attained. The entire assembly was then
inserted in the dewar and the excess plastic cut away. The remaining insulation was
tamped into place. The technical characteristics of this insulation were reported
earlier 1]. One of the dewars is shown in a partially packed state in Fig. 6 and in
final form in Fig. 7. This method permitted the construction of rectangular aluminum
dewars with a wall thickness of only t in. The helium dewars are made of /6-in.
stainless steel. The advantage of using rectangular dewars is obvious in Fig. 4 in
that this geometry gives the maximum clearance from the rectangular guideway.
The design hold time for the cryogens was 8 hr. The actual maximum hold time
during test runs was never carefully determined, but was in every case in excess of
5 hr even though there was slosh and spilling in many tests. The static boil off rate
was determined to be about 1.4 liters/hr of helium and 0.73 and 0.84 liters/hr of
nitrogen in each of the two dewars tested.
e
SRI Magnetic Suspension Studies for High-Speed Vehicles
Fig. 6. Partially completed magnet and dewar
assembly showing compressed fiberglass used for
insulation.
Fig. 7. Completed dewar before installing terminals on counterflow-cooled current leads.
143
144
H. T.Cofl'ey
or-----.------r-----,------.-----.-----~----_,
MAGNET lENGTH ell AND WIDTH eWI
- - L-o.32m,W-o.27m
9
I
- - L-o.30m.W-o.30m
6
aeoN (200 Ibl AT 4 x
104
------.--\--\-'~-+T-
V-20m/.
V· 6 mI.
v-
4 mI.
O~O----~=---~~----~----~~----L-----~----~O.14
SUSPENSION HEIGHT
Fig. 8. Lift force as a function of suspension height for two magnet designs
considered. The straight lines indicate assumed vehicle weights (divided by
four) and intersect the curves to yield the suspension height vs. velocity.
The forces generated by a variety of rectangular magnets was calculated using
theoretical formulas before the design choice was made. The lift force as a function
of the suspension height (measured from the center of the magnet coil to the top
surface of the guideway) is shown in Fig. 8 for two of the sizes considered. The lift
forces appeared to be adequate in either design assuming 4 x lQ4 A-turns in the
magnet. The guidance force, however, was marginally better in the rectangular
magnet; therefore, this magnet was chosen for construction.
The curves in Fig. 8 were used to obtain the assumed suspension height of the
levitated vehicle for two different vehicle weights as shown in Fig. 9. Wheels were
used for support at a suspension height of 6 cm; therefore, all calculations were
145
SRI Magnetic Suspension Studies for High-Speed Vehicles
0.12r-------r-------~------,_------_r------_r------~------~------_,--------r_----~
0.10
0.08
150 Ib {
0.06
0.04
MAGNET LENGTH ILl AND WIDTH IWI
- - L-o.3m,W-o.3m
- - L-O.32m.W-o.27m
0.02
NI - . x 1~ A-I
10
VELOCITV
20
14
m/s
Fig. 9. Calculated suspension heights vs. velocity for two magnet designs and two assumed vehicle weights
per magnet.
3.2r-----,-----_r----~r_----~----_r----_,------~----_r----_,r_----~----~----_,
L • 0.32 m, W • 0.27 m
2.8
2.4
2.0
""s:
1.6
N
:a~
Zo • 0,06 m
1.2
Zo • 0.08 m
0.8
Zo ·0.10 m
0.4
oL-____
o
~
____
~
_____L____
~
____
~
____
10
~~
____
12
VELOCITY
~
14
____
~
16
____
~
18
_____L_____L____
20
22
~
~
mIl
Fig. 10. Calculated drag force as a function of velocity and suspension height for the 0.32 x 0.27 m magnet.
146
H. T. Coffey
terminated at this point. The calculated drag force exerted on each of the magnets
is shown in Fig. 10 as a function of the vehicle velocity and suspension height.
The variation of the guidance force as a function of the vehicle velocity is similar
to that of the lift force but differs in details. This calculation required a modification
of the original formulas, but was successfully completed by Chilton and has been
reported elsewhere 2 ]. The measurement and interpretation of the guidance force
caused some difficulty in the earlier experiments. This seems to have been caused
by the electrical discontinuity between the vertical and horizontal plates of the earlier
guideway. With this seam unwelded, and the magnet shifted toward the side of the
guideway, the force exerted on the magnet tended to force the magnet further to the
side of the guideway, i.e., the guidance force was unstable. The direction of this force
was reversed when this seam was welded.
There are two additional distinctions between the lift and guidance forces.
Once the vehicle is suspended, the lift force remains constant-any change in force
alters the suspension height. This does not occur in the case of the guidance force
since the vehicle and guideway widths are fixed. The second distinction is caused
by the generation of equal but opposite forces on each side of the vehicle. Thus, when
the vehicle is centered in the guideway, there is no net guidance force on the vehicle.
A net force is only produced when the vehicle deviates from its equilibrium position.
e
1.6
.-------.---------,---.------.----~--____,
14
V · 10 m Il
1,
1.0
,
o
~
0 ,8
"
~ 06
~
-
t;;
z
0.4
0'
- - L • 031 m. W • 021
.
-- L
o E-_ _
o
__ ___
0.0\
0 .02
_ __ L_ _
O.OJ
LATERAL DISPLAC EMENT
TTl
• 03 m . W • 0 .3 m
__
0 .06
Fig. 11. Calculated net guidance
force produced by a pair of
magnets interacting with the
side rails as a function of the
lateral displacement. Two magnet designs are shown. Curves
show the force generated for
displacements at two velocities
and several equilibrium spacings Yo between the magnet and
the vertical surface of the guideway.
SRI Magnetic Suspension Studies for High-Speed Vehicles
147
The magnitude of this force varies with velocity, displacement, and the equilibrium
spacing between the magnet and the vertical surface of the guideway. The equilibrium
spacing, measured between the center ofthe windings on the closest side of the magnet
and the guideway surface, is denoted by Yo and the net guidance force produced in
the test vehicle is shown in Fig. 11. This force is only slightly affected by the suspension
height, and can be comparable to or greater than the lift force. This is achieved by
increasing the current in the magnets to increase the guidance force to the desired
value. The suspension height automatically adjusts itself so that the lift force equals
the vehicle weight. The maximum lateral displacement is limited by the contact of
the dewar with the guideway, occurring at Yo = 4 cm in this case. Thus, if the initial
magnitude of Yo is 5 cm, the maximum displacement is 1 cm. Figure 11 shows that
by increasing the distance between the dewar and the guideway, the maximum force
can be increased but the spring constant is decreased. This type of tradeoff will have
to be determined for each application of the system. In these tests, Yo was usually
set at approximately 7 cm, but variations in the track width caused by warpage
during welding caused this to be a system variable.
DAMPING OF VEHICLE MOTIONS
The ride comfort requirements placed on magnetically levitated vehicles are
quite stringent and are specified in terms of the power spectral density of the vertical
and lateral accelerations. The most simple and obvious way to damp the motion
of a vehicle of this type is by means of an electromagnetic damping plate placed
between the superconducting magnet and the guideway. Any motion in the vehicle
then causes a change in flux and induces current in this plate. The energy dissipated
by the induced current is removed from the kinetic energy of the vehicle and causes
the motion to be damped.
The analysis of the damping achieved by this type of system (using a single-turn
copper coil rather than a plate) has been reported elsewhere [12] and thus only the
major results will be given here. Letting the subscripts 1 and 2 refer to the real levitation
coil and the real damping coil and the subscripts 3 and 4 refer to the magnetic images
of the damping and levitation coils, respectively, the following one-dimensional expression is obtained for the displacement of the vertical position from equilibrium:
'z' + az + hi + cz =
0
(1)
where a = lire, b = wo 2 + (2/rOTe), and c = W0 2 Te· Further, letting Leff and Meff
be the effective self- and mutual inductances of the damping coil, respectively, and
R be its resistance, then
(2)
To
= -
R dM 14Id(2zo)
2
2g(dMldz)eff
(3)
148
H. T.Coft'ey
2g d2 M 14/d(2zo)2
dM I4 /d(2z o)
(4)
In these expressions, Zo is the equilibrium suspension height, Z is the displacement
from equilibrium, and g is the gravitational acceleration. The expressions for
(dM/dz)err and Lerr depend on whether the superconducting magnets are operated in
constant-current or constant-flux (persistent) mode. In the constant-current mode,
(5)
and
where Zd is the height of the damping coil above the guideway. The corresponding
expressions in the case of constant-flux operation are
(6)
3.5
..p • Conuam Flu:!! Opera'ion
3.0
Iv • Connan' Current OPHalion
10 • Velocity of 10 m/sec
GV
•
Infinilll Vek)cJtv
2.'
.
.p. 10
~
..
Q
Z
2.0
h
.'
1-
'6
~
z
..."
~
,.5
0
w
~
;:
CJ
Z
~
g
1.0
0 .5
l__________
-L______
'00
::~======~~~~~~~~~~~T,~.~'~O~~~~ ,.
"
200
'50
TEMPeFtATURf
RJ .........."... ...................
250
JOe)
K
Fig. 12. Damping time constants and frequency of damped oscillation as a function of temperature for
the damping coil installed in the vehicle.
~
149
SRI Magnetic Suspension Studies for High-Speed Vehicles
and
dM24(~)1
d~
_
~=ZO+Zd
MI2 -
M24 (ZO + Zd) dMI4WI
LI - M 14(2zo)
d~
(7)
~=2zo
The solution to (1) is then
(8)
where A, B I , and B2 are constants.
These expressions were derived in the limit where the lift force reaches its asymptotic value. At lower velocities, a correction must be made for the incomplete image
in the guideway. This has been approximated by letting the currents in the image
magnets vary proportional to the lift force FL' Thus
(9)
I(v) = f(v)I(v = 00)
FREQUENCY I
HERTZ
:I
0.3
10
30
100
i~+--'-----rl~--~~~-----r~~~---'+----r--~~---,
z '
o
.
~
u
Z
::l
URBAN TACV
"- - SPECI F ICATION
- .
10
E
eiUl
Q
..J
~
..,
'"
10
_ 6
Ul
0:
~
:z
o
L
10
~0.
_7
Fig. 13. Measured and calculated power spectral densities (PSD) and the Urban TACV
specifications. Constant A of equation (11) arbitrarily chosen as 1.5 x 10- 6 m. (Data
include accelerations caused by un welded joints.)
150
H. T. Coffey
wheref(v) = FL(v)/FL(oo). This modifies the parameter b in (1) to
(10)
and
'0
of (3) is divided by f(v).
.
TII
(II
1.
VERTICAL POSITION ---"""=-"LONGITUDINAL
ACCELERAT I ON
VERTICAL ACCELERATION
;=.;m;;:;;:::.=
1......
~!!!i~!li~~jlillllllll~lllilll
la,
1
DATA FROM TEST 15-7; VERTICAL OFFSET
LONGITUDINAL ACCELERATION MEASURED
VERTICAL POSITION
LATEIlAL ACCELERATION
T....
"
i~w:t~~0~~r:11Jj
(II
1.
!!i1lHIIlIii.....~-7!....-:'f!l------~---.-----~-......:!,....---'.,J;..,M~
T..
....
VERTICAL ACCELERATION
II.'
DATA FROM TEST 15-3; VERTICAL OFFSET
Fig. 14. Data from two tests in which a 1.9-cm vertical offset was placed in the guideway. All data was
taken at the c.g. Sharp accelerations were experienced when the front and rear magnets passed over the
unwelded joints in the guideway.
SRI Mapetie Suspension Studies for High-Speed Vebides
151
At first sight, it appears that the performance ofthe damping coil can be increased
indefinitely by making the cross section larger or by cooling it to reduce its resistance.
The induced voltage, however, is not purely resistive and with a low enough resistance
the current is limited by the inductive reactance. The calculated effect of cooling the
single-turn coils used in the vehicle on the parameters of (8) is shown in Fig. 12.
The effect is shown for constant-flux cp and constant-current Iv operation at 10 m/sec
and at infinite velocities.
This shows that in all cases the first, exponential term in (8) decays quite rapidly,
and that the time constant in the oscillatory term reaches a minimum at approximately 100 K. The coil in the actual vehicle was cooled to approximately liquid
nitrogen temperatures, its construction and installation having been completed
before these calculations. Obviously, the same damping could have been achieved
with a coil at room temperature, but the required cross section would have been
about seven times as great.
The power spectral density of the vertical accelerations has been calculated in
terms of the above parameters and is given by
The parameter A of this equation characterizes the relative roughness of a particular
guideway and has not yet been determined for the present guideway. A value of
1.5 x 10- 6 m corresponds to good runway construction and was assumed for the
calculation shown in Fig. 13. The calculated curve is in reasonable agreement with
the PSD determined from the data in Fig. 14(a). The latter figure, however, indicates
a serious problem encountered in these experiments.
Expansionjoints were placed in the guideway at three places, and a strong vertical
acceleration was experienced by the front and rear magnets as they traversed these
joints. Two additional joints were left unwelded around the center expansion joint to
permit the movement of plates for experimental purposes. This caused a total of six
acceleration spikes in the middle of the guideway. These accelerations plus a 1.9-cm
vertical displacement in the guideway were included in the data in Fig. 13. The peak
PSD over a short unwelded segment of guideway was more than an order of magnitude
lower than the one shown. The peak in the measured PSD at approximately 10 Hz
is caused by the resonance ofthe chassis of the vehicle. The guideway has now been
completely welded and the chassis has been reinforced to remove these difficulties.
Instrumentation is being added to the vehicle to acquire a complete set of
dynamic data on the levitated vehicle and testing will begin very soon. A completely
nonlinear, six degree-of-freedom computer program has been written and will be
verified in these tests. Since it now appears that passive damping of this type will
not be sufficient, active damping control will be incorporated into the vehicle and
tested as a part of this program.
152
H. T.Coffey
ACKNOWLEDGMENTS
The author would like to acknowledge the work of L. Hoppie, F. Chilton, J. Colton, K. Mahrer, and
R. Singleton in the performance of this program, and R. Wesolek and I. Sauer for their assistance in the
construction and operation of the vehicle.
REFERENCES
I. E. Bachelet, The Engineer, 1972 (October 18):420.
2. C. Guderjahn, Atomics International Rept. AI-68-149 (January 20, 1969).
3. C. A. Guderjahn, S. L. Wipf, H. J. Fink, R. W. Boom, K. R. McKenzie, D. Williams, and T. Downey,
J. Appl. Phys., 40:2133 (1%9).
4. T. W. Barbee, Jr., G. N. Bycroft, E. G. Chilton, F. M. Chilton, and H. T. Coffey, "The Hypervelocity
Rocket Sled-A Design Analysis," Stanford Research Institute Report on Project 7014 (July 1968).
5. H. T. Coffey, F. Chilton, and T. W. Barbee, Jr., J. Appl. Phys., 40:2161 (1969).
6. H. T. Coffey, T. W. Barbee, Jr., and F. Chilton, in: Proceedings Low Temperature and Electric Power,
Intern. Inst. Refrigeration, Commission I, London (1969), p. 311.
7. F. Chilton and H. T. Coffey, in: The Helium Society Symposium Proceedings, The Helium Society,
Washington, D.C. (1971), p. 288.
8. H. T. Coffey, F. Chilton, and L. O. Hoppie, Applications of Cryogenic Technology, Vol. 4 (R. W. Vance,
ed.), XYZYX Information Corporation, Los Angeles, California (1972), p. 275.
9. J. R. Reitz, J. Appl. Phys., 41 :2067 (1970).
10. J. R. Reitz and L. C. Davis, J. Appl. Phys., 43: 1547 (1972).
II. H. T. Coffey, F. Chilton and L. O. Hoppie, "The Feasibility of Levitating High Speed Ground
Vehicles," SRI Rept. on Project 1080 (Task I) under Contract DOT-FR-lOOOI (February 1972);
Federal Railroad Administration (FRA) Rept. No. FRA-RT-72-39, National Technical Information
Service (NTIS) Rep. No. PB-210505.
12. H. T. Coffey, J. D. Colton, and K. D. Mahrer, "Study of a Magnetically Levitated Vehicle," SRI
Rept. on Project 1080 (Task II) under Contract DOT-FR-IOOOI (February 1973); FRA Rept. No.
FRA-RT-73-24; National Technical Information Service Rept. No. PB-221696.
13. J. R. Reitz, L. C. Davis, D. F. Wilkie, and R. H. Borcherts, "Technical Feasibility of Magnetic
Levitation as a Suspension System for High Speed Ground Transportation Vehicles," Ford Motor
Company Rept. on Contract DOT-FR-lOO26 (Task I) (February 1972); FRA Rept. No. FRA-RT72-40, NTIS Rept. No. PB-210506.
14. L. C. Davis, J. Appl. Phys., 43:4256 (1972).
15. R. H. Borcherts and L. C. Davis, J. Appl. Phys., 43:2418 (1972).
16. H. T. Coffey, F. Chilton, and L. O. Hoppie, in: Proceedings Applied Superconductivity Conference.
IEEE Publ. No. 72CH0632-5-TABSC (1972), p. 62.
17. P. L. Richards and M. Tinkham, J. Appl. Phys., 43:2680 (1972).
18. J. R. Reitz, R. H. Borcherts, L. C. Davis, T. K. Hunt, and D. F. Wilkie, "Preliminary Design Studies
of Magnetic Suspension for High Speed Ground Transportation," Ford Motor Company Rept. on
Contract DOT-FR-lOO26 (Tasks II and III) (March 1973), FRA Rept. No. FRA-RT-73-27.
19. L. O. Hoppie, private communication.
20. R. L. Byers, R. F. Begley, and G. R. Stewart, "A Magnetically Levitated Merry-Go-Round," to be
published in Am. J. Phys.
21. L. O. Hoppie, F. Chilton, H. T. Coffey, and R. C. Singleton, in: Proceedings Applied Superconductivity
Conference, IEEE Publ. No. 72CH0632-5-T ABSC (1972), p. lB.
22. E. Ohno, M. Iwamoto, and T. Yamada, Proc. IEEE, 61: 579 (1973).
23. T. Yamada, M. Iwamoto, and T. Ito, in: 1972 Digests of the Intermag Conference, IEEE, New York
(1972), p. 48.3.
24. S. Yamamura, Y. Ishikawa, and T. Hayashi, in: 1972 Digests of the Intermag Conference, IEEE,
New York (1972), p. 48.1.
25. J. R. Powell and G. T. Danby, Mech. Eng. 89:30 (1%7).
26. J. R. Powell and G. T. Danby, in: Recent Advances in Engineering Sciences, Vol. 5 (A. C. Eringen, ed.),
Gordon and Breach, New York (1970).
27. J. R. Powell and G. T. Danby, Cryogenics, 11: 191 (1971).
28. J. R. Powell and G. T. Danby, in: Proceedings 4th IECEC Conference, ACS, Washington, D.C.
(1%9), p. 953.
29. J. R. Powell and G. T. Danby. Cryogenics and Industrial Gases, 4: 12 (1969).
30. J. R. Powell and G. T. Danby, in: Applications of Cryogenic Technology, Vol. 4 (R. W. Vance, ed.),
XYZYX Information Corporation, Los Angeles, Citlifornia (1972), p. 299.
SRI Maguetie SuspeDsioa Studies for High-Speed Vebkles
153
31. J. A. Ross, Proc. IEEE, 5:617 (1973).
32. R. D. Thornton, Proc. IEEE, 5:586 (1973).
33. H. H. Kohn and R. D. Thornton, in: Proceedings Applied Superconductivity Conference, IEEE Pub.
No. 72 CH0632-5-TABSC (1972), p. 76.
34. D. L. Atherton, "Study of Magnetic Levitation and Linear Synchronous Motor Propulsion," Queen's
University Rept. No. 6.71.72 (December 1972).
35. G. Bogner, in: Proceedings Applied Superconductivity Conference, IEEE Pub!. No. 72CH0632-5TABSC (1972), p. 214.
D-3
AC LOSSES IN MULTIFILAMENTARY
SUPERCONDUCTING COMPOSITES FOR
LEVITATED TRAINS UNDER AC AND DC
MAGNETIC FIELDS
T. Satow, M. Tanaka, and T. Ogama
Mitsubishi Electric Corporation
Amagasaki, Hyogo, Japan
INTRODUCTION
High-speed trains using superconducting magnets for levitation are presently
in the planning and developmental stages in several countries. The superconductor
used in such magnets is a composite consisting of normal metal matrix such as
copper or aluminum with many superconducting filaments embedded in the matrix.
The superconducting composite is influenced not only by a relatively strong dc
magnetic field, but also by an ac field resulting from irregularities in the tracks or
"heaving motions" of the train. The ac field produced in this manner has amplitudes
below several hundred Oe and frequencies from a few Hz to several hundred Hz.
To date, there have been several studies which have investigated the ac losses in
superconducting composites under ac fields of relatively large amplitudes suitable
for superconducting accelerators, but few reports have been presented covering the
ac losses in the range mentioned above. This presentation reports on the ac loss
measurements and the theoretical investigations which have been made on twisted
and untwisted superconducting composites for use in levitated trains.
EXPERIMENTAL PROCEDURES
Two types of superconducting composites consisting of rectangular copper
matrix and Nb-Ti filaments were investigated in the study. The dimensions of the
composites are listed in Table I. Each sample was wound into a one-layer coil with
an ID of 4.6 em and a length of 6 em. The coil was electrically insulated between
turns, and both ends of the sample were open. The lengths of samples A and B were
5.92 and 1.75 m, respectively.
An ac field, Ho cos (2nfr) (Ho = 20 to 1200 Oe,f = 5 to 500 Hz) and a dc field
H" (0 to 10 kOe) were applied to the sample coil by two superconducting coils with
bore diameters of 7 and 10 em, respectively, placed coaxially with the sample coil.
The power supply for the ac field was a 300-W power amplifier with an ac oscillator.
Losses were measured by a boiloft' calorimeter technique. The relationship between
the losses and the flow rate of helium gas was calibrated by using a standard heater
wound on the sample coil. The lower limit of the ac loss measurements was a few
mWand was comparable to the heat leak along the nylon pipe used for the flowing
helium gas. The reproducibility of the measured values was within ± 5 % at 500 mW
and ±10%at5mW.
154
155
AC Losses in Mu1tifilamentary SuperconductiDg Composites
Table I. Composite Parameters
B
A
Overall size a x b, mm 2
Size over filaments a' x b', mm 2
Number of filaments n
Filament diameter d, mm
Filling factor '1
Twist pitch, mm
Resistivity of matrix p, ohm-em
1.0 x 1.4
0.81 x 1.1
361
0.034
0.254
12.7
3.02 x 10- 8
2.4 x 4.8
1.7 x 3.7
36
0.25
0.152
None
9.51 x 10- 9
LOSSES IN TWISTED COMPOSITE
The observed losses vs. ac field characteristics for sample A (twisted) under
zero dc field are shown as solid lines in Fig. 1. In order to account for these losses,
let us consider a model for a rectangular multifilamentary composite which is twisted
enough for individual filaments to behave independently. This model assumes an
infinite slab of normal metal containing superconducting filaments distributed
uniformly across the width a' as shown in Fig. 2(b). The filling factor of the superconductors in the slab should be equal to that in the real composite.
The magnetic field distribution in the normal metal is assumed to be the same
as that without any superconductors. This assumption is considered valid for a
composite with a twist pitch sufficiently shorter than a critical length corresponding
to the time rate of change of the applied field [1].
g
10- 3
~
a:
rr:a~~
vf
en
o..J
U
<X
0:: 0
0 00 •
Applied
field
i
(0 ) Composlt.
000·
00 0 0
0"0°
0000
00 0 0
00°0
:::=
Normal metal
SupercondUCflno
filement
0"0.
0 00 0
0°0.
(b) Model
50 100
200
AC FI E LD, He,O.
500
1000
2000
Fig. 1. The ac loss characteristics of sample A
(twisted), Hd = O. Solid lines, observed; Broken
lines, calculated from model.
L,
Fig. 2. Schematic description of model for twisted
composite.
T. Sato", M. T...u, ... T. ()p8Ia
156
When an ac field Ho cos(21t/'t") is applied parallel to both surfaces of the slab,
the amplitude of magnetic field can be calculated from the diffusion equation as
follows:
IH I = Ho{coSh[(a - 2x)/t5] + cos[(a - 2X)/t5]} 1/2
cosh(a/t5)
z
+ cos(a/t5)
(1)
for 0 :$; x :$; a/2. Therefore, the eddy current losses· in the normal metal per unit
area of the slab surfaces are found to be
(2)
where
k (L\ ~)
e'
= _1 sinh(2~/M - sin(2~/L\)
L\~ cosh(2/L\) + cos(2/L\)
(3)
The first term of (2) indicates the entire loss in the slab when the superconducting
filaments are assumed to be replaced by the same metal as the matrix; the second
term gives the eddy current loss in the replaced sites.
The hysteresis loss per unit surface area of an individual filament is calculated
from the Bean model [2]
(4)
where IHzl is given by (1). Then the hysteresis loss per unit area in both surfaces of
the slab is given by
(5)
where
(6)
the critical current density J e is about 2.0 x 10 5 A/cm 2 for sample A and it is assumed
that the flux penetrating into the individual filaments does not reach the center of
these filaments.
The total losses are the sum of (2) and (5), and are shown as broken lines in Fig. 1.
The observed ac losses are in relatively good agreement with those predicted, except
for those at frequencies above 200 Hz.
The disagreements may be explained as follows. The Rutherford group [1]
showed that all filaments are coupled for fI > fIe' where fIe is a critical time rate of
change of field corresponding to a twist pitch. This relation between fIe and a twist
pitch derived on the basis of constant II is not exactly available for the case where
an applied magnetic field is sinusoidal. However, assuming a constant fI of 4fH o ,
the filaments are not fully decoupled but partially coupled at frequencies above
200 Hz since the time rate fI is comparable to the critical time rate fIe of sample A.
According to the model for twisted composites, the ratio of hysteresis loss to
eddy current loss for sample A is found to be 0.07 at I = 500 Hz, Ho = 100 Oe, and
2.6 atl = 5 Hz, Ho = 100 Oe. This shows that the hysteresis and eddy current losses
are dominant in the lower- and higher-frequency ranges, respectively.
* The MKS rational units are used for the theoretical treatment in this paper.
157
AC Losses in Multifilamentary SupercooductiDg Composites
LOSSES IN UNTWISTED COMPOSITE
Consider now a model for an untwisted composite in which the filaments are
fully coupled. this model assumes an infinite slab of normal metal for region (1)
of the composite in Fig. 3(a), and an infinite slab consisting of three normal layers
and two superconducting layers for region (2) as shown in Fig. 3(b). The thickness
of each superconducting layer d z is determined in such a way that the product of
2d 2 and b' is equal to the total area of all filaments.
When an ac field is applied parallel to both surfaces of the untwisted composite,
the losses per unit surface area are calculated from the following relationships:
P = (1 - b'/b}Pel
P"2 =
+ (b'/b}Pe2 + (b'/b)Ph
(7)
pH 02k e(/l, l)/a
(8)
and
Pe2 = 1(pH//2{)(d t/{)3 = 1n 2(fJ.oHo)2d 1 3f2/p
() » d 1
(9a)
Pe2 = pH//2{) = (n/4fJ.03)l /2(fJ.oHo)2(pf)l /2,
() « d 1
(9b)
Ph = (2/3fJ.o 2 Jc)(fJ.oH o)3f,
() »d 1
(lOa)
Ph = (2/3fJ.o 2 JJ(fJ. oH o)3f exp( - 3dt/{),
() « d 1
(lOb)
and
Equation (9b) is easily given by a theoretical treatment similar to that for the twisted
composite. The derivation of (9a), where the magnetic field distribution in the normal
surface layer is affected by the superconductor, is described in the appendix. It is
assumed in the above treatment that the applied ac field is too small to penetrate
into the central normal layer shown in Fig. 3(b), namely, that Ho ::; J c d2 • This
assumption was always valid for the experimental conditions of the study.
Figure 4 shows the measured and calculated losses vs. ac field characteristics
for sample B (untwisted) under zero dc field. For f ::; 20 Hz, the theoretical values
on the assumption of () » d 1 agree well with the observed values. On the other hand,
the theory in the case of () « d 1 seems to be applicable for f ~ 100 Hz. These results
mean that equations for () » d 1 and f> « d 1 can be used in the ranges of dt/{) < 0.3
and dt/f> > 0.5, respectively.
The ratio of hysteresis loss to eddy current loss for sample B is estimated to be
10- 4 at 500 Hz and 10- 2 at 5 Hz as calculated from relations (7) through (lOb).
The hysteresis losses are negligibly small.
Normal ",efol
SuP.' eonduclor
(G I Composife
(.,) Mod. , of
I
Fig. 3. Schematic description of model for
untwisted composite.
L.
r.~pon
2
158
~
T.
Saw", M. Tuaka, aad T. Ogam.
o·
~
cL
vi
'"0
..J
U
<l
10- 4
10' ~'L----'---'---'--::--c:':-:"--'---,--:'
10
20
50
100 200
AC FIE LD.Ho. Oe
500
1000
2000
Fig. 4. The ac loss characteristics of sample B
(untwisted), Hd = O. ( - ) Observed; (---)
calculated from model for (j» d 1 ; ( - - - )
calculated for (j « d 1 •
At lower frequencies, the losses in the untwisted composite are less than those
in the twisted, as shown in Figs. 1 and 4. This occurs because the hysteresis loss
calculated from (lOa) is much less than that from (5), and because the eddy current
loss from (9a) is below 10 %of that from (2). The latter may be explained by considering
that the magnetic field in the normal layer near the boundary with the superconducting
layer is increased by addition of the field induced by diamagnetic magnetization in
the superconductor and that the eddy current loss in normal metal is decreased.
also reported that the losses produced during pulsed operation of a
Dahl et al.
magnet wound with twisted, multifilamentary superconducting wire of 0.5 mm
diameter were seven times as large as those in a magnet wound with untwisted wire
when peak fields were zero to 4 kG.
eJ
Effect of DC Magnetic Fields
Partial penetration hysteresis losses in type II superconductors increase with
dc magnetic fields since the area of the minor loop in the magnetization curve is
greater at a higher field. Eddy current losses in normal metal are varied through the
change in resistivity with dc fields. Since the eddy current loss is dominant for sample
B, the dc field dependence of the losses is obtained from (8), (9a), and (9b) as follows:
P/Po = (p/PO)1/2
at higher frequencies
(1Ia)
P/Po = Po/p
at lower frequencies
(lIb)
The change in P of the copper matrix is due to the magnetoresistance effect and is
given by the following empirical formula derived by the authors according to Kohler's
rule:
(12)
where P6 is the resistivity at the Debye temperature (lJ = 333 K, P6 = 1.99
ohm-cm for copper).
X
10- 6
AC
~
in M~tary Supercoaducting Composites
159
0.6
0.4
rE
ii>
d.
Fig. 5. The change in ac losses in sample B (untwisted) with dc magnetic fields. Solid lines:
theoretical; circled points, crosses: observed.
0.2
0
ioE==--+-----t-
Hd,"Co
-0.2
-0.4
-0.6
The changes in ac losses in sample B with dc fields below 10 kOe are shown in
Fig. 5. The measured values agree approximately with the values calculated from
(lla) and (llb). The changes for the most part are within 30 to 40%, and the losses
decrease as dc fields increase in the lower-frequency range.
CONCLUSION
Two models have been proposed to explain the ac losses in multifilamentary
superconducting composites. The superconducting filaments are assumed to behave
independently in the model for twisted composites and to act as one superconductor
in the model for untwisted composites. The losses in the twisted and the untwisted
composites are reasonably well explained by the two models. In the higher-frequency
range (200 to 500 Hz), a major part of the losses in both composites are eddy current
losses in the surface copper layers. The ac losses in twisted composites under low
ac fields have been found to be greater than those in untwisted composites at lower
frequencies. It should be noted that the eddy current losses in surface copper layers
calculated by using the model for the untwisted composites are much less than those
given by the model for the twisted composites when the skin depth is much greater
than the thickness of the surface layer. The changes in losses for the untwisted composite under a dc magnetic field of 10 kOe are 30 to 40 % at most. The losses at lower
frequencies are decreased as dc fields increase.
NOTATION
a
= thickness of composite
= thickness of part that filaments occupy
b
= width of composite
b'
= width of region (2), Fig. 3(a)
d
= filament diameter
d.
= thickness of each surface normal metal layer of model for untwisted composite, (a - a')/2
d2
= thickness of each superconducting layer of model for untwisted composite, nd 2 n/8b'
f
= frequency of ac field
H0
= amplitude of applied ac magnetic field
H.
= z component of magnetic field at x
H
= time rate of change of field
He
= critical time rate of change of field corresponding to a twist pitch
Je
= critical current density of superconductor
k.(l1, e) = parameter of eddy current loss defined by equation (3)
k.(l1, e) = parameter of hysteresis loss defined by (6)
n
= number of filaments
P
= total losses per unit area
p.
= eddy current loss per unit area in normal metal
= eddy current loss per unit area in region (I), Fig. 3(a)
p. 2
= eddy current loss per unit area in region (2), Fig. 3(a)
a'
p.,
160
T. Satow, M. TIlllUa, and T. Ogama
Pf~
P~
= hysteresis loss per unit surface area of superconducting filament
= hysteresis loss per unit area
x
= distance from surface of normal metal
Po
= total losses per unit area without dc magnetic fields
Greek letters
(j
= skin depth of matrix, {p/n/Jof)1/2
~
1'/
/Jo
~
p
Po
Pe
r
= dimensionless skin depth, 2/j/a
= filling factor of superconducting filaments, nd 2 n/4ab
= magnetic permeability of vacuum, 4n x 10- 7 (H/m)
dimensionless parameter, a'/a
resistivity of matrix
= resistivity without dc magnetic fields
= resistivity at Debye temperature
= time
=
=
APPENDIX
The eddy current loss in the normal layer of region (2) (see Fig. 3a) of the untwisted
composite for b » d 1 is calculated as follows.
The magnetic and electric fields in region (2) should be generally represented
in the form of a sum of harmonics. The magnetic field distribution in the normal
layer including integration constants is calculated from the diffusion equation of
the field. Using a boundary condition Hz = Hocoswratx = Oand an approximation
to the first power of x/b, the magnetic field is given by
Hz<x,r)
=
Ho[(l -
~C1) coswr + ~(C1' -~) sinwrJ + m~2 function(mwr)
(AI)
where C 1 and C l' are real integration constants. The electric field is then calculated
from (AI) and one of Maxwell's equations by the relation
pH 0 [ C 1 cos wr Ei x , r) = -b-
(
C 1,
2X).sm wr
7)
-
J+
~
m~2
.
functlon(mwr)
(A2)
Assuming the Bean model as the magnetic field distribution in the superconducting layer, the electric field at the boundary plane between the matrix and the superconductor is given by
E (d
y
)
1,r
=
Bp
± B1(r) oB1(r)
2 J
J1.o
c
(A3)
~
ur
where B1(r) is the ac flux density at the boundary plane and Bp is the amplitude of
B1(r). The plus and minus signs of (A3) correspond to the plus and minus values of
oBdor. Naturally B 1(r) should be represented in terms ofthe sum of the harmonics.
The tangential components of both magnetic and electric fields are continuous
atthe boundary plane, so that C 1 = 0 and C l' = 2ddb and the terms ofthe harmonics
for m ~ 2 vanish, to a first approximation in ddb. Here, it is assumed that Ho « 2Jcd 1 •
Finally, the magnetic and electric fields in the normal layer are given by
Hz(x, r) = Ho [ cos(wr)
+ (2d1b2-
J
x)x.
sm(wr)
(A4)
AC Losses in Multifilamentary Superconducting Composites
161
and
E,(x, -r) = -
2pHo(d1 - x) .
()2
sm(an)
(AS)
and the eddy current loss is represented as (9a).
REFERENCES
J. Superconducting Applications Group, Rutherford Laboratory, J. Phys. D: Appl. Phys., 3(11): 1517
(1970).
2. C. P. Bean, Phys. Rev. Letters, 8(6):250 (1962).
3. P. F. Dahl, G. H. Morgan, and W. B. Sampson, J. Appl. Phys., 40(5):2083 (1969).
DISCUSSION
Comment by M. S. Walker, Westinghouse Research Laboratories: Similar results have been obtained
in measurements over nearly the same frequency and alternating field amplitude range in our laboratory.
The results were well explained using an anisotropic continuum model for the eddy current and hysteresis
loss in the composite which has been developed by W. 1. Carr, Jr. at Westinghouse Research and is to be
published soon in the Journal of Applied Physics (presently available in the form of Westinghouse Papers
73-9J2-MACON-PI and 73-9J2-MACON-P2.)
One must be careful not to draw "across-the-board" conclusions regarding the choice of single-core
or twisted multifilament superconductors, since the magnitude of the alternating field and its frequency
and the magnitude of any bias field seriously affect not only the quantitative but the qualitative form of
the losses. Large bias fields, for example, can cause increased or decreased hysteresis losses since they affect
the critical size of the conductor or filament. The amplitude of the alternating field also determines this
critical size. The bias field also causes a magnetoresistive change in high-conducting matrices which can
cause decreased or increased ac eddy current losses, depending upon whether the skin depth at the frequency in question is large or small compared to the overall composite size. Losses are decreased and
increased with shortening twist pitch, depending upon which of these two skin-depth limits are controlling. Finally, a low loss may be achieved under some ac conditions, but the conductors sometimes
must be capable of being energized to high-field operating levels without instability and designs for stable
energization may not be the lowest-loss designs.
Comment by author: We found experimentally and theoretically that the losses in the untwisted
composite were lower than those in the twisted composites at frequencies of about 5 Hz and at amplitudes
about several hundred Oe. Even if some shielding techniques are adopted in levitation coils, it is going to
be more difficult to shield the ac field with lower frequencies. So, we must be careful of low-frequency losses.
D-4
SHAPED FIELD SUPERCONDUCTIVE
DC SIDP DRIVE SYSTEMS
T. J. Doyle
Naval Ship Research and Development Center
Annapolis, Maryland
INTRODUCTION
A ship drive system provides the power transmission path between the prime
mover, which may be a gas or steam turbine, diesel engine, or other source of rotary
power, and a propeller, fan, jet pump, or other propulsor generally operating at
much lower speed. When multiple prime movers or propulsors are used, the drive
system should also permit power-combining and cross-connection.
If the full potential of advanced hull forms now under consideration is to be
realized, several additional drive system features must be available [1]. Advanced
gas turbine-powered catamarans, hydrofoil craft, and surface effect ships are characterized by large turbine-propeller separations, constrictive machinery spaces, and
demanding takeoff loadings. In addition to speed reduction and power distribution
functions, therefore, drive machinery must be compact, easily located, reversible,
and capable of variable-torque ratios.
Many of the features required in advanced ship drives are inherent in electric
transmissions which include motor-driven propellers powered by turbogenerators.
The alignment-free electric linkage provides great machinery arrangement flexibility
when compared to hard-coupled mechanical drives. Control and maneuverability
advantages are also evident in dc transmissions which permit variable reduction
ratios and electric reversals. These location and control benefits have seldom been
available in propulsion applications because of the prohibitive size and weight of
conventional motors and generators.
The advent of superconductivity, however, promises to bring these electric
coupling benefits to many naval applications. Superconductors will support very
high currents without resistance losses when their temperature is reduced below a
critical value. With superconducting field windings, it is possible to produce very
intense magnetic fields with small quantities of material and negligible electric power
loss. It should now be possible to produce power with generators which are smaller
than the prime movers driving them and to transmit it to motors which are smaller
than the thrusters with which they are used.
The availability of practical superconducting materials in the last decade,
coupled with the need for improved drive system characteristics, therefore provided
impetus for development of superconductive machinery tailored to shipboard
S ]. One product of the many investigations at NSRDC was the
applications
evolution of a new dc superconducting machine concept [6]. Termed the "shaped
e-
162
163
Shaped Field Superconductive DC Ship Drive Systems
field machine" because of the unconventional manner in which the magnetic field
geometry is established, this unique approach provides small-diameter, shockresistant, magnetically shielded designs with high efficiencies and modest refrigeration
req uiremen ts.
A 400- to lOOO-Hp motor has been designed and constructed to verify the predicted size and performance advantages of the shaped field configuration. Presently
the laboratory machine is undergoing no-load evaluation. Load testing up to the
4OO-Hp level will be conducted with a rectifier power supply and a water-brake load.
Subsequently, system performance will be determined when the rectifier is replaced
by a laboratory-built, superconductive generator (of different design) powered by
a lOOO-Hp gas turbine. Successful motor-generator operation in the laboratory will
be followed by shipboard evaluation in a small test vehicle.
This presentation describes the design and construction of the shaped field
laboratory motor. Loss mechanisms are treated and machine performance estimated.
The advantages that a superconductive dc drive can provide future high-performance
ships are illustrated in a hypotheticaI80,OOO-Hp, small water area, twin hull (SWATH)
ship and 40,OOO-Hp hydrofoil (HF) craft.
SHAPED FIELD CONCEPT
The size and efficiency advantages of the shaped field acyclic machine accrue
directly from the unique magnet/rotor arrangement, illustrated conceptually in Fig. 1.
The superconducting solenoidal magnet and helium vessel, or dewar, are the innermost machine elements. The intense magnetic flux generated in the bore of the solenoid
is attracted by the ferromagnetic shield, forcing virtually all of the magnetic flux to
radially transverse the rotor twice. When current is passed through brushes and
axially down the copper rotor drums, the resulting Lorentz interaction provides
motor action. Conventional arrangements of superconducting dc machines include
the same magnet/rotor/shield elements, but the rotor is located in the magnet bore
and a lesser portion of the generated flux is utilized in the power production process.
Similarly, high-flux utilizations are possible with internal magnet disk rotor
machines which incorporate current collectors at inner and outer disk perimeters.
The shield is sliced to provide narrow passages for the disk conductors which rotate
when the radial current couples with the axial flux. The drum configuration is
preferred, however, since it permits the reduced stray fields possible with a nonsegmented shield and is more easily adapted to high-conductivity liquid metal current
collectors. The improved flux utilization of an internal magnet design, coupled with
the high current levels in compact liquid-metal brushes, accounts for the attractive
power density characteristics of the shaped field arrangement.
r~~~!!!~;AGHEnC
.~.
SHI(LD
TUMIKALS
\JI\'"T C,,!.llCT"R
MAIJHLU,lMTH
.:;, _ _t-""'U NOIDAl M ..GHlT
Fig. I. Shaped field machine concept.
164
T. J. Doyle
The capacity and maintenance benefits possible with liquid metal collectors
point up a second desirable feature of the shaped field arrangement, the ability to
locate brush sites in low-magnetic-field regions. In conventional arrangements, the
collector fluids, located in the magnet bore where the field is most intense, will experience severe field-induced torques with correspondingly large viscous drag losses.
This major source of inefficiency can be virtually eliminated in the shaped field
machine and an improved efficiency characteristic results.
The opportunity to use a small, centrally located magnet--<iewar assembly
provides additional benefits. The cylindrical dewar, without a warm hole as in
previous dc designs, can be built with a very low heat leak, reducing refrigeration
requirements and further improving efficiency. It is also protected by an almost
invulnerable set of steel and copper shells and inherently rugged machine construction
results.
LABORATORY MACHINE DESIGN
Major design features of the shaped field laboratory machine are shown in Figs. 2
aQd 3 in cross-sectional and isometric views. The magnet system (supplied by Intermagnetics General Corporation) includes a solenoidal superconducting magnet
enclosed in the double-walled, vacuum-insulated, helium dewar supported by the
rotor at one end through an idler bearing. At the opposite end, the dewar neck, which
contains the helium inlet and exhaust lines and power leads, is rigidly attached to
one bearing housing. The winding includes 13,340 turns ofO.030-in. OD, I80-filament,
niobium-titanium superconductor with a 1.8: 1 copper to superconductor ratio,
providing magnetomotive forces approaching 2 x 106 A-turns.
Two inline sets of four concentric drum conductors are located symmetrically
about the magnet center plane and epoxy-bonded to the outer surface of the stainless
steel rotor shaft. The eight copper rotor drums are insulated from each other and
from the shaft, permitting connection in electrical series through liquid-metal brushes
with a similar arrangement of copper current return drums in the stator. Stator
drums and terminal rods are bonded to the ferromagnetic shield.
Fig. 2. Motor cross section with flux plot.
Shaped Field Superconductive DC Ship Drive Systems
165
Fig. 3. Shaped field laboratory motor isometric. (I) Input or output leads; (2) stator drums; (3) rotor drums;
(4) magnet and dewar; (5) collector rings; (6) brush disks; (7) input or output shaft; (8) iron shield;
(9) shaft, rotor section; (10) epoxy; (11) bearings and seals.
Flux generated in the solenoid by the circumferential field current is attracted
to the steel shielding. Entering the iron radially, flux proceeds axially through the
iron to the other end of the machine, returning radially to the magnet bore to complete the path. Although fields of 55 to 70 kG exist in the magnet, the superimposed
flux map in Fig. 2 indicates extremely effective flux containment, with stray fields
outside the shield limited to less than 100 G. Field lines are concentrated in the active
drum regions and widely spaced at the collector sites, illustrating the flux utilization
and loss suppression benefits of the shaped field arrangement.
When voltage is applied at the terminal rods, the resultant axial current flow in
each rotor drum interacts with the radial field, producing a circumferential torque.
The cumulative torque developed in each drum is coupled to the output shafting
through the drum-to-drum and drum-to-shaft insulated bonds. Reaction torque
produced in the stator drums is transmitted through the shielding to the machine
mounting points.
The brush work includes sixteen copper rotor disks, two on each drum, each
rotating in stator channels with the disk/channel gap bridged by liquid sodiumpotassium eutectic (NaK). The collector geometry can be seen in Fig. 4, which illustrates the machine assembly. The stator channels are grooves cut into copper stator
rings, two annular rings electrically bonded to each stator drum. An unflooded
operating mode is employed with no external liquid-metal circulation. Small quantities
of liquid metal, sufficient only to assure uniform contact around the disk/channel
annulus, are injected into each channel with centrifugal and magnetohydrodynamic
forces providing for liquid-metal distribution. The fluid pools at the bottom of each
channel during periods of nonoperation. A dry, oxygen-free cover gas maintains
long-term liquid-metal purity.
A nonconducting coolant fluid is circulated through parallel tubes in the stator
rings to remove resistance- and viscous-drag-generated heat. Most of the heat
originating in the rotor is transferred by conduction to the stator rings across the
166
T. J. Doyle
Fig. 4. Rotor placement in
bottom stator half.
liquid-metal collectors, with a lesser portion radiated and convected to the stator
drum assembly and the fluid-cooled dewar surface.
The completed machine weighs 1050 kg (2300 Ib) and displaces 0.17 m 3 (6 ft 3 ),
with a (maximum) 51-cm (20-in.) diameter at the machine center plane. The shield
is 76 cm (30 in.) long and the bearing housings add an additional 10 cm (4 in.) at
each end of the machine. Output shafting is 10 cm (4 in.) in diameter.
MACHINE LOSSES
Loss mechanisms in the laboratory acyclic motor can be conveniently divided
into two categories: Ohmic losses in the armature circuit and drag losses from collector, windage, and bearing effects reducing output torque.
Ohmic losses are proportional to circuit resistance, estimated at 30 j1!l in the
laboratory motor. This extremely low resistance results from the use of large copper
cross sections (30 to 35 cm 2 ) and liquid-metal collectors, whose contact voltage drops
can be rendered unmeasurable by proper surface preparation.
The preponderance of drag losses is attributable to the liquid-metal collectors,
where three separate and distinct loss mechanisms can be identified. The normal
viscous shear loss becomes a cubic function of rotor speed since the relatively narrow
collector gap assures turbulent flow; these will not exceed a few kilowatts in the 1800
to 2400 rpm range appropriate to motor operation.
The two remaining losses result from the presence of axial and radial magnetic
fields in the collector regions. The axial field component interacts with the radial
transport current with a consequent tangential magnetohydrodynamic body force
experienced in the collector fluid. The resulting acceleration of the liquid metal
provides an additional fluid shear loss proportional to the square of the local axial
field. This axial field in the motor configuration, however, is of the order of 1000 G
and the predicted losses of a few watts per site can be safely neglected. In contrast,
similar collectors operating in the 40- to 60-kG fields present in a conventionally
arranged superconductive machine could generate Lorentz losses of 200 to 400 kW.
167
Shaped Field Superconductive DC Ship Drive Systems
The final collector loss mechanism considered results from the radial magnetic
field in the collector region. The radial field component generates an axial voltage
differential across the brush tip in the same manner as in the drum conductors. This
collector disk voltage, however, is shorted to the stator ring through the liquid metal,
and an entropy-generating current loop is established across the collector gap. In
the laboratory motor, the radial fields in the collectors are also quite low, of the
order of 3 kG. With proper disk insulation, these circulating current losses are limited
to a few kilowatts.
OPERATING POINT PERFORMANCE
Two operating points have been selected for motor load testing, one dictated
by rectifier current limits and the other consistent with turbogenerator capacity
and propeller speed requirements in the test bed vehicle system. Estimates of machine
performance at these two points are summarized in Table I, with significant electric
Table I. Shaped Field Motor Operating Point Performance Estimates
Powered
Machine performance
Input
Output
Power density
Efficiency
Magnetic circuit data
Magnet
Rotor
Shield
Electric circuit data
Current density
Circuit
Machine losses
Electrical circulation
collector
Windage
Bearings and seals
Total
Terminal voltage, V
Load current, A
Power, kW
Rotor speed, rpm
Torque, N-m
Power, Hp
Hp/m 3
Hpjkg
%
Rectifier
Turbogenerator
31.74
10,000
317.4
2,400
1,215
410.2
2,400
0.39
96.4
30.11
25,000
752.6
1,800
3,865
976.5
5,750
0.93
96.8
Field current, A
Generated flux, We
Effective flux, We
Flux utilization, %
Maximum field, kG
Stray field at 6 in., G
121
0.139
0.098
70
14.4
15
Rotor drum, A/em 2
Stator drum, A/em 2
Terminal drum, A/cm 2
NaK brush, A/em 2
Resistance, 10- 6 n
Back emf, V
320
280
275
515
30.3
31.35
805
695
685
1,290
30.3
29.25
3.03
4.29
0.Q2
3.27
0.Q7
0.70
11.38
18.94
1.81
0.21
2.85
0.04
0.52
24.37
Ohmic,kW
Viscous, kW
MHD,kW
Circulation current, kW
kW
kW
kW
150
0.172
0.120
70
18
19
168
T.J.Doyle
98
~ 97
....
..
,:
<.>
z
2
a
.....
'"
I:i
~
z
~
a
96
9 5 0 ' - - - - - ' - - - . L . - - - - - L - -•....Iooo---L...--•....Jsoo
OUTPUT POWER, HP
Fig. 5. Estimated motor performance (l40-A
field current).
and magnetic circuit parameters included. Losses, discussed above, appear quite
manageable, permitting efficient operation at high-power density. The predicted
motor performance over a wider range of power, speed, and efficiency conditions is
included in Fig. 5.
ADVANCED HULL APPLICATIONS
Two advanced vehicle types were conceptually fitted with superconductive
dc machinery to assess the size and performance potential of full-scale drive systems.
The first vehicle considered was a 4OOO-ton SWATH ship equipped with 4O,OOO-Hp
contrarotating screws in each hull driven by twin 5000-Hp cruise turbines and four
20,OOO-Hp boost turbines for high-speed operation. The SWATH can be described
as a football field, strut-supported on two submarine hulls. Wave action along the
relatively thin struts provides minimal pitching and rolling moments and consequent
high platform stability. The difficulty of efficiently coupling the high-speed turbine
prime movers to remote, low-speed contrarotating propellers is apparent.
One possible solution is illustrated in Fig. 6. Motors and generators, sized at
the same current and flux densities and tip speeds as the laboratory machines, are
quite compact, permitting close coupling. Each 1.7-m 00 motor includes concentric
contrarotating rotors designed for 24 x 106 N-m of torque. Motor and generator
unit efficiencies, including refrigeration, exceed 98 % and low-loss, highly maneuverable systems with minimal machinery space requirements are predicted. The total
drive system weight, including superconductive machinery, coaxial transmission
lines, helium refrigerators, controls, and auxiliaries, is estimated at 150,000 kg.
Hydrofoil craft with submerged, strut-supported propellers and dynamic lift
foils provide platform stability similar to that of the SWATH and high-speed capability. Requirements for pivotable struts make an electric linkage an attractive
alternative to geared drives, with the additional variable reduction ratio capability
of dc machinery well suited to the demanding takeoff torque schedule. One machinery
arrangement for a 750-ton, 4O,OOO-Hp hydrofoil is illustrated in Fig. 7. The compactness and modest helium requirements of the shaped field design permit location of
Fig. 6. 80.000-Hp SWATH ship superconductive
drive arrangement.
DeIIUTOt
Fig. 7. 40,OOO-Hp hydrofoil craft supercondu ctive
drive arrangement.
170
T. J. Doyle
the motor and its cryosections in small-diameter, hydraulically efficient pods. The
1.0-m-diameter motor illustrated is designed for 20,000 Hp at 1200 rpm and an
efficiency of 98.5 %, including refrigeration. The drive train with all auxiliaries is
estimated to weigh 42,000 kg.
REFERENCES
I. W. J. Levedahl, in: Proceedings Applied Superconductivity Conference, IEEE Publ. No. 72 CH0682-5TABSC (1972), p. 26.
2. A. D. Appleton, in: Proceedings Applied Superconductivity Conference, IEEE Publ. No. CH0682-5TABSC (1972), p. 16.
3. G. R. Fox and B. D. Hatch, in: Proceedings Applied Superconductivity Conference, IEEE Publ. No.
CH0682-5-TABSC (1972), p. 33.
4. E. F. McCann and C. J. Mole, Naval Eng. J., 84(6):35 (1972).
5. W. J. Levedahl and T. J. Doyle, "Superconductive Machinery for Naval Propulsion Systems," paper
presented at the 1972 IECEC, San Diego, California (September 5, 1972).
6. T. J. Doyle, US Patent No. 3,657,580 (April 18, 1972).
E-1
ALTERNATING CURRENT LOSSES IN
SUPER CONDUCTING CONDUCTORS FOR
LOW-FIELD APPLICATIONS
M. A. Janocko, D.
w. Deis, and W. J. Carr, Jr.
Westinghouse Research Laboratories
Pittsburgh, Pennsylvania
INTRODUCTION
The loss characteristics of type II superconducting conductors carrying an ac
current in the absence of an applied field, either ac or dc, is one of the principal
factors which will determine the feasibility of ac power transmission lines constructed
with these materials. These self-field conductor losses depend not only upon the
intrinsic superconductor characteristics, such as the shape of the critical current
density-critical field curve [1], but also upon the conductor geometry, and the
amount, type, and distribution of the nonsuperconducting matrix material. In
order to obtain a better understanding of these losses and their mechanisms, an
experimental investigation of ac losses has been undertaken in a number of types of
commercial state-of-the-art superconducting conductors. Conductors investigated
include Nb 3 Sn tapes, Nb-Ti multifilament and single-core conductors in round
and rectangular configurations, and a Nb multifilament conductor.
All measurements were performed at 4.2 K and 60 Hz using a helium boiloff
calorimeter capable of measurements between 0.010 and 50 W with sample currents
up to 1500 A (rms).
EXPERIMENTAL TECHNIQUE
Proposed operating current densities for transmission lines employing type II
materials are typically about two orders of magnitude below the critical current
density J c ' so that an acceptable loss level and overload capability are obtained [2].
To perform self-field loss measurements on samples in this operating region requires
relatively long samples (about 10 to 100 m) and a calorimeter sensitivity of approximately 10 mW. To test these same samples under conditions corresponding to
overload conditions requires the calorimeter to be capable of handling power
inputs of approximately 10 W or more.
These requirements have necessitated the construction of a helium boiloff
calorimeter with a large working volume as shown in Fig. 1. This calorimeter is
constructed almost entirely from commercial ultrahigh-vacuum copper-gasketed
flange components. Flanges and feedthroughs of this type are highly reliable upon
thermal cycling and allow for a modular design which can easily be assembled,
disassembled, or modified as necessary. The details of the construction and operation
of this probe are reported elsewhere [3]. The essential part is the large (20 cm ID,
171
172
M. A. Janocko, D. W. Deis, and W. J. Carr, Jr.
Fig. 1. Alternating current loss apparatus for calorimetrically measuring long lengths of conductors in a
bifilar coil arrangement.
35 cm long) vessel at the bottom of the probe. The wire sample on a Micarta mandrel
is mounted inside this vessel. The vessel and surrounding dewar are filled with
liquid helium. Current is fed into and out of the dewar via several vapor-cooled
leads in parallel, and into and out of the vessel via massive OFHC copper feedthroughs. In operation, the boiloff from the dewar is throttled to balance the dewar
and vessel pressures (and thus their temperatures). This control and the vessel
vacuum jacket minimize interbath heat flow. The heat transfer effect of a pressure
imbalance was determined to be 5.6 mW/Torr of pressure differential. The balance
condition was always within ±O.l Torr during measurements. The helium vapor
evolving from the vessel is ducted up a pipe inside a support tube, the vacuum jacket
between them serving to minimize the effects of changing helium level in the dewar
when low flow rates are being measured. The evolved gas is first passed through a large
heat exchanger and then is metered by variable-area float-type flowmeters calibrated
to 2 %accuracy for helium gas. The helium flow rate is translated into heat input by
correcting for the temperature-dependent heat of vaporization and for the temperature-dependent gas/liquid density ratio (to allow for the gas remaining to fill the
space left by the evaporated liquid). The heat input thus calculated was found to
agree with the known heat input of a calibrating heater to about 2 % (the accuracy
obtainable in reading the flowmeters) over the whole range of operation from 50 W
down to about 10 mW, at which point the residual heat leak into the vessel, about
25 mW,limits the observable heat input.
The sample mandrels are Micarta cylinders 25.4 cm long by 7.8 cm 00, into
which a double helical wire groove is machined. The wire center-to-center distance,
Alternating Current Losses in Superconducting Conductors for Low-Field Applications
173
between wires carrying current in opposite directions, was kept at five times the wire
diameter, resulting in field changes due to adjacent wires of less than 2 % of the selffields. For noncircular conductors, the maximum dimension face was wound flat
onto the Micarta mandrel and this dimension was used to determine the conductor
spacing. The mandrel with current terminal blocks is shown schematically in the
lower part of Fig. 1. Data points were taken for each sample over a range of currents
from that which produced heat inputs of about 10 mW up to the level at which
either the sample went normal or the flowmeter capacity was exceeded.
During operation, the joint resistances between the superconducting conductor
and the copper feedthroughs were always checked by measuring the voltage drop
across the joint for a 100- to 200-A dc current to ensure that any loss occurring at
these points was negligible as compared to the conductor loss.
RESULTS AND DISCUSSION
Nb 3 Sn
Materials such as Nb 3 Sn are of interest principally because of their relatively
high transition temperature, possibly allowing operation at temperatures above
4.2 K. Their main disadvantage is their brittleness, making handling and device
fabrication difficult in some instances. Bulk self-field ac losses in strong type II
materials such as Nb 3 Sn are principally due to magnetic hysteresis. Such h-ysteresis
losses are predicted to vary as Hn, where H is the surface magnetic field due to the
transport current and 3 < n < 4, depending upon whether J c equals a constant
(Bean-London model) or J c x H equals a constant (Kim-Anderson model) is
used in the critical state model loss calculation [1,4,5]. Conductors containing significant fractions of low-resistivity normal metal will also have a loss contribution due to
normal eddy currents varying as H2 if the normal metal sees an alternating magnetic
field. For magnetic fields below H p , the field at which flux enters the bulk of the
specimen, surface hysteresis losses can occur with n predicted to be approximately
equal to two [6].
The sample information on the Nb 3 Sn conductors is listed in Table I. The
conductors are all commercial tape samples* produced by vapor deposition ofNb 3 Sn
on stainless steel substrates. The Nb 3 Sn forms a continuous sheath completely
enclosing the substrate. Copper stabilizing material is soldered onto each side of this
tape.
Table I. Nb 3 Sn Sample Data
Sample
No.
H-49
H-51
H-52
H-53
H-54
Overall
dimensions,
cm
0.127
0.127
0.229
0.229
0.229
x
x
x
x
x
Nb 3 Sn
circumference,
cm
0.0203
0.0297
0.0127
0.0165
0.0305
• Canada Superconductor and Cryogenics, Ltd.
0.267
0.269
0.467
0.470
0.472 .
Critical
current
(lOOkG),
A
Copper
thickness
(per side),
cm
40
120
90
140
240
0.005
0.0102
0.0025
0.005
0.0102
174
M. A. Janoeko, D. W. Deis, and W. J. Carr, Jr.
0----0 H- '/'!9
0----0
0.. - . 0
H·51
H-~2
.. - -0 H ·)J
1)0 ••••
101
-0
H- 5,d
104
1 S P'iiIImperes ~NI~J I c.m
Fig. 2. 60-Hz ae loss data for Nb 3 Sn-tape conductors. The units of loss P, are W/cm 2 of Nb 3 Sn
surface area and I,p is in units of peak ac A/cm of
Nb 3 Sn circumference. The conductor specifications are detailed in Table I.
The loss data for these conductors are plotted in Fig. 2. The ordinate, Ps ' is the
power dissipation per unit surface area, where the area is taken as the total Nb 3 Sn
surface area. The abscissa, Isp, is the peak current divided by the Nb 3 Sn circumference. The slopes one would obtain for Ps ex I=p for n = 2 and n = 4 are also shown
on the figure. In general, for Isp > 700 A/em, the measured losses correspond to
4 < n < 5 and for Isp < 700 A/em, n is approximately equal to 2. The value of n for
low currents may indicate the presence of a significant eddy current losses in the
copper stabilizing material, or a transition to a surface loss region, or a combination
of both.
Chant et al. C] and Garber et al. 8 ] have previously reported data on similar
types of Nb 3 Sn conductors. The data of Chant et al. obtained at 50 Hz, cover the
region 4 x 10 2 < Isp < 103 A/em and vary as n approximately equal to five, the
magnitude of the losses being close to those reported here in the region where Isp
has a value of about 1000 A/em, but being significantly less for lower currents due to
the continued dependence of n close to a value of five in this region. The data of
Garber et al., where 150 < Isp < 500 A/em, correspond to n between 2 and 2.7, in
general agreement with the data reported here, and have a magnitude four times
lower.
The data of Garber et al. were obtained from test specimens wound in a bifilar
pancake coil, with small separation between conductors carrying current in opposite
directions. Chant et al. used a straight configuration, again with only small separation
e·
Alternating Current Losses in Superconducting Conductors for Low-Field Applications
175
between adjacent flat faces of conductors carrying current in opposite directions.
The geometric differences between these configurations and the one used in this work
may be a cause of the discrepancies in both n and the magnitude of the loss.
Nb, NbTi
A variety of commercially available NbTi conductors with different values of
current capacity, copper-to-superconductor ratio, geometry, filament number and
size, and twist pitch have been tested. In addition, one Nb multifilament conductor was
also investigated. The relevant parameters which characterize these conductor
samples are given in Table II.
The data for the Nb-Ti and Nb wires are presented in Fig. 3 in a form useful for
practical design calculations, with Pv' the loss per unit volume of total conductor,
plotted against JI' the current density in the total conductor cross section. On this
basis, the small wires (long-dashed lines) have the lowest losses, with the niobium
wire being the lowest of these. The four large, rectangular, small-filament wires (solid
lines) exhibit changes in slope, going from an n = 2 dependence at low currents to
2.3 < n < 3 at high currents. It is interesting to compare H-13 and H-70, which are
almost geometrically identical, but which were supplied by different manufacturers,
and which thus may have slightly different Nb-Ti alloy compositions, number of
filaments, and metallurgical differences (cold working, heat treatment, etc.). The
same n = 2 line fits the points for both wires at the low-current end while the losses
are slightly different at the high-current end.
The remaining three curves are for the large, circular cross-section wires (shortdashed lines). H-101 and H-102 are physically identical, large-filament wires, except
for twist. Their data points are well represented by straight lines with nearly equal
values of n (close to two), but the losses in the twisted wire are more than twice as
-....-...... H-l
0---0
H-4
H-6
H-8
00--"""0
H - l~
0--.......0
0- -
-<)
_
_
0-. -
~
f
H-Il
c
1f
1/ I 17
t; ! {I
H-22
t----o H-l0
<>- --- 00
H- IOI
&- - - - -00
H-1D'Z
00- ---01]
H- 10)
Il
rI
H-li
II
II
't"
t
dII 1 0
l'
I
/I
Fig. 3. 60-Hz ac loss per unit volume of
conductor, p., as a function of the
overall conductor current density J,
for Nb and NbTi composite conductors. The conductor specifications are
detailed in Table II.
10-1
104
J 1,
._,e.
"",k l'cm2
"
t
11
I
of
v;
,
0.67/1
0.67/1
om 73 diam.
O.046diam.
0.3048 diam.
1.75/1
2/1
0.254diam.
1.8/1
0.099 x 0.140
2/1
1/1
0.0208 diam.
0.254diam.
2/1
1.35/1
0.183 x 0.183
0.051 diam.
1.8/1
1/1
0.020diam.
0.1 02 x 0.145
2/1
0.168 x 0.335
Copper/
superconductor
No.
0.780
0.228
4.6 x 10- 3 0.175diam.
2.8 x 10- 3 0.239diam.
2000
7225
Untwisted
U81
Untwisted
4.6 x 10- 3 0.175diam.
4 x 10- 3 0.088 x 0.112
6.5 x 10- 4 0.0194diam.
0.787
0.394
4 x 10- 3 0.0415 diam.
1 x 10- 2 0·155 x 0·155
U81
2 x 10- 3 0.086 x 0.115
26,000:1:
15,000:1:
15,000:1:
5,OOOt
177·
1O,000t
550t
5,OOOt
740t
3.6 x 10- 2 0.036diam.
176·
15,ooot
ISO·
3.937
0.394
Twist pitch,
T/cm
Critical
current, I,
(H = 0)
A
1.27 x 10- 2 0.0134diam.
7.5 x 10- 4 0.0194diam.
3 x 10- 3 0.142 x 0.305
Size,
em
Filament
bundle
dimensions,
cm
2000
300
400
121
61
1340
400
2133
• Measured directly.
t Extrapolated from higher field data.
:I: Calculated from manufacturer's specifications.
§ Calculated from superconductor area.
H-I
(NbTi)
H-4
(NbTi)
H-6
(NbTi)
H-8
(NbTi)
H-13
(NbTi)
H-14
(NbTi)
H-18
(NbTi)
H-22
(Nb)
H-70
(NbTi)
H-101
(NbTi)
H-102
(NbTi)
H-103
(NbTi)
Sample No. Dimensions,
(material)
cm
Filament
849
2395
670
9.5 x 105
6.4 x 105
9.0 x 105
851
492
503
438
1.012 x 106
8.88 x 105
8.88 x 105
9.79 x 105
5741
2606
7.5 x 105
1.039 x 106
7653
5563
377
31.97
25.47
24.93
11.78
1.956
22.4
4.85
12.5
4.28
1.79
1.80
21.2
1.752 x 10- 5
2.012 x 10- 5
1.968 x 10- 5
3.404 x 10- 5
2.296 x 10- 4
2.68 x 10- 5
9.58 x 10- 5
3.396 x 10- 5
1.042 x 10- 4
3.061 x 10- 4
2.225 x 10- 4
1.508 x 10- 5
Total
Length, I volume, F" = 2ft x 10- 9/3
cm
cm 3
W/A 2
Sample
1.07 x 106
1.08 x 106
8.0 x 105
Critical
current
density J,,§
A/cm 2
Table ll. Sample Information for Nb and NbTi Conductors
Alternating Current Losses in Superconducting Conductors for Low-Field Applications
177
large. The third circular cross-section wire, H-I03, has smaller filaments and a thinner
copper surface sheath, and exhibits the change of slope characteristic of the large
rectangular wires.
The loss data for the Nb and NbTi conductors are shown in Fig. 4 normalized
in the same manner as the Nb 3 Sn results in Fig. 2. The surface area used to normalize
the measured power Pm to obtain Ps is taken as either the surface area of the superconductor core in a monofilament wire or the surface area of the filament and copper
bundle interior to the surface sheath of copper for a multifilament wire. The circumference used to normalize I p and obtain I sp is the circumference of the filament or
of the bundle cross section. Thus, the filament bundle for multifilamentary conductors
10- 1
P
-..... - ,
-----0
-0--0
H-I
H-4
Cu rrent
H-6
0--0
H·i
H·1l
Exponent n
H- li
10-2
<>--.
0-
H· 18
H·22
H·l0
~ -- .. c
H· IOI
0---....0 H· I02
0- - - ...0 H·IOJ
------
10- 1
-0
0 - -0
--.
t--------O
,
~e
10- 2
~
H-I
H· 4
H·6
H- 8
H: IJ
H- 14
H-18
H· 12
H-l0
H· IOI
H-I02
H- I03
2. 29'
3. 12
3-36
1.01
3.04'
1.25
2.S2'
4.18
2.82'
2. 16
2. 18
2.41'
of'
10-)
102
101
1(1'
I.~p, ampeft5 ~1( ) /( m
Fig. 4. 60-Hz ac loss data for No and NbTi
composite conductors. The units of loss Ps are
W/cm 2 of superconductor bundle surface area.
For single-core specimens, this is the core surface
area; for multifilament conductors, it is the
surface area of the conductor taken at the
interior surface separating the copper surface
sheath from the rest of the material. Isp is in units
of peak ac A/cm of circumference, where the
circumference is taken to correspond to the
surface used to calculate P,. The conductor
specifications are detailed in Table II.
Fig. 5. Normalized ac loss data as a function of
normalized current for the conductors listed in
Table II. The current exponent n is the same for the
curves of each conductor, respectively, in Figs. 3,
4, and 5. For curves with slope changes (H-I,
H-I3, H-IS, H-70, H-I03), the value of n for the
high-current end is given, the low-current points
in each of these cases being fitted by a line with
n = 2.
178
M. A. Janocko, D. W. Deis, and W. J. Carr, Jr.
is regarded as a continuous superconducting surface. This IIlIJY be done if there is
complete coupling between filaments and if no transposition of filaments occurs [9. I0].
A comparison of the data of Fig. 4 with the data of Fig. 2 for Nb 3 Sn shows that
the five small-diameter samples, four NbTi (H-4, H-6, H-8, and H-14) samples,
and the one Nb sample (H-22) have losses lower than the minimum values obtained
for the Nb 3 Sn conductors.
In Fig. 5, the loss data are presented in terms of a normalized specific loss P"'/F,Ie 2
(derived below) for each conductor, plotted vs. I,/Ic' the peak ac current normalIzed
as a fraction of the dc critical current. This normalization allows direct comparison
between various conductors and with the n = 3 theoretical loss curve obtained from
the Bean-London model [1.4.5]. This model assumes the presence of magnetic fields
strong enough so that bulk flux penetration occurs and the superconductor is in the
"mixed state." This critical state model has been used to obtain an expression for the
self-field losses in a wire carrying a small (lP.« Ie) impressed ac current [1.4.5]. In the
case where J e is assumed not to vary with 11,
Peale
=
jfl(l//1CR2Je)
X
10- 9
(1)
where Peale is the theoretical power loss in W,J is the frequency in Hz, I is the conductor length in em, I, is the peak value ofthe sinusoidally varying current in A, R is the
superconductor radius in cm, and J e is the dc critical current density in the superconductor in A/cm 2 • For the situation in which JeB is equal to a constant, the losses
are predicted by the Kin-Anderson model to vary as 1,4. In both of these cases, a
J e- 1 dependence is predicted. However, it should be noted that there is experimental
evidence for a J e-1.6 dependence in various similar Nb-Ti alloys [11]. Assuming the
r I dependence and substituting 1CR 2Je = I c' the dc critical current, we have
(2)
where F, is a constant for a particular sample length and frequency. All loss data were
taken at 60 Hz. The lengths and F, values for each sample are given in Table II. We
may now think in terms of a "specific loss" Peale/F,Ie 2 for a conductor, which eliminates geometric, frequency, and critical current factors. Ie is known from direct measurement, from extrapolation of the Je-H curve to zero field, or from wire specifications.
Thus, we obtain the equation of the theoretical loss curve of Fig. 5, in which
n = 3,as
Peale/F,I/
=
(I,/Ie)3
(3)
In order to exhibit n, and to allow comparison with the theoretical curve, the loss
data (the measured loss Pm in W, and I,) is plotted as P"./F,I/ vs. the ratio I,/Ie'
Figure 5 shows that there is good agreement with the n = 3 theoretical curve at
high I,IIc' the small conductors for the most part having losses one-half to two-thirds
ofthe theoretical loss, and the upper parts ofthe curves for those ofthe large conductors whose curves show slope changes having about one and one-half to three times
the theoretical loss.
The small wire data points are all well represented by straight lines, with n
values (given in Fig. 5) between 3.0 and 3.7 for the Nb-Ti wires, which are consistent
with the results of Kudo et al. I] (n = 3 to 3.5) and Eastham and Rhodes [10] (n =
3.13). The n value of 4.78 for the Nb multifilament wire is consistent with values of
5.11 ± 0.42 (the variation being due to differing metallurgical treatment) obtained
by Beall and Meyerhoff [12].
e
Alternating Current Losses in Superconducting Conductors for Low-Field Applications
179
The lower parts of the curves for the large conductors approach or enter the
region of low self-field where the magnetic flux does not penetrate into the bulk of
the superconducting core. The critical state model is not applicable in this region
and bulk hysteresis losses are not expected. Magnetization measurements on NIr Ti
wire samples show that there is no well-defined point at which flux penetration occurs
[13]. There is a broad, rounded maximum on the magnetization curves, and the initial
deviation from linearity occurs at 0.030 and 0.070 T, for large filaments (360 jLm) and
small filaments (5 to 10 jLm), respectively. Because of the ill-defined nature of the
point of flux penetration, more refined loss theories, such as that of Dunn and Hlawiczka [14], which depend on well-defined flux entry points on the hysteresis loop,
cannot be applied. Again, since the point of flux penetration is poorly defined, an
abrupt change in the loss curve, with the hysteresis losses falling abruptly, is not to be
expected. In fact, no falloff (n going to > 3) in losses is observed. Those curves that do
change slope shift over to an n = 2 line at low currents. As previously stated for the
Nb 3 Sn data, such an /2 loss dependence has two conceivable sources, surface hysteresis
and eddy currents. All of the wires have surface sheaths of low-conductivity matrix
material. Sheath eddy current loss estimates were made, using simplifying assumptions
such as an equivalent bundle radius for the rectangular wires and spatial magnetic
field uniformity in the sheath, which indicated that for the small wires, sheath losses
were negligible. For the large rectangular wires, large sheath loss contributions would
be obtained which would be significant at high currents and dominant at low currents.
Complicating the eddy current estimates is the fact that for twisted wires, additional
eddy current losses may occur in the bundle interior, because of the time-dependent
axial magnetic field produced by the spiraling current in the filaments 5]. Such
twist-associated losses are difficult to estimate because of the unknown magnitude of
the transverse conductivity of the filament bundle. Twist also complicates the loss
picture through additional hysteresis losses due to the axial magnetic field, geometrically decreased J c due to the filament angle with the axis [16], and possible metallurgical
J c enhancement due to the additional cold-working [10].
For the two large, round wires H-I0l (untwisted) and H-102 (twisted), identical
except for twist, the loss data fall on straight lines with no slope changes and low n,
approximately equal to 2. The outer sheath for those samples is relatively much
thicker than for the other wires and the dominating sheath losses cause n to be close
to two over the whole current range. The twist in H-I02 causes additional interior
eddy current losses and results in a parallel but higher loss curve. No loss decrease
caused by a J c increase due to the additional cold-working is seen because the untwisted
wire is already extremely cold-worked.
The estimates of eddy current losses indicate that they alone may be large enough
to cause the /2 dependences observed, without invoking surface hysteresis losses. Loss
measurements as a function of frequency, which are planned, will definitely separate
the two effects.
e
CONCLUSIONS
The ac losses of stabilized type II superconducting conductors are characterized
by hysteresis losses in the superconducting material and eddy current losses in the
normal matrix. The relative magnitudes of these two contributions depend upon the
conductor design. Frequency dependence measurements and an investigation of
transposition effects would help to increase understanding of losses in conductors of
these types.
180
M. A. Juocko, D. W. Dels, IUId W. J. Carr, Jr.
ACKNOWLEDGMENTS
The authors would like to thank M. P. Mathur for performing the magnetization measurements on
several of the conductors used in this work. The technical assistance of H. N. Sopko and P. J. Steve in
performing this work is gratefully acknowledged.
REFERENCES
1. R. Hancox, Proc.IEE, 113: 1221 (1966).
2. E. B. Forsyth, "Underground Power Transmission by Superconducting Cable," Brookhaven National
Laboratory Rept. No. 50325, Brookhaven, New York (March 1972).
3. M. A. Janocko and D. W. Deis, Cryogenics, to be published.
4. H. London, Phys. Letters, 6: 162 (1963).
s. H. R. Hart, Jr. and P. S. Swartz, "The Calculation of AC Losses Using the Critical State Model,"
Air Force Materials Laboratory Tech. Rept. No. AFML-TR-6S-BI, Section IIA (March 1966).
6. T. A. Buchhold, Cryogenics, 3: 141 (1963).
7. M. J. Chant, M. R. Halse, and H. O. Lorch, Proc.IEE, 117: 1441 (1970).
8. M. Garber and W. B. Sampson, in: Proc. 13th Intern. Congress of Refrigeration, Intern. Inst. Refrig.,
Paris (1973), p. 393.
9. D. R. Salmon and J. A. Catterall, J. Appl. Phys., D3: 1023 (1970).
10. A. R. Eastham and R. G. Rhodes, in: Proc. 3rd Intern. Cryogenic Engineering Conference, Iliffe Sci.
and Tech. Pub!., London (1970), p. 167.
II. M. Kudo, K. Aihara, K. Kuroda, and T. Doi, in: Proc. 4th Intern. Cryogenic Engineering Conference,
Iliffe Sci. and Tech. Pub!., London (1972), p. 143.
12. W. T. Beall, Jr. and R. W. Meyerhoff, J. Appl. Phys., 40:2052 (1%9).
13. M. P. Mathur, private communication.
14. W. I. Dunn and P. Hlawiczka, Brit. J. Appl. Phys. (Ser. 2), Dl: 1469 (1968).
IS. W. J. Carr, to be published in J. Appl. Phys.
16. M. S. Walker, in: Proceedings Applied Superconductivity Conference, IEEE Pub!. No. 72CH0682-STABSC, p. 477.
DISCUSSION
Question by M. Garber, Brookhaven National Laboratory: Eddy current losses are suggested as
possibly playing a role for the region where P goes as r2. Calculation of the magnitude of the effect to
be expected shows this to be not likely for the kinds of fields and thicknesses of copper used (11 to 2 mils,
for example, in commercial Nb 3 Sn tape.) Snowden and co-workers found a region of r 2 behavior in
Nb 3 Sn which was not due to eddy currents. This might be worth referencing.
Answer by author: The point concerning the cause of /2 losses is well taken, but we do not believe
simple estimates of eddy current losses can be made in these tapes, because of the complications introduced
by edge effects, particularly if the critical state model is applied at the low current densities where the /2
dependence is seen It is our opinion that eddy current losses may be significant due to the relatively large
amounts of copper present on the tapes. We have already included the possibility that /2 surface losses
may also playa part, and have made reference to the work of Garber and Sampson, [8] on Nb 3Sn tapes.
The work of Snowden evidently refers to unpublished work.
E-2
THE APPLICATION OF LOSS MODELS TO
SUPERCONDUCTING SOLENOIDS
J. T. Broach and W. D. Lee
u. S. Army Mobility Equipment Research and Development Center
Ft. Belvoir, Virginia
INTRODUCTION
The use of inductors, both cryogenic and superconductive, for the storage of
energy in pulsed power systems of various types has recently attracted considerable
attention. The storage element is usually operated in a pool of liquid cryogen but for
certain applications, integrated refrigeration using helium gas is desirable. In either
case, an estimate of the losses generated during transients is an essential part of the
design. The principal loss mechanisms in the superconducting case are eddy currents
in the stabilizer and hysteresis in the superconductor, both of which depend upon the
magnitude and frequency of the transient field. This presentation presents the results
of an application of loss models to a particular configuration of multifilament
superconductor. The eddy current loss model is compared separately to loss measurements made on a coil wound from commercial copper wire. Losses calculated using a
simple linear model ofthe field in the winding are compared with those obtained from
a more sophisticated calculation using digital computer techniques.
EDDY CURRENT LOSSES
The power losses due to eddy currents caused by a time-varying self-field in a
normally conducting solenoid will first be calculated. The magnetic field is assumed
to be axial and the eddy currents are assumed to flow in circular paths within the
wire.
Applying the integral form of Faraday's law to the circular geometry of an eddy
current with a maximum radius rm. we obtain the electric field driving the eddy
currents
(1)
From Ohm's law, the power density can be written, using (1), as
(2)
Working in spherical coordinates, taking the field along the (J
rm = r sin (J,
q=
[(r2 sin 2 9)/4p]B2
181
=
0 axis so that
(3)
182
J. T. Broach and W. D. Lee
Integrating to get the average loss per unit volume
Q
V=
J~J~" J~- [B2r2(sin 2 O)/4p] dV
41trw 3/3
(4)
results in
Q/V =
iJ2 rw 2 /10p
(5)
Assuming a sinusoidal transport current, this becomes
Q/V = Bm2(J)2rw2/20p
(6)
To check this model, measurements of the effective resistance of a copper coil
were taken at various frequencies with a Maxwell bridge [1]. To obtain the frequency
dependence, the dc resistance was subtracted from the measured value at each
frequency. The measured resistance was converted to power loss by using ]2Rerr .
Values ofthe power loss obtained in this way are compared in Fig. I with the results
calculated from (6). The value of resistivity, 1.6 x 10- 10 Q-m, used in the loss calculation was determined experimentally at 4.2 K. An expression for B which varies
linearly across the winding was used. The departure of the measured values from a
straight line at the higher frequencies is due to the fact that the experiment was run at
constant voltage across the bridge. As frequency is increased, the impedance of the
coil increases; therefore this arm draws less current and the ]2 R decreases below the
expected value.
In order to derive an expression for eddy current losses in the copper stabilizer
of multifilament superconducting wires, the method used in the solid copper wire
10·
Fig. L Power loss vs. frequency for copper
coil at 4.2 K.
Fig. 2. "Unit cell" for multifilament conductor.
The AppHcatioo of Loss Models to Superconduc:ting Solenoids
183
derivation was adapted to a geometry in which the filaments are distributed throughout the wire cross section. The eddy currents are confined to the regions of copper
between the filaments. Considering a cross section of the composite, we assume a
"unit cell" defined by superconducting filaments of circular cross section (see Fig. 2).
The composite is taken to consist of "hcp" cylinders which represent the filaments;
the spacing of the filaments is taken as a multiple of the filament radius, allowing
application to a variety of conductors. The integration of the eddy current paths was
taken in the regions between the filaments so that the power loss per unit volume of
copper was calculated. The spacing of the unit cell was estimated from the number of
filaments, the copper-to-superconductor ratio, and measurement of photomicrographs. Applying (3) to the geometry of Fig. 2 and integrating yields the loss per unit
volume
g=
V
2nB f.x rm
2
°
Vp
4
dx'
4
(7)
Referring to Fig. 2, rm is equal to one-half the distance in the y direction between the
outer boundaries ofthe unit cell taken in such a way as to exclude the region occupied
by the superconductor filaments. The expression obtained for a single "unit cell"
is
i=B::2(~5 +n4+n3+~ +i+1!O)/(n; +n2+~+214)
(8)
HYSTERESIS LOSSES
The hysteresis losses in a solenoid wound from composite superconductor can
be estimated using a model due to Bean 3 ]. A long solenoid can be approximated
as a multilayer scroll wound from a sheet of superconducting material. The local power
dissipation per unit volume derived by Bean [3] is given as
e·
q = llcd'Bo( a2
4r
-
r)
a2 - at
(9)
This expression assumes a simple linear variation of field in the winding as a function
ofr and assumes no variation with z (displacement parallel to the axis). A coil design
based on NIOMAX TC· composed of 101 turns of a single layer of 144-strand braid
having at = 0.194 m, a 2 = 0.202 m, 1= 0.792 m, A. = 0.3, and designed for Bo =
1.4 T and an inductance of 2 x 10- 3 H with 105 J stored energy resulted in calculated
losses of 1.24 Jjcycle.
For many solenoid configurations, the simple linear field profile assumed above
is a good representation. To determine the error which might be introduced in design
estimates through the use of (9), a more refined estimate of the field profile for the
case described was made with the aid ofa digital computer. The algorithm was based
on the superposition of field contributions due to a finite number of current loops.
The field was found to be approximately linear with r near the midplane with the field
reversing at a point near the outer winding edge. Near the end of the coil, the field
was proportional to r2. To simplify the estimate based on (9), the coil was divided
into regions and the field profile in a region was fitted to a simple algebraic expression.
Calculated hysteresis losses based on this method were found to be approximately
5 % lower than the previous estimate.
* Imperial Metal Industries, Ltd., Birmingham, United Kingdom.
J. T. Broach and W. D. Lee
184
32r---------------------------------,
2.1
2.4
.1
.4
m
~
M
~
~
n
~
~
FRfQUEIICY. Hz
B
3
n
~
~
~
Fig. 3. Power loss vs. frequency for a small
solenoid. Circles are the experimentally measured
values and the line represents the calculated loss.
The method described above has been compared to calorimetric measurements
of solenoid losses made at low frequencies by JUngst [4]. Eddy current losses were
negligible at these frequencies. The results are shown in Fig. 3. There is fair agreement
between the calculated and experimental results in this frequency range, although the
slope of the curve differs from that of a curve drawn through the data points. The
solenoid upon which the calculation is based has an ID of 0.024 m, an OD of 0.065 m,
and a length of 0.053 m. The conductor has a diameter of 0.0004 m and is composed
of 61 35-Jlm Nb-Ti filaments with a twist rate of four turns/in. The copper-to-superconductor ratio is 1.4: 1. Central field is 5.2 T at 46.9 A.
NOTATION
at
a2
=
=
=
d
=
=
B
Bm =
Bo =
E
I
Jc
I
n
Q
q
r
Reff
rm
r.,
V
=
=
=
=
=
=
=
=
=
=
=
inner winding radius
outer winding radius
magnetic flux density
sinusoidal peak flux density
central field of solenoid
superconductor filament diameter
electric field strength
rms transport current
superconductor critical current density
winding length
scale factor (not necessarily integral) for filament radius specifying spacing of composite
power
power density
radial position
elJective resistance
maximum radius of eddy current loop
radius of wire
volume available for eddy current paths
Greek letters
A.
p
r
= ratio of superconductor volume to winding volume
w
=
=
=
resistivity
charge (or discharge) time
angular frequency
The Application of Loss Models to Superconducting Solenoids
185
REFERENCES
I. W. M. Schwarz, Intermediate Electromagnetic Theory, John Wiley and Sons, New York (1964).
2. C. P. Bean, Phys. Rev. Letters, 8:250 (1962).
3. C. P. Bean, "A Research Investigation of the Factors that Affect the Superconducting Properties of
Materials," Tech. Rept. No. AFML-TR-65-431 (March 1966).
4. K. P. Jiingst, G. Krafft, and G. Ries, "Measurements on Pulsed Superconducting Magnets," paper
presented at Third International Conference on Magnet Technology, May 19-22, 1970, Hamburg,
Germany.
E-3
INVESTIGATION OF THE DYNAMIC
PROCESSES OCCURRING IN
SUPERCONDUCTING WINDINGS
V. A. Altov, M. G. Kremlev, V. V. Sytchev, and V. B. Zenkevitch
Institute for High Temperatures
Moscow, USSR
Dynamic processes of propagation or contraction of a normal zone in composite
superconductors are of special interest when studying the operation of superconducting devices under different conditions. It was therefore desirable to supplement the
results of previous investigations [1.2] of thermal equilibrium of the normal zone in a
composite superconductor with a study of the behavior of the normal zone under
nonequilibrium conditions when the normal zone for some reason propagates (or
contracts) along the conductor.
The specimen investigated, having a length of 200 to 300 m, was wound in the
form of a bifilar winding (see item (1) in Fig. 1). Since this winding had no inductivity,
it did not create a self magnetic field. A plastic filament was used to form the channels
used for liquid helium cooling of the winding. A microheater was attached to the
inner tum of the winding. This solenoid was placed within another superconducting
solenoid (2). The latter was used to provide a uniform magnetic field of required
intensity.
The use of a specimen in the form of a bifilar winding placed into an external
magnetic field makes it possible to reference the experimental results to some fixed
value of the intensity of the magnetic field H. This, in turn, provides a possibility of
comparing the supplemental results with the data .2] obtained earlier for the case
of thermal equilibrium of the normal zone.
The solenoio under investigation was connected in series with an ordinary
superconducting solenoid (3) located beyond solenoid (2). The energy-storing solenoid
(3) served as a power source for the solenoid being studied. The transfer of the combined system (solenoid (I) plus solenoid (3» to a regime of "frozen" magnetic flux was
made possible with the aid of the superconducting shunt (9).
e
9
6
3
Fig. 1. Experimental arrangement for investigating the dynamic processes occurring in superconducting windings.
186
Investigation of the Dynamic Processes Occurring in Superconducting Windings
187
The potential taps (4) were connected to an X-Y recorder (5) to measure the
potential difference across the solenoid under study. Since the winding ofthis solenoid
was made bifilar, the measured potential difference was equal to the voltage drop
associated with the active resistance of the normal zone. The second channel of the
two-coordinate recorder (5) was fed with a signal from a magnetoresistive Bi probe (6)
located within solenoid (3); the value of this signal was proportional to the current
intensity in the combined system (solenoid (1) plus solenoid (3)). Thus, the currentvoltage dependence of solenoid (1) was obtained on the recorder chart.
Besides the above-mentioned quantities, the voltage drop occurring across the
ends of a small (1 cm) auxiliary section was also measured. A pulse microheater was
located in the center of this auxiliary section for initiating the appearance of a normal
zone. The voltage measuring leads (7) from this auxiliary section were connected to
another X- Y recorder (8). The second channel of this recorder was also supplied
with a signal from the Bi probe (6). Thus, a current-voltage dependence for solenoid
(1) ,was obtained on this recorder. The measurements were conducted for various
values of the "frozen" current exceeding the minimum propagation current 1p'
Figure 2 illustrates the current-voltage dependence for the case when solenoids
(1) and (3) were made of an eight-core composite conductor having a diameter of 1 mm
in an external magnetic field of 40 kOe. The ratio of the copper matrix to the superconductor in the composite conductor was equal to 3: 1. Solenoid (1) with a bifilar
winding had an inner diameter of 30 mm, an external diameter of 80 mm, a height of
40 mm, and 346 turns; the power source solenoid (3) had an inner diameter of 30 mm,
an external diameter of 60 mm, a height of 70 mm and 363 turns.
Each curve in Fig. 2 corresponds to the dynamic processes which occur after
the appearance of a normal zone in the composite conductor at different values of the
initial current loin the solenoid. From the general qualitative considerations of the
earlier investigations [1,2], it is obvious that the disappearance (collapse) of the
normal zone in the described dynamic process should occur when the current is
equal to the minimum current necessary for the existence of the normal zone, 1m ,
regardless of the value of the initial current in the solenoid. t This a priori conclusion
is in good agreement with the experimental data, i.e., for a definite interval of initial
currents (see Fig. 2), all the lines of the dynamic processes terminate practically at the
same point on the X axis; this point should be interpreted as the minimum current of
existence for the normal zone, 1m.
Then, in accordance with the above arguments, it is clear that the resistance of
the portion of the conductor reverting to the normal state during the dynamic
process reaches its maximum value upon attaining a current equal to the current
of propagation 1p for the normal zone. Therefore, in Fig. 2, the points of intersection
of the current-voltage curves for the dynamic processes with the straight lines drawn
from the origin must lie on the vertical 1 = 1p plane, since
Rmax
= (VI l)max
(1)
Figure 2 shows this to be true over a fixed range of initial currents and the value of 1p
can be located on the plot without any difficulty. However, when the initial current is
t In principle, when the values of the initial current I in the solenoid do just barely exceed the minimum
current of propagation of the normal zone, Ip' the disappearance of the normal zone may occur at
current values lying in the range of 1m < I < Ip' As 10 increases, the value of the residual current quickly
approaches the minimum current of existence of the normal zone, 1m'
v. A. Altov, M. G. Kremlev, V. V. Sytchev, and V. B. Zenkevitch
188
!
o.s
f-----.---H-f-f---\l---;----1
0.4
I--+--f/ff".,""-++-+----l
0 .2
1-,.-+*H7hI->rr\-\-\t-\----i
"0
>
w
C>
;!:
~
0 .1
Amps
Fig. 2. Voltage as a function of current for an eight-core composite conductor in an external magnetic field of 40 kOe.
increased still further, the points corresponding to the maximum resistance of the
normal zone on the current-voltage plot begin to deviate from the I p value.
This effect can be explained in the following manner. Figure 3 illustrates the
current-voltage relationship obtained in this experiment. (Here the voltage is the
potential difference across the ends of the auxiliary section.) It is clear that after the
normal zone is sufficiently removed (distance-wise) from the ends of this auxiliary
section, the v(I) dependence may be regarded as a terminal characteristic of this
section for the composite conductor maintained (at each given moment) under
practically isothermal conditions. The stationary current-voltage characteristics
describing the state of thermal equilibrium for a composite conductor under isothermal conditions were considered in detail in an earlier publication [1]. We
recall the basic conclusions obtained in this paper with the help of the schematic
diagram given in Fig. 4. The current-voltage curve shown characterizes the states of
thermal equilibrium of a composite conductor with a parameter of stabilization
3 r--.-;;- , - ----,
6 1-_+-+----i~_+-+-~
~
!! 5 I-_+-+---)i<--_+----'t-~
g
!: 2 r---fr--;-- --,I
g
E 4 1-_+--hl~-;oL-_+~H~
w
~ 3
c5
> 2
r--+----t'-fl-+--1-,.<i-+-----"<+\-+-l
1----+..t....-f+i<~H':::.Ro~\:tt-t+!
Im I' Ie
CURRENT , Amps
CURRENT , Amps
Fig. 3. Current-voltage relationship obtained for
the experimental arrangement.
Fig. 4. Current-voltage relationship used to
characterize the states of thermal equilibrium
for a composite conductor with a parameter of
stabilization ex < I.
Investigation of the Dynamic Processes Occurring in Superconducting Windings
189
< 1 with the assumption of nucleate boiling of helium. The branch of the currentvoltage characteristic curve lying below the critical temperature 1'" of the superconductor (section a-b-c) relates to the resistive states of the conductor. The branch
lying above 1'" (section c-d-e) relates to the normal state of the conductor. The specific
appearance of the current-voltage characteristic of the conductor at a normal state is
stipulated by the temperature dependence of the specific resistance of the matrix of
the composite conductor.
In this case, within the temperature range from 1'" to T* the normal states of the
conductor are stable (the derivative dvidl for section c-d has a positive sign). On the
other hand, section d-e ofthe current-voltage curve relates to unstable normal states.
The basic feature of the current-voltage characteristic under consideration is that
above a certain threshold current 1*- (the so-called maximum equilibrium current),
the conditions of thermal equilibrium for a conductor in the normal state are completely absent.
Now let us return to the current-voltage curves shown in Fig. 3. It is evident
that the value vi I is the resistance of a unit length of the composite conductor at any
given instant in time. In this case, the point on the dynamic process line corresponding
to the given value of the initial current 10 where the resistance for the section of the
composite conductor viI has a maximum magnitude is the point of the equilibrium
state.
From this, it is evident that the points associated with the equilibrium states for a
given composite conductor can be found on a current-voltage diagram by drawing
tangents from the origin to each dynamic process line. These points can then be used
to determine the line of equilibrium states for the isothermal case (dashed line in
Fig. 3) of the composite conductor being considered. By comparing the equilibrium
curve obtained in this fashion with the diagram depicted in Fig. 4, one can easily
recognize that the character of this line completely corresponds to the section of the
current-voltage characteristic describing the states of equilibrium of the conductor
when T> 1'". In this case in Fig. 3 one can easily identify the stable and unstable
branches of the equilibrium curve corresponding to the sections c-d and d-e of the
current-voltage characteristic in Fig. 4, a value of maximum equilibrium current 1*
for the conductor under investigation can also be found above which no conditions of
thermal equilibrium of the conductor in the normal state can be found in principle.
It should be remembered that the concepts for the characteristic currents 1m
and I p are introduced on the basis of models constructed under the assumption of
the existence of stable equilibrium states of the conductor. It is clear that the concepts
of maximum current for the existence of the normal zone, 1m , and the minimum
current for propagation of the normal zone, I P' lose their meaning in those dynamic
processes where current-voltage curves have no points corresponding to the stable
states of the thermal conductor and the models .2] of the composite conductor
for describing these dynamic processes cannot be used. This analysis permits us to
understand the anomalous behavior of the dynamic current-voltage curves under
conditions where the initial currents substantially exceed the maximum equilibrium
current 1*.
Let us now turn to another important consideration in this study. Earlier it was
noted that the solenoid employed in this experiment was wound in the form of a
bifilar winding; therefore, between two turns of the winding having the same direction
of current there is always a turn of wire in which the current flows in the opposite
direction. As a result, when the normal zone appears in one of the turns of such a
winding, the following phenomenon takes place. Having completely occupied one
0(
e
190
V. A. Altov, M. G. Kremlev, V. V. Sytcbev, aad V. B. Zenkevitcb
tum, the normal zone starts spreading through the next turns; in this specific situation,
adjacent turns that are in a normal state are separated from each other by turns that
are in a superconducting state. Consequently, the turns of the winding that are in a
normal state are at a comparatively large distance from each other. This is in contrast
to the winding of a single solenoid of a comparatively small diameter, where, during
the propagation of the normal zone, the turns that are in a normal state are much
closer to each other (adjacent turns).
What is the difference between these two cases from the point of view of the
conditions of propagation of the normal zone along the winding? In these two cases,
the winding tum may encounter substantially different conditions of heat transfer
from the surface. Once the peak nucleate heat flux (burnout) is attained during boiling
heat transfer, a film of vapor forms separating the conductor surface from the liquid
helium with a resultant increase in thermal resistance. In the case of a bifilar (or similar)
winding, the surface films of vapor on the adjacent normal turns are separated from
each other by a sufficiently large gap; whereas in the case of an ordinary winding
of small diameter, the vapor films located close to each other can apparently join together. Such joining ofthe vapor films from the adjacent turns (equivalent to a growth
in the effective thickness ofthe film) results in an increase in the vapor layer separating
the surface of the composite conductor from the liquid helium. This increases the
temperature of the conductor and, naturally, has a significant effect on the parameters
influencing the propagation process of the normal zone.
In order to verify this assumption, a special experiment has been performed. In
the experimental solenoid, which was similar to that described above, diagnostic
pulse heaters were arranged on adjacent turns of the bifilar layer (i.e., on the turns
with opposite direction of the current).
Before conducting the basic series of experiments, two preliminary series of
experiments were carried out. The current-voltage dependence, using the same
values of the initial current 10 , was measured under two different operating conditions,
i.e., introduction of the normal zone into the winding first by means of one of the
microheaters and then by means of the second microheater. Within the accuracy
of the two-coordinate recorder available, the current-voltage dependence was
identical for both series of experiments. This result indicates that, first, the composite
conductor specimen used for making the bifilar layer was homogeneous; second, the
power input to each of the microheaters was correctly selected; third, the conditions
of heat transfer were identical for the two turns of the winding under examination.
Two basic series of experiments were then carried out. In the first series of
experiments, the current-voltage dependence was determined for the case when only
one microheater was switched on, namely, the heater located inside the measuring
section. The results of this series of experiments are represented by the solid curves in
Fig.5.
In the second series of experiments, both microheaters were switched on. The
experiments were conducted using the same values of initial current as in the first
series; in this case, the current-voltage dependence was recorded on the same chart
as for the first series of experiments. The results of the second series of experiments are
indicated by the dashed lines in Fig. 5. The experimental data seem to confirm the
assumption of a possible overlapping of the vapor films from adjacent turns of the
winding since the equilibrium curves obtained from the second series of experiments
(two microheaters are energized) correspond consistently to lower currents than the
equilibrium curves obtained from the first series of experiments (only one microheater
is energized).
Investigation of the Dynamic Processes Occurring in Superconducting Windings
191
B
7
.I!
6
-0
> 5
w
~
Fig. 5. Current-voltage relationship obtained under different
heating conditions of turns in solenoid.
2
o
\
, ,S-
4
~ 3
~
./
/\
I
X
.-
,
\
,
\
"
" -''\
' / .-- / ~ 1
r(~
IT!
100
200
CURRENT . Amps
This presentation would not be complete without mentioning a number of other
characteristic features of the normal zone propagation process that are unique to
actual superconducting windings as compared to bifilar windings. These features
stem from the fact that in actual windings, the magnetic field intensity is uniquely
associated with the value of the current of the "frozen" solenoid. Consequently,
each value of the initial current 10 is related to a definite intensity of the magnetic
field in the solenoid at the moment the incipient normal zone is introduced. In the
dynamic process under consideration, the magnetic field intensity varies with the
current in the solenoid. Thus, in this case, different curves of the dynamic process
relate to different values of the magnetic field intensity. (This situation is complicated
still further because different regions of the solenoid are, at any given moment, in
magnetic fields of different intensities, since in any solenoid there occurs a gradient of
the intensity of the magnetic field). This fact leads one to expect that the different
dynamic process curves will not converge at a point corresponding to the current 1m
and that, when plotted, the maximum values of the dynamic process curves in the
diagram will not be located on a vertical line.
These assumptions are well confirmed by the experimental data. To illustrate
this point, Figs. 6 and 7 depict the V(1) and v(1) dependences obtained with a sevencore twisted-cable (four superconducting cores plus three copper cores) solenoid
impregnated with pure indium. These figures show that the value of 1m for the sevencore wire is uncertain over a current range from 63 to 68.5 A; also, the points on the
current-voltage curves corresponding to the maximum resistance values for different
10 are noticeably displaced relative to each other.
With respect to the data given in Fig. 7, the following should be noted. As in the
previous case, the dashed line in the diagram is drawn through the points corresponding to the equilibrium states of a composite conductor when T > 1'.:. While in the
case of a bifilar winding, each such curve can be related to a definite value of intensity
of the external magnetic field (produced independently by means of the external
solenoid (2) in Fig. 1), the curve representing the equilibrium states as defined from the
experimental data for a real solenoid corresponds to a situation where the field
produced by the winding varies with the current. It is clear that the equilibrium
state curve fixed in such an experiment is correlated with the equilibrium state curve
associated with different values of constant magnetic field intensity. This is schematically demonstrated in Fig. 8, where the dashed lines correspond to the constant values
of the magnetic field intensity (H 1 > H 2 > H 3) and the solid line corresponds to the
magnetic fields varying in the experiment with a real solenoid. In this case, it is
obvious that the higher the 10 , the higher the 1 (and therefore H) at the point where
the dynamic process line corresponds to the equilibrium states.
192
V. A. Altov, M. G. Kremlev, V. V. Sytchev, and V. B. Zenkevitch
0.5
OA
~
o
I
IL~ \
\
0.3
>
w
~
~2 ~--+---~4-++~~
-
!oJ
~l ~--~~~~++~~~
~>
\\
~
/
0.2
~
~
~ 3
0 .1
o
90
70
50
110
o
40
80
120
160
CURREN T. Amps
Fig. 7. Current-voltage relationship for a seven-core twistedcable solenoid showing the equilibrium states of the composite
conductor when T> 7;.
130
CURRENT. AMPS
Fig. 6.Current-voltage relationship
for a seven-core twisted-cable
solenoid.
.
~ ~ ~
884
££f
\
\
\
I
\
\
\
,
\
'
\
\
\
\
,'
I
I
CURRENT
Fig. 8. Schematic representation showing how the magnetic field (solid
line) produced by the winding in a real solenoid varies with the current.
ACKNOWLEDGMENTS
The authors are indebted to A.I. Korolkov, N . A.Kulysov, and G. N . Simakov for their assistance with
the measurements.
REFERENCES
1. V. V. Sytchev, V. B. Zenkevitch, V. A . Altov, M . G . Kremlev, and N. A. Kulysov, Cryogenics, 12(5): 377
(1972).
2. V. V.Sytchev, V. B. Zenkevitch, V. A Altov,
.
M. G . Kremlev, and N. A. Kulysov, Cryogenics, 13(1) : 19
(1973).
E-4
ANALYSIS OF CRYOGENIC CURRENT LEADS
WITH NORMAL CONDUCTORS AND
SUPERCONDUCTORS IN PARALLEL
B. B. Gamble
Research and Development Center
General Electric Company
Schenectady, New York
and
J. L. Smith, Jr. and P. ThuUen
Massachusetts Institute of Technology
Cambridge, Massachusetts
INTRODUCTION
Superconducting electrical equipment requires cryogenic current leads and connections from superconductors to normal conductors. The literature contains many
discussions on the design of cryogenic current leads [1-4]. The superconductor-normal
conductor connections can be made at one point, or the superconductor can be run
in parallel with the normal conductor to form a long connection. The object of this
discussion is to show the advantages offorming the connections by the latter method.
The problem to be analyzed is the coolant requirement of a vacuum-insulated,
vapor-cooled, cryogenic current lead constructed from normal conductors and superconductors in parallel. The length of the normally conducting portion of the lead is
variable with movement of the superconducting-normal transition point. Figure 1
is a schematic of the lead to be analyzed.
CALCULATIONS
Boundary Conditions
The temperature at the beginning of the normally conducting portion of the lead
(x = 0) is by definition the transition temperature of the superconductor. Assuming
negligible temperature difference between the coolant and lead, the heat conducted
past x = 0 must be sufficient to vaporize a fraction ex of the coolant which is delivered
to the lead as liquid and to warm the coolant flow to the transition temperature of
the superconductor:
(1)
T(x = 0) = 1;
Q(x = 0) = m[exhfg
193
+ c(1; - 4.2 K)]
(2)
J94
B. B. GUIbIe, J. L. Smith, Jr., .... P. Dullea
+-I- QI.-O)- am"ft· m.,tT." 4.Z
I
K)
I
I
,I
L
I
.. " 0
I
I
SUPERCONDUCTOR
Fig. 1. Schematic of a lead constructed from
superconductor and normal conductor.
.... L
T-T.
T.l"
The solution is not ~xtremely dependent on the warm-end temperature of the lead.
The warm-end temperature can be arbitrarily set at 300 K :
T(x
=
L)
=
= 300 K
Tw
(3)
Differential Equations
Heat transfer in the lead consists of heat convected to the coolant stream, heat
conducted down the lead, and heat generated due to the electrical resistivity of the
normal conductor. Assuming negligible temperature difference between the coolant
and the lead, the governing differential equation is
dQ
dT
J2 r
--mc-+-=O
dx
dx
A
(4)
With x and Q defined in the directions shown in Fig. 1, the conduction equation
can be used to remove position dependence from (4) by
dx
kA
dQ
dT
= Q dT,
=
mcpQ - J2 rk
Q
(5)
Nondimensionalizing, (5) yields
d(Q/mcpT,) _ (Q/mcpT,) - (J2 rk/m 2c/T,)
d(T/T,) Q/mcpT,
(6)
Integrating this equation subject to the bottom and end boundary conditions
for a given material yields a functional relationship for conduction,
Q/mcpT, = F 1(T,
T" a., J/m)
(7)
Equation (4) can be written with the position dependence removed from the
mass flow term as
dQ _ mcpQ + J2 r = 0
(8)
dx
kA
A
m dx _
~
d(Q/mcpTt )
- c p (Q/mcpTt ) - (I2 rklm 2c/Tt )
A
(9)
By replacing Q/mcpT, in (9) by the function F1 defined in (7), a function F2 can
be derived for the integral from zero to x of the length term m dx/A:
z -
IA
x
0
m dx
-
-
fQ(X)/"",pT'
Q(O)/_pT,
z = F 2 (T, T" a., J/m)
k d(Q/mcpT,)
cp[(Q/mcpT,) - (Prk/m 2cp2T,)]
(10)
(11)
Analysis of Cryogenic Current Leads with Nonnal Conductors and Superconductors in Parallel
195
The integral referred to as z in (10) and (11) can be thought of as a transformation
of position having the dimensions of kg/m-sec. For the given material, the length of
the normally conducting portion of the lead can be calculated from (11) since
foL (m dxlA) = F2(Tw , 7;, rx, lim)
(12)
Equation (11) allows functions to be developed for temperature and its slope:
T = F3 (z, 7;, lim, rx)
(13)
dTldz = (Alm)dTldx = F4 [z(or T), 7;,Jlm,rx]
(14)
Convective Heat Transfer Limitations
If the analysis based on the assumption that the lead and coolant are at the same
temperature is to be valid, the temperature difference must be shown to be small
enough to not significantly affect the solution.
Equation (4) can be altered to include the temperature difference,
dQ _ mc
dx
p
[(1 _
I1T) dT _ T d(I1TIT)]
T dx
dx
+ J2 r =
A
0
(15)
If 11 TIT« 1, the effect of 11 T on the previous equations can be neglected. To
calculate this ratio, the convective heat transfer per unit surface area must be
calculated:
Q =mcpdT =m 2 c p dT =hl1T
c
P
dx
AP dz
(16)
In (16), the temperature gradient has been replaced by (14).
If the convective heat transfer coefficient is known, the ratio of temperature
difference to temperature can be calculated:
I1T
T
=
Qc
m 2cp dT
hT = hAPT dz
(17)
SOLUTIONS
The differential equation (4) can be solved subject to the boundary conditions
[equations (1) through (3)] as an initial-valued finite-element problem. This was done
for several values of rx and 7; for OFHC copper.
Figures 2 through 4 are solutions for (12) for rx of 1, 0.5, and 0, respectively. No
steady-state solution was found for rx = 0 and 7; = 4.2 K. The solution for rx = 0.1
and 7; = 4.2 K is given in Fig. 4.
Figure 5 presents several solutions to (14).
DISCUSSION
Leads Constructed without Superconductor
The curves for a transition temperature of 4.2 K in Figs. 2 through 4 can be used
to show the operation of a lead constructed from only normal conductors. A lead
constructed from only normal conductors can be operated at the maximum currentto-mass-flow ratio yielding a steady-state solution for a specific current. If such a lead
196
B. B. Gamble, J. L. Smith, Jr., and P. lbuDeo
T••
.
,
,
...'0'"
....'"
~
.5.0
7.5
.0.0
'.2
~
f
.
'0
~
~
OFHC
COPPE
11.,.1.0
0
0.3
0.5
0.'
0.6
JL 1lljA.
0.7
0.8
0
0.3
0.9
0.5
0.'
0.6
0.7
0.8
0.9
L
f m..tJ.
kg/m-sec
0
kg/m-sec
o •
Fig. 2. Solutions for equation (12).
Fig. 3. Solutions for equation (12).
is designed for operation at the maximum current-to-mass-flow ratio at current 11
and it is desired to operate at a current 12, the current-to-mass-flow ratio must be
lowered from the design value. The following examples show this for a lead which is
supplied with saturated liquid.
Example 1. 12 < 11' From Fig. 2, a design point with a large current-tomass-flow ratio is selected for the design current 11 :
1 tim 1 = 2.45 x 107 A-sec/kg,
miL/A = 0.72 kg/m-sec
L/A = (1.76 x 10 7 A/m)/1 1
Operating a lead with this aspect ratio at a lower current requires lowering the
current-to-mass-flow ratio. To show this, another point is selected from Fig. 2:
= 1.5 x 107 A-sec/kg,
m 2L/A = 0.57 kg/m-sec
L/A = (0.855 x 107 A/m)/12 = (1.76 x 107 A/m)/11
12/m2
0
0
IS.D
'0.0
0
7.5
T.
-4.2
.
I
---10.0 1
,
,
~
~
::-
...'".
20 tr-"-"'~"I~O=.O_0=-t-_ _ _--T_ _ _- I
E
'0
v
....
E
v
OFHC COPPER
11m" 2.45.107 lA-seC/kg)
.... 1 0 h l - - - - r - - - - T - - - - - 1
0
0 .•
0.5
0 .•
f m.p..
0.7
0.8
0.9
1.0
L
kOI m-$IC
0
Fig. 4. Solutions for equation (12).
TEMPERATURE.
K
Fig. 5. Solutions for equation (14).
Analysis of Cryogenic Current Leads with Normal Conductors and Superconductors in ParaDel 197
Thus,
12//1 = 0.485
Example 2. 12 > 11 . Operating a lead with this aspect ratio at a higher
current also requires lowering the current-to-mass-ftow ratio. Figure 6 shows a
possible solution for higher currents. The resistive heating of the length Ll in which
boiling occurs provides 1 - oc of the latent heat of boiling. The heat conducted at
the point x = 0, where boiling is completed, is ocmhfg . To show this, a point can be
selected from Fig. 4 (oc = 0.1):
12/m2 = 1.6 x 107 A-sec/kg,
L2/A
=
(1.6 x
m2L 2/ A = 1.0 kg/m-sec
107
A/m)/12
The length Ll can be found from a heat balance
l/rLdA = (1 - oc)mhfg
Ld A = (1 - oc)hfg/(I 2/m2)1 2r = (0.68 x 107 A/m)/1 2
L/A = (LdA) + (L2/A) = (2.28 x 107 A/m)/12
=
(1.76 x 107 A/m)/Il
Thus,
Leads Constructed with Superconductors and Normal Conductors
Figures 2 through 4 also show the operation of a lead with normal conductors
and superconductors in parallel. The aspect ratio from the warm end of the lead to
the transition temperature is determined by the current-to-mass-ftow ratio and the
magnitude of the current applied to the lead. As long as the physical length of the
lead is not exceeded and the assumption of negligible temperature difference between
the lead and the coolant is valid, the lead can be operated near the maximum currentto-mass-ftow ratios shown in the figures. As the current is varied, the normally conducting length of the lead can be adjusted for the optimum tradeoff between resistive
heating and conduction heat leak.
The solutions indicate that for leads with transition temperatures higher than
7.5 K, the maximum current-to-mass-ftow ratio is not a strong function of the quality
of the liquid delivered to the lead. The superconductor permits the latent heat of
vaporization to be used to intercept other heat leaks to the cold space.
\OPPER
\
: - 0(."0) " .mh'Q
1
~====~'~:============~~l
~LI--~:'-----L2----~':
: (BOILING)
I
4.
I
I
I
I
T"4.2
• -0
LIQUID
VAPOR
I
TaT•
x· LZ
Fig. 6. Schematic of a normally conducting lead.
198
B. B. Gamble, J. L. Smith, Jr., and P. 11udIen
The superconductor also allows the design of a high-current lead which has a
low-conduction heat leak when not in use. During current flow, only a small length
of the lead is normally conducting and when not in use, the entire physical length of
the lead is available to act as a thermal standoff.
Example 3. Figure 4 can be used to design a lead with a 10 K transition temperature which consumes only vapor (0( = 0). An operating point can be selected
from Fig. 4 such that
Ijm = 2.45 x 107 A-sec/kg,
mLjA = 0.77 kgjm-sec
LjA = (1.88 x 107 Ajm)jl
This equation is then used to calculate the normally conducting aspect ratio
necessary to operate at the selected operating point for any given current. The lowest
anticipated current determines the actual aspect ratio of the lead, and for higher
currents, the normally conducting length of lead can be decreased with movement
of the superconducting-normally conducting transition point. The lead cross section
and therefore the length are chosen from convective heat transfer considerations.
Convective Heat Transfer
The convective heat transfer coefficient in helium is an increasing function of
temperature for both laminar and turbulent forced convection. This and the solution
given for (ljT) dTjdz in Fig. 4 indicate that !1TjT given by (17) is a decreasing function
of temperature.
Comparing the mass-flow term in (4) to that in (15) shows that the effective mass
flow cooling of the lead is decreased by a temperature difference between the lead
and the coolant. The mass flows calculated by the previous method of assuming
negligible temperature difference between the lead and coolant should be adjusted
to a higher value m' given by
m < m' < m/(1 -
!1T(~t= Tt»)
/(1 - h:;~t ~:(T =
=m
Tt)
(18)
Equation (l8) indicates there is an advantage to having a large cross section
and perimeter. Increasing the cross section increases the length of lead necessary for
operation at a given current. Therefore, it is advantageous to have a large perimeterto-cross section ratio. This suggests constructing the lead from braided strands of
superconductor coated with copper. In addition, it may be advantageous to vary the
cross section of the lead to provide the cross section needed for operation at high
currents and the large aspect ratio required for low currents.
STABILITY
The transient characteristics ofthis type of lead have not been investigated. An
important question is the stability of this type oflead to perturbations in flow. During
a momentary loss of coolant, the lead will heat up and the normally conducting length
of lead will increase. If the flow after a perturbation is not sufficient to return the lead
to its initial temperature distribution, the lead will continue to heat up until failure
results. The closer the lead operating point is to the maximum in Figs. 2 through 4,
the smaller is the perturbation in flow that will cause the lead to become unstable.
Aulysis of Cryogenic Current Leads with Normal Coodudors uti Supercoadudors in ParaDeI
199
CONCLUSIONS
There appears to be an advantage to running superconductors in parallel with
normal conductors in cryogenic current leads for most applications. Most of the
advantages are a result of the superconductor's ability to make the normally conducting length of lead independent of the physical length of lead. This allows operation
at high current-to-mass-ftow ratios over a wide range of currents. This also allows
a high-current lead to be constructed with a low-conduction heat leak with zero
current. Leads constructed with superconductor have the additional advantage of
being able to operate at high current-to-mass-ftow ratios without consuming liquid
helium.
NOTATION
A
IX
= cross section, m 2
= [Q(x = 0) -
mcp(T, - 4.2)]/mhl ,
cp = specific heat of helium gas, J/kg - K
FI = function defined in equation (7)
F2 = function defined in equation (11)
F3 = function defined in equation (13)
F4 = function defined in equation (14)
h = convective heat transfer coefficient, W/m2-K
hi. = latent heat of boiling of helium, J/kg
1 = current, A
k = thermal conductivity of copper, W/m-K
L = length from T, to Tw ' m
LI = length defined in Fig. 6
L2 = length defined in Fig. 6
m = mass flow calculated assuming l!J. T = 0, kg/sec
m' = mass flow calculated with a finite heat transfer coefficient, kg/sec
P = perimeter, m
Q = conduction heat flux, W
Q. = convective heat flux per unit surface area, W/m2
r = resistivity of copper Q-m
T = temperature, K
T, = transition temperature of the superconductor, K
Tw = temperature of the warm end of the lead, K
l!J. T = temperature difference between the lead and coolant, K
x = position measured from T" m
z = transformed position defined in equation (10), kgfm-seC
REFERENCES
1. P. Thul\en, in: Advances in Cryogenic Engineering. Vol. 16, Plenum Press, New York (1971), p. 292.
2. J. P. Scott, in: Proceedings 3rd Intern. Cryogenic Engineering Conference. Iliffe Sci. and Tech. Pub\.,
London (1971), p. 176.
3. M. Rauh, in: Proceedings 3rd Intern. Cryogenic Engineering Conference, Iliffe Sci. and Tech. Pub\.,
London (1971), p. 182.
4. J. W. L. Kohler, G. Prast, and A. K. J onge, in: Proceedings 3rd Intern. Cryogenic Engineering Conference,
Iliffe Sci. and Tech. Pub\., London (1973), p. 192.
£-1
A DECADE OF INVOLVEMENT WITH
SMALL GAS-LUBRICATED TURBINES
M. E. Clarke
British Oxygen Company Limited
London, United Kingdom
INTRODUCTION
The decade in question is that which brings us to the present day and in point of
fact largely concerns activities in the field of commercially available helium refrigerators. Work on the small gas-bearing turboexpander commenced about five years
earlier when research work being undertaken by Sixsmith at Reading University on
a machine for a small air liquefaction plant was encountered. During those five years
of sporadic development, the potential market for a small gas-lubricated turbine
was for air separation duty, particularly when some of the special attributes of gas
lubrication were realized. In particular, the absence oflubricating oils in the presence
of oxygen and oxygen-enriched air as a contaminant in the process fluid is well known
to be dangerous. However, at this time, liquid helium commenced to be used in
increasing quantities and the advantages of gas bearing turboexpanders were also
examined for use in helium refrigerators. The more obvious advantages are:
1. The machinery is rotating, not reciprocating, and hence there are no vibrations
and a very low level of noise.
2. The bearings allow very high rotational speeds needed for high efficiency
due to the high sonic velocity of helium.
3. There is simplicity of construction, for high reliability.
4. There are no partitioning seals between the ambient temperature bearing
and process gas, resulting in simplicity and reliability.
5. There is no metal-to-metal contact, and hence absence of wear and absence
of formal maintenance.
It is perhaps an obvious comment that the advent of superconductivity for commercial applications is responsible for the intensifying interest for reliable refrigerators.
As would be expected, the early designs of the gas-lubricated turbine were
primarily concerned with the bearing performance and therefore somewhat unsophisticated with regard to heat conduction, vacuum integrity for thermal insulation,
and ease of assembly. The first commercial turbine using helium was run in 1964 in
a refrigerator that produced 73 W at 3 K for the Rutherford helium bubble challiber.
The turbine used in this plant was designed with the premise that only warm seals
could be relied upon and therefore cryogenic seals were not used. The design of
this turbine was such that it could not be assembled first as a unit and then installed
in the plant. Instead, it had to be assembled in the plant with all the hazards associated
with the presence of dirt and dust. It must be said, however, that the design proved
200
A Decade of Inyolyement with Small Gas-Lubricated Turbines
2(11
NOCUS CAS
IN
....... CAl
OUT
Fig. I. Cryogenic expansion turbine---early design.
to be more tolerant in this respect than one might suppose. Figure 1 shows the construction of this series of machines where the diffuser assembly is built into the coldbox, thus separating the machine in such a way as to make it impossible to pretest the
turbine and subsequently install the undisturbed unit into the plant, a feature now
considered essential.
PRESENT DESIGN
The design concept of the early machines dictated that the center section containing the running gear also contain almost all the precision parts so as to avoid
difficult alignment problems. This feature has been retained in the present units.
Design changes, however, have been concentrated on developing a machine which
would be capable of being removed from a test bed as a unit and similarly installed
in a plant. Figure 2 shows the design details of the latest units.
The design features a plug and socket concept which provides a very simple
and therefore rapid means of installing or removing the expander. If necessary, a unit
can easily be removed from a cold plant. If standby units are fitted, they can easily
be removed from a plant that is in use. Just as simply, a turbine can be installed in
an operating plant. The time required to do this is typically less than 1 hr.
Commercially available cryogenic seals are used but not relied upon. In the case
of the two seals guiding the process gas through the guard filter, dimensional tolerances
ensure that the lower seal seats preferentially. The upper seal is backed by a warm
O-ring. Experiments to determine the preferred cryogenic seal ring have shown that,
for the conditions prevailing in the turbine, completely clean, lapped faces are as
good, if not better than, the best commercially available cryogenic reusable seal
rings of the PTFE-coated metal type.
20.2
M. E. Clarke
Fig. 2. Cryogenic expansion turbine-----<:urrent design.
TURBINE ENVIRONMENT-DEVELOPMENT
So far, the differences in mechanical design detail between the early and present
units have been outlined. The development that results in current designs is now
discussed.
Bearing Gas Circuits
Hydrostatic bearings, as used in these turbines, have advantages oflarge running
clearances, very high speed limitations, and reasonable tolerance to dirt inclusions.
They do, however, require a continuous supply of gas above a critical pressure. This
has to be maintained at all times the turbine is operating and in particular during
startup and shutdown. All the designs discussed in this presentation are arranged so
that the bearing pressure is the same as the process, thus obviating auxiliary compressors. The gas at ambient temperature is filtered to 1 J1.m but is otherwise untreated.
In early plants, the gas supply was maintained during emergency stops by means
of a high-pressure gas bottle, a pressure regulator, and a nonreturn valve. This system
was found to be inconvenient and sometimes dangerous since the gas could slowly
leak away and the pressure would not be available when required in an emergency.
Present systems employ a gas receiver working at plant process pressure. In the event
of main compressor failure, this volume is sufficient to keep the bearings lubricated
through the turbine shutdown.
Since gas bearings have no built-in capacity to survive even short interruptions
in the lubrication supply, it is essential to know the limiting parameters and to ensure
that these boundaries are never exceeded even under transient conditions. As an
example, Fig. 3 shows how the turbine limiting speed is affected by the back pressure
on the bearing drains. Early failures were encountered when drain pressures unknowingly rose to unacceptable levels while changing plant conditions.
A Decade of Involvement with Small Gas-Lubricated Turbines
Fig. 3. Relationship between seizure onset speed
and bearing exhaust back pressure.
2000
3000
203
4000
5000
SHAFT SPEED, REV/SEC
Protection Equipment
Safety valves in the bearing drain circuits now fully protect the turbines against
the type of abuse described above. Interlocks are provided so that under no circumstances can a turbine be started if any of the limiting bearing supply parameters are
exceeded. In addition, should the turbine be inadvertently operated near or exceeding
these boundaries, the turbine stop valve is immediately tripped. The effectiveness of
these developments was recently demonstrated by a customer who in a plant acceptance test carried out about thirty separate, carefully conceived fault trips designed to
produce turbine bearing failures. The turbines survived with no trouble whatsoever.
The protection equipment is simple and of low cost, is backed up by automatic
speed control, and permits pushbutton start-stop of the turbines.
Pipework Constraints
In order to achieve good thermodynamic performance, certain parts of the
turbine body have to be designed as heat breaks. These parts have been found to be
incapable of carrying heavy static loads and still maintain the agreed accuracy of
alignment, so all connections to the machines are now flexible.
TURBINE DEVELOPMENT
The capacity of the turbine can be defined by a single dimensionless number
referred to as the flow number, or
F = Q(T/p)1 /2
(1)
where Q is the helium throughput in sefm, P is the turbine inlet pressure in atm,
and T is the turbine inlet temperature in K. The gas flow is independent of rotational
speed, largely due to the axial flow wheel configuration. Thus, with either a plant
design point or one flow measurement, the turbine throughput can be predicted at
all other operating conditions.
It has also been determined that the flow number concept applies to other process
fluids provided the F number is corrected for gas density. Thus
F oc (l/density ratio)1/2
(2)
The flow number for the machine under discussion may be varied from about 100
to 300 by fitting different stationary nozzles, turbine wheels, and brake wheels. The
204
M. E. Clarke
Table I. Relation between
Flow Number and
Efficiency of a Turbine
Flow number
Efficiency
100
150
200
250
0.66
0.69
0.71
0.73
60
50
30
Fig. 4. Relationship between UIC o and turbine
efficiency for a specific turbine.
Table II. Thermodynamic Losses
Experienced in a Turbine
Component
Stationary nozzle
Vaneless space
Radial to axial turning
Rotor passage
Exit
Leakage
Disk friction
Efficiency drop,
1.02
1.75
2.05
18.30
UO
1.20
4.55
30.00
%
A Decade of Involvement with Small GIIIl-Lubricated Turbines
20S
latter change results from the differing work output. However, the shaft and bearings
sizes remain unchanged.
Studies also indicate that there is a relationship between the flow number and
the maximum obtainable thermodynamic efficiency based upon enthalpy drop. For
about a 7: 1 expansion ratio the values in Table I are typical. Typically, the turbine
with a flow number of 100 would be used in the lower-temperature stage ofa liquefier
having a capacity of about 20 liters/hr.
Another important design parameter involves changing rotational speed to
obtain optimum efficiency with varying inlet temperature and pressure to the turbine.
The ratio U/C o relates the tangential velocity of the turbine wheel U with a velocity
which would be attainable if the entire enthalpy drop through the machine were
converted into velocity; the latter is sometimes referred to as the "spouting velocity"
and is given by
Co = 2gJ(AH)1/2
(3)
where g is the gravitational constant and AH is the change in enthalpy. The turbines
described here operate at optimum efficiency when Co equals 0.65, which is a typical
value for a 50 % reaction turbine. Figure 4 shows the relationship between U/C o
and turbine efficiency for a specific turbine. Note that the rotational speed must be
maintained reasonably constant for any set of operating conditions if maximum
efficiency is to be realized.
Studies have also been made to proportion the thermodynamic losses in one of
these units having an overall efficiency of 70 %, with the results shown in Table II.
The rotor passage loss accounts for approximately 60 % of the total and is believed
to be due to the finite thickness of the blading in the region of the nozzle causing gas
breakaway at the trailing edge. In such a small machine, it is difficult to do much
about this, but different blade forms are under development in an endeavor to reduce
this particular loss.
Materials
The main body of the turbines, which was originally constructed of brass and
nickel silver, is now manufactured of 18/8 stainless steel, which is also used for the
socket. All early units had 18/8 stainless steel shafts with shrink-fitted aluminum
alloy brake wheels and removable aluminum alloy turbine wheels held on to a
precision register with a socket head cap screw. The soundness of the designs can be
appreciated when it is realized that this has changed little since 1964, with the exception of a change to titanium alloy for the shaft to avoid dimensional changes in
service due to a partial change from austenite to martensite. Titanium has additional
advantages of lightness and lower thermal conductivity.
Bearings and fixed nozzles are of nickel silver, used largely for ease of accurate
machining and compatibility with respect to its coefficient of contraction in cooling
down.
Gas Bearings
Operational facts have emerged during the development of these units and
these are discussed with reference to Fig. 5. The latter shows an elementary loaded
hydrostatic thrust bearing and the type of pressure distribution patterns arising
from different surface contours. Gas is fed to the center of the thrust area and is metered
with an orifice. When pressure is applied, the pad lifts off until the integrated pressure
M. E. Clarke
IURFACIl CONTOUfII KEY
,. CONYEX
2. FLAT
1 CONCAVE
Fig. 5. Pressure distribution on elementary thrust
bearing.
over its area a just equals the applied force W. Bearings of this type have a touchdown
performance p, where
p
=
(0.3 to 0.5) Wa
(4)
Typically, we design for about a p of0.25Wa.
The operating clearance t is a function of the gas flow through the bearing and
for given loading conditions and gas supply, this is related to the size of the metering
orifice. In order to conserve bearing gas, it is tempting to run at the smallest possible
values of t. Taken to extremes, it has been found that a heavily loaded bearing can
suddenly fail due to even slight thermal distortions. If these occur in a manner that
cause the surfaces to convex, the load-carrying capacity rapidly falls due to the changing shape of the pressure profile.
Occasionally, when faced with having to carry extreme thrust loads, a concave
surface is generated which gives a better pressure profile than flat surfaces. Great
care is, however, necessary since the dynamic behavior of the bearing can be seriously
affected and, in the extreme, self-sustained oscillations promoted. Thus, it is essential
to manufacture turbines with extreme precision not only in terms of absolute accuracy
but in terms of profile and shape also.
With the use of the Sixsmith type of bearings, which basically consist of two
orifices in series, an intermediate pressure within the antiwhirl pockets is produced.
The magnitude of this pressure depends on the selection of the sizes of these orifices.
It has been found possible in practice to select a value of pressure which approximately equals the turbine wheel tip pressure, so that warm gas from the bearings
cannot enter the turbine wheel space, nor can the cold helium leak into the warm
bearings. The fact that no shaft seal is necessary is a further justification for the use
of antiwhirl bearings. Development work has shown that the very high speeds required
can also be obtained by much simpler configurations, but additional shaft seals may
be required for optimum thermodynamic efficiency.
Bearing development work has been carried out on simple shafts simulating the
same mass as the final turbine shaft system. It has been found, however, that this
gives no indication of the dynamic load-carrying capacity of the bearings and the
results can therefore be very misleading in that rotational speeds can sometimes
be obtained which are far higher than ultimately achieved with full gas loads on
the turbine and brake wheels.
Gas consumption for the bearings is generally 4 to 5 %ofthe turbine throughput
and is considered to be a small penalty to pay for extreme reliability and much higher
load-carrying capacity of the hydrostatic gas bearing systems compared with selfacting bearings.
A Decade of Inyoh'ement with Small G_Lubricated Turbines
It has been necessary to carefully control the dynamic balancing process. Early
rotors were balanced to about 1 x 1O- 5 0z-in. It was found that in some circumstances, this was insufficiently precise and a range of balancing machines an order
of magnitude better were developed. These support the shaft in gas bearings and
use gas drive to provide high-speed rotation.
During the course of turbine development, an attempt has been made to correlate
known limiting speeds of hydrostatic gas bearing systems with rotating mass, with
the result shown in Fig. 6. Thus, there is some evidence to indicate the importance
of low weight when the ultimate in rotational speed is required.
SERVICE EXPERIENCE
Many of these machines are now in service, the majority in helium refrigerators
and liquefiers, but others are on air separation duty and one larger machine is
operating on argon. Reliability is of a very high order, as shown recently where a
turbine was removed from a helium refrigeratorjliquefier after five years of service.
It was found to be dirty but only needed a solvent wash before refitting to the plant.
Another helium plant which has now run for a considerable period was accidentally
filled with air while it was cold. Additionally, the turbine was stop-started over nine
hundred times during two days of commissioning. The turbine survived this abuse.
Turbine failures have been experienced due to the buildup of water behind the
turbine wheel resulting from excessive moisture in the bearing gas supply. Concentrations in excess of 150 vpm should therefore be avoided. The tiny turbine wheels are
capable of operating without damage while liquefying impurities. Erosion of the
turbine wheel can occur by solid nitrogen when crossing the triple point during
'0
'.,
\
,
,
\
,
,,
,
,
\
\
\
...
.0'
Fig. 6. Limiting shaft speed vs. rotating mass.
1\
,
,
10
20
SHAFT SPEED .10"""' RPM
30
.
40 50
,
,
leo,..
70 10
M. E. Clarke
cooldown. This has caused no trouble in practice since the situation is transient and
only occurs under conditions of gross neglect of the purification system.
FUTURE TRENDS
This class of expansion machinery will probably be utilized more in the future
for reasons of compactness, reliability, and lack of maintenance requirements. Even
now it is not unusual for specifications for helium liquefiers/refrigerators to demand
rotary expansion machines and this demand is almost universal where continuous
operation of superconducting machinery is required.
The most recent liquefier embodying turbines will produce 12 to 15 liters/hr,
needs no nitrogen precooling, and will deal with makeup helium containing large
quantities of air. The turbines provide silent, trouble-free operation with no wearing
parts. Fully automatic operation is provided.
REFERENCES
1. H. Sixsmith, in: Advanced Cryogenics (C. A. Bailey, ed.), Plenum Press, New York (1971), p. 225.
F-2
GAS BEARING CRYOGENIC EXPANSION
TURBINES
J.-c. Villard and F. J. Muller
L'Air Liquide-Centre d'Etudes Cryogeniques
Sassenage, France
INTRODUCTION
Present trends indicate that industrial cryogenic equipment will soon gain
recognition because of its reliability and resistance to wear during long-term operations. The reliability and durability of cryogenerators depend on the expander, the
key component of these installations. The thermodynamic performance of this
expander also has an important effect on the efficiency of the system. Reciprocating
expanders are presently being replaced by turboexpanders in very high-power
installations. The function of these turboexpanders in a given cycle is to provide the
polytropic expansion of a cold gas with recovery of the resulting work.
Some important research projects, undertaken in this laboratory since 1967,
have resulted in the perfecting of a group of high-performance cryogenic turboexpanders capable of treating a small flow rate of helium gas with a high expansion
ratio. The range of turbines actually developed at this time allows the treatment of
helium flow rates between 2 and 450 x 10- 3 kg/sec. (The helium flow rate depends
on the inlet temperature and the expansion ratio.)
Fig. 1. Gas bearing cryogenic expansion turbine.
209
210
J.-c. Villard and F. J. Muller
The turboexpanders consist of a centripetal expansion turbine and a centrifugal
brake compressor with an aero static-type high-speed gas bearing (see Fig. 1). The
high-pressure gas of the cycle, collected in the inlet pipe (1), is expanded in part in
the fixed blade distributor (2) and the turbine wheel (3). The gas, accelerated in the
distributor and then expanded in the turbine wheel, imparts a torque to the turbine.
This torque is transmitted to a compressor brake (6) by means of a gas bearing system.
The gas, which has been cooled and expanded, escapes through the diffuser (4) and
the exhaust pipe (8). The expander work is transformed into heat in the compressor
brake by compressing a gas in a closed system and the resulting heat is removed
afterward by a water-cooled heat exchanger (7). In the case of turboexpanders where
large enthalpy changes are encountered, high rotation speeds, necessary to obtain
high thermodynamic performances, are obtained by means of a gas bearing system
consisting of journal and thrust bearings. This gas bearing system practically assures
an unlimited lifetime to the rotating system because of the complete lack of contact
between moving parts.
The journal bearings are equipped with a stabilization system which dampens
the transverse oscillations [1] and function by means of a pressurized gas at room
temperature which comes from an exterior source (the compressor of the cycle).
Oiagram 01 Ihe principle of an
auiOmobl1 1E ~ho c lt otlosorbtr
(0 )
,
Condui t lc ading to 0
t ixf d ~ u urt: drop
l"l V COf'ln lt ch c CO'\Clul cs
( c )
O.ogrom 01
Q
ilmpOhcd
_ _ __ _ _0_ompl"9 sys tem
E.amp'le of a dam p.!!' gJ~t~~.
\
I
t p e la il ot Itt,
(gYl lln
Interconntc:t cd c:oPldllollts leod1f'l 9 to
G tilC r d
prtSs.,ut:
dr op
Fig. 2. Gas bearing damping system.
Gas Bearing Cryogenic Expansion Turbines
211
This stabilization system consists of a damping device composed of cavities interconnected by conduits which lead to a fixed pressure drop (see Fig. 2).
At the outlet of the gas bearing system, the gas used for the bearings is collected
and sent back to the low-pressure side of the cycle. The use of gas of the same composition as the expanded gas to provide the lift of the rotor and the braking of the
turbine avoids any pollution of the installation. The optimum design of each gas
bearing, in terms of its size, the nature of the gas used, and the speed required, results
from an experimental study at the test bench.
On the other hand, a no-contact pressure drop sealing system (10) (see Fig. 1)
located on each side of the turbine wheel limits the leakage of cold gas from the cycle
toward the bearing box. The leakage of cold gas depends on the dimensions of the
sealing system, the nature of the gas, and the pressure and temperature conditions.
For example, for a turbine treating a helium flow rate of 15 g/sec under 15 atm at
22 K, a leakage toward the bearings of about 0.1 g/sec has been measured. On the
other hand, for a larger turbine treating 180 g/sec of helium under 15 atm at 30 K,
the leakage measured is about 0.3 g/sec. The leakage can be reduced significantly
by reducing the difference in pressure on the two sides ofthe sealing system. However,
a certain amount of cold gas permitted to flow toward the bearings limits the entry
of heat by conduction along the rotor. Thus, this loss of cold gas flow, which is
negligible in comparison with the flow in the turbine, slightly increases the overall
efficiency of the turbine.
RECENT GAS BEARING STUDIES
Resistance of the Gas Bearing System to Shocks
The gas bearing system developed in this laboratory and being used on the
turboexpanders is characterized by a great resistance to shocks. Tests have recently
been performed on a gas bearing system rotating at 90,000 rpm and using helium
under 7 atm pressure. Figure 3 is a schematic diagram of the shock test bench. The
bearing system to be tested was suspended by a wire in such a way as to swing freely.
An accelerometer and a displacement sensor were fixed rigidly on the frame of the
bearing system, which supports the fixed parts of the bearings. The sensors were
connected flexibly to the measuring and recording instruments while the gas feed line
of the bearing system was connected flexibly to the gas source. A second pendulum,
consisting of a block suspended on a wire, was used to communicate various shock
intensities to the bearing system.
Figure 4 shows, for longitudinal and transversel shocks, the axial and radial
displacements of a 365-g rotor in relation to its bearings when the bearing system
was subjected to strong acceleration impulses which were either less than or equal
to 1000 m/sec 2 . The results clearly show the short duration of time (a few milliseconds)
required for the reestablishment of the steady state after the initial shock. Obviously,
this type of bearings system shows great potential in the realization of a turboexpander
destined for use in a mobile installation.
Tests of the Gas Bearing System Using Hydrogen
These turbines are designed specifically for expansion of helium at low temperatures but can also function with other gases such as hydrogen and nitrogen, etc.
Figure 5 indicates the results obtained with the larger bearing system operating with
nitrogen, hydrogen, and helium. The solid lines on the figure give the speed at which
the instability of the rotor appears for the different fluids as a function of the gas
feed pressure to the bearing system. The broken lines show the required gas flow rate.
212
J.-C. Villard and F. J. Muller
1(51 OF GAS lHRUSTS
Fig. 3. Diagram of the shock test bench.
Displa.u.mt'nl of ~oloybC'Or~gs in mkron:i
OiSptoc cmcni ~ rOior
.J.lA.A
o
'.
/ bcaring\ in microns
lA
10
20
)0
I:
.
Xl
S
Ac.c«"luation imparted to I u"'b in~ ){ 10· 2nv's. "2
10
20
JO
""'-,,-~
Sires" on th e bca rin9S
20
f eed pressure :
7aim
O~
D
Rotat ion spud : 90 ,000 rpm
o«c l (I"QI;OI"l
.....:L.
-
Stress on
'"
thrusts
Fig. 4. Displacements of 365-g rotor due to shocks.
JO
t im
213
Gas Bearing Cryogenic Expansion Turbines
~.
/
~),L ~ .
lQD
t--
~(
'I
/
/
"./
".
/
Fe"
2
".
,
de'edOf'"
~v
/
/
/
~
~Iorl.'"
System
".
DIAGRAM OF T~T BEHOt
-- -•
I
Spo.~
".
".
Pl"'d. Ut"t'
sl
/
".
/
.!!J.-
--
"'m.ab • .
6
, .
o
of 90'1 btarnJ Iyste m
,
Fig. 5. Comparative test bench
study of a large-capacity bearing
operating with helium, hydrogen, and nitrogen.
COLD HE LlUM
12 (
TURB INE
0 <I S'I()'Kg I S
12 .( ~~~.I: (15
i nle t iF> , 15.5
70
~
-
I~
U QI iq
60
0
-.
~
--- ..
Of"
~f--
5 of»
~
r-- ?:::
f-
Co
17
18
IS
20
21
23
2'
25
26
27
Inl.. t
tt:mpt:ralurr
Fig. 6. Percentage of isentropic efficiency as a function of the inlet temperature for a
cold helium turbine.
K
J.-c. Villani and F. J . MuDer
214
Figure 5 also shows the diagram ofthe test bench. These tests have demonstrated
that it is possible, with this type of bearing system using helium, hydrogen, or nitrogen,
to obtain rotation speeds sufficient to obtain maximum isentropic expansions of
400 Ji g in a single-stage turbine.
Performance
Cold Turbine. Work is now being directed toward improving the performance
of cold helium turbines having a small flow rate, subjected to inlet temperatures of
30 K or less, and operating with a high expansion ratio. Figure 6 shows the results
obtained in 1972 with a turbine handling a flow of 12 to 18 glsec and an expansion
ratio varying between 12 and 15. The maximum efficiency of the turbine was attained
when the ratio of vleo was equal to 0.59. The degree of effective reaction is then
equal to !!Hwheel/!!Htotal = 0.3. The performance of each turbine was determined
under normal operating conditions on a test bench which was equipped to measure
temperatures very precisely.
Figure 7 shows the technical details of a rotor for a low-capacity cold turbine.
The shaft, as well as the distributor, is constructed of a titanium alloy. The blades
of the distributor are cut out of a single piece and then polished manually. The two
parts of the turbine wheel are machined out of an aluminum alloy and then brazed
together. The rotor unit is dynamically balanced manually with a delicacy which
surpasses that obtained with classical balancing machines.
Various installations are equipped with this type of turbine and the total operating
time is now about 40,000 hr. Some larger-capacity turbines have already operated
for about 8000 hr while being subjected to numerous thermal shocks [2J.
Hot Turbines. Turbines that operate with an inlet temperature of 60 K or higher
are generally connected with cold turbines to form an autonomous refrigeration
system operating without outside assistance (e.g., liquid nitrogen). These centripetal
CR'I'OGE NIC
T\R:IO EXPANDER
Fig. 7.Small-capacity cryogenic turbine expander. Helium flow rate, 15 g/sec;
inlet temperature, 22 K; inlet pressure. 15.5 atm abs: expansion ratio, 14 ; and
rpm, 180,000:
Gas Bearing Cryogenic Expansion Turbines
215
50
HOT HELIUM TURBINE
9.104 <Flowra.te Q
Fig. 8. Comparison of isentropic efficiencies of "closed" and "open" turbine
wheels as a function of inlet temperature;
expansion ratio of 14.
~4--r
E"pens .
<
12.10. 1 KSVS
r.1 ;o :
14
4--r~--~+-~
40 L..--I._'--.......................---'_.L-...L..._.L.........................- - - ' _ . . . . . . . ,
1
20
40
60
80
'00
Inlet temperatl,Ar.., K
turbines expand helium or hydrogen with a high expansion ratio. One of the difficulties encountered with small, hot turbines, however, is the attainment of high
tangential velocities on the turbine wheel (V ~ 500 to 600 m/sec). These speeds are
attainable with an open-type turbine wheel by using an aluminum alloy with a high
tensile strength.
The purpose of this study was to compare the performances obtained using an
open turbine wheel with the results obtained using a classical, closed turbine wheel.
The disadvantage of the open turbine wheel is the leakage which occurs just in front
of the turbine wheel between the blades of the turbine and the frame. One of the main
objects of the study was to show the influence of this leakage on the efficiency. For
this study, the test machine was dimensioned for a maximum tangential velocity of
400 m/sec, which is compatible with the closed wheel. The height of the blades on
the periphery of the wheel was 0.44 mm, while the gap between the turbine wheel
and the face plate was reduced to a few hundredths of a millimeter.
Figure 8 shows the results obtained on a turbine with a flow rate between 9 and
12 g/sec and operating with an expansion ratio of 14. The results obtained with this
machine were similar to those obtained with the classical, closed turbine wheel.
CONCLUSION
These studies have contributed to the development of a series of high-performance
cryogenic turboexpanders. Their reliability, compactness, and continuous operation
without maintenance make these turboexpanders choice elements for use in industrial
cryogenic installations.
REFERENCES
1. French Patents No. 1,535,725, No. 2,118,292, and No. 94,739.
2. J. Gass, J.-C. Villard, D. Marinet, and P. Solente, "Gas Bearing Cryogenic Expansion TurbinesApplication to a Refrigeration Plant Delivering 10 kW between 10 and 24.5 K," presented at 4th
Intern. Cryogenic Engineering Conference, May 24-26, 1972, Eindhoven, Netherlands.
£-3
PNEUMATICALLY DRIVEN SPLIT-CYCLE
CRYOGENIC REFRIGERATOR
S.B. Horn, M. E. Lumpkin, and B. T. Walters
u.s. Army Night Vision Laboratory
Ft. Belvoir, Virginia
INTRODUCTION
The split-cycle cryogenic refrigerator with a remote cold finger offers many
advantages for use in cooling infrared systems. The packaging flexibility of the remote
cold finger, because of its small size and weight, offers the dual advantage of use in
rapidly scanning gimbal aerial systems where gimbal weight is restricted to a minimum and in periscope-mounted ground systems where packaging space is restricted
to a minimum. Although this type of cooler design offers many advantages, it previously has not been attractive for use in thermal imaging systems because of its low
efficiency. This presentation describes a valveless split-cycle cryogenic refrigerator
with a remote cold head which has demonstrated high efficiencies for a miniature
cooler.
DYNAMICS OF THE COLD FINGER
This valveless split-cycle cryogenic cooler consists of a mechanical compressor
and a pneumatically driven remote cold finger as shown schematically in Fig. 1.
The dynamics of the cold finger is based upon the force balance of a displacer-piston
which seals off two gas volumes [1]. A pneumatic "spring" volume is located in the
ambient end of the remote cold finger, and this volume is maintained at the average
cycle pressure due to long-term gas leakage past the seal. This pneumatic volume at
the ambient end of the cold finger eliminates the large, cold "dead" volumes which
COLD
R.EG[~EIFIATOR
Fig. I. Split-cycle cryogenic rerrigerator.
TO MOTO"
216
PneumaticaDy Driven Split-Cycle Cryogenic Refrigerator
217
have characterized many previous miniature designs [2]. The working gas pressure,
which is modulated by the motion of the mechanical compressor piston, provides
the second active force acting on the displacer-piston. As the working gas pressure
increases, an increasing force is applied to the displacer which tends to move the
displacer toward the ambient end of the cold finger. This occurs because the working
gas pressure in the cold volume is acting upon the cross-sectional area of the displacer.
This working gas pressure is, of course, also being exerted upon the displacer in the
ambient end of the cold finger. At this end of the displacer, however, the working gas
pressure is acting upon less area, due to the pneumatic piston which is located on
the ambient end of the displacer. Hence, the net force caused by the working gas
pressure tends to move the displacer toward the ambient end of the cold finger.
When this net force exceeds the seal frictions and fluid frictions, the displacer begins
to move. The restoring force acting on the displacer is the force caused by the pneumatic "spring" volume pressure acting on the pneumatic piston area. When the
working gas pressure is at its maximum, the displacer has been pushed all the way
to the ambient end of the cold finger. At this time, the pneumatic volume is, of course,
at its minimum, and hence, the pneumatic pressure is at its maximum. As the working
gas pressure begins to decrease due to the recession of the mechanical compressor
piston, the pneumatic force becomes greater than the force supplied by the working
gas and the displacer begins to move back toward the cold end of the cold finger.
Because of the small displacer stroke required (typically 0.050 to 0.100 in.), and
because of the small mass of the displacer-regenerator-piston (typically 2 to 12 g),
the time required to sweep out the volumes is small compared to the cycle time.
With the proper design, the phase angle between the working gas pressure wave
and the motion of the displacer can be adjusted such that an almost square P- V
curve can be generated.
THERMODYNAMICS OF THE CYCLE
Figure 2 shows an idealized P-V diagram for the cooling cycle. Starting with
the compressor piston at midstroke and the free displacer at bottom stroke, the
series of operations by which refrigeration is achieved is shown in Fig. 3 and described
below.
Point 1 to point 2: The compressor piston moves upward, increasing the pressure
in the cold finger. When the net force acting on the displacer caused by the working
gas pressure exceeds the force exerted on the displacer caused by the pneumatic
pressure acting on the pneumatic piston area, and exceeds seal and fluid friction, the
I
Fig. 2. Idealized P-V diagram of cycle.
218
S. B. Horn, M. E. Lumpkin, and B. T. Walters
COMPRESSOR
POINT 3
POINT 4
Fig. 3. Schematic of refrigerator operation.
dis placer begins to move toward the pneumatic space. During this phase of the cycle,
the gas in the cold finger does work on the gas in the ambient end of the cold finger
resulting in cooling at the refrigeration end of the cold finger and heating at the
ambient end.
Point 2 to point 3: The working gas pressure decreases from its maximum to
some intermediate value due to the motion of the compressor piston. During this
time, the displacer dwells at X = X max .
Point 3 to point 4: When the pneumatic force acting on the displacer exceeds
the working gas force and the seal and fluid frictions, the displacer begins to move
toward the refrigeration end of the cold finger. During this phase of the cycle, the
gas in the ambient space of the cold finger does work on the gas in the refrigeration
end. This transfers thermal energy back from the ambient end into the refrigeration
end.
Point 4 to point 1: The compressor piston now moves upward, causing the
working gas pressure to increase. During this time, the displacer dwells at X = O.
It should be noted that from point 3 to point 4, thermal energy is transferred
back into the cold end of the refrigerator. It is important to note, however, that the
cold volume gas does work on the ambient volume gas when the system is at a high
pressure. Conversely, the ambient gas does work on the cold end gas while the working
gas pressure is at a low value. Therefore, while some thermal energy is transferred
into the cold end during the cycle, the net result is a flow of heat out of the cold end
and into the ambient end. In other words, the cold end gas does more work on the
ambient end gas during the cycle than in the reverse process. This is accomplished
by choosing a design which produces a good phasing between the working gas pressure
wave and the motion of the displacer. This net thermal energy may then be rejected
from the ambient end of the remote cold finger. Figure 4 shows the relationship
between the displacer dynamics and the working gas pressure wave.
Pneumatically Driven Split-Cycle Cryogenic Refrigerator
219
DvRIPfG 'MISI "RlSS!JIIIl$ IHl CI~LoIoC(R SHOuL08.(
MOVI"I G AlSVl TING IN' COO\I~ At 'HE IIIEFRII;IR,!,HON
'""D "ND t<IIIoII,,(j I'" lHl AMBltN'1 ~INO
010., THi.:5.(.,\£SSUIll(S Htl
0II5I'LA([fII~LD6E
,.,
OW(L LI~
AI EthUR THE
COLO (I'IIDOA tH[ AMl8J(NT
Fig. 4. Displacer motion during
pressure cycle.
1- 0
IT
EXPERIMENTAL MODEL
An experimental split-cycle cryogenic cooler with a pneumatically driven remote
cold head was fabricated to investigate this design. The refrigerator is illustrated in
Fig. 5. The compressor contained a I-in. 00 piston with a stroke of 0.32 in. The
cold finger employed a ?6-in.-diameter glass epoxy displacer with a O.l-in. stroke.
The regenerator was fabricated from 200 mesh stainless steel screens which were
closely packed into the displacer. Rulon @) clearance seals were used in the displacer
and Rulon@ spring seals were used in the compressor. The heat rejection area of the
compressor was water-cooled. The heat rejection from the ambient end of the cold
finger was accomplished by natural convection. The cold finger was separated from
the compressor by 18 in. The connecting line was copper tubing with an 00 of
in. and an 10 of /6 in.
t
Fig. 5. NVL split-cycle cryogenic
refrigerator.
220
S. B. Hom, M. E. Lumpkin, and B. T. Walters
u,----------------.
iil.o
iii
II
lU
T, • K
III
III
100
Fig. 6. Cooling load vs. refrigeration temperature.
EXPERIMENTAL RESULTS
Figure 6 shows the heat load vs. refrigeration temperature for the cooler. Both
experimental and theoretical curves are given. Agreement between theory and
experiment seems quite good. Seal blowby of clearance seals was not included in
the theoretical model and can account for the higher theoretical predicted values.
The best performance (ratio of net refrigeration to input power) achieved from the
cooler was 0.87 W of net cooling at 73 K with an input power of 48 W for a cold head
separation distance of 18 in. from the compressor.
CONCLUSIONS
The split-cycle cryogenic refrigerator with a remote cold finger can achieve lowtemperature refrigeration with high cryogenic performances. It should be possible
to fabricate these coolers in production quantities for military applications less
expensively than the standard crank linkage-driven models.
REFERENCES
I. W. Higa, Cryogenic Technology, 1:203 (1965).
2. W. E. Gifford and E. M. Withjack, in: Advances in Cryogenic Engineering, Vol. 14, Plenum Press,
New York (1969), p. 361.
F-4
THEORETICAL ANALYSIS OF PNEUMATICALLY
DRIVEN SPLIT-CYCLE CRYOGENIC
REFRIGERATORS
s. B. Horn and M. E. Lumpkin
U.S. Army Night Vision Laboratory
Ft . Be/voir, Virginia
INTRODUCTION
The split-cycle cryogenic refrigerator has been under investigation for the last
several years for use in infrared devices. This refrigerator, which uses a pneumatically
driven remote cold head, offers the advantages oflow production cost, remote cooling
capability, and flexibility of packaging in both ground and aerial thermal devices.
A cycle that has been exploited recently in this laboratory is one in which a pneumatic "spring" volume, located in the ambient end of the displacer, is sealed off from
the working gas. This cycle offers a positive retarding force on the displacer to prevent
This cycle also
displacer bounce and avoids the use of cold end dead volumes
employs no valves ; hence, the device is able to recover most of the potential energy
stored in the gas during the compression part of the cycle.
eJ.
v"".
""'" p(1t
rOCO LO F1NO£A
StAL
S£Al
THERMAL COMPRESSOR
~
MECHANICA.L COMPRESSOR
INPUT PRESSURE
W4.vE FROM COWR ESSOR
COLD fING£R
v.
.0. 0 "
T.
PNEUM ... lIC -..rlr-=~
GASSII'ACE
Fig. I. Split-qcle cryogenic
refrigerator.
221
S. B. Hom and M. E. Lumpkin
222
DESCRIPTION OF IDEALIZED CYCLE
The split-cycle cryogenic refrigerator (Fig. 1) consists of a remote cold head and
a compressor (either mechanical or thermal). The compressor generates a periodic
pressure wave which travels down a connecting tube to the cold finger housing which
contains the free displacer. A sealed pneumatic volume is located in the ambient end
of the remote cold finger and is pressurized at the mean cycle pressure to account
for long-term leakage past the displacer-piston seal. The pressure in this sealed
pneumatic volume provides the second active force on the free displacer-piston.
Other pressures which influence the cold finger dis placer dynamics are the pressure
drop through the regenerator, the pressure drop through the connecting line, and
(for the thermal compressor only) the pressure drop in the hot cylinder. Frictional
forces also affect the displacer dynamics. In the present study, the dynamic and
thermodynamic heat balance analysis will be given. Since the dynamic and thermodynamic problems are now interrelated, a closed-form solution cannot be found;
thus, a numerical solution to the problem can only be presented.
For a mechanical compressor, the instantaneous volumes of the free dis placer
and the compressor at any time i are given by
l';; = XjAc
+
(1)
~c
l';a; = (Xmax - Xj)(Ac - An)
aVmmaJ1 +
Vm ; =
cos(wt j
+
(2)
~a
+ <pm +
~m
(3)
For a thermal compressor, the instantaneous volumes of the free displacer and the
compressor are given by
l';; = XjAc +
l';a;
(1a)
~c
(Xmax - XJ(A c - An)
=
v"a; =
v"h, =
+
~a
aVmmaJl + cos(wt + <p)J} + ~ha
aVmmaJ1 - COS(wtj + J} + VV
j
<P
hh
(2a)
(3a)
(4a)
The term <P defines the initial position and velocity of the compressor piston.
To determine the instantaneous pressures in both the pneumatic and working
fluid spaces, it is necessary to determine the mass distribution in the two spaces.
Since the pressure in the pneumatic gas space is the same as the pressure in the
working fluid spaces at zero time, for the first iteration
(4)
where Fj =1 is the beginning mass fraction in the total working volume ~Ol' Since
the total mass of the working fluid is constant and equal to the sum of the masses
in the separate subvolumes, and since the pressure is uniform throughout,
Pj =1,j ( l';
Mgw = ~ Tc
l';a
~g
~
Vm)
+ Ta + T.rg + Ta + T.m
(5)
for the mechanical compressor and
M
gw
=
Pj =l,j(l';
R
for the thermal compressor [2J.
r::
+
l';a
T"
+ ~g + ~ + v" + v"r g )
T,g T" 7;. 7;.rg
(Sa)
Theoretical Analysis of PneumaticaDy DriTen Split-Cycle Cryogenic Refrigerators
223
Substituting (1), (2), and (3) into (5); (la), (2a), and (3a) into (Sa); and using the
initial relation
M gw = Pch( ~ot)/R J;,
(6)
the instantaneous pressure in the working space is
P'- l ; =
[Pch(~ot + v"maJ]Fj = 1
)-,
aX; + b cos(wt; + tjJ) + c
(7)
where
a = (J;,Ac/J;) - (Ac - An)
(8)
b = ZmaxAmJ;,/2Tm
(9)
C
= v"t J;, + X max(Ac - An) + (ZmaxAm J;,/2 Tm)
(10)
where
(11)
for the mechanical compressor and
a
= (J;,A./J;) - (Ac - An)
= ZmaxAm(J;, - 7;.)/27;.
C = v"tJ;, + Xmax(Ac - An) + [VmmalJ;, + 7;.)/27;.]
b
(8a)
(9a)
(lOa)
where
(11a)
for the thermal compressor.
It is important at this point to note that the i subscripts denote iterative values
over one cycle for one particular mass fraction. The j subscripts denote the iterative
values for the different mass fractions.
The instantaneous pressure in the pneumatic space is
(12)
Using the fact that
(13)
and that
(14)
then
(15)
S. B. Ron .... M. E. Lumpkin
224
The pressure drops in the cold regenerator and connecting line are given by [3]
MCi = fc(t;)G c2(t;)Lrg/2Pc(t;)rc
(16)
= 2J,(t;)G/(t;)L1/P1(t;)D 1
(17)
M~i
and for the hot cylinder
(18)
SOLUTION OF DYNAMICS AND HEAT BALANCE
The equation of motion of the cold displacer is now given by
~
MCdA(t;)
Mgc(t;)
= An[P(t;) - Pit;)] - Ac L\Pc(t;)IMgc(t;) I
Mit;)
X(t i )
- An L\P,(t;)IM,(t;) I - SIX(t;)1
(19)
for the mechanical compressor and
..
McdX(t i )
= An[P(t i )
-
Pn(t i )]
-
Mgc(t;)
Ac L\PAt;)IMgc(t i) I
M1(t i)
L\
Mgh
X(t;)
- An L\~(t;)IMI(t;)1 - An Ph(t;)IMghl - SIX(t;)1
(19a)
for the thermal compressor.
Because of the computational complexity involved in finding a solution for
these coupled dynamic-thermodynamic equations, computer program models, which
employ a fourth-order Taylor series expansion, were developed to simulate the
dynamic operation and thermodynamic performance of the device. To calculate the
correct mass distribution in the pneumatic volume and working space, the following
condition must be met:
(20)
If we define the pressures in the (j
Pj+1,i
+
=
l)th iteration to be
Pj,~Fj+1/F)
(21)
and
(22)
then we can express the (j + l)th mass fraction in terms of the jth mass fraction by
substituting (21) and (22) into (20), resulting in
F.
J+1
=
(1 - FjL~P Pj,i + 1)-1
F."~
)L,
(23)
"j.i
Note that (23) is independent of the position of the displacer and depends only on
the pressure.
Theoretical Analysis of Pneumatically Driven Split-Cycle Cryogenic Refrigerators
225
Thus, the general expressions for the instantaneous working and pneumatic
pressures are given by
p.. =
J,'
aX i
Pch(l';ol + v"maJFj
+ b COS(wti + cjJ) + c
(24)
and
= Pch(l';ol + v"maJT,,(1 -
p
Fj )
T"(v,,max - AnXi)
nj,i
(25)
The number of mass fraction iterations needed is determined by the speed at
which the F j approach and satisfy the convergence criterion
Fj + 1
-
Fj ~ A
(26)
where A is the maximum allowable difference. The final solution ofthe thermodynamic
and dynamic problems is found when the mass fraction convergence criterion is
satisfied.
One must consider the initial conditions for both the cold displacer and the
compressor piston. The phase angle cjJ defines the compressor piston position and
velocity at t = O. The other initial condition that must be chosen defines the position
and velocity of the free displacer at t = O. Since the displacer is not mechanically
coupled to the compressor, after choosing a set of displacer initial conditions, one
must then determine an appropriate phase angle cjJ; i.e., determine the appropriate
compressor initial conditions. One choice for the initial conditions for the free
displacer was that X = 0 and X = 0 at t = O. From these initial conditions, 4> = 90°
was then chosen for the mechanical compressor and cjJ = - 90° for the thermal
compressor. The phase angle was chosen such that during displacer dwell (both at
X = 0 and X = x max), the pressure is at some intermediate value. For an optimum
design, the displacer should be moving when the working pressure is at the extrema
and the displacer should dwell during the intermediate pressures. To further investigate designs that were not periodic (those in which the displacer did not return to
X = 0 after one cycle) and also to verify that the steady-state solution is independent
of the initial conditions, a multicyclic computer program was developed. In this
program, the final conditions of the dis placer after each cycle were used as the initial
conditions for the following cycle:
=
X(t =
X(t
T)(cycle
T)(cycle
= 1) = X(t = O)(cycle = 2)
= 1) = X(t = O)(cycle = 2)
(27)
(28)
Convergence to steady-state operation, namely
=
X(t =
X(t
0)
0)
= X(t = T)
= X(t = T)
(29)
(30)
for designs which were not periodic after one cycle (because of improperly chosen
initial conditions) was so rapid that all designs converged to steady-state operation
in less than five cycles. A computer study was done to determine whether the steadystate solution is independent of or dependent upon the initial conditions. Results
have shown that for a particular phase angle, the cooler always returns to the same
steady-state operation regardless of the initial conditions for the displacer. This
226
S. B. Hom and M. E. Lumpkin
Table I. Thermal Loss Analysis [4]
Thermal loss
Shuttle
Regeneration
heat
transfer
Cold finger
Q
'<
= 0.1861tK.cDcc X!.,xCT"
LCdrc
Qht< = NCitJ.t)(T;. - 7;)
Regeneration
temperature
swing
Q
Pressure drop
QN
Thermal compressor
.
<
=
N Ac(t.t)
.
=
"
L Ic,(t)Mc,(t)
= N(Mcmax - Mcmj(Cp);(T;. -
IS<
Q
- 7;)
O.l861tK,hDhcZ !.,x(T"
Lhdrh
Qht. = NCitJ.t)(T" - T;.)
Q
7;)
2Mr.(C p )r.
IS.
L tJ.Pc,(t)Xi(t)
=
.
QPdh = N Ah(tJ.t) L tJ.Ph,(t)Z,(t)
Table II. Theoretical Heat Balance Analysis for the
Mechanical Compressor Split-Cycle Cryogenic
Refrigerator (Miniature Design)
Pmax
P min
Pressure ratio
Temperatures, K
Compressor temperature
Ambient temperature
Pneumatic temperature
Cold end temperature
.
L Ih;!t)Mh,(t)
21tN 2(C p );(Mh ma• - Mhmj
AhrgKhr.{[21tN p(Cp)r,J/ K hr.} 1/2
Conduction
Pressure, psia
- T;.)
591.4
402.8
1.47
300
300
300
77
Resulting thermal predictions, W
Gross cooling
Shuttle loss
Heat transfer loss
Temperature swing loss
Total conduction loss
Pumping loss
Pressure drop loss
Net cooling
Net compressor work
2.44
0.07
0.61
0.34
0.25
0.00 (cold seal employed)
0.24
0.96
14.83
Thermal efficiency, %
Carnot efficiency
Gross efficiency·
Net efficiency·
34.53
16.48
6.46
• Excluding motor inefficiency, seal friction, and blowby.
Theoretical Analysis of Pneumatically Driven Split-Cycle Cryogenic Refrigerators
227
means there is a unique steady-state solution which depends not on the initial
compressor-displacer phasing but only upon the design.
Assuming constant temperatures for the gas spaces, the expression for the gross
cooling is given by
Qg
a
=
NAl~t) L Pj,iXi
i
(31)
The thermal losses associated with the cooler are given in Table I. The net
cooling is obtained by
(32)
The work extracted from the compressor is given by
Qm =
a
N(~t)Am L Pi~'
(33)
i
Finally, the gross coefficient of performance (COP) is given by
(34)
DESIGN CONSIDERATION
Two refrigerator designs were analyzed and will be presented: One employs a
mechanical compressor and one uses a thermal compressor. Both designs have
miniature remote cold fingers with identical free displacers, charge pressures, speeds,
pressure ratios, and cold finger separation distances. Figure 2 shows the displacer
position, thermodynamic pressure, and pneumatic pressure, all vs. time over one
cycle for the split-cycle cooler with the mechanical compressor. The displacer position,
thermodynamic pressure, and pneumatic pressure curves for the cooler with the
thermal compressor are very similar to the curves in Fig. 2 except that the pressure
values are higher. Figure 3 shows the computer prediction for the P-V curve for
the refrigerator with the mechanical compressor. Furthermore, the theoretical P- V
curve for the thermal compressor device was almost identical to the curve in Fig. 3.
To verify the theoretical predictions in Figs. 2 and 3, pressure and position transducers
were connected to an actual prototype of this cooler design. Photographs of the
resulting oscilloscope traces proved to be in close agreement with the theoretical
predictions. Finally, Tables II and III show the theoretical heat balance predictions
for the refrigerators driven by the mechanical and thermal compressors, respectively.
CONCLUSION
The analytical model for the split-cycle mechanical and split-cycle thermal
cryogenic refrigerators that has been developed is shown to be useful in analyzing
and understanding the basic characteristics of these cycles and in predicting the
operation and cooling of such devices theoretically before they have been built.
The model also shows that such a cooler can operate efficiently without valves and
without losses due to the nonrecovery of work absorbed by the working gas.
S. B. Hom and M. E. Lumpkin
228
Table III. Theoretical Heat Balance Analysis for a
Thermal Compressor Split-Cycle Cryogenic
Refrigerator (Miniature Design)
Pressure, psia
p_•.
765.7
514.6
1.49
Pml•
Pressure ratio
Temperatures, K
Compressor hot side temperature
Ambient temperature
Pneumatic temperature
Cold end temperature
977.6
320.0
305.0
77.0
Resulting thermal predictions, W
Compressor end
Shuttle loss
Heat transfer loss
Temperature swing loss
Total conduction loss
Net compressor work
Required heat input (total of above)
Cold finger end
Gross cooling
Shuttle loss
Heat transfer loss
Temperature swing loss
Total conduction loss
Pressure drop loss
Pumping loss
Net cooling
Net motor work
5.92
39.66
2.64
14.66
15.18
78.06
3.10
0.08
0.96
0.61
0.25
0.37
0.00 (cold seal employed)
0.82
2.74
%
Thermal Efficiency,
Carnot efficiency
Gross efficiency·
Net efficiency·
21.32
3.84
1.02
• Includes all compressor losses.
~ '-------~----------~-------------------' I~
510 /
~
.
~ I/
/---f.-~ . . . ,
I
I
i
I
410
·
I
m
~
!
~
i
i
Fig. 2. Theoretical normalized displacer position, pneumatic pressure,
and working gas pressure vs. time
for one cycle for the mechanical
compressor device.
I
m
~5OD
I'
/
i
"0
I
j
i
i
I
\
\
\
\
\.
\
- - _WPIIlSSIJIl
\
\\
~1
\,
\
\
\
U
\
\--------------f-d
\
\
\
i
\
\
..
\
OJ
\
\
\
'\
/
I
/
" /
'-4,_ __/
/
I
O~~
OA
I
OJ
02
0.1
~~
I
Theoretical Aulysis of Pneumatically Omen Split-Cycle Cryogenic Refrigeraton
229
~r----------------------------------.
Ii!
I
..
812
_L-~
Fig. 3. Theoretical P-V curve for the mechanical
compressor device.
______________________________
NOTATION
A
= area
COP = coefficient of performance
Cp
= specific heat
D
= diameter
F
= mass fraction
f
= fluid friction factor
G
= mass flux
I
= inefficiency
K
= thermal conductivity
L
= length
M
= mass
N
= speed
P
= pressure
R
= gas constant
= hydraulic radius
r
S
= frictional force
T
= temperature
t
= instantaneous time
V
= volume
= displacer pOSItion
X
Z
= compressor position
Greek Letters
r
= annular separation between displacer and cylinder
c/J
p
w
= phase angle between compressor and displacer motion
= density
= angular frequency
Subscripts
a
ex
c
ca
cc
=
=
=
=
=
ambient
total number of increments used in numerical solution
cold end of the cold finger
ambient end of the cold finger
cold cylinder
CGUlSWlEPWIIUM
~
s. B. Hom aod M. E. Lumpkin
230
cd
ch
g
gc
gh
gn
gw
ha
hc
hd
hh
hrg
m
n
rg
tot
= cold displacer
= charge [Pell = charge pressure]
= working fluid (gas)
= gas in the cold end
= gas in the hot end
= gas in the pneumatic volume
= gas in the working system
= ambient end of the hot finger (thermal compressor)
= hot cylinder
= hot displacer
= hot end of the hot finger (thermal compressor)
= hot regenerator
= iterative value over one cycle for one particular mass fraction
= connecting line or tube to cold finger
= compressor
= pneumatic part of cold finger
= cold regenerator
= total in closed working system
Void Volume Subscripts
va
vc
vha
vhh
vhrg
vm
vrg
=
=
=
=
=
vI
=
vtot
=
=
=
ambient end of the cold finger
cold end of the cold finger
ambient end of the hot finger (thermal compressor)
hot end of the hot finger (thermal compressor)
hot regenerator
mechanical compressor
cold regenerator
total temperature-weighted void volume in the working space
total void volume in the working space
REFERENCES
I. W. E. Gifford and E. M. Withjack, in: Advances in Cryogenic Engineering, Vol. 14, Plenum Press,
New York (1969), p. 361.
2. T. T. Rule and E. B. Qvale, in: Advances in Cryogenic Engineering, Vol. 14, Plenum Press, New York
(1969), p. 346.
3. F. N. Magee and R. D. Doering, "VuiIleumier Cycle Cryogenic Refrigerator Development," Hughes
Aircraft Company Tech. Rept. AFFDL-TR-68-2, Wright Patterson Air Force Base, Ohio (August
1968), p. 28.
4. B. Shelpuk and M. Crouthamel, "VM Thermal Actuation and Burner Investigation," USAECOM
Report, Ft. Belvoir, Virginia (January 1972), p. 13.
F-5
REFRIGERATION FOR THE CULHAM
SUPERCONDUCTING LEVITRON
D. N. Cornish and R. E. Bradford
Culham Laboratory
Abingdon, England
and
A. J. Steel
British Oxygen Company Limited
London, England
INTRODUCTION
The Culham Superconducting Levitron has a number of unusual features which
will be of interest to those working in the area of cryogenics and superconductivity.
It is, as far as is known, the first machine of appreciable size in which the superconducting coils have been specifically designed to operate in gaseous helium. The
use of fine-filament, twisted niobium titanium conductor, vacuum-impregnated in
suitable epoxy resin results in coil performances very close to that measured on
short samples of the wire.
Separate circuits are required to provide 4.4 K gas at 1.2 atm and 3 K supercritical fluid at 2.5 atm. The design of the Levitron and the refrigeration plant has
been coordinated so that (1) a standard cold box could be used, (2) the Levitron itself
is as simple as possible from the viewpoint of maintenance and operation, and (3) all
the nonstandard, very low-temperature equipment is separately housed in an auxiliary
cryostat.
DESCRIPTION OF LEVITRON COIL SYSTEM
e-
The main purpose of this presentation is to discuss the cooling aspects of the
S ] for further
Levitron and the reader is therefore referred to existing literature
information on this and similar systems.
The heart of the equipment is a room-temperature vacuum chamber in which
a short-circuited superconducting winding, sealed into a steel case, 9 cm minor and
60 cm major diameter, can carry an induced current of up to 0.5 MA-turns while
being supported only by magnetic forces. The ring is cooled to its operating temperature by conduction from four pairs of cold clamps which can be brought into
contact with it under high pressure and which can then be withdrawn, leaving the
ring floating in the vacuum space.
The possibility of cold-welding occurring between the clamps and ring, and the
rate at which cooling could be obtained between the surfaces in contact in a good
vacuum, were investigated on a model. The stainless steel case of the ring is copper231
232
D. N. Cornish, R. E. Bradford, and A. J. Steel
Fig. 1. Levitron ring ready for assembly in machine.
plated to a thickness of about 0.038 mm to improve the thermal conductivity around
the ring, and it is then gold-flashed to prevent oxidation and maintain a good
emissivity. The annular space between the case and the winding is filled with helium
gas at room temperature and 150 atm to provide good thermal conductivity between
the case and winding, and added thermal capacity at the normal operating temperature. The top and bottom surfaces of the case are machined flat, as shown in. Fig. 1,
to provide a good bearing face for the cold clamps, which are fabricated from copper
and faced with indium pads. The model was built to simulate this arrangement, and
the effects of clamp pressure, thermal and mechanical cycling, and thermal contact
resistance were investigated. The tests showed that a force of 680 kg on a 20 cm 2
surface was adequate and could be applied many times without the surfaces becoming
damaged. Figure 2 shows the heat transfer rate obtained on the model plotted against
ring temperature, the cold clamp being maintained at 3.5 K. Using these data, it was
estimated that it would take less than 2 min to recool the ring from 5 to 4 K.
"
~~--=-~----~------~
Fig. 2. Heat transfer on model cold clamp. Force on clamps
maintained at 3.5 K is 680 kg over a 20 cm l surface.
Refrigeration for the Culham Superconductiog Levitron
233
In practice, cooling of the ring is close to that predicted. The ring has not yet
been energized above 300 kA in the Levitron, although during the proving tests, it
was run at 550 kA in a separate cryostat. It has been quenched many times at 300 kA,
when the mean temperature is estimated to rise to 40 K. Recooling from this temperature takes about 15 min.
With the ring and vertical field coils energized, a high-temperature plasma is
created in the vacuum space around the ring, and the purpose of the machine is to
study the behavior of the plasma in the presence of the magnetic field surrounding
the ring. Heat from the plasma and radiation from the surrounding surfaces will
cause the ring temperature to rise, and the levitation time before recooling is necessary
is determined by the thermal capacity of the ring.
The critical temperature of the ring at full rated current is 6 K. To allow some
safety margin and a reasonable levitation time, it was decided to cool the clamps to
3.5 K, using supercooled helium at 2! atm to ensure single-phase operation under
all conditions. The levitation time at full current, during which time the ring temperature will rise from 4 to 5 K, is estimated to be about 1 hr. This is perfectly acceptable since the ring can be recooled in the 2 min mentioned above while the currents
remain unaltered.
The vertical field coils, which are used to induce the ring current and also to
provide the levitating field, are shown schematically in Fig. 3. For the sake of clarity,
only the coils and the main low-temperature structure have been shown in this
diagram. In fact, there is very little free space, and that occupied by the coils must
be kept to the absolute minimum. A high current density in the winding is vital,
and space for helium reservoir and expansion volume necessary to limit the pressure
in the event of a quench must be minimized. It was considered that the best way of
meeting these conditions would be to use windings similar to that in the ringintrinsically stable Nb-Ti, vacuum-impregnated in epoxy resin for strength, elimination of interturn movement, and cooling with helium gas. The use of gas instead of
the more usual liquid has the advantage of requiring no reservoir in the coil cases.
Thus, instead of liquid level controllers in each coil, only one is required in the main
Fig. 3. Illustration of central chamber of superconducting Levitron.
234
D. N. Cornish, R. E. Bradford, and A. J. Steel
bath in the auxiliary cryostat. Also, the helium distribution is very simple since
the gas flows through all the coils in series. Finally, in the event of a quench, the
pressure rise in the coil case is very small.
LOW-TEMPERATURE SYSTEM
The low-temperature system comprises three separate units, the cold box,
auxiliary cryostat, and Levitron. The cold box houses the standard refrigerator items
and produces cold gas which is fed into the auxiliary cryostat through a vacuuminsulated transfer line. All the special low-temperature items such as liquid sumps,
level controllers, and low-temperature valves are assembled together in the auxiliary
cryostat. Thus, any item which might require maintenance is removed from the
Levitron itself, and space is not required in the Levitron for any auxiliary cryogenic
equipment. This not only eased the design of the Levitron main chamber, but also
enabled a relatively simple separate unit to be designed and manufactured completely
independent of the Levitron. A six-core vacuum-insulated transfer line from the
auxiliary cryostat to the Levitron carries the coolant to and from the three separately
controlled circuits, the vertical field coils, the cold clamps, and the cryoplates.
REFRIGERATION REQUIREMENTS
The normal running refrigeration loads were estimated to be:
1.
2.
3.
4.
8 W at less than 4.5 K using gas at 1.2 atm for the vertical field coils;
0.15 g/sec of helium gas at less than 4.5 K for current lead cooling at 1.2 atm;
10 W at less than 3.5 K using supercritical fluid at 2.5 atm for the cold clamps;
1.5 W at less than 3.2 K using supercritical fluid at 2.5 atm for cryopumping
in the Levitron chamber;
5. 2 W at 4.5 K to compensate for heat inleak down the blowoffreliefvalve lines.
To allow for the actual losses being higher than estimated and to ease the
operation and control of the cryogenic system, the refrigeration plant was designed
for loads 50 % higher than the above figures. System losses, such as those associated
with control valves, transfer lines, and liquid sumps, comprised an additional load.
The masses to be cooled by the refrigerator were estimated to be 1054 kg for the
field coil and support structure, 126 kg for the cold clamps and levitated ring, and
5 kg for the cryoplates.
CRYOGENIC DESIGN OF COILS
The coil system is arranged so that the gas enters each coil at a point diametrically
opposite to the exit point and is thus split into two circumferential parallel paths.
There are cooling channels surrounding each coil and a manifold at the top of the
coil case ensures that the gas must flow through these channels from the inner to the
outer diameter.
The maximum pressure drop through all the coils in series occurs with the
initiation of the cooling process at 300 K when it is estimated to be 0.5 atm for a
mass flow of 15 g/sec. This decreases to 0.011 atm under normal operating conditions.
Intercoil forces are absorbed by a set of four substantial support arms fabricated
from stainless steel. Temperature gradients in this structure during cooldown are
reduced by copper-plating the arms to a thickness of approximately 0.5 mm.
235
Refrigeratioo for the Culham Superconducting Levitroo
A number of 0.025-mm-thick washers are required to adjust the total height of
the coil assembly inside the case to the required value. These washers are cut from
sheet copper and are positioned at the coil end adjacent to the support structure so
that the heat conducted into the coils at these points is spread out and prevents the
formation of local hot spots.
The support structure is mounted on a strong ring which also serves as the
liquid nitrogen reservoir (Fig. 3) for the radiation shields between the coils and the
room-temperature vacuum tank. There is an independent liquid nitrogen dispensing
system for supplying this and other parts of the machine.
REFRIGERATION SYSTEM
The refrigeration system employed is essentially a standard machine similar to
that used on the Fawley motor project [6], and is based on a Claude cycle with a
single expansion turbine and liquid nitrogen precooling. Reference to the ftowsheet
(Fig. 4) shows that helium is compressed in the two-stage, dry-lubricated compressor
to 10 atm and, after cooling and filtration, passes to the cold box in which successive
cooling occurs in the aluminum plate-fin heat exchangers. The final product, leaving
the cold box at about 5 K and 9.5 atm, is split in the auxiliary cryostat into two streams
for the clamps and cryoplate circuits. After partial expansion to 2.5 atm, each stream
is further cooled to 3 K by liquid helium baths at 4.4 and 2.9 K.
These supercritical streams are passed to the Levitron through the six-core line,
and after absorbing heat inside the Levitron, return to the auxiliary cryostat. Expansion in the two pressure control valves results in a two-phase mixture, and the liquid
Current leads
(0.239/S)
Superconducting COlts
(12 watts)
~~~~--wt.--,Ijj
"=-.
Cryoplate
-----""-----+--+t.....
13W":'_'_ _
Levitron
Auxiliary cryostat
Refrigerator coldbox
Reference point
A
Flow
C
Clamps
D
G
H
37.4
18
6.7
12.7
6.7
12.7
19.4
19.17
2.12
1.90
36.95
Temperature (K)
300
18
3.0
3.0
3.2
3.S
4.4
4.5
295
300
295
Pressure (atma)
10
9.S
2.S
2.S
2.S
2.S
1.2
·1.2
0.13
1.23
102
(g/s)
Fig. 4. Simplified ftowsheet of the helium cryogenic system. Key to components:
(1-7) heat exchangers; (8) nitrogen boiler; (9-10) adsorbers; (11) expansion turbine;
(12) compressor; (13) vacuum pump; (14) 4.4 K sump; (15) 2.9 K sump.
D. N. Cornish, R. E. 81'11df'onl, and A. J. Steel
is collected in the 4.4 K bath. The gaseous fraction, together with boiloff from the
4.4 K bath, is returned to the Levitron to cool the magnet coils and current leads.
The cold gas from the 2.9 K bath is heat-exchanged against a return stream in
an auxiliary exchanger, and is pumped by a noncontaminating vacuum pump.
Peripheral equipment includes a complete automatic recovery system, a small drier
for initial system purging, and adsorbers at 80 and 12 K inside the cold box for longterm operation. The recovery tank is sufficiently large to store the helium supply
from the entire plant, thus enabling it to operate as a sealed system. The impurity
level should therefore be low, and any losses are made up with grade A helium. If
required, the refrigerator is capable of performing the alternative duty of liquefying
helium which can be syphoned from the 4.4 K sump, where it is produced at the rate
of 35 liters/hr.
Plant Layout
The refrigeration equipment is concentrated in three areas: the compressor
house; the Levitron area; and the control cubicle. Due to restrictions in access of
personnel to the immediate Levitron area, the control cubicle is sited about 20 m
from the cold box, while the compressor is in a separate building 50 m away.
Control
Plant control is completely automatic after initial cooldown, with automatic
control loops being provided on the following items:
1.
2.
3.
4.
Turbine inlet temperature, achieved by throttling of the inlet valve.
Liquid level control in the two helium baths.
Pressure control on high- and low-pressure sides of the plant.
Automatic recovery/makeup from the recovery tank by means of special
helium servo valves.
Turbine stop/start is carried out by push button from the control cubicle while
the main compressor and helium vacuum pump can only be stopped, starting being
provided at the machines themselves.
Machinery
The compressor is a two-stage V-machine made by Broom Wade, and became
available as surplus from an earlier project. It operates at the relatively slow speed of
300 rpm which, from the point of view of ring wear, is advantageous. Piston rings and
seals are of carbon/PTFE composite. The throughput is 40 g/sec (500 scfm) and the
compressor is driven by a 125-Hp motor.
The pump used to reduce the pressure of the 2.9 K sump is a two-stage Northey
nonlubricated rotary piston machine. Special gas-packed seals are fitted on the shaft
to prevent the ingress of air. It is capable of pumping 2.5 g/sec of helium with an inlet
pressure of 0.13 atm and is driven by a 30-Hp motor.
The expansion turbine is a BOC helium-lubricated machine running at 6000 rps
with an isentropic efficiency of over 65 %, protected with the standard BOC system.
COMMISSIONING AND OPERATING EXPERIENCE
The refrigerator has now operated for nearly 7000 hr, initially against dummy
electrical heaters, and lately in conjunction with the Levitron. Some of the more
noteworthy experiences can be summarized as follows:
Refrigeration for the Culham Superconclucting Levitron
237
1. Liquid helium was produced on the first attempt. Similarly, the first cooldown
2.
3.
4.
5.
6.
of the Levitron was carried right through to 4.4 K although the initial intention
was to cool to 80 K only.
No turbine failures have occurred in spite of severe air contamination on two
occasions, once due to a pipe coupling on the compressor suction manifold
becoming disconnected, and second, due to shortage of feed gas which
resulted in subatmospheric suction and air inleakage.
Due to an oversight, the plant was left running over a weekend with the turbine
inlet valve control coupled to the liquid level in the 4.4 K sump via a highgain loop. This gave rise to a short on-off cycle to the turbine-an estimated
3000 start-stops without any damage. This is more than a plant normally
performs during its lifetime, and illustrates the reliability of the hydrostatic
bearings.
The compressor piston rings and sealing glands were examined after about
3000 hr, when there was a convenient break in the program. Wear was not
excessive and only the glands on the high-pressure piston rod were replaced.
The compressor was due for a full service after 5000 hr, but the physics
demands on the Levitron have been such that this has not yet been carried out.
The machine is due for a shutdown in the near future for the addition of more
windings, and it is hoped that this service can be delayed until then. The
helium leak rate has increased considerably, but this is to be expected under
these conditions.
It has been possible to carry out a substantial part ofthe Levitron commissioning program without liquid in the 3 K sump. Under these conditions, the field
coils operate at their normal temperature of 4.5 K, but the clamp and cryoplate circuits drift up to 5 K. These conditions are satisfactory for "floating"
the ring at 300 kA and for initial plasma physics experiments to be carried out.
The low-pressure, 3 K system has therefore only been in operation for 1000 hr,
during which time there have been no serious problems.
It was estimated that the refrigerator should be capable of cooling the Levitron
from 300 to 4.5 K in less than 6 hr. The cooling, however, is throttled so that
the maximum temperature difference across the coil system does not exceed
30 K, which has been rather arbitrarily chosen to prevent damage due to
differential contraction. Under these conditions, it should be possible to have
the plant ready for operation two days after cooldown is initiated. In fact,
the plant is left unattended overnight, even during the cooldown stage, and
the normal allowance for cooling is three days. However, this presents no
great hardship since the plant normally remains cold for many weeks at a
time.
ACKNOWLEDGMENTS
The authors wish to acknowledge the assistance of all those associated with this project, in particular,
J. W. Mart of BOC and J. Alsford of this laboratory.
REFERENCES
I. D. N. Cornish, in: Proceedings Intern. Conference on Magnet Technology, Deutches Elektronen-
Synchroton, Hamburg, Germany (1970), p. 847.
2. C. E. Taylor, T. J. Duffy, T. L. Rossow, D. R. Branum, J. H. Sexton, and R. L. Leder, "The Livermore
Superconducting Levitron," Lawrence Radiation Laboratory, UCRL Rept. 73076, Livermore, California (1971).
238
D. N. Cornish, R. E. Bradford, and A. J. Steel
3. D. N. Cornish, in: Proceedings Symposium on Electro-Magnetic Suspension, Southampton University,
United Kingdom (1971), p. BI.
4. J. File and P. Bonanos, "Operation of the Levitated Superconducting Ring in the Princeton Floating
Multipole Machines," in: Proceedings 1972 Applied Superconductivity Conference, Annapolis, Maryland (\972), p. 354.
5. D. N. Cornish, Cryogenics, 12:89 (1972).
6. F. Tinlin and J. S. H. Ross, in: Proceedings Conference on Low Temperatures and Electric Power,
IIR, Pergamon Press, London (1970), p. 277.
DISCUSSION
Question by R. R. Conte, M.V.E. Cr'yogenics: Is the 30 K temperature difference from inlet to outlet
gas flow noted in this presentation constant from room temperature to the lowest temperature?
Answer by author: The 30 K temperature difference is between the aluminum shell inside the toroidal
field coils and the aluminum plate that supports the field coils. This difference is maintained constant from
room temperature to about 120 K in order to keep the shell hotter and thereby prevent mechanical
interference between the shell and the toroidal field coils.
F-6
HEAT LOAD DUE TO ORTHO-PARA
CONVERSION IN A CLOSED-LOOP
HYDROGEN REFRIGERATOR
R. L. Pubentz and D. A. VanGundy
Argonne National Laboratory
Argonne, Illinois
INTRODUCTION
The hydrogen refrigerator, that is part of the 12-ft bubble chamber complex at
the Argonne National Laboratory, is a closed-loop system. When the chamber is in
operation, the refrigerator supplies liquid hydrogen to cooling loops for temperature
control of the chamber liquid. This refrigerator takes gas from a compressor through
cold boxes to a Joule-Thomson valve. Liquid and saturated gas are routed to a
12,OOO-gal dewar, where the liquid is retained. This liquid goes to the chamber cooling
loops, where it is vaporized. The gas goes back through the cold boxes, to the suction of
the compressor, thereby completing the loop.
In the Fall of 1972, after a seven-day period of operation, the refrigerator was
shut down and the dewar pressure was observed to rise rapidly, requiring venting.
The usual checks of dewar vacuum, liquid level gauge calibration, and possible
sources of warm gas entering the dewar showed nothing wrong.
INVESTIGATION
The dewar, containing 7400 gal, was sampled five days after system shutdown.
The sample had an orthohydrogen concentration of 50 % as determined in a thermal
conductivity cell. A secondary storage dewar that had been filled with parahydrogen
from a tanker was found to contain in excess of 98 % parahydrogen.
The observed rate of pressure rise in the dewar indicated that the orthohydrogen
concentration in the dewar liquid was indeed approximately 50 %. The heat evolved
from the ortho-para conversion was about 1 kW, with 800 W being stored in the
liquid and about 200 W used to vaporize liquid resulting in the observed pressure
rise [1].
In order to verify this ortho-para conversion within the system, the dewar was
filled with parahydrogen just before the next period of operation late in 1972. After
three days of refrigerator operation, the sample analysis showed 60 % orthohydrogen
in the dewar and 75 % orthohydrogen (normal hydrogen) in the suction line at the
compressor.
The balance of gas flows, measured by orifice flowmeters in the refrigerator circuits, shows an internal heat load on the dewar that varies with the time. Prior to the
next running period, the dewar was again filled with parahydrogen. Data were taken
239
240
R. L. Pubentz I11III D. A. VanGundy
o
TIME, dOl'
Fig. 1. Dewar heat load as a function of time.
for five days to determine the amount of excess gas returning from the dewar to the
cold box. This is shown in Fig. 1 as the dewar internal heat load vs. time.
The last point corresponds to a heat load caused by the conversion of75 %orthohydrogen (normal hydrogen) in the dewar at a level of approximately 9500 gal.
CONCLUSIONS
The existence of an internal dewar heat load has been verified during the period
following refrigerator shutdown and also during refrigerator operation. The system
has a carbon steel line about 210 ft long from the cold box to the compressor, and
this line is known to be internally coated with rust. The initial conclusion was that
conversion to normal hydrogen was taking place in this line. Another conclusion,
just as valid, is that conversion is taking place in the compressor and in the deoxo
downstream of the compressor. Conversion to normal hydrogen has been cited as
commonly taking place in compressors [2]. The situation of seeing normal hydrogen
at the compressor suction may be the result of feeding normal hydrogen to the
refrigerator with its return back to the compressor.
The total heat load caused by the conversion of normal liquid hydrogen to parahydrogen is a function of the quantity of liquid in the dewar. Reduction of this heat
load on the refrigerator can be accomplished by simply operating with a minimum of
liquid in the dewar.
The design of the 12-ft bubble chamber refrigerator was based upon chamber
cooldown at a maximum load of 20 kW, while the normal chamber heat load is
about 7 kW. Ortho-para conversion heat loads have been experienced as high as
4 kW when the dewar is nearly full. The magnitude ofthis additional heat load can be
such that it needs to be considered in the design of closed-loop systems.
REFERENCES
I. R. F. Barron, Cryogenic Systems, McGraw-Hill Book Company, New York (1966), p. 62.
2. R. B. Scott, W. H. Denton, and C. M. Nicholls, Technology and Uses of Liquid Hydrogen, Macmillan
Company, New York (1964), p. 60.
£-7
PROTOTYPE TESTS ON A 200-W
FORCED CONVECTION LIQUID
HYDROGEN/DEUTERIUM TARGET*
K. D. Williamson, Jr., J. E. Simmons, F. J. Edeskuty, J. H. Fretwell,
J. T. Martin, and H. Ficht
Los Alamos Scientific Laboratory, University of California
Los Alamos, New Mexico
INTRODUCTION
A forced convection target system is being developed for use at the Clinton P.
Anderson Meson Physics Facility (LAMPF) which will produce neutrons by bombarding liquid deuterium with a proton beam via the process p + d -+ n + 2p. Incident
energies will vary in the range from 300 to 800 MeV. This neutron source is favorable
from two points of view: (1) At an angle of 0°, the emitted neutrons have a small
energy spread and good separation from lower-energy neutrons associated with
meson production, and (2) at an angle near 26°, the neutrons are emitted with a spin
polarization averaging 30 % and with a degraded but usable energy spectrum. To
obtain an 800-MeV neutron flux of 107/cm 2 -sec at 8 m from the source will require
an 84-pA proton beam passing through 15 cm of liquid deuterium, which implies
500 W dissipation in the deuterium. This value of beam dissipation has been used as a
design goal in the preliminary planning. For various reasons this dissipation value
has now been reduced and a 150-W beam dissipation capability at negligible density
variation is the present goal.
In the past, relatively few accelerators operating at medium and high energy had
sufficiently intense external beams to require special consideration for heat removal
from liquid hydrogen targets. An exception is the SLAC high-energy electron linear
accelerator, where external beams of 15 pA are accelerated. Since the area of the
SLAC beam is small, approximately 0.25 cm 2 , high power density is created in a
target substance, and for cryogenic liquid hydrogen targets, reduction of liquid
density occurs because of vaporization in the path of the beam. Two approaches
have been used at SLAC to circumvent this problem Anderson [1] described a target
in which enhancement of natural convection was used to remove heat from the interaction region. Subsequently, Bell et al. [2] described a forced convection target in
which liquid hydrogen is driven through a closed loop by a fan. Part of the loop
comprised the target cell and part was in thermal contact with a liquid hydrogen
reservoir where the accumulated heat was extracted. With this system, density
variations were reduced to less than 1 %.
Kaiser
has made calculations ofthermal conduction in liquid hydrogen being
driven by heat input from 6O-MeV electron beams. He investigated the temperature
e]
* Work performed under the auspices of the U. S. Atomic Energy Commission.
241
242 K. D. W~, Jr., J. E. Sm.-, F. J. Edeskaty, J. H. FretweD, J. T. Martin, ud H. Fidlt
rise in cylindrical geometry in which the outer radius is that of the target cell, maintained at a fixed reservoir temperature. Effects of convection were not considered.
He found that relatively small average beams of approximately 1 J1.A could cause
lOoC temperature increases in the interaction cylinder, for reasonable geometries.
An implied conclusion of this work is that thermal conduction alone is not adequate to
remove the heat accumulated in liquid hydrogen by relativistic electron beams
greater than 1 J1.A in intensity.
In this presentation, the characteristics of a prototype target designed to operate
with a heat input in the neighborhood of 100 to 200 Ware discussed. Data are
presented on the operating characteristics ofthe unit using liquid hydrogen with heat
accumulation from both electrical heaters and from an electron beam of 25 MeV
which has a time structure similar to the LAMPF beam. In the electron beam experiments, energy depositions of approximately 25 W/cm 3 were achieved. Using data
obtained on these tests, the final target system was designed and will be discussed
briefly.
The basic objectives of the prototype flow loop tests were the following:
1. Confirm the design ofthe heat exchanger between the liquid hydrogen and the
gaseous helium heat sink.
2. Provide operational data on the pumps used for liquid hydrogen circulation.
3. Study the effect of energy deposition by an electron beam, with emphasis on the
question of vaporization at high power densities.
EQUIPMENT AND INSTRUMENTATION
The major components of the system are the compressors, purifier, gas holder,
Collins-type refrigerator, vacuum-jacketed transfer lines between the flow loop and
refrigerator, the prototype target loop, and associated beam diagnostic equipment.
Figure 1 provides a schematic plan view of the parts of the system in the beam line,
including certain beam measuring and shaping components, the target cell through
which the electron beam passes, the flow loop with heat exchanger, the graphite
beam scraper, which measures beam scattered from the windows and contents of the
target cell, and the graphite beam stop.
The 200-W helium refrigerator used was a Collins unit model CHC-14. The
incoming warm, high-pressure helium is first progressively cooled within the main
heat exchanger by the cooler outgoing helium gas and by liquid nitrogen used in a
precooling coil. It is then expanded to its final temperature (14 K) in two reciprocating expansion engines. This cold gas is piped to the load, where it picks up heat and
is then returned to the refrigerator. After going through several heat exchangers, the
warm, low-pressure gas is compressed and the cycle repeated.
A representation of the prototype flow loop is shown in Fig. 2. The major components of this system are the liquid hydrogen-gaseous helium heat exchanger, the
liquid hydrogen blowers, a target cell which may be replaced by a venturi test section
for flow calibration, and a series of three heaters to provide a load for testing the
refrigerator when not in use at the Electron Prototype Accelerator (EPA) and for
balancing the beam load with the refrigeration capacity. Instrumentation consisted
of a microphone to monitor pump operation, pressure gauges, carbon resistor temperature sensors, and carbon resistor level sensors. The stainless steel flow loop held
approximately 13 liters of liquid hydrogen-this volume being partly a result of the
fact that the pumps were fans capable of only a 12 mm Hg pressure differential. The
system was sized to keep within this pressure drop limitation.
Prototype Tests on a 200-W Forced Convection Liquid Hydrogen/Deuterium Target
243
To rQel Cell with
Wlndow$
I Alum,(1um
/ GfOOl'l llt 8t'om SlOP
GUlj)t'lIlf
Scrope r
H.O' ....
Beom
M2 Flow In
HeGl E .chonQe r
Fig. !. Schematic diagram of the prototype target
system, as assembled at the Electron Prototype
Accelerator (EPA).
Annulu s
He f"- w
thru Tubes
Flange
Au. mbly
f
.
.::;
0
r
Stoekec3
Vo lume
------+!,,-!.}
,I
Flo w
Fig. 2. Schematic of the prototype flow loop
(elevation) showing heat exchanger, liquid hydrogen blowers, target section with aluminum window, and heaters.
The pumps used were submersible 3-in. blowers* and utilize three-phase motors
that operate from 10 to 40 V depending on torque requirements. Speeds vary from 500
to 3500 rpm. The ball bearings are nonlubricated with a nonmetallic retainer.
Two cell sections were used. The first was a venturi used to calibrate the pumps.
The second was the thin aluminum window test section. The windows were hydroformedt from flat 1100 aluminum sheets and each had a final thickness of 0.0076
cm. The window apertures were rectangular in shape, being 13 cm long and 2.5 cm
wide, and were dish-shaped to provide a relatively smooth flow passage for the liquid
hydrogen moving through the cell. The seal between the window and flow channel was
made by compressing a 0.16-cm-diameter Cerroseal (low-temperature indium solder)
O-ring. A new O-ring was required each time the window was replaced. Self-energized,
Teflon-coated O-rings were used to attach the cell sections to the flow loop. These
rings leaked on several occasions and will not be recommended for use on the final
target system.
In the countercurrent liquid hydrogen-gaseous helium heat exchanger, the cold
helium gas passes through twenty-two l-cm ID copper tubes in parallel while the
liquid hydrogen flows in the annulus around the outside of the tubes. This annulus
is bounded by the ID ofthe flow loop and a solid cylindrical core. Also attached to the
heat exchanger section were three 300-W heaters which were used to simulate beam
heat input during the initial testing. The calculated heat transfer coefficients were
0.21 W/cm 2 -K on the liquid hydrogen side for a flowrate of 38 kg/min and 0.014
W/cm 2 -K on the helium side for a flowrate of 0.015 kg/min. The overall heat transfer
* Type VAX-3 blowers, Globe Industries, Dayton, Ohio.
t Units made by the Metal Fabrication Section ofCMB-6, Los Alamos Scientific Laboratory, Los Alamos,
New Mexico.
244 K. D. WiUiamIon, Jr., J. E. Simmons, F. J. Edeskuty, J. H. FretweU, J. T. Martin, and H. Ficbt
coefficient U was 0.013 W/cm 2 -K. With a log mean temperature difference of 3 K, this
should provide a heat removal capability of 260 W. It should be noted that the controlling coefficient is on the helium side. With a larger refrigerator and compressor,
the heat removal capability of the heat exchanger should improve as a result of
increased helium flow until the latter coefficient becomes comparable to that on the
liquid hydrogen side. The heat exchanger was designed so that it could be removed
easily from the loop without disassembly of the loop.
The gas handling system provided the capability of filling the loop directly
with liquid hydrogen or by gas condensation on the heat exchanger. Safety relief
valves were installed both on the flow loop (25 psig) and vacuum jacket (2 psig). Most
of the gas handling valves were l00-V, remotely operated solenoid units. The vent
lines were sized to minimize the pressure rise in the system in case of a vacuum failure
(3-in.ID).
Since 7 hr was required to fill the loop with condensed hydrogen, this filling
technique was not used after the initial runs. In general, the loop was purged and
maintained under positive gaseous helium pressure, the refrigerator started, and the
cooldown allowed to proceed until liquid nitrogen temperatures were attained. At
this point, the gaseous helium was replaced with liquid hydrogen from an external
storage vessel. As the liquid hydrogen began to subcool, the electrical heaters were
adjusted to maintain the desired degree of subcooling.
Throughout this experimental program, difficulty was experienced with the
operation of the refrigerator. Premature shutdowns resulted from repeated seizures
of the pistons in the expansion engines. Installation of a liquid hydrogen trap on the
gaseous helium line to remove neon and other possible contaminates did not solve the
problem.
SYSTEM PERFORMANCE
The operating characteristics of the blowers are shown in Fig. 3 for both a
single unit and two units in series. To achieve maximum flow in the series configuration, a vane straightening section was placed between the blowers. The fluid velocity
required to move a hydrogen particle across the beam path between 8.3-msec beam
pulses is 100 cm/sec. It is desirable to operate at velocities in excess of this to ensure
that new fluid is in the path of the beam for each pulse. The gaseous helium, liquid
hydrogen heat exchanger performed as expected. At a maximum experimental
total heat removal rate of 175 W, no degradation of performance was noted. With
175 W being removed by the heat exchanger, the liquid hydrogen temperature was
19 K. This temperature value decreased to 18.5 K at a 135-W removal rate. Depending
upon the helium overpressure applied to the top of the loop, several degrees of subcooling could be achieved.
LH2 Velocily thru Target. m/sec
0.5
1.0
1.5
2.0
.
'"
~2
"0
>
~IO
~
6
LH2
10
12
8
Flow. kg Imin
14
16
18
Fig. 3. Pumping capability of the Globe VAX-3
blowers, showing RMS voltage across each unit
vs. flow, for one pump only and for two pumps in
series.
Prototype Tests on a 200-W Forced Convection Liquid HydrogenfDeuterium Target
245
The 0.0076-cm aluminum windows were rupture-tested to determine the pressure
rise in the system following window failure. It was necessary to provide a puncture
unit since the windows maintained their integrity at pressures in excess of 160 psig.
Following rupture ofthe window, the vaporization of liquid hydrogen caused a peak
of 30 psig in the containment vessel approximately 1 sec after the break. The containment vessel had a volume of 530 liters and was equipped with a I-psi relief valve.
TESTS IN THE 25-MEV ELECTRON BEAM
The Electron Prototype Accelerator (EPA) was built at Los Alamos to test the
side-coupled linear accelerator cavity design and to provide a target capability at
high beam power. The time structure of the EPA electron beam was the same as the
LAMPF proton beam, namely 500 x 10- 6 sec macro pulse length at 120 Hz repetition rate and a superimposed micro pulse structure at 200 MHz radio frequency.
Beam intensity was likewise comparable, ranging up to 1 mA average.
Figure 1 shows a schematic of the installation of the EPA. The electron beam
enters a collimator from the right-hand side of the figure, directly from the accelerator
without focusing. The graphite collimator has a diameter of 0.9 cm, its purpose being
to confine the beam to a fixed region of the target. Two beam measurement devices are
placed after the collimator, namely an x-y wire scanner for profile measurement and a
toroid beam transformer for measurement of the peak current in the macropulse.
The beam then enters the cryostat vacuum enclosure and passes through the target
cell. A fraction of the beam will be scattered by the windows and hydrogen of the
cell and will be collected on an annular graphite electrode called the beam scraper.
The electrical signal from this electrode may readily be measured on an oscilloscope,
and has the well-known R-C rise and decay time characteristic of a rectangular
current pulse charging a cable capacitance with resistive termination. That part of the
beam not collected on the beam scraper is stopped in the final graphite beam stop.
Both scraper and stop were water-cooled.
The system was aligned by optical observation of burn spots made by the beam
and by observations on beam tube centers. The center position of the x-y scanning
wires was observed and calibrated. The shape of the beam was that of an elongated
oval with the long axis in a vertical position. The area of the beam was approximately
0.5 cm 2 at a section corresponding to one-tenth of maximum intensity.
The electron beam releases heat in the target cell by energy loss in the two
0.0076-cm aluminum windows and in 5.72 em of liquid hydrogen. Using standard
tables [4] for collision energy losses at 25-MeV incident energy, we have llE = 1.94
MeV in liquid hydrogen and llE = 0.073 MeV in both foils, for a total energy loss of
2.013 MeV. Thus, 1 J1.A average beam deposits 2.01 W into the target cell.
In passing through the target, a fraction of the beam is scattered by multiple
Coulomb interactions. A major part of the scattering is due to the mass of hydrogen,
which represents a signal related to the average density of hydrogen in the beam path.
By use of an approximation formula [51, the multiple scattering can be estimated as
follows: ORMS = 2.38 0 from 5.72 cm liquid hydrogen, and ORMS = 1.200 from 0.Q15 cm
aluminum, for 25-MeV electrons incident in both cases. We define the angle 0 1 to be
the angle subtended by the inside radius of the annular beam scraper, which, for the
geometry used, was 0 1 = 4.19 The probability of multiple scattering at or beyond the
angle 01 is given by exp( -0/ /OiMS)' This probability gives the fraction f. of incident
beam scattered to 01 or greater. A measurement of f. gives 0iMs, which can be related
to the average liquid hydrogen density by the multiple scattering formula.
0
•
24(i
K. D. Williamson, Jr., J. E. Simmons, F. J. Edeskuty, J. H. FretweU, J. T.
Mama, and H. Ficht
0.20r----,r-----,---,----r----,
0.10
~
ODe
~ 0.06
8
en
c: 0.04
.!!
u
D
~
0.02
•
. / TOrliJet empty measurement,
,/
Fig. 4. Calculated values of scattered fraction of
0.01 0'-------'0.'-2---'0.4---0...1..6---0-'-.e---'1.0
Fraction of LH2
~~~r~:o~:~rf t~;e~el~::u~:!~~~:not~:~~~t
empty scattering fractions.
Densily
Figure 4 shows the calculated values of the fraction of the beam scattered !.
as a function of the fraction of normal liquid hydrogen density in the path of the
beam. The curve includes the effects of the windows, as measured. Also shown are
three measured values of!. for the empty target cell. The curve indicates two things.
First, there is a reasonable sensitivity of!. relative to variations in the average liquid
hydrogen density. Second, the scattering due to the liquid hydrogen alone is an order
of magnitude greater than that due to the target windows. It will be seen below that the
calculations of!. for full liquid hydrogen density are in reasonable accord with lowbeam measurements. The calculated value of!. for 0.015 cm aluminum, namely for
an empty target, is a very small number, ofthe order of 10- 5 . The measured values
for this quantity are much higher, namely of the order of 10- 2 , as shown in Fig. 4.
The reasons for this are unclear, but are related to the presence of a paint film on the
windows which was used to locate the passage of the electron beam. This situation
does not prejudice the scattering measurements as a result of the large hydrogen
effect. In any case, the calculations of multiple scattering given are not meant to be
more than reasonable estimates of the true scattering.
The significant results ofthe EPA measurements are shown in Fig. 5. The results
are sparse due mainly to trouble with the refrigerator, as noted above. Figure 5 shows
measured values of scattered fraction!. vs. average beam intensity I B in p.A. Data
for four runs are shown. In each case, the runs were aborted after short periods;
however, the data obtained in the last two runs are considered the most reliable.
The measured values of!. for low beam intensity are in reasonable agreement with
the calculated values shown in Fig. 4. In runs 3 and 4, data are available at 24 and
"0
0.20
••
~
~
<;
u
en 0.10
•••
•
J
••
•
.~ 0.08
'2
0.06
~
or
T
o
10
40
20
30
Average Beam Current.,.A
50
Fig. 5. Scattered fraction of electron beam f. vs.
average beam intensity for full liquid hydrogen
target loop. Symbols are indexed as follows: ••
run 1 ; e, run 2; •• run 3, ~. run 4. Data for runs
3 and 4 wen: obtained at 88 % or greater effective
liquid hydrogen density.
Prototype Tests on a 200-W Forcecl Convection Liquid Hydrogen/Deuterium Target
247
36 JlA and in run 3, one measurement is available at 48 JlA. At 36 JlA, the average
beam dissipation is approximately 72 W. The effective volume is 2.85 cm 3 , which
implies 25 W/cm 3 power density. By reference to Fig. 5, we can conclude that the
data of runs 3 and 4 were obtained for 88 % or greater effective hydrogen density in
the path of the beam. This is not a very strong conclusion, but it was sufficient to
provide the necessary confidence that the prototype system was operating as designed.
Looking at run 4 alone would have permitted a stronger conclusion.
Using the data obtained from these experiments, a new target system has been
designed for experiments at LAMPF. Briefly, the new system consists of a forcedconvection flow loop very similar to that described here. Instead of Teflon O-rings,
Aeroquip Conoseal joints will be used throughout and the target section will again
use indium seals for the aluminum windows. The loop can be moved vertically to
present new aluminum window surfaces to the proton beam when radiation damage
has occurred. A CVI turbine refrigerator* will provide 250 W total refrigeration
power. The latter is limited by the compressors now in use with the system. With
proper compressors, the refrigerator itself has a capacity of 700 W at 20 K. It is anticipated that the system will be used on some of the first experiments at the LAMPF
accelerator beginning in September of this year.
ACKNOWLEDGMENT
The authors would like to acknowledge their thanks to the EPA accelerator crew under 1. Busick for
their support. J. C. Martin provided much of the help in setting up the equipment at the EPA.
REFERENCES
I. R. L. Anderson, Nucl. Instr. Methods, 70:87 (1969).
2. R. Bell, H. Clay, J. Mark, and W. Pierce, IEEE Trans. Nucl. Sci., NS-16(3):631 (1969).
3. H. F. Kaiser, IEEE Trans. Nucl. Sci., NS-12(3):519 (1965).
4. Publication 1133, National Academy of Sciences, National Research Council, Washington, D.C.
(1964).
5. R. M. Sternheimer, Methods of Experimental Physics, Vol. 5A, Academic Press, New York (1961).
• Manufactured for the National Bureau of Standards and Argonne National Laboratory by CVI Corporation of Columbus, Ohio (1965). The authors are indebted to D. B. Chelton and R. O. Voth of NBS, Boulder,
Colorado for help with the system.
F--8
LIQUID HYDROGEN PUMPING FOR
HYDROGEN TARGETS·
J. W. Mark
Stanford Linear Accelerator Center
Stanford, California
INTRODUCfION
Forced circulation of liquid hydrogen through hydrogen targets is necessary
when the geometry and heat load exceed the capacity of normal convection systems.
Three small pumps, one centrifugal and two positive displacement, were built for that
purpose. Hydrogen flows of one-half to two liters/min were attained while the
pressure head developed was 10 to 40 ft of liquid hydrogen.
HYDROGEN TARGET
Physicists using the SLAC 2 Meter Streamer Chamber [1] last year needed a
hydrogen target inside the chamber for studying hyperon production in KP interactions. The hydrogen supplied the protons to interact with the K particles that are
produced by the SLAC linear accelerator. For the experiment to be productive,
the distance from the liquid hydrogen to the neon-helium gas mixture in the streamer
chamber had to be less than 2 mm, and the liquid hydrogen in the cell had to be free of
bubbles. The first target consisted of a tape-wound Mylar straw, 8 mm in diameter
and 460 mm long. It was supported inside a lO-mm ID Mylar tube serving as a vacuum
jacket.
The flask was constructed with an internal tube 2 mm in diameter to introduce
the hydrogen at the tip ofthe target where it could flow back. (See Fig. 1). No electrically conductive parts were within 700 mm of the end of the target.
The reason for the long nonconductive parts requirement is the high voltage
field of the streamer chamber. The target operates between the electrodes of the
chamber, which are pulsed to plus and minus 600,000 V over a gap of 60 cm. The high
voltage gradient of 20 kV/cm causes streamers and sparks within the chamber, so a
conductor inside this field would be untenable even though it is located very near the
ground plane. During operation, streamers and sparks resembling lightning bolts
appeared very close to the target, but none punctured the Mylar vacuum jacket. As
added insurance against electrical breakdown in the target, pulsing of the electric
field was prevented when the vacuum was poorer than 1.5 x 10- 6 Torr.
The Mylar tubest used for the feedlines, target cell, and vacuum jacket are
purchased items, probably the same as used for electrical insulation. They were
tape-wound to design specification of size and wall thickness. Vacuum leak checking
any Mylar containers with a helium leak detector is not practical because helium
• Work supported by the U. S. Atomic Energy Commission.
tPrecision Paper Tube Company, 1033 S. Noel Ave., Wheeling, Illinois.
241
249
Liquid Hydrogen Pumping for Hydrogen Targets
LEXAN
C~NTRI'UG.L
PU MP
BEAM
BEAM
Fig. I. Hydrogen target assembly for streamer chamber.
diffuses through the Mylar film fast enough to exceed the capacity ofthe leak detector.
By careful bonding techniques, the system was assembled with sufficient integrity that
vacuum leaks were not a problem.
Refrigeration was supplied by a reservoir of liquid hydrogen. Hydrogen gas was
condensed into the heat exchanger and circulated by a pump. Inlet and outlet temperatures from the heat exchanger were measured by hydrogen vapor pressure bulbs.
These bulbs could not be located near the cell, but with fluid circulating and both
vapor pressures lower than the fluid pressure, there were no hydrogen bubbles in the
cell.
CENTRIFUGAL PUMP
The first pump started out as a model 10-50-316 Micropump.* It was modified
to the extent that only the ii-in. impeller and impeller thrust plate were in the final
version. By driving the pump with an air motor,t the speed could be controlled up to
10,000 rpm. The pump assembly is shown in Fig. 2. A special ball bearingt with a
filled polyimide retainer was used. The first whirl speed of the shaft is greater than
12,000 rpm. An air drive motor was used because it was safe, inexpensive, and would
function well in the magnetic field present.
There was no method of measuring the flow through the pump, but the head vs.
speed curve is shown in Fig. 3. The curve was not always reproducible. Since the
flow was restricted by the 2-mm straw, only a small flow passed through the system .
• Micropump Corporation, 1021 Shary Court, Concord, California.
t Gast Manufacturing Corporation, Benton Harbor, Michigan.
t Bemol Feuralon AW, Bemol Corporation, Newton, Massachusetts.
J. W. Mark
STANO.ll.RO
- BALL BEARING
Fig. 2. Cryogenic centrifugal pump.
Heat generated by the pump and bearing probably caused some cavitation that
kept the output pressure below theoretical. The target filled and remained free of
bubbles as long as the pump pressure was 30 in. of water or more. The high speed
resulted in bearing life of a few hundred hours. Startup was also slow since the gas in
the cell had to be forced through the small feed straw. Another disadvantage to the
high speed was the eddy current heating caused by the proximity of a large (5 MW)
magnet. (Stray magnetic field was about 500 G).
BELLOWS PUMP
In order to be able to develop higher pressures in the same space occupied by the
centrifugal pump, a double-acting bellows pump was hurriedly put together. Figure
4 shows the pump. The welded metal bellows were taken from SLAC stock and the
valves are air-compressor reed valves. * Initially, the pump stroke was! in. Since the
net area was 2 in. 2 , 1 in. 3 of hydrogen was displaced each half-stroke. Sixty to 120
strokes per minute provided ample flow.
6C
0
W[llSlJqED W • H Ll'.z
IO[ AI..
50
40
30
20
to
0
0
8
p'o"" .. I 00
to
Fig. 3. Cryogenic pump performance with liquid
nitrogen; total head at low flow rate.
* Bell and Gossett. No. C87040, C87041, ITT Bell and Gossett, Morton Grove, II1inois.
Liquid Hydrogen Pumping for Hydrogen Targets
251
Fig. 4. Double-acting cryogenic bellows pump.
The inverted design with pressure outside the bellows is more involved to
build than with pressure inside, but the bellows will not move about and can stand
greater pressure that way. An air cylinder* with internal pilot valves provided
the reciprocating motion necessary to make it pump. It pumped very well until a
bellows failed after some two to three million cycles. Thereupon, the pump was rebuilt with a shortened stroke of i in. No data are available on this unit.
Before the bellows failed, the experimenter determined that he needed a slightly
larger target, so one 10 mm in diameter and 460 mm long was built with a 13-mm
vacuum tube. This target had a 3-mm ID feed straw with 0.OOO5-in. walls to transmit
the liquid hydrogen to the extreme end of the target (see Fig. 5). The new design
had less impedance to flow, which allowed the centrifugal pump to work well at the
* Mead Fluid Dynamics Division of Stanray Corporation, 4114 N. Knox Ave., Chicago, Illinois.
Fig. 5. Streamer chamber target with the 13-mm-diameter vacuum jacket removed.
252
J. W.Mark
more reasonable speed of 6000 rpm. The lower impedance resulted from using a
larger feed straw and a bypass in the target block.
PISTON PUMP
Long, small-diameter targets will probably be required regularly at SLAC, so
another positive displacement pump has been built. It has a 2t-in.-diameter double
acting piston (see Fig. 6). It uses the same type of valves as the bellows pump. The air
cylinder providing the reciprocating motion is a I-in. stroke version of the one used
for the bellows pump. The piston rings are graphite-filled Teflon Tec Seals.· The
piston rod and cylinder bore were hard-chrome-plated in the areas of sliding motion.
Testing to date indicates that the pump works well with no obvious areas of high
wear after 30 days of continuous operation. Continued use of this type of pump is
anticipated because it is reliable, relatively easy to build and maintain, and is readily
accessible should it need maintenance. Less heat input is observed from this pump
than from the centrifugal pump or the fans with a submerged electric motor [2]. Flow
is easily controlled by adjusting the air pressure and flow rate.
.
ACTUATING
SHAFT
Fig. 6.Double-acting cryogenic piston pump.
SUMMARY
Several small pumps have been built which provide sufficient throughput to
operate poorly insulated targets or those with unfavorable aspect ratios. A vacuum
space of only 1.0 to 1.5 mm was sufficient to allow the pencil target to work. Heat
transfer to the vacuum tube was sufficient to cause condensation on humid days, but
no ice formed. Tape-wound Mylar straws can serve as carriers ofliquid hydrogen or as
vacuum vessels. A liquid hydrogen target can operate in a streamer chamber without
electrical problems if the vacuum is good.
ACKNOWLEDGMENTS
The building of the targets and pumps described in this paper was accomplished by the efforts of every
member of the SLAC Hydrogen Target Group. B. Denton, A. Koula, 1. Nicol, and T. Van Hooydonk were
especially patient and persevering with all the revisions.
REFERENCES
1. F. Bulos, A. Odian, F . Villa, and D. Yont, "Streamer Chamber Development," SLAC Rep!. 74 (June:
1967).
2. 1. Mark and W. Pierce, "Hydrogen Targets at SLAC," SLAC-PUB-833 (March 197\).
• Tec-Seal Corporation, 22624 Avalon Blvd., Wilmington, California.
G-1
EFFECT OF WEATHERING OF LNG IN
STORAGE TANKS
J. M. Shah and J. J. Aarts
Chicago Bridge and Iron Company
Oak Brook, Illinois
INTRODUCTION
Recent advances in liquefied natural gas technology, especially in the storage
of liquefied natural gas, have created new demands for accurate prediction of the
thermal behavior of the stored product. With LNG being a mixture of methane,
heavier hydrocarbons, nitrogen, and some lighter gases, the composition of the
liquid can be substantially different depending on the source, time in transit, and
time in storage. With the importation of LNG from various locations about the
world, it is important to more accurately predict LNG properties such as density,
composition, heating value, etc.
LNG weathering is a gradual change of the composition of the stored liquid over
a period of time. As the more volatile components evaporate, the remaining liquid
increases in heating value and the compatibility of this liquid with the normal gas
supply must be evaluated. A mathematical analysis is developed to calculate these
effects as a function of time based on a mass and energy balance. Also, a simplified
procedure is shown to solve this problem by the use of a thermophysical properties
prediction method. Examples of weathering in a typical flat bottom, above-ground
LNG tank and weathering during the cargo voyage of an LNG ship are given using
the proposed analysis.
The weathering problem in a base load LNG terminal is minimal compared to
the peak-shaving installation. Refrigeration associated with continuous pumpout
cancels a major portion of the heat leak or driving force for weathering and as a
result the change in liquid composition during pumpout is minimal.
DEVELOPMENT OF THE WEATHERING ANALYSIS
Assume that a system with initial number of moles M L 0 undergoes an isobaric
change resulting only from a heat input dQ in a differential time increment dt. Assume
that the system is initially in a state of thermodynamic equilibrium and includes only
liquid. At the end of the process, it is also in a state of thermodynamic equilibrium
but now includes a homogeneous liquid phase in equilibrium with the ensuing vapor.
A mole balance on the system yields
M LO = MLl + dM
253
(1)
J. M. Shah ..... J. J. Aarts
where the superscript 0 refers to time t, superscript 1 to time t + dt, and dM denotes
the number of moles of vaporized gas. An energy balance on the system from time t to
t + dt yields
dQ = MLl(HLl - H LO) + dM(H y 1 - H LO)
(2)
The latter can be rewritten as
dQ
=
MLl(dHd
+ L*(dM)
(3)
If the terms in (3) are defined as
dQw
dQB
=
=
MLl(dHL)
(4)
L*(dM)
(5)
then (3) can be rewritten as
dQ = dQw
+ dQB
(6)
The dQw term in (6) can be designated as the "weathering effect" on the system due
to the change in the enthalpy of the liquid which is accompanied by a change in the
composition and the temperature of the system during the time increment dt. For a
single-component liquid at a constant system pressure, there is no change in liquid
enthalpy or bubble point temperature, so the term dQw is equal to zero and all the
heat added in time dt goes to vaporizing the liquid.
For a constant heat leak rate q,
dQ
=
qdt
(7)
Substituting (7) into (3) results in
q dt = ML 1(dH L)
+ L*(dM)
(8)
Dividing by dt and rewriting (8), we get
dM
dt
q
MLI dH L
-----L*
L* dt
(9)
Solution of the above differential equation becomes very complex and time consuming
because M r" H L , L*, and Tare functions of time, but with the help ofa digital computer, (9) can be written in a finite-difference form and solved. The procedure for this is
shown here.
Rewriting (9) for a finite time step ll.t, we obtain
ll.M
q
MLI ll.HL
ll.t = L* - L* At
(10)
Substituting
we get
(11)
(12)
Effect of Weathering of LNG in Storage Tanks
255
Let F represent the mole fraction of liquid not vaporized in time step Ilt. Then
MLI
= FM L O
From a mass balance and (13), we find
IlM
= (1
- F)ML
(13)
°
(14)
Substituting (13) and (14) into (12), we get
(1 - F)MLO(H/ - H LO) = IlQ - FMLO(HLI - H LO)
(15)
Rearranging (15) results in
FHLI
+ (1
- F)H y l = (IlQ/M LO)
+ HLO
(16)
By defining
HMI = FHLI
+ (1
- F)H y l
(17)
+ H Lo
(18)
equation (16) can be rewritten as
HMI = (IlQ/M LO)
This mathematical model can be solved by a digital computer, provided that an
accurate prediction method for thermophysical properties of LNG mixtures is
available. A particular problem can be solved following the steps outlined below:
1. From the given initial liquid composition X i O, system pressure P, and the
bubble point temperature TL 0, calculate the liquid enthalpy HL 0.
2. With the initial mass ofthe liquid M L and given heat leak rate q, calculate IlQ
for a time intervalllt.
3. Use (18) to calculate the total system molar enthalpy HM at the end of the
time step Ilt.
4. From the given system pressure P, mixture enthalpy HMl, and feed composition XiO, calculate the temperature TLl, liquid composition XiI, density,
enthalpy, heating value, and liquid fraction at the end of time step Ilt. This
calculation involves an iterative procedure beginning with an assumed value of
temperature TL 1 and then calculating the other variables.
5. Repeat the above procedure for each time step Ilt until the desired final time is
reached.
°
CALCULATION OF HEAT LEAK RATE IN STORAGE TANKS
The accuracy of the results from the mathematical model to predict the weathering effect depends upon the accuracy of the estimation of the heat leak rate into the
storage tank. Figure 1 shows a cutaway model of a typical flat bottom, above-ground,
double-wall storage tank with suspended deck insulation. More details on aboveground metallic LNG storage tanks are available in various literature articles [1-4].
The well-defined geometry of an above-ground LNG tank and the accurately
determined thermal properties of the insulation materials permit a close estimate
ofthe heat leak rate into above-ground storage tanks. For a partially filled tank, the
problem of predicting heat leak rate and wall temperatures as a function of liquid
level depends on the relative contributions ofthe different modes of heat transfer to the
liquid such as radiation, gas convection, wall conduction, etc. A mathematical model
has been established and computerized for the prediction of heat leak rate as a
J. M. Shah and J. J. Aarts
Fig. I. Double-wall LNG tank with the suspended deck insulation.
function of the liquid level based on an analysis similar to that developed by Neill et
al. [5]. The actual field data have correlated quite well with this mathematical model.
PREDICTION OF THERMOPHYSICAL PROPERTIES OF LNG MIXTURES
The other important factor in calculating the LNG weathering effect is the
capability to predict thermophysical properties of natural gas mixtures. The required
thermophysical properties include phase equilibria, density, and enthalpy of natural
gas mixtures at operating conditions.
To provide experimental data on the variation of enthalpy with pressure and
temperature, two calorimeters, an isobaric calorimeter and an isothermal-isenthalpic
calorimeter, were constructed. The details of construction of the isobaric calorimeter
with typical isobaric data on natural gas mixtures have been reported by Laverman
and Selcukoglu [6]. The construction details of an isothermal-isenthalpic expansion
calorimeter and a tabulation of some reported experimental data are given by Alkasab
et al. CJ. Both calorimeters are designed to operate over a temperature range from
- 300 to + 100°F with pressures up to 2500 psia. To obtain accurate phase equilibrium
data, an equilibrium cell was also constructed. This apparatus is suitable for operation
over a temperature range of - 320 to + 100°F. The construction details of this device
and some typical experimental data are reported by Djordjevich [8J.
Utilizing measured thermodynamic data at cryogenic temperatures, modifications were made to the Bennedict-Webb-Rubin (BWR) equation of state. A computer
program was developed to predict the thermo physical properties of LNG mixtures
based on this modified BWR equation of state. The BWR coefficients for methane at
cryogenic temperatures are reported by Lee et al. [9] and those for ethane are reported
by Alkasab et al. [10].
ElI'ect of Weathering of LNG in Storage Tanks
RESULTS
With the help of the weathering analysis and a thermophysical properties
prediction method, a computer program was developed to study the weathering
effects in flat bottom LNG storage tanks and in LNG ship tanks during the cargo
voyage. Sample calculations have been worked out using this computer program and
the results have been plotted in Figs. 2 through 6.
Figure 2 is a plot of the volume of LNG left in a tank as a function of time for a
300,OOO-bbl, flat-bottom LNG tank, with a boiloff rate of 0.05 %/day. Figure 2 also
shows the change in the temperature of the stored LNG with respect to time. The
initial steep rise in the liquid temperature is due to the rapid removal of volatile
nitrogen which was originally present.
>
~
UJ
1050
~
;kI
~ 1000
Fig. 2. Weathering time vs. :::;
volume and temperature of ~
LNG in a 300,OOO-bbl, flat ~
bottom tank with a boiloff rate ;<;
of 0.05 %/day. Storage condi- (!)
tions: heat leak rate, 200,000 3
Btu/hr; tank pressure,15.0 psia; ~
initial LNG composition in UJ
mole %, CH 4 88.0, C 2 H 6 7.0,
C3HS 2.5, C 4 H IO 0.5, and N2 §;
2.0.
kL
~Iitll
'" 1025
....
u
975
950
/'
925
3
900
1/1
o
30
V
60
.~
/
120
150
180
-256
-258
.I.J..
UJ
'"~
'""-
-260
UJ
~
~
90
-254
210
240
1:i
...
-262
(!)
z
-264 -'
-266
270
WEATHERING TIME. DAYS
Figure 3 indicates the effect of heat leak rate on the heating value of the LNG
as a function time for a 300,OOO-bbl, flat bottom LNG tank. The curves in Fig. 4 show
the change in heating value as a function time for various initial liquid compositions.
These curves indicate that the presence of nitrogen in the liquid has a predominant
effect on LNG weathering.
I I <y
'I; 1150
i
• 1140
'"
z
~
:.- '
., 1130
z
-' 1120
Fig. 3. Weathering time vs. heating :5
value of LNG in a 300,OOO-bbl, flat
bottom tank with a boiloff rate of ~
0.05 %/day. Storage conditions: storage tank pressure, 15.0psia; initial
LNG composition in mole %, CH 4
88.0, C 2 H 6 7.0, C 3Hs 2.5, C 4 H 10 0.5,
and N2 2.0.
'3
1110
I 100
",~'f.,
fff
~
~
(!)
bt~t~
~
1090 0
30
60
90
120
e~""r
oQ 000..--
\\",~t ~'2;
150
180
WEATHERING TIME. DAYS
210
240
270
J. M. Shah and J. J. Aarts
':;
. :::. 1150i---t----l---,-::-::-:±-;;-.
~
~ 1100i---t---t--+---t---t---+~
<!: 10501--Tl.:.:'::';'=F~¥~=F=="F====F==1
(!)
z
~ 1000j==t:;~~~~~~~==:x===1
(!)
z
~
UJ
:I:
9001---+=
Fig. 4. Weathering time vs. heating value for
various LNG compositions. Storage conditions:
heat leak rate, 200,000 Btufhr; tank pressure,
15.0 psia; and initial tank volume, 300,OOO-bbl.
Figure 5 shows the volume and the liquid temperature as a function oftime for a
cargo voyage of a typical 125,OOO-m 3 LNG ship with a boiloff rate of about 0.25 %/day.
(The geometry and number of tanks used are immaterial as long as the total volume
and boiloffrates match the values stated.) Figure 6 shows the change in liquid density
and heating value during a twelve-day voyage. These curves show that the variations
in density, heating value, and temperature of LNG are less than 1.0 %during a normal
ship voyage for the case considered.
At the receiving base load LNG terminal, the liquid level is decreasing at a
constant rate between arrival of the ships. The refrigeration produced by evaporization in the tank as the liquid level drops will offset tank heat leak. The driving force
for weathering is thus minimized, if not essentially eliminated.
....z
2610
-256.0
UJ
:i 2600 b...
-256.5
S
-257.0
>
:;
If)
2590
~
~
---f~
~ 2580
'"
::!
::!
,;, 2570
z
~
~ 2560
'"<!:
(!)
z
2550
...02540
--'
.J
>
1/
/'
...
;~\E
I""
'\
o
2530
o
246
8
WEATHERING TIME. DAYS
10
12
-257.5 c
-258.0
UJ
0:
::J
~
0:
UJ
0..
-258.5 ~
....
(!)
z
-259.0 --'
-259.5
-260.0
Fig. 5. Weathering time vs. volume and temperature of 125,000 m 3 of LNG in a ship tank with a
boiloff rate of 0.25 "Jday. Storage conditions:
heat leak rate, 2,490,000 Btufhr; ship tank pressure, 15.6 psia; and initial LNG composition in
mole %, CH 4 87.55, Cl H6 8.14, C3HS 2.30, iC 4 H IO 1.16, and Nl 0.85.
Eft'ect of Weatheriag of LNG iD Storage Tub
29.5 6
1134
29.5 4
..,
/
., 29.52
!"...>Ci)
z
o
,0""
29.5
29.4
"'
~ 29.4
o
:J
V
"\
'=/ 1J< '"'"
6
29.4 2
29.4
°°
~
1133
2
4
6
W~ATHERING
....
.:::.
-f-
1132 ~
CD
1131
'"....z
"o
1130 "'
....
:J
~
1129,-,
z
;::
"
29.4 4
Fig. 6. Weathering time vs. density and heating
values of 125,000 m 3 of LNG in a ship tank with a
boiloff rate of 0.25 %/day. Storage conditions in
mole%: CH 4 87.55, C 2 H 6 8.14, C3HS 2.30,
i-C 4 H,o 1.16, and N2 0.85.
/
/
""
«
1128
I~I
8
10
TIME, DAYS
12
!I!
127
1126
CONCLUSION
The results from this analysis show that the design of the storage vessel (in
terms of heat leak), the initial composition of the stored liquid, and the length oftime
of storage all have considerable impact upon the ultimate utilization and the interchangeability ofthe stored liquid. These effects can readily be evaluated following the
analysis described.
NOTATION
dH L =
dM =
dQ =
dQw =
dQB =
dt =
F
HL
Hy
HM
L•
ML
P
q
=
=
=
=
=
TL
=
=
=
=
=
Yi
=
=
=
Q
t
Xi
differential liquid enthalpy
moles of liquid vaporized in time dt
heat input in system for differential time dt
weathering part of heat input
boiling part of heat input
time increment
mole fraction of liquid not vaporized
liquid enthalpy
vapor enthalpy
mixture enthalpy
pseudo-latent heat of vaporization
moles of liquid
system pressure
heat leak rate
total heat input
liquid temperature
time
liquid composition
vapor composition
Greek Letter
Ii
= finite increment
Superscripts
o
1
= denotes quantities at time t
= denotes quantities at time t
+ dt
J. M. Shall .... J. J. Aarts
REFERENCFS
I. D. T. Lusk and D. C. Dorney, in: Progress in Refrigeration Science and Technology, Vol. 1, L'Institut
International du Froid, Madrid, Spain (1969), p. 2SI.
2. I. L. Wissmiller, "Above-Ground Storage Tanks for Liquefied Natural Gas," paper No. ASME
66-WA/PID-4, presented at ASME Winter Annual Meeting and Energy Systems Exposition, New
York (November 1966).
3. C. C. Hanke, "New Developments in Above-Ground Metal LNG Containers," presented at American
Gas Association Operating Section Proceedings, Arlington, Virginia (1968).
4. Y. A. Selcukoglu, in: Proceedings 2nd Intern. Cryogenic Engineering Conference, I1iffe Sci. Tech.
Publ. Guilford, Surrey, United Kingdom (1968), p. 131.
S. D. T. Neill, H. T. Hashemi, and C. M. Sliepcevich, Chem. Eng. Progr. Symp. Series,64(87): III (1968).
6. R. J. Laverman and Y. A. Selcukoglu, in: Progress in Refrigeration Science and Technology, Vol. 2,
L'Institut International du Froid, Madrid, Spain (1969), p. 749.
7. K. A. A1kasab, R. J. Laverman, and R. A. Budenholzer, "Experimental and Calculated Specific Heats
and Joule-Thomson Coefficients in Methane-Ethane Mixtures," paper No. ASME 71-WA/HT-46,
presented at ASME Winter Annual Meeting, Washington, D.C. (November 1971).
8. L. Djordjevich, Ph.D. Dissertation, Illinois Institute of Technology, Chicago, Illinois (1968).
9. T. W. Lee, S. B. Wyatt, S. H. Desai, and K. C. Chao, in: Advances in Cryogenic Engineering, Vol. 14,
Plenum Press, New York (1969), p. 49.
10. K. A. Alkasab, J. M. Shah, R. J. Laverman, and R. A. Budenholzer, 1& EC Fund., 10:237 (1971).
DISCUSSION
Question by D. Zudkevitch, Esso Research and Engineering Company: What do you mean when you
speak of a "modified BWR"? The dependence of Co with temperature has been suggested by many authors,
starting with Benedict's original papers.
Answer by authors: We mentioned a "modified BWR equation of state" in our paper, but what we
should have enphasized was the modified BWR coefficients. The form of the BWR equation of state which
we used is the same as that proposed by Benedict with the addition that Co is a polynomial in temperature.
References 9 and 10 indicate the new BWR coefficients which were used. In addition, interaction coefficients
were developed from available vapor-liquid equilibria data, but these new coefficients have not yet been
published. The primary purpose of this paper was to present the procedure that should be used in making
LNG weathering computations. As an example of using this procedure, we showed the results that could
be obtained if one used thermophysical properties predicted from the BWR equation of state. However,
any other suitable thermodynamic property correlation could also be used with the same procedure.
G-2
DESIGN OF LNG RECEIVING TERMINALS
D. B. Crawford and C. A. Durr
The M. W. Kellogg Company
Houston, Texas
INTRODUCTION
Liquefied natural gas (LNG) has become increasingly important as a source of
energy in recent years. In the United States, LNG has been used for "peak-shaving"
purposes; i.e., natural gas from domestic sources is liquefied during "off-peak"
demand periods, and then vaporized when peak demand for gas exists. This type of
peak-shaving plant usually liquefies of the order of two to twenty MMscfd and
vaporizes when required (about ten operating days per year) at a rate of 50 to 400
MMscfd.
However, base load LNO receiving terminals are vastly different from peakshaving plants because the LNG for the receiving terminal generally comes from
foreign sources and the flows for the terminal greatly exceed the flow through a peakshaving unit. For example, typical plans for LNG terminals call for delivery of 1000
MMscfd of gas to the pipeline. Also, liquid unloading flow rates of about 50,000 gpm
are required.
The purpose of this presentation is to consider the basic components of a
receiving terminal and to analyze several of the technical problems encountered in
designing such terminals; emphasis will be placed on the ship unloading and vapor
handling systems.
PROCESS FLOW SCHEME
Figure 1 is a simplified flow scheme for a LNG receiving terminal. LNG is
delivered via specially designed cryogenic ships. The LNG is unloaded from the ship
by submerged pumps in the cargo tanks and flows to the LNG storage tanks. During
the unloading period, vapor is physically displaced from the storage tanks; part of this
vapor flows through a vapor return line to the ship in order to replace the LNG
being pumped out. Since there is usually more vapor displaced from the storage tanks
than can be returned to the ship, the excess vapor is used for fuel, is flared, or is
compressed and sent to the pipeline.
Fig. 1. Simplified fiowsheet for a LNG receiving terminal.
261
262
D. B. Crawford aud C. A. Dorr
Table I. Typical Design Parameters for LNG Receiving Terminals
Ship
Size
Draft
Boiloff rate
750,OOObbl
35 ft
0.25 %/day based on full ship
Unloading System
Unloading rate
Design unloading time
50,OOOgpm
12 hr
Land Storage
Capacity
Heat leakage
Two to three times ship's volume
0.05 %/day to 0.1 %/day
Send out
Sendout rate
Pressure
Temperature
1000 MMscfd (8300 gpm)
Set by pipeline (usually 800 to 1200 psig)
40 to 60°F
The LNG in the storage tanks is pumped to an intermediate pressure via booster
pumps and is then pumped to the pipeline pressure via the sendout pumps. After the
LNG is vaporized, it is directed to the pipeline.
Typical design parameters for LNG receiving terminals are presented in Table I.
LNG UNLOADING SYSTEM
The unloading system must be able to handle LNG flows as high as 50,000 gpm
in order to unload a 750,000-bbl ship in a design time of 12 hr. The unloading line
size is chosen to accomplish these flow rates consistent with the ship's pumping
capabilities, length of unloading line, and elevation of storage tanks. After the line
size is chosen, strict attention must be given to the transient conditions possible in
the unloading system. For example, the piping system must be capable of dealing
with fluid hammer which occurs whenever there is a sudden change in velocity of the
moving fluid. This change in velocity results from numerous causes, such as valve
closing or pump failure. When the fluid velocity is suddenly reduced, the change in
kinetic energy manifests itself in compressing the fluid and stressing the pipe walls and
supports. The resulting pressure surges from velocity change may establish the pipeline design pressure.
In order to determine if fluid hammer is a problem, the following equation can be
used to determine the maximum excess pressure (above the normal hydrostatic
pressure) due to fluid hammer in a rigid pipe:
= ySVo/l44 g
(1)
S = (gKI44/y)1/2
(2)
Po
where
The above equation represents the most extreme case of instantaneous reduction
of flow, a completely rigid pipe, and complete reflection of the wave to the point of
flow reduction. The condition of instantaneous reduction of flow applies as long as
the time of reduction (valve closing, for instance) is less than the period for the piping
Design of LNG Rec:eiviDg Terminals
system defined as 2H/S. As an example of the above equation, the excess pressure Po
developed in closing a valve against LNG flowing at 20 ft/sec is about 550 psi; this
would require a valve closure of about 2 to 3 sec in a mile-long unloading line.
If the excess pressure Po indicates that the fluid hammer sets the design pressure
for the unloading line, then the wall thickness of the unloading line should be set by
this pressure. However, it is economically desirable to specify as thin a pipe wall
thickness as possible. To accomplish this, a positive means of dealing with pressure
surge can be used to reduce the magnitude of the pressure surge. For example, the
valve closure time may be selected to ensure that the time of closure exceeds the time
period 2H/S of the system. The longer the time of valve closure, the less severe will
be the resulting pressure buildup. Another consideration is that the pressure surge
will depend upon the closure characteristic of the valve. As shown in Fig. 2, gate
valves have a different percentage opening vs. percentage valve stroking characteristic
than that of a butterfly or globe valve. Since the change in fluid velocity is more pronounced in the region of 0 to 20 % valve opening (as compared to 50 to 100 % valve
opening), gate valves are the least useful in minimizing the pressure surges. It is better
to use a butterfly valve because these valves are capable of dissipating the kinetic
energy over the full closure time. Figure 3 clearly shows the difference between butterfly and gate valves in reducing the maximum surge pressure.
Since the calculation of the pressure surges during valve closing is important, a
reasonably accurate method of calculation must be used. The method should be
capable of dealing with fluid friction, effect of piping orientation, and inclusion of
proper boundary conditions. For example, the method of characteristics given by
Streeter and Lai [2] could be used.
Another consideration in unloading line design is that rapid valve closure may
cause a vacuum on the downstream side of the valve. This possible vacuum condition
may also set the minimum thickness required for the line.
RECIRCULATION SYSTEM
The operation of the unloading line must also be examined for the situation in
which a ship is not unloading. In general it is advisable to maintain a LNG recirculation loop, as shown in Fig. 4, during this period. LNG from the storage tanks is
pumped to a pressure of about 50 psia and a portion of this liquid is directed down
the unloading line to the pier. Here the LNG crosses over into the recirculation line
and returns to the storage tank area; it rejoins the major portion of the LNG from the
first-stage pump and flows to the suction of the sendout pumps. In this manner, the
heat leak in the unloading line is removed from the terminal in the form of a slight
rise in sensible heat of the liquid.
100
200
%
OP~~~~G
MAX PRESS.' 300 %
t
%
~~~M
50
AREA
100
PRESSURE
50
100
% STROKE MOVEMENT. TIME
Fig. 2. Percentage of valve opening as a function of
stroke movement for various valves.
o ~O1IIIIICi--1OIIIIIII!5O~=----J,00
% STROKE MOVEMENT. TIME
Fig. 3. Maximum excess surge pressure caused by
fluid hammer effect.
D. B. Crawford ..... C. A. Durr
TO
IIU'ORIZER
PIER
SHORE
Fig. 4. Simplified Ilowsheet for LNG recirculation loop.
The advantages of such a recirculation loop are twofold. First, if there were no
recirculation, the LNG in the unloading system would warm up due to heat leakage.
The net result would be that when a ship starts to unload, the warm inventory in the
unloading system would be displaced into the storage tanks and cause excessive
flashing as the liquid enters the tanks; this could cause the operating pressure of the
storage tanks to be exceeded Additionally, there could be flashing in the LNG unloading line itself as the static head on the LNG is reduced. With a recirculation loop,
these problems are minimized because the LNG inventory of the line is maintained
as cold as possible. The second advantage of recirculation is that it is a positive means
of preventing the phenomena of geysering. Geysering is defined as a rapid expulsion of
a boiling liquid and its vapor from piping. Geysering occurs in vertical or inclined
piping systems where sufficient heat is added to the liquid to cause it to boil. The
formation of bubbles in the liquid reduces the static pressure on the remaining fluid
in the piping. Since the liquid below the bubbles is also saturated, the reduction in
static pressure results in additional bubble formation which again reduces the static
pressure. This situation can eventually lead to a rapid expulsion of both liquid and
vapor from the piping system.
Some LNG storage tanks have top fill connections; as a result, vertical sections
of large piping may exist that extend about 120 ft to the top of the storage tanks.
These lines may be subject to geysering. The result of such a geyser is that large
volumes ofvapor could be released in the storage tanks, resulting in excessive pressure.
If the storage tanks (having bottom fill lines) are significantly above sea level and
require a long, inclined unloading line, geysering into the storage tank also becomes
possible. After the geyser effect, the LNG will flow back into the line, causing the
previously described fluid hammer effect. This type of behavior has been experienced
in the missile industry, where the fluid hammer effect has damaged both lines and
valves. By maintaining a LNG recirculation loop, this geysering can be avoided
because the recirculation prevents the heat buildup in the line that is a prerequisite
for geysering.
VAPOR HANDLING SYSTEM
A typical vapor handling system for a terminal is shown in Fig. 1. Essentially,
the system consists of a vapor return line to the ship and a method of using the excess
gas that cannot be returned. A cold blower is used as the driving force in returning
the vapor to the ship and also in supplying enough pressure to feed the fuel system
and/or the pipeline compressor.
During ship unloading, the vapor space pressure of the ship experiences a
decrease because liquid is being removed. Part of this volume depletion is made up by
ship boiloffvolume. However, at maximum ship unloading rates, the generated boiloff
volume is insufficient to replace the liquid being removed. Thus, vapor return from
Design of LNG Receiving Terminals
shore is necessary to provide the volumetric differences. Unfortunately, the vapor
flow from the storage tanks during ship unloading is larger than the vapor return
requirement. As shown in Fig. 1, the resulting excess gas must go to fuel, flare, or be
compressed to pipeline pressure.
The solution to this excess gas disposal problem is not simple. Flaring is usually
not desirable because of the waste of valuable gas. The alternative of using this gas as
fuel is not always present. For example, heat for LNG vaporization may be obtained
through use of sea water exchangers or required power may be purchased instead of
generated on site. Even if the gas is used as fuel for the vaporizers and/or power
generation, there are two problems associated with this approach. First, the amount of
excess gas available during ship unloading can easily exceed the fuel requirement for
the vaporizers and power generation. Second, the nitrogen concentration of the boil off
gas may be high, which will make this gas difficult to use in burners that normally
use the pipeline gas as fuel. The remaining alternative of compressing the gas to
pipeline pressure is useful, but expensive. It is cheaper to compress the gas than to
flare it; however, it usually requires expensive reciprocating compressors that will
be used only 10 to 20 % of the time.
VAPOR HANDLING STUDIES
In designing the vapor handling system, one very important study must be made
that has a major bearing upon the system. This is to determine the effect the storage
tank design pressure has upon the vapor handling system flow rate. The factors to be
considered in this study will now be examined in detail.
The storage tank design pressure has a major influence upon the vapor flow rate
that the compressors must handle. For example, consider the following equations:
~=~+~
W
FD
=
(4)
F
= ~(Q
H
(Pv/PL)L
A.
QT = Ql
_LCB LlP)
p
T
+ Q2
where FH
~
0
(5)
- Q3
(6)
Table II. Typical Vapor Flow Rate for Situation in which AP = 0.0 psi
Ship's pressure, psia
Storage tank pressure, psia
l:1P, (storage pressure - saturation pressure of LNG on ship), psia
LNG unloading rate, gpm
Sendout rate, gpm
QT, net energy input into LNG, Btu/hr
F D, vapor flow rate due to displacement, Ib mole/hr
F H' vapor flow rate generated due to energy input in unloading
LNG, Ib mole/hr
F T' total vapor flow rate out of storage, Ib mole/hr
Vapor return to ship, Ib mole/hr
Excess vapor flow rate to fuel and/or pipeline compressors, Ib mole/hr
Horsepower to compress 3636 Ib mole/hr of vapor to 1000 psia
pipeline pressure, BHp
15.7
15.7
0.0 psi
50,000
8,300 (one billion scfd)
10,000,000
2,517
2,847
5,364
1,728
3,636
9,800
D. B. Crawford .... C. A. Burr
Table m. Typical Vapor Flow Rate for Situation in which AP
= 1.0 psi
15.7
Ship's pressure, psia
16.7
Storage tank pressure, psia
1.0
AI' (storage pressure - saturation pressure of LNG on ship), psia
50,000
LNG unloading rate, gpm
8,300 (one billion sefd)
Sendout rate, gpm
10,000,000
QT' net energy input into LNG, Btujhr
2,678
F D' vapor Bow rate due to displacement, Ib mole/hr
F H' vapor Bow rate generated due to energy input in unloaded LNG,
o
Ibmole/hr
2,678
F T' total vapor Bow rate out of storage, Ib mole/hr
1,728
Vapor return to ship, Ib molejhr
Excess vapor Bow rate to fuel and/or pipeline compressors, Ib mole/hr
950
Horsepower to compress 950lb mole/hr of vapor to 1000 psi pipeline
2,700
pressure, BHp
Table II shows the results for a possible unloading system in which the pressure
differential is zero; i.e., the saturation pressure of the LNG in the ship is identical
to the vapor space pressure in the storage tanks. The unloading rate is 50,000 gpm,
the ship's boiloff is 0.25 %/day from a 750,000-bbl ship, and the vapor is returned at
-200°F. The net energy input into the unloaded LNG is about ten million Btu/hr,
3500?-------~----~~----~------~------~----~
ASSUMPTIONS
3000
II:
II SENOOUT RATE IS 8300 GPII (I BilLION SCFD I
2) ""POR IN STORAGE TANK I S AT
111.7 PSIA AND -258"F
3) VAPOR RETURN AT SHIP IS AT -200"F
AIID 15.7 P SIA
4) BOil OFF AT SHIP IS .25 'Yo/DAY
BASED ON A 750,000 III SHIP
2500+-------t-------r------+------~--~~4_----~
2:
~
oJ
o
2
m
~oo+-------t-------r------+----~~--~--4_--~~
oJ
UI
III
~
II:
•
ooJ
1500+-------+-------~--~~------~
"-
II:
o
Go
~
IOOO+------If----
500
O·~----~~~--_+------~----~~----~----~
o
10
20
40
50
LNG UNLOADING RATE. GPM liN THOUSANDS)
Fig. 5. Vapor Bow rates as a function of LNG unloading rate.
Design of LNG Receiving Terminals
which is typical for this unloading rate. For this situation, about half of the vapor
flow rate out of the storage tanks is attributed to the flashed vapor formed as a result
of the net energy input to the unloaded LNG. Taking into account the vapor return
flowrate to the ship, this leaves about 3636lb mole/hr of vapor that must flow to fuel
and/or pipeline compressors.
Table III shows the results of the same calculation except that in this case the
pressure differential is maintained at 1.0 psi. Notice that by increasing the pressure
differential, the vapor flow rate out of the storage tanks is cut in half because there are
no flash vapors generated due to the net energy input into the unloaded LNG. Instead,
this energy input manifests itself as a slight rise in the saturation temperature of the
stored LNG. The net result ofthis calculation is that the vapor flow rate to fuel and/or
pipeline compressors is reduced to 950 lb mole/hr.
The principle to be emphasized here is that the vapor handling system flow rate
can be reduced drastically by raising the design pressure of the storage tanks. Table II
assumes a 1.0-psig design pressure, while Table III assumes a 2.0-psig design pressure
for the storage tank. An economic evaluation needs to be made in order to determine
the optimum tank design pressure; if the cost increase for a 2.0-psig design pressure
tank as compared to a 1.0-psig design tank is small, then it is probably best to reduce
the size of the vapor handling system as much as possible.
Figure 5 is a plot of FD vs. unloading rate. In addition, the vapor return flow rate
to the ship and the excess gas flow rates are given at the conditions specified. This
figure can be used in calculating the vapor flow rates for other ship unloading rates.
CONCLUSION
The unloading and vapor handling systems have been studied and certain
problem areas have been indicated. Care must be taken in designing these systems to
ensure that the optimum solution is chosen and that all possible transient conditions
are analyzed to ensure the proper design of terminals.
NOTATION
B = slope of vapor pressure vs. temperature, psij"F
C p = specific heat of LNG, Btujlb-mole-oF
F D = vapor flow rate out of storage due to physical displacement, lb mole/hr
F H = vapor flow rate generated due to energy input into the unloaded LNG, lb mole/hr
F T = total vapor flow rate out of storage, lb mole/hr
g = acceleration due to gravity, 32.2 ft/sec 2
H = length of pipe from valve to inlet end, ft
K = bulk modulus of ela-sticity of liquid, psi (123,000 psi for methane at - 260°F)
L = LNG unloading rate, lb mole/hr
Po = excess pressure due to water hammer, psi
S = velocity of wave travel in rigid pipe, ft/sec
Q1 = energy input via ship's pump, Btu/hr
Q2 = heat leak into unloading system, Btu/hr
Q3 = credit due to static head, Btu/hr
QT = net energy input into LNG, Btu/hr
Vo = reduction in fluid velocity causing hammer, ft/sec
y = density of LNG, Ib/ft 3 (29Ib/ft 3 for LNG at - 260°F)
Greek letters
l1P = storage tank pressure minus saturation pressure of LNG on ship, psi
A. = latent heat of LNG, Btujlb-mole
PL = density of unloaded LNG, lb mole/ft 3
Pv = density of vapor in storage tank, lb mole/ft 3
D. B. CnwfonlMd C. A. Durr
REFERENCES
1. R. C. King and S. Crocker, Piping Handbook, McGraw-Hill Book Company, New York (1967),
pp.3-190.
2. V. F. Streeter and C. Lai, J. Hydraulics Div., ASCE, 88:79 (1962).
3. D. J. Wood and S. E. Jones, J. Hydraulics Div., ASCE, 99: 167 (1973).
4. S. K. Morgan and H. F. Brady, in: Advances in Cryogenic Engineering, Vol. 9, Plenum Press, New York
(1963), p. 206.
5. D. W. Murphy, in: Advances in Cryogenic Engineering, Vol. 11, Plenum Press, New York (1965), p. 353.
6. D. Hale, Pipeline and Gas J.,2OO(7):21 (1973).
DISCUSSION
Question by R. N. DiNapoli, Private Consultant: Is it possible to add compressed boilolf gases between
primary and secondary pumps for reliquefaction as a means of reducing compression horsepower?
Answer by author: This is certainly a very valid method of reducing the horsepower requirements for
the vapor handling system. The boilolf gas and displacement gas (during ship unloading) can be injected
into the sendout liquid at the discharge of the first-stage pump (usually 50 to 100 psig~ Thus, this vapor
will only have to be compressed to 50 psig instead of the pipeline pressure of 800 to 1200 psig. There are
two problems associated with the approach, however; first, a device must be designed to ensure that all the
vapor is recondensed before the liquid reaches the final sendout pump; second, another use must be found
for the gas when the terminal sendout rate is zero or else this gas must be vented. These two technical
problems can be handled adequately and as such this method of vapor disposal is very attractive. As a
matter of interest, this solution to the vapor disposal problem is being designed into the FOS LNG
receiving terminal in France.
Question by R. N. DiNapoli, Private Consultant: Is ship heat leak considered in vapor boilolf calculations during unloading?
Answer by author: Ship heat leak is considered in designing the vapor handling system. During ship
unloading, the boilolf volume created in the ship due to heat leak is equivalent to about 10,000 gpm of
LNG unloading rate. Thus, when the ship unloading rate is SO,OOO gpm of liquid, the vapor return volume
to the ship will be about the equivalent volume of 40,000 gpm. The remaining 10,000 gpm is from ship
boilolf.
G-3
SOME IMPORTANT FACTORS IN LNG
TANKER DESIGN SELECTION
R. C. Ffooks
Conch Methane Services Limited
London, United Kingdom
INTRODUCTION
Although it is true to say that there are at least fifteen different LNG tanker
design technologies available today, the position is not quite so confusing as this
statement implies. The differences in the systems lie almost entirely in the containment
techniques; the cargo handling arrangements vary very little from design to design.
From Fig. 1 it can be seen that containment systems-which include the insulation and secondary barrier requirements, since they are inextricably combined with
the primary containment----<:an be divided into two basic concepts: (1) self-supporting,
where, in principle, the tank rests on the hold bottom and is free to expand and contract
relative to the structure of the ship, and (2) membrane, where the tanks essentially
rely on the ship's structure for their support over the entire tank surface. Each ofthese
two categories is subdivided into a number of individual techniques which have, for
I LNG
CONTAINMENT SYSTEMS
I
SELF
SUPPORTING
I
MEMBRANE
=~':~!> ;~~j'pL ~~9"'OAAY
REQIJ,RES FULL
SEC,,"OARY BARRIER
OEPEHOIHG OH DESIGN
L
I
1
SPHERlg~L
I.' .
Supported by contl"_
ay"ndr'Of,I
t
(2)TECHNrGAZ
Support«l by 'Inkag.
,ylt._
J
t,..,. oyllndrlc.'
~,.tlldbyop."
box
skirt
OD
l.-rIa
1
('2)ZELLE~TAHK
----l
Multi hDrl_"1
oyllnde,.. *'PPOrted
1
TRAPEZOIDAL
FLAT SIDED
1
(1,.£2!:!£!:!...
!(,JWORMS/C. ere FR
largl dl,. _to eyll . . .,..
... "o,.lyeupport..d
0 _ _ lOll bolt-
IL
on .,ddt ..
O)~
(.04JC.B. &.L
Svpported by
I
CYLI~:ICAL
ULn CYLINDRICAL
SPHERICAL : 1
TYPE
(1) I(VAERHER /MOSS
I
.
---
S'ngl . . . II divided
Supported by
kl .. ,.nlla dl,trlbuted
boV.a- .,....
IlWulltlon/_dIof'y
"'b.
•.,.,. '*'.
be,.,.I.,. bal ../PlF
flbfooue. gl •••
l
I
TRUE
MEMBRANE
I
I
I
SEMIMEMBRANE
1
SR10C.ESTOHE
(1)
6.. ("
on ••
9""
fib,.. Insul.tlon lyet ..
(2)1!:!!...
•
dd'"
91 HI . . .bMonl;
!:!;~~~:~7,.!;~
()).!:.!!:!2f.
Multi_H.'
0)'11..0- lupported
_I
,- ~ ,- All. Iingl, w.11
fI.t lid,,:! tanka Gn
ply/PYC Inlulatlon
aecondary barrl H" s,.at_
SEMI-MEMB~ANE
SYSTEMS
HAVE UNSUPPORTED RADIUS
(5) PDM / C..T,
CORNERS
Unlfo ... ly supporttd
0_ conlc.1 bottell
Fig. 1. LNG design systems.
269
270
R. C.Ffooks
the most part, been developed by separate design groups working independently. No
attempt will be made in this presentation to compare the merits and demerits of the
various designs; however, it should be appreciated that commercial operating
experience is limited to the five systems shown surrounded by broken lines in Fig. 1
and, of these, two are no longer commercially viable.
Materials of construction-at least for the primary containment-are now
relatively well known and widely available though they remain quite specialized in
nature; however, careful control techniques, both as to weld quality and procedures,
are required in order to maintain the necessary standard of construction.
Operating experience with LNG tankers is still relatively limited. The two
Conch design ships entered service in 1964 and have therefore each seen nine years of
regular operational service between Algeria and the United Kingdom-longer in
shipyears than all the other systems put together. The Gas Transport membrane ships
entered service between Alaska and Tokyo at the end of 1969 and early 1970 and have
thus seen nearly four years of regular operation. The Esso ships were completed in
1969-70 and have been in operation intermittently since then due to project difficulties. The Technigaz membrane first entered service in LNG in 1971 and has traded
intermittently in LNG since then-the second Technigaz ship entered regular service
between Brunei and Japan at the end of 1972. Generally speaking, operational
experience of all LNG designs to date has been good and individual experiences have
been published.
FACTORS TO BE CONSIDERED IN SHIP SELECTION
The final selection of the ship(s) for a specific project is a reiterative process
involving many interdependent factors. (Much of the basic optimization task can
now be handled on the computer; in fact, the computer is a virtual necessity for the
final contract negotiations, as is evident from Shell's comments after signing the
Brunei-Japan project, which involved two discharge ports and three customers.)
These factors, important in ship selection, are considered briefly in the following
discussion.
Ship Size
The size of ship selected for an LNG operation depends on:
Project Idealization. This defines the theoretically optimum economic ship size,
taking into account shore storage, capital and operating costs, and service speed.
Shipyard Capability. The main considerations here are berth or dock size,
willingness to guarantee hull structure and/or machinery if these are much larger
than or different from previously built ships, and special lifting requirements, if any.
Operational Limitations. These include principally length and draft restrictions
imposed by ports for the proposed project; they also cover limitations in hours of
port entry, speed limits, etc. Severe draft restrictions may also impose service speed
restrictions (e.g., the EI Paso project).
Interchangeability. Considers interchangeability with ships of other projects.
Design System. This may impose some dimensional restrictions if port facilities
are marginally adequate.
Project Reliability. Where the optimization process indicates that the minimum
shipping cost is achieved with, say, one or two ships, the effect of one unit out of
service could be critical; ship numbers might therefore be increased to two or three,
respectively. Alternatively, spare capacity or speed can be built into each ship, or
Some Important Factors in LNG Tanker Design Selection
271
greater design margins can be included such as twin screw or lower operating stress
levels.
Boiloff. Higher boiloff requires a larger or faster ship for given deliver ability ;
alternatively, relique faction of boil off may be viable (see below).
Drydocking Facilities. This may limit application of larger ship sizes for specific
projects.
Ship Speed
The factor of ship speed is directly related to ship capacity in order to obtain
specified annual deliverability, but other considerations will apply, such as:
Maximum Practicable Power on Single Screw. This will also depend on draft,
to obtain sufficient propeller immersion, and shipyard engineering capability.
Relationship between Trial and Service Speed. Ship speed relationships need
to be established for weather and sea conditions expected on specific runs.
Anticipated Falloff in Speed. This is largely attributed over the duration of the
project to bottom roughening or fouling; some port locations are more prone to
fouling than others.
Hull Form. This factor may affect vibration, slamming, and also economic
application of the specific design system selected.
Number of Ships
In addition to considerations already mentioned, particularly those of project
reliability and interchangeability, the following may influence the final decision on
the number of ships to be employed:
Avail ability of Existing Ships. Obtaining existing ships speculatively at low cost
should not be overlooked. The recent charters for the Abu Dhabi/Japan project
took advantage of several small ships being available at very competitive rates when
very much larger ships would clearly have been preferred from a pure project optimization exercise.
Matching Liquefaction Plant Streams. This calls for efficient utilization of annual
refit and plant overhaul periods.
Project Phasing, Multiport Discharges. This item calls for efficient matching of
ship and shore facilities.
Ownership. Options for project participants to own a percentage of the shipping
need to be considered.
Design Selection
Obviously, there is still the difficult problem of deciding which containment
system to adopt. This presentation makes no attempt to make this selection, only to
provide a check list of the factors to be considered. Individual owners will wish to
apply their own "weighting" to these items. The main factors to be considered include:
Costs (Capital and Operating). Capital cost is relatively simple to define; longterm operating and repair costs are much more difficult; system reliability has
considerable economic significance in a LNG operation.
Availability. Shipyards tend, for a number of different reasons, to be tied to
specific systems; thus, individual shipyard order books may influence the availability
of that system. Similar considerations may also apply if the shipbuilding industry of a
country is influenced by political considerations.
Guarantees. These vary from yard to yard and system to system. Furthermore,
standard shipyard-type guarantees are in some respects of little interest to LNG
272
R. C. Ffooks
shipowners, e.g., a nominal penalty if the trial speed falls x knots below the specification or the right to reject the ship if it falls 2x below the specification. The owner
needs the specified ship on the specified day.
Proven Operating Experience. As already mentioned, long-term operating
experience, an important factor in any shipowner's mind, is still an extremely rare
commodity in the LNG business.
Proven Materials of Construction. These include aluminum, 9 % nickel, stainless
steel, and Invar (36 % Ni) for tank construction. Balsa, PVC, polyurethane foam, glass
or rock-wool, plywood, and others have been used for insulation along with some
adhesives.
Proposed IMCO Regulations. These regulations are still not finalized, but
enough is known to provide general guidance. The final draft is intended to be ready
by early 1974.
Boiloff. This varies considerably among designs and has a significant effect on the
project economics; reliquefaction may soon become an economic reality.
Reliability. This is an item of considerable importance and will be considered
more carefully later.
Ease of Construction. The ability of shipyards to meet design tolerances on a
production basis will affect first cost, delivery, and performance.
Access to Critical Areas. An important consideration during operation and
survey periods.
Sensitivity to Operational Malfunctions. This aspect involves the reliance on
instruments and controls; some designs are more sensitive to damage in the event of
malfunction than others.
Repair and Maintenance Facilities. Consideration must be given to the need for
special servicing techniques.
Extent of Classification Approval. Some designs have approval in principle,
some have full approval of all engineering detail, some fall somewhere in-between.
Operational Factors. These include such factors as visibility from the bridge,
voyage maintenance, ready access to operating (cargo handling) equipment, etc.
Operating Compatibility. Worthy of consideration particularly if more than
one system is used for the same project, e.g., nitrogen requirements, warmup facilities,
speed, interchangeability of parts, etc.
Size and Draft Limitations. Some designs impose greater length and draft
requirements than others, which may become significant in some locations.
Resistance to Major Accidents. This includes such items as collisions, fire,
grounding, etc.
There are undoubtedly other considerations, but the length of this list so far
indicates the difficulty in making a choice. Unfortunately, there is a temptation to be
unduly influenced by the factors which can most easily be quantified.
Several of the above-mentioned factors deserve more detailed treatment, and
indeed, in the ship selection process, must be studied in detail. One is the effect on
the overall economics of the boiloffrate. The author's company has already presented
a short study [I] on this which concludes that a reduction of 0.1 %/day in boiloffwill
result in a saving in annual operating cost of up to $300,000 for a typical current
project; this in turn can be equated to a reduction in capital cost of $3.6 x 106 per
ship. These figures are clearly large enough to justify a careful examination of this
particular item for any project under consideration.
Reliability is more difficult, and indeed is perhaps the most difficult single factor
to define. Again the author's company has published an article on the subject and
273
Some Important Factors in LNG Tanker Design Selection
4.0
1.1
•.0
1! t.O
tn.
1-1:
0
4
\J : 1.1
Q
./
V
1.0
.9
.10
Fig. 2. Cost and horsepower for
reliquefaction plants.
./
V
I.S
1.0
9QOO
lDOO
a
...
_O~
o
./
/
.ooo~
on
_a
o
%
'000
-_..... .....11
1.1
./
.t!
I
"0
t.o
t.I
mrmNUY"sr·
BOllOH
.10
.~I
4.0
the effect of unscheduled down time [2] ; there is therefore no point in repeating the
arguments here. It is sufficient to say that the effect of down time on an LNG project,
regardless of the way one looks at it, is very significant.
The economics of shipboard reliquefaction is the third point, and one which is
being given an increasing amount of attention today. The current cost relationship
between fuel oil prices and delivered price of LNG, together with what appear to
be realistic proposals from manufacturers of reliquefaction equipment, indicate that
serious consideration should now be given to fitting this equipment. In doing so,
there are a number of factors to evaluate, which include:
1. Cost of equipment installed. Figure 2 provides a general guide based on
current indications.
2. Daily operating and maintenance costs.
3. Standby arrangements to meet regulatory requirements.
4. Additional oil fuel bunkers; also, increased ship displacement and draft.
The third item has a significant impact on the previous two. In the author's
opinion, it seems that the most realistic solution will be to retain the dual fuel burning
system as standby.
This solution leaves only the current U. S. Coast Guard requirement (which
forbids venting to atmosphere in port) to be met. This is currently done by installing
an oversize condenser so that excess steam can be released directly to the condenser.
This procedure could be retained as a final standby for the period when the ship is
stationary, in a port area, not connected to shore, and the reliquefaction plant inoperative. It seems that the likelihood of this combination of events occurring is so
remote that the U. S. Coast Guard might be persuaded to relax their present restriction
on a procedure which in any case appears to present little or no hazard; this would
then obviate the need for an oversize condenser.
274
R. C. Ffooks
DOLL . . . .
I
'0' 'lUi,
SilES PRICE OF lNG
Fig. 3. Relationship of fuel oil cost and
LNG sales price for two boiloff rates
and ten-year payout.
A recent variation of the reliquefaction theme is partial reliquefaction; this of
course, reduces the installed plant and operating costs significantly but introduces
problems with regard to standby equipment. Figure 3 shows the beneficial effect of
lower boil off rates on the break even point when reliquefaction becomes economical.
OPERATING EXPERIENCE-SOME LESSONS LEARNED
The previous section of this paper summarized the major factors which the
shipowner must consider in deciding the size and type of ship to engage in LNG
traffic. An equally long list could be supplied for the benefit of the shipbuilder, but
space is insufficient. The owner will, however, be interested in a short summary of
Conch Methane Services' experience as a shipowner. What has this experience to
date taught us?
The first and most important fact is that we do not yet know all the answers.
In addition, the following thoughts are offered:
1. Do not ignore the so-called "conventional" parts of the ship; these can be
the cause of expensive down time if not built to the highest specifications.
All current LNG technologies rely to a greater or lesser extent on the integrity
of the conventional ship structure or equipment.
2. In the life of the ship, there will be many staff and crew changes, most planned,
but some unplanned. Simplicity of operation is a factor which becomes
increasingly important as the years pass: Effective passing on of know-how
is essen tial.
3. A clear understanding by the operating personnel of the reasoning behind
the design and all equipment is essential to ensure that malfunctions are
correctly interpreted and the proper remedial action taken.
4. Instruments and control systems are still not completely reliable. There is
still no substitute for a well-trained human on the spot.
Some Important Factors in LNG Tanker Design Selection
275
5. Provision of a detailed operating manual and its strict observance at all
times is absolutely essential.
6. Theory and statistics of the marine environment are still quite new.
7. Ease of access for inspection and maintenance during annual overhauls,
particularly of critical areas, is important. Time out of service is an expensive
commodity and the more so when it is unscheduled.
SUMMARY
Relatively soon, we shall begin to accumulate experience with larger LNG ships
and new designs. In addition, the industry will be guided by a set of internationally
agreed upon design and operating guidelines. This will, of course, simply provide a
solid jumping off point for further advances in ship size and technology, all aimed at
reducing, or at least slowing the rise in, delivered cost of gas.
There is much fascinating work on hand upon which this paper has not attempted
to touch and there are still great challenges ahead. But we must be sure that we
carefully note and benefit from the relatively short, though certainly successful,
experience we have gained.
ACKNOWLEDGMENT
The author wishes to acknowledge the assistance of the 1. 1. Henry Company in preparing some of the
material for this paper.
REFERENCES
I. "Boil-off as an Economic Factor in the Selection of LNG Tankers," Conch Methane Services Ltd.
publication, Villiers House, Strand, London, England.
2. P. L. L. Vrancken, Petroleum & Petrochemical Intern., 12(8): 26 (1972).
G--#
NEAR-TERM TRENDS IN LNG TANKSIDP DESIGN
J. L. Howard
Kvaerner-Moss, Inc.
New York, New York
INTRODUCTION
The transportation of natural gas in its liquid state aboard ocean tankers began
just about fifteen years ago, the first trade being done in a converted dry cargo ship.
Today there are still only about fifteen LNG tankers in service with an average
capacity of less than 40,000 m 3 . The total operating history amounts only to about
seventy-five ship years. In very rough terms, the total ocean LNG volume of trade
over these past fifteen years is equivalent to sixteen ships of 125,000-m3 capacity
operating for one year.
Contrast this with the fact that presently on order or building in the world's
shipyards are forty LNG tankers, or the equivalent of about thirty-three ships of
125,000-m 3 capacity each. All of this says just three things: our operating experience
is small, our construction program is extensive, and the unit size ofthe fleet is growing
very rapidly.
On the basis that what is past is prologue, it should be beneficial to examine the
design and operating history of LNG tankers before attempting a look ahead. The
earliest ships used internally stiffened, structurally self-supporting cargo tanks of
basically prismatic shape. Gradually, however, for ships with delivery during 19641974, especially in the latter half of that period, the membrane tank configurations
became predominant. The reasons for this change in design preference are rather
elusive, but can probably be reduced to: (1) nationalistic preferences of owners for
domestic shipyards and domestic designs, and (2) apparent lower initial capital cost
compared to then existing competitive designs. For ships now building or on order
with deliveries after 1974 (i.e., contracts placed mainly after 1971), the preference of
owners has shifted strongly back to the rigid, self-supporting tank designs. Among
the reasons for this second shift are owner concern for life cycle cost, apparent shipyard preference for certain lower labor, self-supporting tank designs, and labor cost
increases driving membrane ship prices above some self-supporting tankship prices,
especially as in the case of the newly introduced spherical self-supporting designs.
Figure 1 shows the past design trends.
Although the operating experience of the world LNG tankship fleet is relatively
short and probably insufficient from which to draw conclusions, it is fair to say that
the ships have been run with a very gratifying overall safety record. Those cargo
containment system failures that have occurred have been discovered early enough
so that neither major port safety incidents nor vessel losses have resulted.
In the membrane systems, several failures in primary barriers have been caused
by design deficiencies; however, the total containment system in each case remained
r76
Near-Tenn Trends in LNG Tankship Design
277
TOTAL CARGO CAPACITY (M3)
250,000
Self-supporting
Before 1970
Membrane
Self-supporting
1970-1912
Membrane
Self-supporting
1973-1974
Membrane
----
500,000
750,0001,000,000
~
Self-supporting
1975-1976
Membrane
After 1976
Self-supporting
,0
Membrane
I
Fig, 1. Trends in tank system design, LNG ship tankage capacity built of
self-supporting and of membrane systems.
sufficiently intact to avoid external release of cargo or hull structural damage. With
one exception in a prototype vessel, membrane fatigue failures have not been reported
in large numbers, although this record may change with added sea life. Similarly,
these ships have not been troubled as yet with hull fatigue cracking which could
result in moisture freezing out in the secondary barrier insulation.
The self-supporting tank systems have performed relatively well in service,
with the exception of some fatigue cracking in the stiffening structure of the earliest
tanks, without release of cargo. Hull fatigue cracks and the resultant water leakage
have also caused the partial loss of the cargo insulation fixed to the inner hull of one
ship.
Taken as a whole and in the context that this is a dramatically new marine
transportation concept, one must conclude that LNG tankship performance to
date has been successful, though we have learned that there are design features that
must be avoided and others that can be improved.
Now, what of the future? Let us consider what are believed to be the criteria and
boundary conditions which must be recognized and accommodated in any successful
future LNG ship designs:
1. The vessels will be employed in liner-type trade, on long-term fixed routes
and fixed schedules. Operational reliability is of utmost importance. There
will be little speculation in the traditional tanker sense of short-term or spot
charters.
2. LNG ship prices will not be reduced below their present levels and, in fact,
will probably continue to rise. This means that transportation costs in any
LNG scheme will continue to account for a very large share of the total
price of delivered gas; as high as 50 % or more. It is highly significant to note
that ship capital costs account for about 80 % of total charter hire costs.
Vessel insurance, which is related, of course, to acquisition cost as well as
27B
J. L. Howard
3.
4.
5.
6.
risk, accounts for over half of the remaining 20 % of the charter cost. In short,
shipbuilding prices are all-important in any LNG import scheme, and in
some cases will be determining.
Though there probably will not be a world shortage of LNG building berths
in the near future, demand will be high. At the same time, the independent
shipowners will probably take an increased interest in LNG tonnage as the
market expands and as LNG project developers' capital is consumed in the
nontransportation ends of the projects. These owners will be willing to
speculate in the sense that they will seek out and block the most desirable
berths with least cost. The effect of all of this may then be that shipyards, as
much or more than owners, will decide which of the accepted LNG designs
will be built.
Shipbuilding labor cost increases will cause labor to account for larger
proportions of total unit ship costs. Shipyards, owners, and designers will
seek ways to reduce the labor intensity of LNG ships.
Gas prices have more room to increase than fuel oils, and will probably
increase more rapidly, because of past forced depressed prices in some countries and forced use of the cleaner fuel in certain consumption areas.
Classification society rules, governmental agency regulations, and international rules and recommendations are now starting to be published specifically for LNG tonnage. As operating experience is obtained in meaningful
quantities, these rules will be amended based on experience and anticipated
problems. The designer must not only recognize existing rules but should
anticipate future changes.
All of these criteria or boundary conditions point to one basic fact: Design
changes in LNG ships are expected to occur slowly and to be an evolutionary process
of improvement. Rather than being willing to risk a high capital, very long-term investment to immoderate or drastic design changes, owners will instead seek ways of
improving what has already been shown to work best.
With this as background, let us now consider where LNG ship design changes
may occur, and where improvements may be most needed.
CARGO CONTAINMENT
In a 125,OOO-m LNG tankship, the cargo tanks and insulation account for
about 20 to 30 % of the ship's price, the variation depending mainly on tank type.
This then is a fruitful area in which to work to reduce capital cost. The trend has
begun, and will probably continue, that the less labor-intensive tank systems will
be favored by the shipyards. At the same time, tank systems that are most amenable
to reliable stress analysis and fracture safe design will have the best promise of least
weight and elimination of the secondary barrier. It is the feeling of the author that
these conditions will be disadvantageous to the membrane systems and will be most
closely met by the structurally self-supporting systems, especially those without
stiffeners and with the least welding. The long-term maintenance and repair costs
also appear to favor the more rigid tank systems, though more operating experience
is probably needed in this regard. The problem posed by the light membrane structure
is not only its susceptibility to damage in service, but its greater probability of secondary damage at the hands of the repair shipyard. When one considers that the daily
hire of a large LNG ship is of the order of S60,OOO, to say nothing of the loss of cargo
revenue, it is not surprising that the owner of a heavy walled structural tank ship
3
Near-Term Trends in LNG Tankship Design
279
involved in a charter rate competition can afford to be somewhat more direct in his
off-hire computation (i.e., anticipate less unscheduled off-hire) than the owner of a
membrane tankship.
In summary, it is felt that the future trend in containment design will continue
toward those systems that can best take advantage of new technology in stress analysis
(including the statistical failure analysis techniques), metallurgy, and fracture mechanics, as well as labor-saving production innovations. All of these point toward
simplified, structurally self-supporting, single-barrier tanks.
CARGO TANK INSULATION
Cargo tank insulation is the single most important area for additional work
and near-term improvement in LNG ships today. There are a number of insulation
systems in service and coming into service which will work; however, the cost spread
of these is broad, perhaps 3 to 7 % of the ship's price, or between 10 and 35 % of the
cost of the entire tank/insulation containment system. The wide range of costs is
due mainly to labor intensity and installation time variations. Those systems that
need only a single containment barrier, and those that use the insulation strictly as
an insulant (rather than as a load-bearing member or as an LNG liquid-tight element
of the containment system) are probably the least complicated and can therefore
benefit most readily from new ideas. Insulants which must be relied upon for over
twenty years of liquid-tight life in ship service will take longer to perfect and cost
large sums to prove out.
There is currently a great deal of interest in "inside" or "wet-wall" insulations.
These could conceivably have the advantage of least heat leak to the cargo and, in
the extreme, could eliminate low-temperature structural material from the hulls and
containment support systems. The problems with such containment systems which
must be realistically resolved are:
1. Installation: They must be capable of being installed in large "ambient
condition" cargo holds. By "ambient condition" is meant the normal working
condition in a cargo hold during ship construction (temperature and humidity
control as well as unusual surface preparation are extremely difficult and
expensive to achieve).
2. Service, maintenance, and repair: The system must be capable of withstanding
long-term service loads and ship hull deflections. It must also be capable of
being quickly and reliably gas-freed, and equally capable of being entered
by shipyard workmen for inspection, repair, and maintenance of the tank
internal outfit. These systems generally still require very extensive testing
and proving, not the least of which must be aimed at assuring the integrity
of the containment, even in the event of possible service derangement of
internal tank hardware (such as ladders, piping, and instrumentation coming
adrift).
In summary, the cargo insulation is probably the single element in the entire
cargo system most in need of improvement. In the near future, we should strive to
achieve changes in insulation materials and installation procedures which will reduce
life cycle costs through lower heat leak and reduced maintenance; which will increase
the safety of the ships because of reduced combustibility; and which will reduce
installation time and therefore initial cost.
J. L. Howard
CARGO HANDLING SYSTEMS
Cargo handling and auxiliary systems in LNG ships today are relatively uniform
among designs (except that the two-barrier membranes require more sophisticated
instrumentation and control to maintain the desired state of equilibrium between
the barriers) and are perhaps even less complicated than in warmer-temperature gas
ships, because of the absence of reliquefaction equipment.
This condition will not last much longer; reliquefaction in LNG ships is at
hand. These systems will add probably 3 to 5 % to the cost of the vessels, depending
on the cycle used, when all side-effect changes are included. At this rate, the liquefaction plant will itself account for about one-third of the total cost of the cargo
handling plant. The go/no-go decision point for on-board liquefaction equipment
depends, of course, on the relative value of the gas cargo vs. fuel oil, the length of run,
and time at sea. Another parameter in this decision, which is at least as important
as the direct cost forecast, is the availability oftrained operating crews and the marine
service reliability of the equipment. In the opinion of many owners, the question of
maintenance of this vital (once installed) and sophisticated plant is of primary
importance. It must be remembered that when this plant is down, not only is the
main propulsion plant probably operating off-optimum, but the vessel may face
regulatory problems with boil off disposal on arrival in port.
To some extent, the possible limitations in crew aptitude may be overcome by
the designer in selecting cycles and equipment which are least sophisticated and most
robust and reliable. In this context, a compromise must be achieved between cycle
efficiency and operating reliability.
Even with the maintenance difficulties which must be recognized with reliquefaction plants aboard ship, it seems quite certain that the first installations will be
contracted in the not-too-distant future, and will eventually become standard outfit,
especially on ships for the longer trade routes.
The final subjects which must be mentioned in any attempt to forecast LNG
ship design trends are size and speed. These parameters tend to go together, since
the larger, longer vessels will be relatively more economical for the higher speeds
than will the smaller, shorter vessels. At present, the "standard ship" seems to be
120,000 to 130,000 m 3 capacity, with about 40,000 SHP available to move the vessel
at about 20 knots. With the very shallow draft in LNG tankships, 40,000 SHP is
just about the limit of power on a single screw. Even so, the designer is unusually
concerned with propeller-induced hull vibration. In order to attain higher speeds
in a 125,OOO-m3 ship, one is forced to go to twin screws, which, of course, causes a
quantum increase in cost. It is important to note also that incremental speed increases
in the neighborhood of 20 knots for ships of this size can be attained only with corresponding cube or fourth power increases in propulsion plant output. The greater
length and deeper draft of the larger (160,000 m 3 seems to be the next size step
frequently mentioned) ships are better suited to higher speeds. Additionally, the
larger ships will be most competitive on the longer runs where, again, higher speeds
will pay.
The main question still basically unanswered with the larger ships is cost. Some
prices have been indicated, but there has n~t yet been sufficient firm interest to
establish competitive contract prices.
Shipyards themselves are somewhat reluctant to move too rapidly to larger
ships because of added risk, conflict with other tonnage for the larger building berths,
and the uncertainty of the sales market compared to an increasingly well-defined
market for the "standard" units.
Near-Tenn Trends in LNG Tankship Design
281
On the owners' side, there are uncertainties not only of first cost, but of insurance
rates, possible severe disruption in cargo delivery schedules caused by off-hire of a
single very large cargo unit, and the innate risk in being first to take a quantum step
beyond the norm.
Even in the face of these uncertainties, plus the added question of regulatory and
port authority approval, interest in the very large, high-speed ships is persistent and
it is certainly only a matter of time, and probably a relatively short time, before a
shipyard, an owner, and a project are suited and matched to proceed.
SUMMARY
This discussion has intentionally focused on the marine applications of cryogenics and has tried to emphasize the severity of service that will test both shipboard
LNG systems and equipment design. In fact, since an LNG tanker is expected to be
in service for at least twenty years, these systems and designs will be tested essentially
continuously during that time in all six degrees of motion.
G-5
STATUS REPORT ON LNG TANKER DESIGNS
A. Pastuhov
Gazocean USA, Inc.
Boston, Massachusetts
and
M. Gondouin
American Technigaz, Inc.
Boston, Massachusetts
INTRODUCTION
The largest LNG project, the EI Paso project, was finally approved by the
Federal Power Commission (FPC) in October 1972, some fifteen months after the
hearings were held; but it took EI Paso Natural Gas until March 1973 to finalize
the liquefaction plant financing and an additional three months to order the three
ships it requires. There are now in front of the FPC three major applications to import
Algerian LNG: the Easco project for an estimated 600 million ft 3/day, the Phase II
EI Paso project for 1.0 billion ft3 /day, and the Distrigas project for 120 million
ft 3 /day. As of August 1973, no hearing date has been set for any of these projects.
Substantial delays in obtaining any FPC order may place all three projects in jeopardy
of cancellation [1]. However, if an optimistic viewpoint is maintained, Tables I and II
summarize the LNG ship construction potential and the LNG ship requirements.
Table I indicates that the United States has the potential to build 72 LNG tankers
between 1977 and 1985 (the same as estimated for Japan but using six berths out of
fourteen as compared to only six out of twenty-four for Japan). Table II shows a
requirement potential of 137 U. S. vessels between 1977 and 1985, or almost twice
as many as can be produced here. However, the two Russian projects alone represent
45 % of the U. S. total requirement. The U. S. shipyard industry consequently faces
two major unknowns in gearing its production capacity-the actions of the Federal
Power Commission and the political uncertainties of the 1981-1985 period, which
requires an estimated 75 ships compared to 62 in the 1977-1980 period. Obviously,
the LNG tanker business must be considered simultaneously with the petroleum
crude oil carriers but, there again, the shipyards face another unknown--Congresswhich mayor may not pass a minimum U. S. shipping law for oil imports.
DESIGNS USED IN COMMERCIAL LNG SHIPS
Integrated Tank Designs--General Features
Two designs used in ships of commercial size, the Gaz Transport Invar membrane
and the Technigaz stainless steel wafBed plate, have been described elsewhere [24].
Their common features are summarized in Table III. They differ, however, in many
ways and Table IV summarizes their major differences.
2Bl
283
Status Report on LNG Tanker Designs
Table I. LNG Requirement Potential
Number of ships·
1977-1980t
1981-1985
Total
6
14
11
13
14
Imports to U. S.t
Algeria
Nigeria
S. America and Alaska
PacificfWest U. S.
USSR/East U. S.
USSRfWest U. S.
U. S. Subtotal
20
11
62
17
20
11
75
19
28
11
17
40
22
137
Imports to
Western Europe§
Japan§
Potential ship requirement
10
25
97
10
25
110
20
50
207
• Ship size 100,000 to 160,000 m 3 capacity.
t NPC Report, December 1972, plus proposed USSR projects.
t Does not include ships in operation or that should be delivered prior to
1976 (approximately 8 U. S., 10 Japanese, 20 Western Europe).
§ M. W. H. Peebles, "LNG Trade in the 1980's," paper presented at LNG-3,
Washington, D.C. September 1972.
Table II. LNG Ship Construction Potential
I. Total worldwide shipyard capability· (120,000 m 3 min. capacity)
Shipyards
Construction berths
U.S.
6
14
Western Europe
37
41
Japan
9
24
52
79
II. Potential for LNG constructiont
Shipyards
U.S.
3
Western Europe
12
Japan
4
19
Berths
6
12
6
24
III. Potential number of LNG ships delivered
1977-1980§
1981-1985
U.S.
32
40
Western Europe
48
60
Japan
32
40
112
140
Ships per yeart
8
12
8
28
Total
72
108
72
252
• Figures approximate for current shipyard facilities plus indicated expansion plans.
t Based on LNG construction experience, bid interest, and licensing
agreements for containment systems.
t Deliveries starting in 1977.
§ Excluding ships already in operation and U. S. and European LNG
tankers that should be delivered prior to 1977.
A. P. . . .v ..... M. GoadouiD
m.
Table
Common Features of
Integrated Tank Designs
Load-bearing insulation
Flexible metal containment
Secondary barrier
Barrier integrity constantly monitored
Designed for fatigue resistance
Liquid motion reduced by chamfers
Prefabricated insulation elements
Both primary and secondary barriers are identical in the former design, whereas
in the latter system, the secondary barrier is made of i-in.-thick sugar maple plywood
glued to the balsa insulation. This is the same material used in the Methane Pioneer
since 1959. It has proved to be a very durable material under cyclic loads. The
primary barrier is the well-known WafHed Plate, attached to the balsa insulation
panels by means of stainless steel anchors welded to the edge of the plates.
The corrugations in the wafHed plates provide sufficient flexibility to reduce all
edge loads due to residual thermal stress to a low value (about 400 lb/ft). A flat membrane, on the contrary, remains subjected to appreciable thermal contraction even
when made of Invar (contraction 2.5 x 10- 4 in./in. under service conditions), giving
an edge load about four times as large as for a wafHed sheet. This thermal load, essentially static, is superimposed on the dynamic loads due to the cargo weight and ship
deformation and local stress concentration in the vicinity of the anchoring points.
Because of the simpler configuration of the Invar flat membrane, automatic
welding was used at an early date, whereas the development of automatic welding
equipment for the wafHed membrane took longer. However, since stainless steel is
Table IV. Integrated Tank Designs
Gaz transport
Primary barrier
Insulation
Technigaz
Shape
Material
Thickness
Tensile strength
@ -26O"F
Residual edge
loads
Forming
Welding
Leak testing
Flat strips
INVAR
20 mils
140 x 103 psi
Waffied
304L stainless
52 mils
250 x 103 psi
1600psi/ft
Aboard ship
Resistance
Freon and halogen
detector
400psi/ft
Material
Loose perlite in
plywood boxes (two
layers)
Solid balsa and plywood
on wooden grounds
and studs
Construction
Mechanical fasteners
Adhesives under controlled
atmosphere
20-mil INVAR
JOmole %
4 x 10- 4 in./in.
l-in. Maple plywood
1 mole %
7 x 10- 4 in./in.
Secondary barrier
Methane in interbarrier space N2
Acceptable hull deflections
Prefabricated
TIGArc
NH3 and reactive paint
Status Report 011 LNG Tanker Designs
285
easier to weld, satisfactory manual welding procedures were used in the construction
of the first large commercial ships based on this principle. Because of higher labor
costs in U. S. shipyards, it is expected that welding of the waffled membrane will
largely be done automatically in this country.
The thinner gauge Invar membrane strips, because of their length, are rather
difficult to handle. They are formed within the hold just before assembly and welding.
The waffled plates (approximately 3 ft x 10 ft), on the other hand, are prefabricated
in a special plant and shipped to the yard, ready to be welded.
Acceptable limits for methane content in the nitrogen of the interbarrier space
have been taken as about 1 mole % in the waffled plate system and about 30 mole %
in the Invar membrane system. Correspondingly, the leak testing procedures used
are very different. For example, a sensitive ammonia gas diffusion test is used for
the waffled membrane, with a reactive paint indicator to detect the slightest defects
in the welds. On the other hand, the leak testing procedure for the interbarrier space
of the flat membrane has used Freon injection along with a relatively low-sensitivity
halogen detector. The same leak testing procedure is used for both Invar membranes.
However, for the secondary barrier (constructed of plywood) in the waffled membrane
system, a simple vacuum test is used continuously throughout the entire installation
of the waffled plates in order to make sure that the secondary barrier has suffered
no damage during all subsequent work in the tank.
The waffled membrane and its solid wood insulation support have been subjected
to extensive fatigue tests which provide reproducible results, making it possible to
reliably use a statistical approach to determine the lifetime of the containment
system under the most extreme operating conditions. The capabilities of the composite structure of plywood boxes and Invar membranes are more difficult to evaluate,
because of the unknown influence of possible sliding motions and relative displacements of the discontinuous elements entering into the structure and subjected to hull
deformation. Because of these uncertainties, an overscantling of the ship's hull is
required so as to limit maximum elongations to 4 x 10- 4 in./in. instead of the
7 x 10- 4 in.jin. value normally allowed by classification societies.
Accidental overpressurizing ofthe interbarrier space, which occurred in both the
Pythagore and the Polar Alaska in the early history of these ships, has led to the
redesign of automatic controls and pressure release equipment which make the
operation of integrated tank ships no more difficult than that of LNG ships with
self-supporting tanks.
Self-Supporting Tank Designs
For large LNG ships, only two different systems have been shown to be economical and they have also been previously described [5]. These are (1) the double-barrier
prismatic aluminum tanks ofthe Esso Research and Conch design and (2) the spherical
aluminum or 9 % nickel steel tanks with a reduced secondary barrier. Two spherical
designs have been used commercially, that of Moss-Rosenberg, using a cylindrical
supporting skirt, and that of Technigaz, using articulated supports. The characteristics
of these four systems are summarized in Table V.
For small ships, cylindrical tanks made of either aluminum or 9 % nickel steel
have been used. These are designed as pressure vessels. They can be arranged vertically
as in the tanker Jules Verne or horizontally as in the barge Massachusetts.
From the standpoint of ship construction, all self-supporting tank designs
present the following characteristics: The tanks can be built separately, brought to
the shipyard and installed in the finished holds. This, however, requires that the
A. PlIStuhov IUId M. GondouiD
Table V. Self-Supporting Tanks
Design criteria
Experience to date
Prismatic
Spherical
Liquid loads
Shear forces at keys
Fracture mechanics
Liquid loads
Bending at skirt edge (Moss)
Tangential shear at supports (Technigaz)
Buckling at partial filling
Fracture mechanics
Limited to one small ship
Good (three ships)
Main Features
Prismatic
Containment material
Secondary barrier
Insulation
Control of liquid motion
Roll and pitch keys
Spherical
Containment material
Secondary barrier
Insulation
Supports (vertical)
Supports (horizontal)
Conch
5083 Al
Maple plywood
Balsa and polyurethane foam
Swash bulkheads
Top and bottom
Moss
5083 Al and 9 % Ni
Limited to drip pan
Plastic foam
Cylindrical skirt
Skirt bending
Esso research
5083 Al
5083 Al
Polyurethane foam
Bottom trunk and bulkhead
Side and end walls
Technigaz
5083 Al and 9 % Ni
Limited or absent
Glass wool or perlite bags
Articulated local supports
Keys or articulated supports
ship's deck be installed at a later stage in the ship construction, and that all deck
piping installation be postponed until the deck is complete. It also requires the use
of cranes in the 600-900-ton class to lift the tanks into the holds, a capability which
no U. S. shipyard presently has.
Prismatic Tanks. Conch prismatic aluminum tanks were used for the first-time in
the Methane Pioneer and have since been used in the Methane Progress and Methane
Princess. The complete secondary barrier, initially made of plywood glued on balsa
insulation panels, was later reduced on the upper side insulation to a splash barrier
of plywood shingles supported by a wooden framework, the balsa panels and plywood
surface forming a catch vessel in the event of failure of the primary containmebt.
Progressively, the amount of balsa used as insulation has been reduced, to be
replaced either by glass wool (on the top and upper sides) or by reinforced polyurethane
foam (on the bottom and lower sides). The tank structure itself has also been refined
using finite-element analysis techniques. However, because these computer models
generally do not represent the structure in its entire detail, the U. S. Coast Guard and
classification societies have maintained the requirement for a full secondary barrier
or a large-capacity catch vessel. Aluminum roll and pitch keys located at the centerlines of both the top and bottom ofthe tank are held in hardwood key ways attached,
respectively, to the bottom insulation and to the underdeck structure.
The Esso Research design uses double-wall 5083 aluminum tanks, the outer
walls being the secondary barrier to which are attached two layers of polyurethane
or PVC foam held in place, respectively, by plywood sheets anchored to welded studs
and by aluminum sheathing secured to the plywood by lag screws.
Status Report on LNG Tanker Designs
1B7
Spherical Tanks. Two designs have presently received acceptance from shipyards
and shipowners: namely, the Moss-Rosenberg design and the Technigaz design.
The common features and differences of these two designs are enumerated in the
following paragraphs.
In the case of large tanks, both designs make use of the concept "leak before
failure" of the primary containment vessel to reduce the secondary barrier to a drip
pan and spray shield over the insulation of the tank. This concept is derived from
the assumption that with spherical tanks made of thick plates of suitable cryogenic
materials, any flaw or crack in the shell would propagate slowly enough at the design
stress level to reach the outer surface of the tank long before any catastrophic failure
of the tank could take place. The small leak thus created would therefore be detected
and repaired long before the integrity of the containment system is significantly
reduced.
Two materials have been used in the construction of the spherical tanks for
LNG ships; 9 % Ni steel and 5083 aluminum alloy. The latter material appears to
be presently economically more attractive than the first. For this reason, a great deal
of work has been performed by suppliers and designers to substantiate the basic
assumption under which the "leak before failure" concept has been derived.
It is well known that cracks have appeared in some of the aluminum prismatic
self-supporting tanks, but they were always attributed to minor design or construction
defects in lccally stressed parts of the structure. No such cracks have ever been found
in the spherical tanks ofthe only LNG ship ofthis kind presently in service. The tanks
of this small ship, however, have been designed as pressure vessels and do not constitute a good application of the "leak before failure" concept. Furthermore, they are
made of9 % nickel with a thickness of9 mm (i in.) only. The plate thickness required
for building aluminum tanks in the larger ships may exceed 2 in. in some parts of
the sphere, and fracture mechanics test data for such thicknesses are still quite limited.
The laboratory tests designed to evaluate the low-temperature toughness of such
thick aluminum alloy plates must faithfully reproduce the cyclic stresses and lowtemperature conditions to which the material is submitted on board an LNG ship.
At present, the data available for such thick aluminum plates are considered insufficient to fully ascertain the validity of the derived concept for large spherical tanks.
It is not until the completion of the extensive testing programs presently undertaken
by the Materials Properties Council and by other research laboratories that designers,
classification societies, and regulatory agencies will have in hand all necessary
information on this subject.
With 9 % nickel steel, the required plate thickness is considerably less because
of the higher allowable stress, so that this material can be used safely for the construction of small to intermediate LNG tanks. Nevertheless, more work is necessary to
determine the full capabilities offered by this steel in the design oflarger LNG spherical
tankships, assuming that material and construction costs using 9 % nickel steel
remain competitive with those using 5083 aluminum.
The main differences between the two spherical tank designs reside in the supporting systems ofthe spherical tanks and in the layout ofthe insulation. In the MossRosenberg design, the horizontal dynamic loads, as well as the vertical dynamic
loads, are transmitted to the ship bottom by means of a rigid cylindrical skirt. The
cargo weight and transverse accelerations due to the ship's motion create bending
moments as well as compressive stresses at both ends ofthe supporting skirt; namely,
at the equator of the sphere and at the junction with the bottom plate of the ship.
A. Pastuhov and M. GoodouiD
The Technigaz design uses a system of articulated vertical supports and tangential roll-and-pitch keys to reduce all bending moments applied to the spherical shell
to very low values, so that the primary stresses in the shell plate are tangential and
in tension. The roll-and-pitch keys apply some bending moments and shear forces
to the hull structure, but these are located at a level close to the centerline of the
ship, where all other stresses due to hull deformation are at a minimum. Thus, risks
of developing fatigue cracks causing water leaks from the ballasts into the insulation
space are minimized.
The insulation in the Moss-Rosenberg design is applied directly against the
outer surface ofthe spherical shell. It is made of polyurethane foam. In the Technigaz
design, the insulation is applied against the double hull of the ship. It consists primarily
of glass wool mats, separated from the sphere by plywood or aluminum shingles
serving as a spray shield to protect the insulation in case of a leak in the tank.
From the standpoint of the ship's operator, all self-supporting tanks, whether
prismatic or spherical, require rather long times for all warmup and cooldown operations because of the large mass of metal to be cooled and because of the need to
avoid any large thermal stresses due to temperature gradients during these operations.
Semimembrane Techniques. These techniques have been considered by various
companies (Chantiers de l' Atlantique, A. G. Weser), but only the Bridgestone Liquefied
Gas Co. of Tokyo has constructed prototype ships and established designs for commercial size ships [6] (Table VI).
The system uses as primary containment a relatively heavy membrane of 9 %
nickel steel (i in. thick). This flat membrane rests against a 20-mm-thick plywood
buffer placed inside a secondary barrier tank made of i-in. stainless steel. All edges
and corners of the prismatic tanks are replaced by large radii of curvature, ranging
from 5 ft for the secondary barrier tank to more than 8 ft for the primary tank. The
continuous plywood buffer sandwiched between the two tanks also has rounded
corners of intermediate curvature (6 ft). The load-bearing insulation on the bottom is
made of a honeycomb framework filled with glass wool and covered with plywood
sheets. On the sides, a poured-in-place light concrete foam-glass aggregate is used.
The plywood mold in which the concrete mixture is poured is left in place and is
therefore in contact with the secondary barrier tank.
Both primary and secondary tanks are built oversize and compressed by means
of jacks spaced between the inner hull of the ship and the plywood mold, before the
side insulation is poured. No details have been given concerning the elimination of
heat transmission through the jacks. The purpose of the precompression of the tanks
is to partly compensate the thermal stresses in the corners of the tanks. Nevertheless,
the stress due to liquid loads on the bottom edges and corners of both the primary
Table VI. Semimembraue Tanks (Bridgestone)
Containment material
Thickness
Secondary barrier
Insulation
Control of liquid motion
Roll and pitch keys
Testing
Installation
9 % Ni steel or aluminum
1 .
4 In.
i in. stainless steel
Glass wool and foam-glass concrete
None
Load transmitted to double hull by secondary barrier
X-ray of welds
Requires precompression of composite tanks
Status ReJNIrt on LNG Tanker Designs
Table VII. General Comparison of Designs
Membrane
Space utilization
Stability
Weight
Manufacturing investment
Ships in commercial service
Prismatic
Spherical
Semi-membrane
GT
TGZ
Conch
Esso
Moss
TGZ
Bridgestone
Good
Good
Low
Low
3
Good
Good
Low
Low
4
Good
Good
High
High
2
Good
Good
High
High
4
Fair
Fair
High
High
0
Fair
Fair
High
High
1
Good
Good
Medium
High
0
and secondary tanks may remain quite large if the support provided by the loadbearing insulation is not totally effective in this part of the tanks. Total support
would require perfect conformance of the radii of curvature of both tanks and of
the plywood buffer when subjected to precompression, cooldown stresses, and liquid
loads. The designer has not provided details as to how this was achieved in practice.
The design, presented for the first time at the LNG-3 Conference in 1972, deviates
appreciably from earlier descriptions reported by various authors and may indicate
that perhaps additional changes may be introduced during actual ship construction.
The design and construction methods released at this time show a high degree
of complexity and seemingly do not warrant the claims of low cost and high reliability
made by the designers. The cost of the tank materials, although less than that for
self-supporting tank techniques, appears considerably higher than for membrane
techniques, and the difficulties of installation are also intermediate between those
of the preceding techniques (Table VII). It is therefore likely that overall costs will
be in line with those of other techniques. From the standpoint of the operator, cooldown and warmup times required to obtain complete settling of the cold composite
tank structure in the ship's hold may be about equal to those for self-supporting tanks.
The risk of tank damage from accidental overpressurizing of the interbarrier space
is about equal to that for integrated tank techniques and requires the same care in
design of the controls and operating procedures. The reliability of this containment
system under the cyclic loads and the hull deformations experienced by a ship at sea
remains to be shown for large commercial ships.
SHIPYARD CONSIDERATIONS
Table I shows that the LNG requirement potential has been matched with the
present situation as far as a number of berths are concerned. It must be noted that
there are at least another five shipyards which may decide to build LNG tankers
as well as crude oil carriers. How the situation may change will depend on political,
economic, and regulatory factors. Nevertheless, it is evident that there is the capacity
to expand if needed.
What are the considerations of a shipyard desiring to enter LNG construction?
Table VIII attempts to list some of the factors which a shipyard will want to weigh
in deciding which design (or designs) is best suited to their own situation. Each yard
will rate differently in importance the factors that have been listed and each proponent
of a design will rate his concept differently than his competitor's. However, an in-depth
evaluation of the kind described in Table VIII needs to be made. Because of the
uncertainties previously described, such as with regard to the FPC, Congress, and
the international climate, the management of each yard will prefer to make such
290
A. Pastuhov lUId M. Gondouin
Table vm. Shipyard COlL'iideratiolL'i
Self-supporting
Membrane
In-house
installation
Yard investment
(tank and insulation)
Subcontracted
Low
<5MMS
Diversity of suppliers
Low
In-house
construction
Subcontracted
High
50MMS
(approx.)
Low to medium,
depending on
crane
Excellent
Poor
Yard manpower level
(tank)
Medium
Low
Low to medium
Low
Yard manpower level
(insulation)
Medium
Low
Sphere,
unknown,
Conch, low
Low
Manpower interference
between hull work and
containment
Elasticity between
deliveries of components
Applicability to size
larger than 125,000 m 3
Ease of repairs
Low
Good
Medium
Fair
Poor
Excellent
Sphere, poor; Conch, fair
High
Poor
investments as required and still have a flexibility in the construction of either LNG
tankers or very large crude oil carriers.
NEW DESIGNS
Although progress is, in general, both good and needed, it is also the feeling of
the authors that there has been an overemphasis on the need to develop new designs.
Actually, for the good ofthe LNG industry, the utmost caution should be exercised
and, until operation experience is gained, new designs should be postponed. The
four new systems that have been described are each essentially untried and not ready
for commercial application until test verification has been completed. It must be
remembered that even today's spherical design has been tested only on a prototype
of 4000 m 3 and that the first commercial ship to be delivered later this year will be
over twenty times larger. Other ships ofthe same design only seven and one-halftimes
the size of the prototype will be delivered in the next eighteen months and, together
with the larger ship, will provide a very good source of data for the engineers for
possible further extrapolation CJ.
The authors are not sufficiently familiar with a spherical design proposed by
the Chicago Bridge & Iron Company to make a meaningful evaluation of it. It is
understood, however, that their design uses a series of vertical leg supports much like
ground storage tanks have around their equatorial circumference. This design,
therefore, falls in between the Moss and Technigaz designs as far as bending moments
Status Report on LNG Tanker Designs
291
in the shell are concerned. However, this design, like the Moss design, is closely
coupled to hull strains and therefore stress levels in the sphere must be computed as
a total structure made up of the double hull and the sphere.
CONCLUSIONS
The choice oftanker design will continue to be influenced by the eventual owners,
assuming competitive prices between designs. Operating experience and ship-handling
characteristics will also playa major role in the selection of a design by both the
owners and the shipyard. There is little doubt that the U. S. ship construction industry
will play an important role in supplying the tankers required for not only U. S.
projects but also for foreign projects. The latest dollar devaluation has made U. S.
construction much more competitive than before, especially when compared to
Japanese shipyards.
The future of LNG, despite difficulties with the Federal Power Commission and
the Environmental Protection Agency, international political considerations, and
startup difficulties in some instances, remains bright.
REFERENCES
I. A. Pastuhov, in: Advances in Cryogenic Engineering, Vol. 18, Plenum Press, New York (1973), p. I.
2. M. Guilhem and L. L. Richard, in: Proc. 1st 1ntern. LNG Conference, Institute Gas Technology,
Chicago, Illinois (1968), paper 34.
3. R. G. Jackson and M. Kotcharian, in: Proc. 1st Intern. LNG Conference, Institute Gas Technology,
Chicago, Illinois (1968), paper 35.
4. A. Pastuhov, in: Advances in Cryogenic Engineering, Vol. 12, Plenum Press, New York (1967), p. 23.
5. W. Dubarry and A. H. Schwendtner, Maritime Reporter & Engineering News, 34(1): 10 (1972).
6. K. Yamamoto, in: Proc. 3rd Intern. LNG Conference, Institute Gas Technology, Chicago, Illinois
(1972), paper IV-2.
7. A. Pastuhov, in: Advances in Cryogenic Engineering, Vol. 17, Plenum Press, New York (1972), p. 71.
B-1
PHASE EQUILIBRIA FOR SYSTEMS CONTAINING
NITROGEN, METHANE, AND PROPANE
D. P. L. Poon and B. C.-Y. Lu
University of Ottawa
Ottawa, Canada
INTRODUCTION
Nitrogen and light alkanes such as methane, ethane, and propane are some of
the major natural gas constituents. Phase equilibria for mixtures containing these
compounds are useful in the design of equipment for storage and for the separation
of natural gas into its components. As part of a continuing program for investigating
the nonideal behavior of mixtures of major constituents of natural gas at LNG temperatures, phase equilibria for systems containing nitrogen, methane, and propane
were investigated in this study.
Vapor-liquid equilibrium data are available in the literature for the binary
systems nitrogen-methane [1-3], methane-propane [4.5], and nitrogen-propane [6.7].
However, the available data for the methane-propane system were measured above
130 K. The nitrogen-propane system exhibits partial miscibility in the liquid phase.
In an earlier article [7], liquid-liquid equilibrium compositions determined at
saturation pressures were reported. It was pointed out that these values differ from
the values obtained by extrapolating the data reported by Schindler et al. [6] to the
saturation condition. On the other hand, the liquid-liquid-vapor locus obtained
agrees well with that reported by Schindler et al. [6]. Vapor pressure curves of liquid
rich in propane for the nitrogen-propane system were shown graphically at three
temperatures C]. In this investigation, all measurements were made at the isothermal
conditions of 114.1,118.3, and 122.2 K, which are identical to the conditions reported
earlier [7]. Vapor-liquid equilibrium data were measured for the binary system
methane-propane and the ternary system nitrogen-methane-propane. Additional
measurements were also made for the binary system nitrogen-propane along the
vapor pressure curve ofthe liquid phase rich in propane. Furthermore, vapor-liquidliquid equilibrium data were also measured for the ternary system nitrogen-methanepropane. The binodal curves together with the plait points at the above-mentioned
temperatures were also established for the ternary system nitrogen-methane-propane.
n
EXPERIMENTAL
The equilibrium apparatus, equilibrium cell, cryostat, and experimental technique employed in this investigation were essentially the same as those reported
elsewhere [3]. The equilibrium cell was made of a l00-ml Jerguson transparent gage.
A dewar flask of 18-liter capacity was employed as the cryostat into which the
equilibrium cell was submerged. Isopentane was used as the bath liquid in the cryostat.
m
Pbase Equilibria for Systems Containing Nitrogen, Methane, and Propane
293
In addition to the equilibrium cell, two variable-speed stirrers, a refrigeration coil,
a heating element, and a resistance-type temperature sensing probe were installed
inside the cryostat. The temperature of the cryostat was regulated by controlling the
evaporation rate of liquid nitrogen in the refrigeration coil, and was controlled to
±0.05 K (the precision of the temperature measurements). Equilibrium temperatures
were measured by means of two thermocouples of the protective type, one for the
liquid phase and one for the vapor phase, in conjunction with a Leeds and Northrup
K-3 potentiometer and a Tinsley SRI galvanometer. Thermocouples were calibrated
against the vapor pressure of methane [8]. System pressures were measured by means
of two pretested 12-in. Heise gages (0 to 500,0 to 1000 psia).* The accuracy of these
gages is 0.1 % of the full scale. However, a mercury manometer was used for lowpressure measurements.
In addition to the cryostat, the equipment assembly [3] consisted of a feed
measuring and charging device, a closed recirculation loop, a volume regulator, and
sampling facilities. An electromagnetic pump was used for recirculating the vapor
through the liquid in the equilibrium cell.
Research-grade gases (nitrogen and methane supplied by Matheson of Canada
Ltd., and propane by Phillips Petroleum Co.) were used without further purification.
The specified minimum purities ofthese gases are as follows: nitrogen 99.997 mole %;
methane 99.99 mole %; propane 99.99 mole %.
Analyses of the liquid and the vapor samples were made with a Microtek gas
chromatograph Model GC-2ooo R, using porapak type Q (80-100 mesh) as the
column packing material and helium as the carrier gas. The response of the gas
chromatograph was calibrated against known synthetic samples. Calibration curves
were prepared by plotting peak height ratios against mole fraction ratios. The
reproducibility of the calibration curves was estimated to be 1 % on the average.
The temperature of the thermal conductivity cell of the gas chromatograph was
maintained at 473 K and the flow rate of helium was maintained at 85 mljmin at
75 psia. A retention time of 5 min was required for separating a three-component
mixture.
RESULTS AND DISCUSSION
The experimental vapor-liquid equilibrium data obtained in this investigation
for the systems nitrogen-propane, methane-propane and nitrogen-methane-propane
are listed in Tables I through III, respectively. Vapor pressure curves of liquid rich
in propane for the binary system nitrogen-propane are shown in Fig. 1, and vapor
pressures are plotted against equilibrium compositions in Fig. 2 for the binary system
methane-propane. An attempt was made to compare the experimental results
obtained in this investigation with that reported by Wichterle and Kobayashi [5]
by means of the liquid activity coefficient values evaluated at infinite dilution.
The liquid activity coefficient of component i at a constant temperature T and
a reference pressure po is given by
(1)
where
Yi -+
1 when
• 1 psia
= (101.325/14.7) kPa.
Xi -+
1.
D. P. L. Poon IIIld B. C.-Y. Lu
Table I. Experimental Vapor-Liquid Equilibrium Data for the Binary System
Nitrogen-Propane
114.1 K
122.2 K
118.3 K
P,psia
xN,
YN,
P,psia
xN,
YN,
21.8
23.6
49.5
95.4
145.2
199.1
232.5
239.6
258.0
261.0
0.0075
0.0086
0.0169
0.0335
0.0486
0.0640
0.0736
0.0773
0.0842
0.0845
0.9995
27.2
61.8
107.4
145.3
177.5
188.2
215.8
280.0
297.5
312.2
0.0079
0.0175
0.0321
0.0433
0.0511
0.0557
0.0621
0.0760
0.0801
0.0854
0.9994
0.9994
0.9998
0.9998
0.9998
0.9999
0.9999
0.9999
1.0
0.9996
0.9997
0.9998
0.9999
0.9999
1.0
1.0
1.0
P,psia
xN,
YN,
44.0
51.8
100.8
150.6
166.1
202.9
280.1
341.7
369.5
400.0
400.0
406.0
0.0107
0.0133
0.0254
0.0380
0.0410
0.0486
0.0672
0.0812
0.0858
0.0869
0.0881
0.0888
0.9989
0.9989
0.9994
0.9995
0.9995
0.9996
0.9999
0.9999
0.9999
1.0
1.0
1.0
In (l)'];L(P, T, x) represents the fugacity of component i in the liquid mixture at
system temperature T and pressure P ;flL(P", T) represents the fugacity of pure liquid
i at the system temperature T and a reference pressure 1'" ; and ~ represents the partial
molal volume of component i in the liquid mixture. In this investigation, 1'" was
arbitrarily taken to be 500 psia Values of YiL were directly evaluated from lL' Values
of iiL and f~L(P", T), together with ~L' were evaluated from the Redlich-Kwong
equation of state [9]
RT
a
P--(2)
- v - b TO. 5 v(v + b)
by means of a modified procedure [3]. The parameters
quantities a and b by means of
a
fia
and
fib
are related to the
= QaR2T.:2.5/Pc
(3)
(4)
b = QbR T.:/ Pc
Table II. Experimental Vapor-Liquid Equilibrium Data for the Binary System
Methane-Propane
114.1 K
122.2K
118.3 K
P,psia
Xca.
YCH.
P,psia
~.
YCHo
P,psia
~.
YCHo
6.1
8.6
11.2
13.0
14.0
16.2
17.8
0.1812
0.2911
0.4102
0.5488
0.6647
0.8812
1.0
0.9990
0.9995
0.9997
1.0
0.9998
1.0
1.0
7.9
11.0
14.1
17.6
19.6
20.6
22.9
24.7
0.1775
0.2717
0.3909
0.5714
0.6540
0.7399
0.9031
1.0
0.9986
0.9993
0.9997
0.9992
0.9999
1.0
1.0
1.0
7.1
9.1
13.1
13.5
18:8
23.1
26.9
28.8
31.0
32.3
0.1130
0.1409
0.2219
0.2253
0.3701
0.5297
0.7090
0.8095
0.8910
1.0
0.9976
0.9986
0.9996
0.9996
0.9999
0.9999
0.9999
1.0
1.0
1.0
Phase Equilibria for Systems Containing Nitrogen, Methane, and Propane
Table III. Experimental Vapor-Liquid Equilibrium Data for the Ternary
System Nitrogen-Metbane-Propane
T,K
P,psia
114.1
118.3
122.2
X N,
X cu ,
YN,
Ycu,
70.9
127.2
193.4
82.8
124.5
165.0
65.8
101.4
124.1
16.6
43.4
41.8
77.6
131.7
158.3
173.0
185.3
220.8
233.7
253.7
260.6
0.0245
0.0446
0.0637
0.0388
0.0620
0.1252
0.0562
0.0989
0.1797
0.0000
0.0503
0.0525
0.1270
0.2999
0.4184
0.5072
0.6106
0.8059
0.8644
0.9415
0.9750
0.0161
0.0159
0.0136
0.2624
0.2606
0.2602
0.6126
0.5779
0.5525
0.9320
0.8352
0.8657
0.7679
0.6180
0.5117
0.4373
0.3566
0.1800
0.1248
0.0547
0.0248
0.9886
0.9901
0.9932
0.8986
0.9314
0.9496
0.7851
0.8644
0.9074
0.0000
0.6241
0.6057
0.7902
0.8866
0.9145
0.9283
0.9353
0.9593
0.9699
0.9811
0.9922
0.0112
0.0098
0.0067
0.1013
0.0685
0.0504
0.2149
0.1356
0.0926
0.9999
0.3759
0.3943
0.2098
0.1134
0.0855
0.0717
0.0647
0.0407
0.0301
0.0189
0.0078
50.8
89.6
127.1
155.9
175.8
191.5
214.4
48.8
78.8
142.0
178.6
19.8
106.8
150.5
169.7
193.8
200.7
11.8
59.0
102.2
123.7
147.0
185.8
24.6
64.3
105.3
140.8
245.9
3.5
30.6
64.8
142.8
15.3
0.0559
0.1501
0.2509
0.3559
0.4403
0.4954
0.5946
0.0207
0.0399
0.0820
0.1221
0.0000
0.0970
0.1683
0.2258
0.2843
0.3033
0.0000
0.0227
0.0470
0.0601
0.0743
0.0967
0.0071
0.0199
0.0324
0.0443
0.0858
0.0000
0.0080
0.0190
0.0505
0.0000
0.9126
0.8217
0.7187
0.6287
0.5481
0.4979
0.4007
0.4689
0.4499
0.4169
0.3954
0.7492
0.6262
0.5672
0.5226
0.4869
0.4792
0.2901
0.2826
0.2713
0.2631
0.2531
0.2383
0.0628
0.0603
0.0578
0.0558
0.0491
0.0543
0.0501
0.0458
0.0435
0.2825
0.5299
0.7491
0.8279
0.8633
0.8896
0.8938
0.9156
0.6498
0.7845
0.8815
0.9022
0.4701
0.2509
0.1721
0.1367
0.1104
0.1062
0.0844
0.3501
0.2155
0.1184
0.0977
0.8147
0.8791
0.8969
0.9122
0.9044
0.1853
0.1209
0.1031
0.0878
0.0956
0.8067
0.8836
0.9098
0.9190
0.9324
0.8782
0.9505
0.9671
0.9795
0.9820
0.1931
0.1154
0.0892
0.0802
0.0672
0.1217
0.0494
0.0328
0.0205
0.0179
0.8948
0.9564
0.9794
0.1048
0.0434
0.0205
295
D. P. L. Poon and B.
c.-y. Lu
Table III-Continued
T,K
P,psia
X N,
67.3
108.6
145.2
186.8
23.2
48.3
88.8
173.8
240.3
26.8
51.8
79.4
199.2
244.6
30.1
83.8
137.3
185.1
255.1
327.4
0.0203
0.0360
0.0500
0.0767
0.0000
0 0. 156
0.0378
0.0824
0.1443
0.0000
0.0215
0.0450
0.1699
0.2678
0.0000
0.0846
0.1853
0.3532
0.5875
0.8468
X CH ..
0.2628
0.2462
0.2388
0.2327
0.5158
0.4876
0.4576
0.4073
0.3789
0.7003
0.6813
0.6528
0.5419
0.4635
0.9042
0.8225
0.7294
0.6102
0.4019
0.1510
YN,
YCH.
0.7711
0.8591
0.8951
0.9207
0.2288
0.1409
0.1048
0.0793
0.5374
0.7492
0.8680
0.8875
0.4625
0.2508
0.1319
0.1125
0.4699
0.6528
0.8696
0.9054
0.5301
0.3472
0.1304
0.0946
0.6484
0.7908
0.8561
0.9023
0.9497
0.3517
0.2092
0.1439
0.0967
0.0503
30
300
':
G
200
.
20
~
~
0:
10
100
o
Fig. 1. Vapor pressure curves of liquid rich in
propane for the binary system nitrogen-propane.
o
Fig. 2. Pressure--{;omposition diagram for the
binary system methane-propane.
Phase Equilibria for Systems Containing Nitrogen, Methane, and Propane
•
)-
•
Il'I i$ wor"-
• •
W.ich1er l.
and Kobayashi
3
. ~.~.~-------+-----+----~----~
~
Fig. 3. Liquid activity coefficients at
conditions of infinite dilution for the
binary system methane-propane.
110
120
•
In Tc~
130
140
150
Temperatur,
160
170
180
190
K
were considered temperature dependent. The calculated binary liquid activity coefficients were correlated by the Redlich-Kister equation 2]. For an ij pair
e
(5)
In YjL
= x/[Bij + Cij(x i
+ Diixi - X)(Xi - 5x)]
3x)
-
(6)
At conditions of infinite dilution,
(7)
and
In
ytt = Bij + Cij + Dij
(8)
The calculated In YL 00 values for methane and propane are compared with those
obtained from the literature values [5] in Fig. 3, using the same reference pressure
and the same calculation procedure. A consistent trend is indicated from both sets
of data.
114 I
Fig. 4. Total pressure-liquid composition and liquid-liquid equilibrium diagram for the ternary system nitrogenmethane- propane at 114.1 K.
c.
K
•
!! q ul et compOS-Ilion
•
lop
•
bolt om
•
plol l POLl'll
laye,
loyer
lJ'.1ILIr.=.::=:!>=-:..:-::..:-=:-=-""'-.::-:..:-::.;-=-:.::-:.==-:.:-::..:-::.;-=-x:.=
- :.:-::..:-::..:-::.;::.;-"" N.
D. P. L. Poon and B. C.-Y. Lu
c,
11 8 3
K
li quid compOS ttlon
fOp - l oyer
boll om - loyer
•
plo ll po ln
C·~~--~----
__~______~____~~____~N.
Fig. 5. Total pressure-liquid composition and liquid-liquid equilibrium diagram for the ternary system nitrogenmethane-propane at 118.3 K.
Total pressure-liquid compoSItIon curves for the ternary system nitrogenmethane-propane are presented at constant-pressure conditions in Figs. 4 through 6
for temperatures of 114.1, 118.3, and 122.2 K, respectively. In these figures, binodal
curves together with tie-lines and plait points are also shown. The plait points were
obtained from the experimental tie-lines using the method of Black and Hartwig [10].
As temperature increases, the mutual solubility increases and the position of the plait
point shifts toward lower propane concentration.
The method developed earlier [11] for the prediction of total pressure and vapor
composition from liquid composition and temperature values was applied to the
experimentally determined vapor-liquid equilibrium data for the nitrogen-methanepropane ternary system. The average absolute deviation between the calculated and
experimental values of vapor composition.1y is 0.005 mole fraction, while the average
absolute deviation between the calculated and experimental values of total pressure
122 2
K
li qu id c omposition
lOP - lo yer
bollQm- loyer
•
C,
plo ll Poini
:3 N.
.t.....IJ.r.u.c.:....:.:--::.=-~
-=-=-=-=-.::.-=.::-:.=-=-::.:--:.:-:.=-:.=-.::.-.::.-.::.-.::.-.::.-.::.-.:;;-.::.-=
-=-.::.-.::.-:..-
Fig. 6. Total pressure-liquid composition and liquid-liquid equilibrium diagram for the ternary system nitrogenmethane-propane at 122.2 K.
Phase Equilibria for Systems Containing Nitrogen, Methane, and Propane
199
is 4 %. The average deviations between calculated and experimental values of vapor
composition and total pressure are less for the methane-propane and nitrogenpropane binary systems than those for the ternary system when the same calculational procedures are used.
ACKNOWLEDGMENT
The authors are grateful to the National Research Council of Canada for financial support.
NOTATION
a, b
B, C, D
f
p
R
T
v
x
y
= Redlich-Kwong equation constants
= Redlich-Kister equation constants
= fugacity
= pressure
= gas constant
= temperature
= volume
= mole fraction in liquid phase
= mole fraction in vapor phase
Greek Letters
= activity coefficient
l'
= Redlich-Kwong equation parameters
Superscript and Miscellaneous
= reference state
= property in mixture
Subscripts
c
= critical properties
Qa' Q b
i,j
L
1,2,3
= component identity
= liquid-phase property
= parameter or component identity
REFERENCES
I. O. T. Bloomer and J. D. Parent, Chem. Eng. Progr. Symp.Ser., 49(6): 11 (1953).
2. M. R. Cines, J. T. Roach, R. J. Hogan, and C. H. Roland, Chem. Eng. Progr. Symp. Ser., 49(6): 1
(1953).
3. S.-D. Chang and B. C.-Y. Lu, Chem. Eng. Progr. Symp. Ser., 63(81): 18 (1967).
4. A. R. Price and R. Kobayashi, J. Chem. Eng. Data, 4(1):41 (1959).
5. I. Wichterle and R. Kobayashi, J. Chem. Eng. Data, 17(1):4 (1972).
6. D. L. Schindler, G. W. Swift, and F. Kurata, Hydrocarbon Process. Petrol. Refin., 45(11):205 (1966).
7. B. c.-y. Lu, S.-D. Chang, I. M. Elshayal, P. Yu, D. Gravelle, and D. P. L. Poon, in: Proc.lst International Con! Calorimetry and Thermodynamics, Warsaw, Poland, August 31-September 4, 1969,
pp.755-766.
8. G. T. Armstrong, F. G. Brickwedde, and R. B. Scott, J. Res. NBS, 55(1):39 (1955).
9. O. Redlich and J. N. S. Kwong, Chem. Rev., 44:233 (1949).
10. C. Black and G. M. Hartwig, Chem. Eng. Progr. Symp. Ser., 63(81):65 (1967).
II. C. Hsi and B. C.-Y. Lu, Can. J. Chem. Eng., 50: 144 (1972).
12. O. Redlich and A. T. Kister, Ind. Eng. Chem., 40:345 (1948).
B-2
LIQUID-VAPOR EQUIUBRIA IN THE
NITROGEN-METHANE SYSTEM BETWEEN
9S AND 120 K*
W. R. Parrish and M. J. Hiza
Cryogenics Division
NBS Institute for Basic Standards
Boulder, Colorado
INTRODUcnON
The development and evaluation of liquid mixture (solution) theory depends
heavily on the availability of precise data for mixtures of simple molecules. The most
useful data are for binary mixtures at closely spaced temperatures over as wide a
temperature range as possible, both above and below the critical temperature of the
most volatile component. The nitrogen-methane mixture, which is technologically
important as one of the more important binary mixtures in liquefied natural gas,
is an excellent compromise between theoretical and practical considerations.
The interest in the phase equilibrium properties of this particular system is
apparent by the large number of studies reported in the literature. A number of
experimental investigations have been conducted to determine the liquid-vapor
equilibrium properties of nitrogen-methane mixtures [1-15]. In addition, several
investigations have been conducted to determine the solid-liquid equilibria and the
three phase (solid-liquid-vapor) locus F6-1~. In the liquid-vapor region, the data
of Bloomer and Parent [1] and of Cines et al. [3] cover most of the methane liquid
range, and consequently, one or the other of these data sets frequently has been used
in correlations and for testing computation methods for the nitrogen-methane system
Since the data of Bloomer and Parent are isobaric and those of Cines et al.
are isothermal, direct comparisons are not possible. However, from comparisons
made by Bloomer and Parent of isothermal crossplots of their data and those of
Cines et al., these data sets appear to be in fairly good agreement at the lower temperatures, but are in less satisfactory agreement at the higher temperatures.
Of the remainder of the data reported prior to 1972, most studies were too limited
to add independently to the description of the liquid-vapor equilibrium behavior
of this system. In addition, the method of investigation, i.e., isobaric dew-point,
bubble-point measurements vs. isothermal, vapor-recirculation measurements, and
the method of reporting data, i.e., graphical vs. tabular, further complicate evaluation
and comparison of all of the data.
The recent investigations of Miller et al. FO] at 112.00 K and of Stryjek et al. F3 ]
from 113.72 K up to the methane critical temperature add new isothermal data which
readily can be compared with some of the previous data sets. In the mutual liquid
eO-23].
* Contribution of the National Bureau of Standards, not subject to copyright.
JOO
Liquid-Vapor Equilibria in the Nitrogen-Methane System Between 9S and 120 K
301
range of the two components, with which we are concerned in the present study,
these newer data tend to highlight the discordant aspects of some of the previous
data more clearly.
In addition to phase equilibrium data, density-or more specifically excess
molar volume-is vital to the development of liquid mixture theory. Liu and Miller
[24] recently reported excess molar volumes for the nitrogen-methane system between
90 and 120 K. Subsequently, Massengill and Miller [25] provided an interesting
theoretical discussion of the effects of adjustments to the combining rules and of
predictions of excess molar volume and excess Gibbs energy as a function of temperature with a modified hard-sphere equation of state. Because of the lack of
consistency of phase equilibrium data for the nitrogen-methane system in the same
temperature region, it is difficult to draw definitive conclusions from their treatise.
Thus, the need for a consistent set of phase equilibrium data for the nitrogen-methane
system between 90 and 120 K was clearly indicated.
The purpose of this study was to obtain liquid-vapor equilibrium data for the
nitrogen-methane system at 5 K increments of temperature between the triple point
of methane and the critical point of nitrogen, and from these data to obtain the
magnitude and temperature dependence of the excess Gibbs energy. These values of
excess Gibbs energy provide the basis of comparison with the corresponding values
derived from other data and with the qualitative temperature dependence of the
excess Gibbs energy predicted from the hard-sphere model by Massengill and Miller
e
5 ].
EXPERIMENTAL
The liquid-vapor equilibrium measurements were made in a closed-loop vaporrecirculating system described previously [26]. Thus, other than some minor differences, discussion of experimental details need not be included here.
The equilibrium pressures were measured with a standard laboratory, doublerevolution, 0 to 20 bar Bourdon gage and a 0 to 100 psia spiral quartz Bourdon gage.
The 20-bar gage has a claimed accuracy of ±0.1 % offull scale and was found to be
consistently better than these limits when compared with the 100 psia gage. The
latter instrument was calibrated against an air dead-weight gage in this laboratory
giving a maximum uncertainty ofless than ±0.1 psia over the full range. The pressure
tap was a small tube connected directly to the top of the equilibrium cell and was
independent of the recirculation system.
Temperatures were controlled as described previously 6 ]. However, the reported
temperatures are those determined by vapor pressure measurements of pure methane
and pure nitrogen as compared with the new methane data ofPrydz and Goodwin 7 ]
and the nitrogen values from the equation of Strobridge [28]. For purposes of this
study, nitrogen vapor pressure values from the equation of Strobridge are in excellent
agreement with the newer data of Weber [29]. Within experimental uncertainties of
the present study, the temperatures reported here are consistent with the IPTS-68
temperature scale.
Compositions were determined chromatographically with helium elution gas
and thermal conductivity detectors. Calibration· gas mixtures were prepared on a
pressure basis, corrected for nonideality, at 4.94, 9.90, and 49.89 mole % methane in
nitrogen. These calibration mixtures and pure methane and pure nitrogen were
analyzed at a number of different sample pressures to determine analyzer response
as a function of component partial pressure. It was determined that a linear calibration
e
e
302
W. R. Parrish and M. J. Hiza
based on the pure species at near-atmospheric pressure (60 cm Hg) satisfactorily
reproduced compositions over the entire composition range of interest, based on
peak areas, within 0.5 % of the actual composition of the minor component. Since
this difference is approximately equivalent to the precision of analysis, the results
reported here are based entirely on pure-component calibrations. The experimental
liquid and vapor compositions are estimated to be accurate within ± 1 % of the
actual composition or within ±O.l mole %, whichever is greater.
Liquid samples were withdrawn directly from the bottom of the equilibrium
cell through a stainless steel capillary tube, while the vapor samples were isolated
in a room-temperature sample loop which included the pump free volume. In all
cases, compositions were determined by analysis of both components in the mixture.
The total sample pressure was taken as the sum of the partial pressures determined
in the analysis. Though total sample pressure was measured, the sum of the partial
pressures was used to compensate for slight variations in analyzer sensitivity and
for any lack of sample thermal equilibrium.
It is also worth noting that the apparatus and chromatograph used in this study
were the same as those used in the study of Miller et al. [10]. However, there were
significant differences. The more precise 100-psia quartz Bourdon gage was acquired
for the present study to improve the measurement of the lower pressures. The
standard mixtures, except the pure component species, were prepared separately for
each study using different methods. The calibration mixtures used in the earlier study
were only equimolar mixtures prepared by weight. However, it was determined in
both studies that, assuming linear response, calibration of the chromatograph with
the pure fluid species represented the compositions of the prepared standard mixtures
within the precision of analysis. In addition, temperatures reported in the earlier
study were taken from the platinum thermometer readings, while in the present
study, the control point and thus the experimental temperature were determined
by the vapor pressures of the pure component species. Thus, the two investigations
are sufficiently different to be considered as two independent investigations.
RESULTS AND DISCUSSION
Equilibrium liquid and vapor compositions were measured at 95.00, 100.00,
105.00, 110.00, 115.00, and 120.00 K. The results of these measurements, along with
the vapor pressures of the pure components, are given in Table I.
A negative departure from Raoult's law occurs on the nitrogen-rich end at the
higher temperatures, consistent with the data of Bloomer and Parent [1], Cines
et al. [3], Stryjek et al. [13], and Miller et al. [10]. However, the isotherm of Chang and
Lu [4] at 122.05 K does not show this negative departure. Extrapolation of the data of
Chang and Lu to the pure nitrogen axis suggests a higher nitrogen vapor pressure
than that for 122 K and thus a slightly higher temperature than that reported, which
could easily account for the discrepancy.
In Fig. 1, the liquid and vapor phase compositions at 100.00 K from the present
investigation are compared with the smoothed values of Cines et al. [3] at 99.82 K.
This is the only direct comparison that can be made with the experimental data of
other investigators. The significant point worth noting is that, even though the data
of Cines et al. are for a slightly lower temperature, their vapor-phase compositions
are lower in nitrogen content than the present data. From the temperature dependence
of the vapor-phase compositions of the present study, the nitrogen content should be
higher at the lower temperature. This is significant since the activity coefficient of
methane is strongly affected by this inconsistency.
Liquid-Vapor Equilibria in the Nitrogen-Methane System Between 95 and 120 K
303
Table I. Experimental Nitrogen-Methane Liquid-Vapor Equilibrium Properties
T,K
xN,
YN,
)IN,
)lCH.
GE , J/mole
95.00
0.0000
0.2679
0.3562
0.4249
0.5271
0.5889
0.7495
0.8271
1.0000
0.199
2.254
2.737
3.018
3.431
3.646
4.247
4.580
5.400
0.0000
0.9216
0.9374
0.9460
0.9573
0.9618
0.9756
0.9823
1.0000
1.5755
1.4433
1.3355
1.2238
1.1621
1.0600
1.0327
1.0950
1.1790
1.2384
1.3271
1.4361
1.7038
1.9007
148.67
186.99
194.24
189.77
187.42
139.92
108.73
100.00
0.0000
0.1329
0.2397
0.3301
0.3831
0.4463
0.5069
0.5638
0.6671
0.8133
1.0000
0.343
1.988
2.960
3.621
3.989
4.396
4.706
5.032
5.598
6.451
7.778
0.0000
0.8336
0.8907
0.9168
0.9277
0.9372
0.9448
0.9508
0.9618
0.9764
1.0000
1.8662
1.6059
1.4436
1.3734
1.2987
1.2241
1.1742
1.1002
1.0320
1.0275
1.0987
1.1286
1.1547
1.2099
1.2614
1.3398
1.4799
1.8111
88.48
153.91
168.15
174.87
184.71
180.44
181.37
161.46
113.52
105.00
0.0000
0.2101
0.2869
0.4108
0.4836
0.6012
0.7548
0.8990
1.0000
0.565
3.649
4.512
5.664
6.298
7.242
8.417
9.805
10.835
0.0000
0.8533
0.8855
0.9149
0.9275
0.9440
0.9626
0.9829
1.0000
1.6242
1.4960
1.3191
1.2444
1.1455
1.0511
1.0148
1.0605
1.0963
1.1835
1.2478
1.3831
1.6684
2.0457
129.47
158.15
186.00
192.13
184.21
142.43
74.64
110.00
0.0000
0.2093
0.2856
0.4061
0.4843
0.5008
0.6056
0.6948
0.7956
0.8978
1.0000
0,884
4.810
5.916
7.450
8.366
8.479
9.603
10.580
11.797
13.179
14.680
0.0000
0.8237
0.8615
0.8950
0.9113
0.9147
0.9319
0.9456
0.9612
0.9790
1.0000
1.6008
1.4743
1.3129
1.2342
1.2112
1.1278
1.0756
1.0364
1.0131
1.0546
1.0850
1.1799
1.2477
1.2516
1.3765
1.5125
1.7208
1.9841
128.53
154.66
190.96
197.60
190.24
181.88
161.82
127.47
74.76
115.00
0.0000
0.1891
0.2868
0.3989
0.5027
0.5789
0.7110
0.8043
0.9040
1.0000
1.327
5.882
7.725
9.580
11.130
12.218
14.170
15.650
17.474
19.389
0.0000
0.7782
0.8363
0.8739
0.8983
0.9129
0.9377
0.9538
0.9750
1.0000
1.6384
1.4712
1.3218
1.2148
1.1515
1.0736
1.0344
1.0116
1.0465
1.0873
1.1606
1.2501
1.3398
1.5208
1.7540
2.0407
124.51
162.93
192.05
199.65
195.85
164.16
131.17
75.40
P, bars
304
W. R. Parrish and M. J. Hiza
Table I-Continued
T,K
XN,
P, bars
YN,
YN,
YCH.
GE,J/mole
120.00
0.0000
0.0977
0.1938
0.2990
0.3978
0.4959
0.5938
0.7055
0.7877
1.0000
1.919
4.984
7.549
9.982
12.100
13.942
15.919
18.029
19.880
25.128
0.0000
0.6122
0.7458
0.8125
0.8512
0.8773
0.9000
0.9225
0.9414
1.0000
1.8083
1.6058
1.4348
1.3177
1.2136
1.1444
1.0748
1.0459
1.0174
1.0490
1.0956
1.1523
1.2381
1.3476
1.5323
1.6785
73.31
130.07
171.53
194.66
203.17
200.81
176.16
144.97
A sensitive method for comparing close-boiling liquid-vapor equilibrium data
is through the derived excess Gibbs energy GE as a function of temperature and
composition. This topic was discussed in some detail by Duncan and Hiza [30). The
composition dependence of GE is obtained from
(1)
where x is the liquid-phase mole fraction and')' is the activity coefficient. The equation
selected here for calculating the activity coefficient for each component is the one
given by Duncan and Hiza [30J and will not be repeated here. This equation is based
on the virial equation of state and includes third virial coefficient effects. The activity
coefficients thus calculated are corrected to a reference pressure, which for the purposes
of this study has been taken as one bar. Second virial coefficients for pure nitrogen
and the interaction second virial coefficients for nitrogen-methane were calculated
from the corresponding states equation of McGlashan and Potter [31). The mixture
characteristic temperature was calculated from
-
Calc from Snide" , HII'I'InQlon, II Ii •
•
•
• 100001(, T
1 _
•
110
.. A Sorow ond ProUln it l (1 9661
. 0 Slry ; ••• Cl'loppe leor,and 1(01)0.,.011'1 1(1912 )
'00'
InWeU t9'll 'lon
999 .
99 92 K, Clnel. Rocu:tI.Koc:Ion,
O. 03~
0 Miller, Kidnoy . ond Hi lO (1973)
Clnu , Rooch , Hooon, ClAd Rol onG 119$3 )
g
ol'ld Ftol(l.nd (1953.
y v C
lio no ond L u tJ 96?)
•
FOllo,,".k ii ol'ld PetrOw-s .. ii
XlO
--- A:aoulr's low
.0
(l 9~7 )
full .. Clnd 8.1 1111'101'1' 11'9 67 ,
4oc.-. . , ,,,,"11 110m BO...., .~>IIIII.t lid I
~rool-________________~
· __- -
.
~
II •
.,;
.:
o
,
•
6
'"
_0
O~~--~--~-L--~
eli..
0 ,2
O~
06
0.8
HI
NIT ROGEN MOLE r RACTION
Fig. 1. Liquid and vapor equilibrium compositions for the nitrogen-methane system at 100.00 K
compared with those of Cines et al. [3].
100
-40
110
TEMPERATURE , I(
.20
Fig. 2. Values of equimolar GF: calculated from
the literature data. The value shown for Fuks and
Bellemans [8) is their reported value.
Liquid-Vapor Equilibria in the Nitrogen-Methane System Between 95 and 120 K
1'.:'2 =
(1 -
kd(1'.:,1'.:i I2
305
(2)
in which the value of 0.03 was taken for k 12 • The characteristic volume of the mixture
was calculated from the arithmetic mean rule
v
e12
= liVl/3
8\ Cl
+ V 1/ 3 )3
(3)
C2
For pure methane, second vi rial coefficients were taken from Goodwin [32]. Molar
volumes for pure liquid methane and for pure liquid nitrogen were taken from Goodwin and Prydz [33] and from Strobridge [28], respectively. Isothermal compressibilities for the pure liquids were taken from Rowlinson [34]. For purposes of the present
calculations, it was assumed that the molar volume and compressibility of each
component in the mixture are the same as those of the pure fluid.
Values of the equimolar GE for much of the previous data on the nitrogenmethane system are shown in Fig. 2. The values given include those obtained using
both experimental liquid- and vapor-phase data (filled symbols) and those obtained
from the liquid-phase data only by the method attributed to Barker [35]. In both
methods, the derived values of GE were fitted by the method of least squares to the
following equation:
GE = XIX2RT[A
+ B(XI
- X2)
+ C(XI
-
X2)2]
(4)
where the subscript 1 refers to the nitrogen. The values shown in Fig. 2 were calculated
from the fit of (4) obtained by each method. The two values shown for Fastovskii and
Petrovskii [6], however, were calculated from the smoothed equimolar liquid and
vapor compositions tabulated in their paper, interpolated from the original isobaric
data.
The curve shown in Fig. 2 was calculated by Massengill [36] from the hardsphere model applied by Snider and Herrington [37]. This curve was adjusted by
Massengill to fit approximately the equimolar GEvalue of Miller et al. eO] at 112.00 K
by including a k12 correction for the a parameter of 0.035, similar to the correction
indicated in (2).
Though values of GE obtained from the two methods are not expected to be
exactly equal, the results from both methods should exhibit the same temperature
dependence if the data are internally and mutually consistent. It is not possible to
conclude from the data presented in Fig. 2 that the temperature dependence of the
excess Gibbs energy predicted from the hard-sphere model is either reasonable or
incorrect.
The values of GE calculated from the present liquid and vapor phase data are
given in Table I also. The constants of (4) obtained from these values and from the
Barker method are given in Table II. The equimolar GE values from both methods
are compared in Fig. 3 with those from the data of Miller et al. [10] and the calculated
curve taken from Massengill [36]. The values of the equimolar GE from the present
study are in excellent agreement with those of Miller et al. Furthermore, the results
obtained from both methods give the same temperature dependence, though the
absolute values differ by about 10%. It thus can be concluded that the calculated
curve is a reasonable representation of the temperature dependence of GE •
The fact that the heat of mixing HE is related to the excess Gibbs energy and its
temperature dependence by
(5)
306
W. R. Parrish and M. J. Hiza
Table II. Constants for Equation (4) from the Barker Method and from Both
Liquid and Vapor Data
T,K
Data used
A
Std. dev. A
95.00
x-p
x-y-p
0.856566
0.989270
0.018181
0.000249
0.048882
-0.039063
0.026564
0.001553
-0.006273
-0.060026
0.030452
0.011018
100.00
x-p
x-y-p
0.834333
0.883896
0.004510
0.012843
0.0001627 -0.064193
0.006889
0.001080
0.019011
0.127218
0.011753
0.005988
105.00
x-P
x-y-P
0.797382
0.877361
0.011029
0.000017
0.036777
0.001976
0.018664
0.000084
0.028248
0.058049
0.026115
0.000487
110.00
x-p
x-y-p
0.769161
0.844515
0.010578
0.000072
0.024975
0.001095
0.017310
0.000416
0.017785
0.018386
0.024585
0.002345
115.00
x-P
x-y-p
0.773121
0.832951
0.006625
0.000019
0.030019
0.014834
0.011782
0.000086
0.04Sll5
0.066556
0.016687
0.000489
120.00
x-P
x-y-p
0.753117
0.819650
0.006021
0.000008
0.048651
0.037988
0.010289
0.000023
. 0.058088
0.086550
0.016019
0.000145
B
Std. dev. B
Std. dev. C
C
also allows a conclusion about the heat of mixing for this system. From approximately
90 to 110 K, the value of GE remains essentially constant. Thus, the temperature
dependence is zero and HE and GE are equal. Above 110 K, the temperature dependence becomes slightly positive; thus, the value of HE would become less than the
value of GE in this region. Unfortunately, there are no calorimetric heat-of-mixing
data for this system to substantiate this deduction.
Additional comparisons are given in Figs. 4 through 6 between the isothermal
GE values from the liquid- and vapor-phase data of the present study at 95.00,115.00,
and 120.00 i<. and the corresponding values from four other investigations at approximately the same temperatures. It is apparent that the GE values from the present study
are more symmetric about the equimolar value than those from the other investiga-
Calc. from Snider'· Hlf"l'lnQton, kit- 0.035
• 0 Thll inve.tigation
..
6
MiII,r', Kidnay and Hiza (1973)
(Open symbols from Sorker', Method)
• 95.00 K. This invllti90tion
• 90.67 K. Sprow ond Prou,nitz (1966)
200
~200·~~:L-~~~~L-~O~6~~~
~
)(
~
100
o~-L-L~~-L-L~~-L-L~
90
100
110
120
TEMPERATURE, K
Fig. 3. Values of equimolar GE from the data of
this investigation compared with the values calculated from the data of Miller et al. [10] at 112.00 K.
02
M
M
M
~
NITROGEN MOLE FRACTION
Fig. 4. Values of GE from the data at 95.00 K
compared with the values calculated from the
data of Sprow and Prausnitz [12] at 90.67 K.
Liquid-Vapor Equilibria in the Nitrogen-Methane System Between 95 and 120 K
•
115.00K, This investi90fion
•
307
120.00K, This investigation
• 122.0SK, Cines, Roach, Hogan,
.. 113.72 K, 51ryjek,Choppeleor,
and Kobayashi (1972)
and Roland (l953J
.. 122.0SK, Chang and Lu (1967)
200
200
100
o
oL-L-L-L-L-~~L-L-~
CH 4
0.2
0.4
0.6
0.8
NITROGEN MOLE FRACTION
Nz
CH 4
0.2
0.4
0.6
0.8
NITROGEN MOLE FRACTION
Fig. 5. Values of G E from the data at 115.00 K
compared with the values calculated from the
data of Stryjek et al. [13] at 113.72 K.
Fig. 6. Values of G E from the data at 120.00 K
compared with the values calculated from the
data of Cines et al. [3] and of Chang and Lu [4]
at 122.05 K.
tions. The values of Stryjek et al. and Cines et al. are in reasonably good agreement
with the present results at these temperatures. Those of Sprow and Prausnitz, however, are significantly lower than the present results. With the exception of two
points, the G E values from the data of Chang and Lu are in reasonable agreement with
values from the present study, which is surprising considering the qualitative difference
apparent in their data in the nitrogen-rich end.
SUMMARY
The results of this investigation provide a single set of closely spaced and consistent liquid-vapor equilibrium data for the nitrogen-methane system between the
triple-point temperature of methane and the critical point of nitrogen. The derived
excess Gibbs energy values substantiate the qualitative temperature dependence
of the equimolar Gibbs energy predicted by the hard-sphere model of Snider and
Herrington as applied by Massengill. From the temperature dependence ofthe excess
Gibbs energy, it can be concluded that the nitrogen-methane system closely approximates a regular solution over much of the temperature range examined, i.e., HE = GE
and SE = O. These conclusions strongly suggest the desirability of calorimetric heat-ofmixing measurements for the nitrogen-methane system, at least in the lower temperature region of this investigation.
ACKNOWLEDGMENTS
The authors wish to express their thanks to R. C. Miller for providing the calculated values of the
equimolar Gibbs energy, given only graphically in the thesis of one of his students, D. R. Massengill.
REFERENCES
I. O. T. Bloomer and J. T. Parent, Chern. Eng. Progr. Syrnp. Series, 49(6):11 (1953).
2. L. W. Brandt and L. Stroud, Ind. Eng. Chern., 50:859 (1958).
3. M. R. Cines, J. T. Roach, R. J. Hogan, and C. H. Roland, Chern. Eng. Progr. Syrnp. Series, 49(6): I
(1953).
4. S.-D. Chang and B. c.-y. Lu, Chern. Eng. Progr. Syrnp. Series, 63(81): 18 (1967).
5. H. Cheung and D. I.-J. Wang, Ind. Eng. Chern. Fund., 3(4):355 (1964).
3011
6.
7.
8.
9.
10.
II.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
W. R. Parrish IUId M. J. Hiza
V. G. Fastovskii and Yu. V. Petrovskii, Zh. Fiz. Khim., 31:2317 (1957).
W. ·Forg and P. Wirtz, Linde Rep. Sci. Technol., 15:46 (1970).
S. Fuks and A. Bellemans, Bull. Soc. Chim. Belg., 76:290 (1967).
H. A. McTaggart and E. Edwards, Trans. Roy. Soc. Can., 13 (Sect. III): 57 (1919).
R. C. Miller, A. J. Kidnay, and M. J. Hiza, AIChE J., 19(1):145 (1973).
V. G. Skripka, I. E. Nikitina, L. A. Zhdanovich, A. G. Sirotin, and O. A. Benyaminovich, Gazov.
Prorn.,15(12):35 (1970).
F. B. Sprow and J. M. Prausnitz, AIChE J.,12(4):780 (1966).
R. Stryjek, P. S. Chappelear, and R. Kobayashi, "Low Temperature Vapor-Liquid Equilibria of the
Nitrogen-Methane, Nitrogen-Ethane, and Nitrogen-Methane-Ethane Systems, Monograph, Rice
University, Houston, Texas (June 30, 1972).
N. S. Torochesnikov and L. A. Levius, J. Chern. Ind. (USSR), 16(1): 19 (1939).
E. Vellinger and E. Pons, Cornpt. Rend., 217:689 (1943).
V. G. Fastovskii and Yu. A. Krestinskii, J. Phys. Chern. (USSR), 15:525 (1941).
M. F. Fedorova, J. Exptl. Theoret. Phys. (USSR), 8:425 (1938).
D. W. Moran, Ph.D. Dissertation, Imperial College, University of London, London (1959).
M. H. Omar, Z. Dokoupil, and H. G. M. Schroten, Physica, 28(4):309 (1962).
H. E. Barner and S. B. Adler, Hydrocarbon Process., 47(10): 150 (1968).
M.-S. Lin and L. M. Naphtali, AIChE J., 9(5): 580 (1963).
H. H. Stotler and M. Benedict, Chern. Eng. Progr. Syrnp. Series, 49(6):25 (1953).
G. M. Wilson, in: Advances in Cryogenic Engineering, Vol. 9, Plenum Press, New York (1964), p. 168.
Y.-P. Liu and R. C. Miller, J. Chern. Therrnodyn., 4:85 (1972).
D. R. Massengill and R. C. Miller, J. Chern. Therrnodyn., 5:207 (1973).
A. G. Duncan and M. J. Hiza, in: Advances in Cryogenic Engineering, Vol. 15, Plenum Press, New
York (1970), p. 42.
R. Prydz and R. D. Goodwin, J. Chern. Therrnodyn., 4:127 (1972).
T. R. Strobridge, NBS Tech. Note No. 129 (1962).
L. A. Weber, J. Chern. Therrnodyn., 2:839 (1970).
A. G. Duncan and M. J. Hiza, Ind. Eng. Chern. Fund., 11(1):38 (1972).
M. L. McGlashan and D. J. B. Potter, Proc. Roy. Soc. (London), A267:478 (1962).
R. D. Goodwin, J. Res. NBS, 74A(5):655 (1970).
R. D. Goodwin and R. Prydz, J. Res. NBS, 76A(2):81 (1972).
J. S. Rowlinson, Liquids and Liquid Mixtures, Butterworth and Company, London (1969).
J. A. Barker, Austral. J. Chern., 6:270 (1953).
D. R. Massengill, M.S. Thesis, University of Wyoming, Laramie, Wyoming (1972).
N. S. Snider and T. M. Herrington, J. Chern. Phys., 47:2248 (1967).
H-3
GAS-LIQUID EQUILIBRIA OF THE C01-CO
AND C0 1-CH 4-CO SYSTEMS
L. J. Christiansen and A. Fredenslund
Instituttet for Kemiteknik
Technical University of Denmark
Lyngby, Denmark
and
N. Gardner
Case Western Reserve University
Cleveland, Ohio
INTRODUCTION
In densely populated areas which are far from natural gas and oil fields but
within easy access to coal deposits, coal gasification may soon become the basis of an
economically feasible, alternate energy supply. The refined gasification reaction
products may be distributed through the existing city gas networks. A high-pressure,
low-temperature gas-liquid contacting process may be employed in refining the
gasification products, the major components of which are hydrogen sulfide, carbon
dioxide, methane, carbon monoxide, and hydrogen. In order to predict the gasliquid equilibrium properties of mixtures of these components, it is necessary to
study all the possible binary mixtures. Data on ternary mixtures of these components
also give some insight to the behavior of the multicomponent mixture. Mixtures
containing hydrogen are not covered in this work, and of the remaining possible
binary systems, accurate and complete gas-liquid equilibrium data do exist for the
CO r -CH 4 [1.2], H 2S-C0 2 [3.4], H 2S-CH 4 [5.6], and CH4-CO systems [7.8]. Gasliquid equilibrium data for the CO 2-CO systems are more scarce, however. Kaminishi
et al. [9] have published a total of eighteen complete rr- T-x-y measurements spread
over five different isotherms ( - 50, - 40, - 20,0, and + 10°C). The dilute and critical
regions received virtually no attention in that study [9], making a thermodynamic
analysis of the data difficult. The H 2 S-CO system has, as far as is known, not been
studied at all. For practical reasons, both with regards to the gasification process and
the experimental procedure, the CO 2 -CO binary system was chosen for further study
in this work, which also includes gas-liquid equilibrium data for the CO 2-CH 4 -CO
system.
EXPERIMENTAL
The apparatus used in this work and the experimental procedure are described
in detail elsewhere [10]. The apparatus is of the vapor-recirculation type, the central
unit being a cell containing a stationary liquid phase. The gaseous phase is circulated
through the liquid using a diaphragm compressor. When equilibrium is attained, the
309
L. J. Cbristiaosen, A. Fredeuhmd aod N. Gardner
310
I
I
I
I
I
I
I
I
I
L
I GSY
L
_______ _
SUPPLY
Fig. 1. Vapor recirculation apparatus. (A) Diaphragm-compressor; (8) temperature; (C) precision comparison bridge; (D) dead weight piston gage;
(E) gas chromatograph; (F) platinum resistance thermometer; (G) stop valve
for compression; (H) ventilation valve; (I) vacuum valve; (J) cooling liquid
nitrogen; (RD) rupture disk; (GSV) gas sampling valve; (LSV) liquid sampling
valve; (DPI) differential pressure indicator.
temperature (± 0.01 0q, pressure ( ± 0.01 atm), and gas- and liquid-phase composition
(± 0.5 %) are measured. The composition measurements are carried out using a gas
chromatograph. The gas sample is obtained by slowly purging a side stream from the
vapor recirculation loop through an ordinary gas sampling valve and injecting the
sample into the gas chromatograph carrier gas.
As indicated in Fig. 1, the liquid sample is injected directly into the carrier gas. A
5-mm-diameter rod protrudes into the liquid phase in the cell. A 0.95-mm hole was
drilled vertically through the part of the rod immersed in the liquid. Activation of a
piston connected to the rod and mounted outside the thermostat causes the cavity
containing 3.5 Jlliters ofliquid sample to be drawn out into the cell wall. Here the liquid
sample comes in contact with the carrier gas and is carried to the gas chromatograph
for analysis.
The use of a gas chromatograph for the liquid and gas sample analysis is by far
the weakest point of the procedure outlined above. Therefore the gas chromatograph
was carefully calibrated by use of gas mixtures of known compositions (±0.1 %)
approximately every other day during the experimental work. Gas mixtures were
prepared by weighing the components into high-pressure sampling bottles. For the
binary system, the peak heights of each component were calibrated as a function of
composition; for the ternary system, peak areas determined using a disk integrator
were used. For the composition measurements for the ternary system, the uncertainties
are of the order of 1.0 %.
The carbon dioxide, methane, and carbon monoxide used were all of research
grade (purity better than 99.99 %) and were used without further purification.
RESULTS
Gas-liquid equilibrium compositions were determined for the carbon dioxidecarbon monoxide system at - 50, - 30, - 10, and + 10°C for pressures ranging from
the vapor pressure of pure, saturated carbon dioxide to near the critical pressure of the
311
Gas-Liquid Equilibria of the COz-CO and COz-CH.-CO Systems
Table I. Experimental Data for the COl-CO System and Activity Coefficients
Based on the Nonsmoothed Data
Equilibrium
compositions
Equilibrium
ratios
%
T,oC
P,atm
Xeo
Yeo
K eo ,
Keo
Inconsistencyof
K's
- 50.00
6.70
8.16
14.04
22.21
37.84
57.55
80.37
99.40
132.99
139.68
0.0000
0.0028
0.0148
0.0318
0.061
0.106
0.161
0.222
0.392
0.463
0.000
0.155
0.491
0.641
0.747
0.790
0.800
0.789
0.700
0.649
1.000
0.848
0.517
0.371
0.269
0.235
0.238
0.271
0.494
0.654
55.2
33.2
20.15
12.18
7.48
4.97
3.55
1.785
1.402
14.07
16.24
23.06
33.70
48.71
76.05
97.73
110.31
120.95
128.51
129.70
0.0000
0.0037
0.0162
0.0360
0.066
0.139
0.205
0.254
0.304
0.364
0.387
0.0000
0.112
0.338
0.506
0.610
0.666
0.667
0.649
0.619
0.570
0.548
1.000
0.891
0.673
0.513
0.417
0.388
0.419
0.471
0.547
0.677
0.738
26.07
29.86
37.34
48.23
63.60
80.67
93.74
100.91
108.40
112.06
0.0000
0.0064
0.0189
0.0387
0.071
0.116
0.155
0.180
0.214
0.240
0.000
0.093
0.226
0.344
0.432
0.477
0.486
0.480
0.463
0.441
44.34
50.31
56.69
66.87
80.68
86.99
95.91
0.0000
0.0106
0.0230
0.0436
0.080
0.099
0.141
0.0000
0.073
0.133
0.202
0.255
0.263
0.242
- 30.00
-10.00
+ 10.00
Experimental activity
coefficients
Yeo,
Yto
30
1.5
4.9
4.6
4.5
4.5
5.1
5.5
1.007
0.981
1.008
1.032
1.067
1.120
1.206
1.634
1.881
0.971
0.964
0.887
0.842
0.714
0.598
0.489
0.307
0.262
30.3
20.9
14.05
9.22
4.80
3.26
2.56
2.04
1.564
1.415
1.9
-4.8
-5.0
0.5
1.9
2.7
3.2
2.7
4.3
1.015
1.007
1.002
1.005
1.073
1.148
1.228
1.326
1.473
1.537
0.986
0.924
0.857
0.754
0.545
0.440
0.378
0.326
0.276
0.259
1.000
0.913
0.789
0.683
0.612
0.591
0.609
0.635
0.683
0.736
14.48
11.96
8.88
6.10
4.13
3.13
2.66
2.16
1.833
0.0
-3.3
-2.4
-2.0
-1.7
-1.0
-0.5
-0.1
1.002
1.00 1
1.002
1.012
1.036
1.068
1.095
1.138
1.180
0.938
0.926
0.839
0.708
0.568
0.483
0.438
0.384
0.346
1.000
0.937
0.888
0.834
0.810
0.818
0.883
6.85
5.77
4.53
3.19
2.66
1.72
1.0
2.6
2.1
1.8
1.6
1.001
1.001
1.000
1.011
1.022
1.060
0.935
0.854
0.764
0.601
0.539
0.417
system at the given temperature. The nonsmoothed experimental results are presented
in Table I together with the equilibrium ratios
(1)
K j == ydxj
The results and those of Kaminishi et al. [9] are shown in Fig. 2. The agreement with
the previous results at - 50 and + lOoC is seen to be fair, although the liquid-phase
compositions determined by Kaminishi et al. tend to be somewhat larger in carbon
L. J. Cbristiauen, A. FredeDsllmd and N. GardDer
312
160
r-----.-----,---.---,.----,---r--...--.,...,
o~ 0.1
0 .2
0 .3
MOLE
FRACTION
0.'
0.5
0 .6
0.7
0.8
CO . XCO
Fig. 2. Gas-liquid equilibria of the CO 2-CO
system.
monoxide than in the present work. As expected, the locus of critical pressures is a
straight line in the P-x, y diagram.
For the CO 2-CH c CO system, data were obtained at constant temperature and
pressure and varying composition. The pressure of the system, after initial charging
of the cell and venting, was adjusted to the nearest +0.01 atm using a 1O-cm 3 volume
hand piston pump which was inserted in the vapor recirculation loop outside the
thermostat (not shown on Fig. 1). Gas and liquid equilibrium compositions were
measured for: T = - 50°C, P = 65.96 atm; T = - 50°C, P = 34.01 atm; T = - 30°C,
P = 68.09 atm; and T = - 30°C, P = 33.72 atm. The original data are presented in
Table II and Figs. 3 and 4.
0 .8
_ _ -66.0 ATM
.------- - 30 .0 ATM
'"o
00.
u
"
eo.•
z
o
..'"
10--
u
0.3
......
o
:I
0.8
0.1
MOLE F RAeTIO" CH,. Xc Hi OR
v CH.
Fig. 3. Gas-liquid equilibria for the CO 2 -CH c CO
system at - 50°C.
=-
=-
T
T
xeo
30.00°C, P = 33.72 atm
0.951
0.006
0.043
0.020
0.037
0.944
0.020
0.931
0.049
0.923
0.066
0.011
30.00°C, P = 68.09 atm
0.835
0.051
0.114
0.808
0.099
0.094
0.764
0.068
0.167
0.203
0.744
0.053
50.00°C, P = 34.01 atm
0.922
0.020
0.058
0.904
0.051
0.045
0.880
0.096
0.024
0.868
0.116
0.016
=-
T
X CH ..
50.00°C, P = 65.95 atm
0.830
0.035
0.135
0.789
0.087
0.125
0.753
0.139
0.108
0.685
0.229
0.086
0.550
0.397
0.053
=-
T
X C02
Liquid mole fractions
0.499
0.495
0.497
0.495
0.344
0.348
0.347
0.349
0.261
0.262
0.264
0.264
0.198
0.199
0.207
0.211
0.218
Yeo,
0.041
0.123
0.303
0.394
0.154
0.271
0.409
0.471
0.116
0.287
0.512
0.591
0.106
0.246
0.357
0.500
0.656
YeH.
Gas mole fractions
0.461
0.382
0.200
0.111
0.503
0.381
0.244
0.180
0.623
0.450
0.224
0.145
0.696
0.555
0.436
0.289
0.127
Yeo
1.020
1.019
1.030
1.032
1.081
1.108
1.139
1.160
1.009
1.025
1.047
1.059
1.079
1.110
1.166
1.246
1.483
'Yeo,
0.947
0.938
0.919
0.895
0.715
0.653
0.582
0.554
0.916
0.874
0.821
0.786
0.705
0.652
0.586
0.497
0.378
'YtH.
'Yto
0.657
0.640
0.627
0.626
0.473
0.442
0.399
0.383
0.681
0.639
0.606
0.572
0.543
0.475
0.440
0.378
0.290
Experimental activity
coefficien ts
1.005
1.006
1.007
1.008
1.060
1.073
1.100
1.114
1.009
1.013
1.018
1.022
1.051
1.078
1.105
1.178
1.434
'Yeo,
0.862
0.854
0.844
0.836
0.633
0.606
0.562
0.544
0.823
0.798
0.766
0.750
0.658
0.603
0.562
0.489
0.369
'YtH.
Van Laar activity
coefficients
Table II. Original Data and Activity Coefficients for the COz-CHcCO System
0.808
0.797
0.783
0.772
0.514
0.479
0.425
0.403
0.786
0.756
0.718
0.699
0.599
0.532
0.483
0.393
0.245
'Yto
~
(M
(M
...
II
rn
0
~
•
g
=
n
...0
=-
0
~
n
...0
i
0
....
ID
:I.
5cr
~
~
,5'
Cl
314
L. J. CIuistiaasea, A. FredeasIaad aod N. Gardner
0.8
o
0.7
u
>0:
Q
3
0.4
o
z
o
...
u
_ _ -61.1 ATM
___ .33.7 ATM
0.3
c
.
0:
...
0.2
....
o
•
Fig. 4. Gas-liquid equilibria for the CO 2 -CH 4 -CO
system at - 30°C.
DATA CORRELATION
The non smoothed experimental data for the CO 2-CO system were subjected to a
thermodynamic consistency test as outlined by Prausnitz [11].
f
X
(K
In-2 + In qJ2 ) dX 2 + f
X
2
X2;O
K1
qJ1
vL dp
RT
2
(LHS)
X2;O
qJ1P
(qJ2
K2)]
(RHS) (2)
[ InK1+ln sal~al+x2In-+lnK
qJ 1 1
qJ1
1 X2
qJi is the fugacity coefficient of component i in the gas phase and is calculated using a
modified Redlich-Kwong equation of state as suggested by Chueh and Prausnitz [12] :
=
P
RT
V- b
a
T1/2v(V
(3)
= - - - ---,-..,.,..-.,------::-
+ b)
For mixtures, a and b are given by
a
= L LYiYjaij;
i
aij=
j
nG
Uai
nG R2T2.5
+ Uaj
cij
2
p ..
e'J
T.:i and Pei denote the pure-component critical temperatures and pressures, respectively; T.:ib and Pcij denote the mixture pseudocritical temperature and pressure;
n~ and Q"i are parameters which are curve-fitted from saturated vapor P-v-T data
for each of the pure components.
The suggested mixing rules for Pcij and T.:ij are
Pcij = zcijRT.:i/Vcij,
T.:ij
=
(T.:iT.:l/ 2 (1 - Xij)
where vcij = i(v:r + VW}3, and Zcij = 0.291 + O.04(Wi + Wj}. Here Wi and Wj are the
acentric factors for components i and j, while Xij is a measure of the deviation of T.:ij
Gas-Liquid Equilibria of the COz-CO and COz-CH.-CO Systems
315
from the geometric mean for the ~ystem. The latter is chosen to be equal to zero in this
study.
Once the volume v ofthe saturated vapor mixture has been calculated, ({Jk may be
obtained from [12]
I
- I _v_ ~ _ 2 Li Yiaik I v + b
n ({Jk - n v _ b + v _ b
RT3i2b n v
+
abk ( v + b
b)
Pv
RT3/2b2 In-v- - v + b - In RT
(4)
The saturated liquid molar volume if in (2) is calculated using the corresponding
states correlation for saturated liquid mixtures by Chueh and Prausnitz [13]
(5)
where the generalized functions v~), v~), and v~) are polynomials in T/T"M' The
coefficients of the polynomials and the method of calculating the pseudocritical
volume and temperature are also given by Chueh and Prausnitz [13]. For TR > 0.93,
vcM and T"M contain two additional binary parameters, which may be found either
from experimental information or general correlations.
The thermodynamic consistency test, equation (2), was applied to each binary
data point, and the results are shown in Table I. The column labeled" %inconsistency
on K's" is defined as 100(RHS - LHS)/RHS of (2). Considering that the consistency
test is applied to nonsmoothed data, the results are seen to be very satisfactory.
Activity coefficients were calculated for each binary and ternary data point
according to the asymmetric convention
[
({JIYI P ]
(P) -
1'1 -
(P)
Xl! 1 ,pure
(6)
T
where component 1 is the condensable, i.e" carbon dioxide
"(P)
[ ,2
=
({J2Y2 P ]
X2
H(P)
2(1)
(7)
T
and component 2 is the noncondensable, either methane or carbon monoxide. The
superscript (P) indicates that the property is calculated at the pressure of the system.
The terms pnure and H<f;h are obtained by applying a Poynting correction factor
to the reference properties at the saturation pressure of component 1 :
! (P)
1 ,pure
_
H(P)
-
-
2(1) -
(Psat)
[ i
[f 1 ,pure] exp
p
Psat
[H(Psat)] [ex p f.P
2(1)
L
~
Psat
RTdP
V20()
RT
]
dP]
(8)
(9)
The activity coefficients calculated with the aid of the data are given in Tables I
and II. The reference fugacities and the Henry's law constants are listed in Table III.
The term v/ is calculated from (5), while v2 0() is calculated from (14) below,
L. J. CIuistiMlea, A. Fredeas...... ad N. Ganlaer
316
Table III. Reference Properties
T,OC
f~:,atm
H~.'co, ' atm
HS:!.co, ' atm
-50
-30
6.23
12.11
21.16
32.73
465.4
508.6
485.2
420.9
173.3
190.6
-10
+10
The binary CO 2 -CO data of this work and the CO 2-CH 4 data of Donnelly
and Katz [1] at - 50 and - 30°C were fitted to the dilated van Laar model [11]
In y<[O) = vc1 [<I>" 2(1
In y!(r)
= vd<l>/(1
In y~(r) = vc3 [<1>/(1
+ 3<1>~ 2)]
+ 3<1>/ + 3<1>/ -
(10)
2'1Mi)<I>~) - 21X~i~1)<I>,,(1
2'1W)<I>q) - 21X~j~1)<I>,,(1
+ <1>/)]
+ <1>/)]
(11)
(12)
Again, component 1 is carbon dioxide, 2 is methane, and 3 is carbon monoxide. The
terms <1>" and <l>q are given by
m
1/2 m
W" = 1X22(1)W2
1/2 m
+ 1X33(1)w3'
1/2 m
1/2 m
= '12(1)w2
+ '13(1)W3
m
Wq
where the <l>i are volume fractions:
i = 1,2,3
i
The lXii(l) term is the so-called self-interaction constant for molecules i in the environment of molecules 1, while the '1i(1) term, the dilation constant, is a measure of how
effectively the light component swells the liquid solution.
As indicated, (10), (11), and (12) are only valid at constant pressure, arbitrarily
chosen here as P". The relationship between y!P) and y!r) is from straightforward
thermodynamics and is shown to be CJ
y!P) = [y!r){ exp
and
f: 1
f: ;;t'
Vi ; ; / dP
y~(P) = [y~(r)f exp
Vi
dPJ '
i
=1
i
= 2,3
(13a)
(13b)
The Vi in (13a) and (13b) are calculated using liquid molar volumes from (4) and then
applying the modified Redlich-Kwong equation of state (3). If this is done, one obtains
RT [
bk ]
-- 1 + --
2 ~ xiaik - [abJ(v + b)]
- ----'------,...".--v(v + b)T1/2
V - b
v- b
Vi =~-~~r--R~T~~a-~2~v-+~b~~-'1------
(v - b)2 T1/2 v2(v
(14)
+ b)2
The constants a and b are calculated in the same way as in the determination of
fugacity coefficients, except that n~ and n~ are replaced by n~i and ~i' respectively,
(i.e., parameters for saturated liquid volumes) and that vapor-phase mole fractions are
replaced by liquid-phase mole fractions.
Gas-Liquid Equilibria of the COz-CO and COz-CH.-CO Systems
317
Table IV. Van Laar Constants
CO 2 -CO
T,oC
CO 2 --CH 4
1X33 (1)
'13(1)
1X22 (1)
'12(1)
0.0155
0.0224
0.0239
0.0325
4.03
4.05
6.67
19.5
0.0105
0.0106
0.778
1.068
0.0179
0.0249
0.0273
0.0338
2.22
1.94
6.69
41.3
0.0107
0.0109
0.910
1.44
po = 67 atm
-50
-30
-10
+10
po = 34atm
-50
-30
-10
+10
Values of ylP were calculated for the CO 2 -CO system at the - 50, - 30, -10,
and + 10°C isotherms choosing po = 67 atm. When X2 = <1>2 = 0, equations (10) and
(12) contain two parameters, OC 33 (1) and 113(1).
The parameters are obtained by fitting the experimental ylr) values to (10) and
(12). The values obtained are shown in Table IV. The iX 22 (1) and 112(1) values for the
CO 2 -CH 4 system (X3 = <1>3 = 0) were obtained at - 50 and - 30°C in a similar
manner using the experimental information from Donnelly and Katz [1]. For the
binary systems, the van Laar activity coefficients proved to be in good agreement
O
)
2.5
....
-0
..~u
II:
0
~O'
~u
......
z
.
.......
u
0.5
0
u
0.5
OA
0.3
0.2
Fig. 5. Pressure-adjusted activity coefficients for the
CO 2-CO system at - 30°C (PO = 67 atm).
0.1
0
318
L. J.~, A. Fpdepe........ N. G ........
Table V. Critical Properties and RedOch-Kwong
Parameters for the Pore Components
7;.K
P•• atm
v•• cm 3 /mole
Wi
~
~
~
~I
COl
CH4
CO
304.19
72.85
94.065
0.224
0.4470
0.0911
0.4184
0.0794
191.06
45.80
98.72
0.013
0.4278
0.0867
0.4546
0.0872
132.92
34.529
93.06
0.049
0.4421
0.0906
0.4707
0.0888
with the experimental pressure-adjusted activity coefficients calculated using (7),
(8), and (13). A comparison of the experimental and van Laar pressure-adjusted
activity coefficients is shown for the CO2 -CO system at - 30"C in Fig. 5.
The parameter values obtained on the basis of the binary data were used to
predict the activity coefficients for the CO 2-CH4 -CO system at T = - 50°C, P =
65.96 atm and T = - 30°C, P = 68.09 atm, where P is approximately equal to po.
The results of the predictions are shown in Table II, and the predicted van Laar
activity coefficients are found to be in good agreement with those calculated from
experiment.
The entire procedure was repeated starting with (10), (11), and (12) for the binary
systems, this time choosing po = 34 atm. The new van Laar constants, also shown in
Table IV, were used in predicting the ternary activity coefficients for T = - 50°C,
P = 34.01 atm and T = 30°C, P = 33.72 atm. These results are also given in Table II.
The predictions are seen to be somewhat less satisfactory than before, especially
for carbon monoxide. This is most likely due to difficulties in calculating molar and
partial molar volumes accurately in the highly critical region. When po is much
smaller than the maximum encountered pressure, the Poynting correction factor for
carbon monoxide becomes very small, down to about 0.2 in this case. The relatively
large uncertainty in the value of Vi in (13) makes the determination of the parameters
equally uncertain.
CONCLUSIONS
Experimental gas-liquid equilibrium data have been obtained for the CO 2 -CO
and the CO 2-CH4 -CO systems, and the binary data have been shown to be internally
consistent. The dilated van Laar model represents the binary data very well, and it has
been shown that the binary van Laar constants may be used in predicting the gasliquid equilibrium of a ternary system. The data and method presented here may thus
be used with reasonable confidence in design calculations dealing with multicomponent gas-liquid equilibria.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the assistance of L. Lauerhass. They also wish to express their
appreciation to Consolidated Natural Gas Service Company for their partial support of this project.
Gas-Liquid Equilibria of the COz-CO aod COz-CH.-CO Systems
319
NOTATION
a
= constant in Redlich-Kwong equation of state
=
=
=
=
=
b
f
H
K
P
R
T
=
=
=
=
=
v
x
y
z
=
constant in Redlich-Kwong equation of state
fugacity, atm
Henry's law constant, atm
equilibrium ratio
pressure, atm
gas constant, cm 3 -atm/mole-K
temperature, K
molar volume, cm 3/mole
liquid-phase mole fraction
gas-phase mole fraction
compressibility factor
Greek Letters
= self-interaction constant
0(
y
=
"
= dilation constant
activity coefficient
= deviation from geometric mean
x
cp
=
fugacity coefficient
= volume fraction
= acentric factor
= constant in Redlich-Kwong equation of state
ell
w
Q
Subscripts
1
2
3
= carbon dioxide
= methane
= carbon monoxide
i,j, k = components
(1) = in solvent, component 1
0(,,, = subscripts for van Laar constants
= critical property for mixture
c
Superscripts
..
= asymmetric convention is used
= infinite-dilution property
= liquid-phase property
00
L
G
(P)
= gas-phase property
Psat
sat
= partial molar property
= evaluated at pressure P
= saturation pressure of component 1
= evaluated at saturation
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
'9.
10.
II.
12.
13.
H. G. Donnelly and D. L. Katz, Ind. Eng. Chem., 45:511 (1954).
J. A. Davis, N. Rodewald, and F. Kurata, AIChE J., 8:537 (1962).
D. F. Sobocinski and F. Kurata, AIChE J., 5:545 (1959).
J. A. Bierlein and W. B. Kay, Ind. Eng. Chem., 45:618 (1953).
H. H. Reamer, B. H. Sage, and W. N. Lacey, Ind. Eng. Chem., 43:976 (1951).
J. P. Kohn and F. Kurata, AIChE J., 4:211 (1958).
L. J. Christiansen, A. Fredenslund, and J. Mollerup, to be published in Cryogenics (1973).
A. Toyama, P. S. Chappelear, T. W. Leland, and R. Kobayashi, in: Advances in Cryogenic Engineering,
Vol. 7, Plenum Press, New York (1962), p. 125.
G. I. Kaminishi, Y. Arai, S. Saito, and S. Maeda, J. Chem. Eng. (Japan), 1: 109 (1968).
A. Fredenslund, J. Mollerup, and L. J. Christiansen, to be published in Cryogenics (1973).
J. M. Prausnitz, Molecular Thermodynamics of Fluid-Phase Equilibria, Prentice-Hall, Englewood
Cliffs, New Jersey (1969), p. 500.
P. L. Chueh and J. M. Prausnitz, Ind. Eng. Chem. Fund., 6:492 (1967).
P. L. Chueh and J. M. Prausnitz, AIChE J., 13: 1099 (1967).
H-4
SOLUBILITY OF SOLID BENZENE, TOLUENE,
n-HEXANE, AND n-HEPTANE IN LIQUID
METHANE
G. P. Kuebler and C. McKinley
Air Products and Chemicals, Inc.
Allentown, Pennsylvania
INTRODUCTION
Reliable phase equilibrium data are essential for the design of effective and economical gas treatment processes such as natural gas liquefaction. Data on cryogenic
mixtures involving relatively simple molecules are also of theoretical interest. An
apparatus and the techniques developed to obtain solid-liquid phase equilibrium
data for cryogenic systems are described here. Results are presented for the individual
solubilities of solid benzene, toluene, n-hexane, and n-heptane in liquid methane.
There are basically two ways to obtain solubility data: the isoplethal method or
the isothermal method. The isoplethal method, in which mixtures of known composition are subjected to temperature changes, is not widely used when the solubility
is very low. The early work of Stackelberg et al. [1] and the more recent work of
Heastie and Lefebvre [2] and Yunker and Halsey [3] are representative of this
approach.
The use of gas chromatography, which allows the accurate and convenient
analysis of vaporized fluid mixtures, has favored increased application of the isothermal or analytical method. Two steps are important in the isothermal method:
making the solution and the sampling of this solution for analysis. A usual technique
of preparing the solution is to agitate the mixture in the presence of excess solute by
mechanical stirring or recirculation of the vapor phase. Jensen and Kurata [4] and
Neumann and Mann [5] used the former, while Preston et al. [6] used the recirculation
technique. Failure to promote equilibration by agitation can result in systematic
errors, as found in the data of Fedorova [']. Solutions can also be formed using a
single-pass mode either with solute precipitating from a feed mixture or a previously
deposited solute dissolved into a pure solvent feed. Cheung and Zander [8] used
the first techni'lue and both techniques have been used in solid-vapor equilibrium
measurements ["]. The resultant solution can be flash-evaporated directly to ambient
temperature [4.6.8] or the solution can be transferred to a separate chamber and the
entire apparatus warmed to ambient temperature [5.'].
Discrete samples taken from a bulk solution tend to be poorly reproducible,
especially in the parts per million concentration range. Also, the presence of minute
quantities of highly soluble contaminants in the solute material can interfere with
analysis by gas chromatography. A major advantage of the continuously flowing
sample stream is that it attains steady state with respect to adsorption effects in the
310
321
Solubility of Solid Benzene, Tolnene, n-Hexane, and n-Heptane in Liquid Methane
sampling system. Also, the stream tends to quickly leach out the highly soluble
contaminants in the solute bed, resulting in an uncluttered chromatogram. Initial
attempts, however, to use the single-pass continuous flow technique at pressures
sufficient to ensure liquefaction of the solvent stream likewise yielded poor reproducibility and ultimate plugging of the sample line. These problems were eliminated
by flowing the solvent above its critical pressure, thus avoiding the liquid-to-vapor
phase transition in the sample line. Subsequent experience [10] has demonstrated
the validity of data obtained with this technique in comparison with solubilities
obtained at the solid-liquid-vapor equilibrium pressure. The technique is also inherently suited to the investigation of the effect of pressure on solubility.
The solvent, CP-grade methane (99.0%), was obtained from Air Products and
Chemicals, Inc., and used without further purification because the technique required
substantial quantities of solvent gas. The actual feed gas was not analyzed but typical
analysis ofthis grade of methane shows 0.5% nitrogen and 0.2% ethane as impurities.
The benzene and n-hexane were Baker reagent grade; the toluene and n-heptane were
chromatoquality reagent grade (99.9%) obtained from Matheson, Coleman and Bell,
with benzene and n-xylene listed as the principal contaminants. All the solute materials
were used without prior purification; however, there is a purification process inherent
in the experimental method, as mentioned above.
EXPERIMENTAL APPARATUS
The flow apparatus is shown schematically in Fig. 1. The solvent, methane
obtained from a commercial high-pressure cylinder, was passed sequentially through
two coils immersed in a temperature-controlled bath. The first coil ensured that the
solvent stream was at the desired temperature while passing through the equilibration
coil which contained the solute material. A 0.8-mm-thick disk of sintered stainless
steel (5 J1.m pore size) was incorporated into a piping union for filtration on the downstream side of the equilibration column. Beyond the filter, the saturated fluid passed
through stainless steel tubing (0.56 mm ID) to a heated throttling valve. The stream
pressure was reduced to approximately atmospheric pressure across the throttling
valve, after which the gaseous stream flowed through the sampling loop of a gas
VENT
0-3000 PSIG
V
HEISE GAUGE
0-1000
eC/MIN
FLOWMETER
TtMOTTLE
NEEDLE
VALVE
O--_-.....:I:'-----;*--. r---(:i<~:t-.L--.... CHRO~~T~~APH
TO ICE BATH AND
POTENTIOMETER
FINE
Ir====~
TEMPERATURE
CONTROLLER
LN2 SUPPLY
STIRRER
4 LITER DEWAR
RESISTANCE THERMOMETER
3 JUNCTION
COOLING COL
HEATER
EQUILIBRATION COLUMN
Fig. 1. Single-pass, continuous flow
solubility apparatus.
CU-CONSTANTAN
THERMOPILE
51NTERED STAINLESS
STEEL FIL TEA
-""":::'::::=:::::::;::0><_ FREON
CONTROLLED
TEMP£R.&T\IRE BATH
-41. OR -13
311
G. P. Kuebler aad C. McKillley
chromatograph. Samples were injected into the chromatograph at intervals regulated
by the retention time of the least volatile component. A hydrogen flame detector
was used to determine the solute concentration. Sample peak areas were compared
to areas obtained with gas mixtures referred to gravimetrically produced standards.
A disk integrator was used in the early work and an electronic integrator was used
for the n-heptane investigation.
Concentrations of the solute in the solvent stream were measured at various
temperatures and pressures. At several temperatures, the solute concentration was
measured as a function of flow rate. At very high flow rates, the contact time is
probably too short to allow phase equilibration; however, no detectable change in
the solute concentration was observed for flow rates as high as 700 standard cm 3 fmin;
the flow rates used were well below this value.
The equilibration column was a 1.2-m-length of copper tubing (6.3 mm 00,
3.9 mm ID) wound in a 2.9-cm 10 helix. Such coils are easily produced and a separate
one was formed for each solute investigated. Crushed firebrick (45-60 mesh) carrying
the solute was loosely packed into the column prior to coiling. Loose-fitting glasswool plugs kept the firebrick in the column. The solute had previously been adsorbed
onto the firebrick by simply placing a few grams of the solute into a small jar containing the firebrick. The crushed firebrick remained free flowing with as much as
30% by weight of liquid adsorbed.
The general techniques described have also been successfully used to study
systems in which both the solvent and solute are gases at ambient conditions [10].
For such systems, a feed mixture of the appropriate concentration is passed through
the apparatus and excess solute is deposited on the walls of the column. It is also
possible to pass pure solvent gas through the column and dissolve the previously
deposited solid. This serves as a check on the validity of the equilibration data.
The bath in which the coils were immersed was liquefied Freon-13 used from
90 to 180 K, and liquid Freon-ll used for temperatures higher than 180 K. The
desired temperature was obtained by passing liquid nitrogen or cold nitrogen vapor
through a separate coil immersed in the bath and providing the required heat balance
with an electric resistance heater. The regulated power supplied to the resistance
heater was provided by a commercial proportional electronic controller. Temperatures were controlled to within ± 0.03 K of the set value. The bath temperature
was measured with a three-junction copper-constantan thermopile with an ice bath
reference. The thermopile output was measured with a millivolt potentiometer. The
estimated accuracy of the reported temperatures is ± 0.1 K. The upstream pressure
was regulated by a high-pressure regulator and was measured with a 0 to 3000 psi
(20.6 MPa),· temperature-compensated Bourdon tube gage.
RESULTS
The measured solubilities are up to several orders of magnitude lower than the
ideal solubility at the boiling temperature of liquefied natural gas. Ideal solubilities
were calculated according to the methods of Hildebrand and Scott [11]. Modifications
of the Hildebrand approach, especially in the determination of the liquid-phase
activity coefficients, have been used to correlate these data [10].
The experimental data are given in Table I and selected points, representing
approximate isobars, are shown in Fig. 2 for comparison with the literature data.
• 1 MPa
=
106 N/m2
=
145.0377 psi.
Solubility of Solid Benzene, Toluene, n-Hexane, and n-Heptane in Liquid Methane
323
Table I. Solubilities of Hydrocarbons in Methane
Temperature,
K
Pressure,
MPa
163.7
163.1
162.6
162.0
160.9
158.1
155.3
149.7
144.2
138.6
133.0
127.4
121.8
121.8
121.8
121.8
121.8
116.2
110.6
105.0
99.4
93.8
9.61
9.61
9.61
10.51
10.51
10.51
10.51
10.48
10.44
10.44
7.06
7.27
13.20
10.54
10.27
8.41
6.99
10.54
10.54
10.54
10.44
10.51
166.5
163.7
160.9
160.9
155.6
155.6
150.1
144.4
138.7
138.7
133.1
127.6
122.0
122.0
118.6
116.4
110.9
110.9
110.9
105.3
99.8
94.2
7.30
6.89
8.99
5.61
9.20
5.61
9.20
9.20
9.20
7.10
9.58
9.06
9.58
7.13
6.89
9.27
9.51
7.06
5.82
8.72
9.51
9.37
166.4
160.9
160.9
155.3
149.7
144.2
13.41
13.30
10.37
10.37
8.41
13.30
Hydrocarbon.
molar ppm
n-Hexane
150,000
148,000
111,000
78,300
57,600
34,800
24,300
13,500
8,160
4,980
2,960
1,780
1,060
1,044
1,043
1,031
1,027
586
317
158
77.0
34.9
n-Heptane
23,700
15,200
12,600
11,200
7,730
7,000
4,820
2,970
1,800
1,750
1,080
621
354
347
232
187
96.7
95.6
94.7
47.1
21.6
9.1
Toluene
2,670
1,930
1,730
1,290
866
683
Temperature,
K
Pressure,
MPa
144.2
144.2
144.2
138.6
133.0
127.4
121.8
116.2
110.6
105.0
99.4
93.8
92.1
10.37
8.37
5.82
8.51
8.41
8.41
8.41
8.41
8.41
8·41
8.41
8.41
8.37
199.8
199.8
199.8
188.7
188.7
188.7
183.1
183.1
183.1
183.1
183.1
177.6
172.0
166.4
166.4
166,4
166.4
160.9
155.3
149.8
144.2
144.2
144.2
144.2
144.2
144.2
144.2
138.6
133.0
127.4
125.1
121.8
121.8
116.2
110.6
105.0
99.4
13.68
12.61
10.54
13.61
12.54
10.54
13.61
10.54
8.61
6.13
5.41
5.41
5.41
13.58
10.54
8.61
5.41
5.41
5.41
5.41
13.58
12.61
10.54
8.37
6.99
6.06
5.41
5.41
5.41
5.41
5.41
13.61
5.41
5.41
5.41
5.41
5.41
Hydrocarbon.
molar ppm
Toluene cont.
637
609
569
422
273
174
106
64.0
35.9
19.6
10.0
4.7
3.6
Benzene
1,400
1,290
1,080
916
868
779
711
614
537
440
419
363
293
295
270
252
226
167
122
81.9
62.1
61.4
58.9
56.5
54.5
55.0
54.7
34.9
20.6
12.5
9.7
7.6
7.1
3.6
2.0
0.90
0.39
324
G. P. Kuebler and C. McKinley
TEMPERATURE.
,J~~~~~~--r-
10'
K
__--~'~~~--~'TOO~--'
6
PRESENT WORK . LOW PRESSURE
•
o
PRESENT WORK , HIGH PRESSURE
NEUMANN ANO MANN t 1970'
•
NEUMANN, [T AL.. .
o
•
SHIM AND KOHN
KOHN
( 972
1
I
(1 962 I
I 1961 I
--- ES TIMATEO
c
(;
e
- 10,)
Z
o
~
cr
II:
....
Z
~ ta Z
z
o
u
w
....
3
o<JI
10
'-0
8 f:NZEH£
RECIPROCAL
TE MPERATURE •
IOOO/K
Fig. 2. Solubilities of hydrocarbons in methane.
The results presented here are compared with the data of Neumann and Mann [5]
and Neumann et al. [12] for benzene, n-hexane, and n-heptane. The agreement is
considered reasonably good for low-temperature solubility work.
Each of the binary systems reported here can be classed in terms of one of three
types of phase behavior exhibited by methane in combination with a heavier hydrocarbon. Characteristic differences are apparent at the higher temperatures for the
different types of system; however, all are qualitatively similar in the low-temperature
region.
The methane-n-hexane binary is representative of a system in which the solute
triple-point temperature is lower than the solvent critical temperature and the solubility curve is continuous from the lowest temperature up to the melting point of the
solute, n-hexane. The present data join smoothly with the high-temperature data of
Shim and Kohn [16].
The methane-n-heptane and methane-toluene binaries ar,e representative of
systems in which immiscible liquids form and combine with the vapor and solid
phases to produce a quadruple point. A quadruple point has been reported by
Chang et al. [13] and Kohn [14] for the methane-n-heptane system and by Chang
and Kobayashi [15] for the methane-toluene system. For such systems, the solubility
curve is not continuous from low temperatures up to the solute melting point. The
apparent discontinuities in the solubility curves shown in Fig. 2 for the methanen-heptane and the methane-toluene systems are the result of this miscibility gap.
Solubility of Solid Benzene, Toluene, n-Hexane, and n-Heptane in Liquid Methane
325
The methane-benzene binary is representative of a system in which the solute
triple-point temperature is much higher than the solvent critical temperature. For
such systems, the solubility is very pressure sensitive near the solvent critical temperature and a family of isobaric curves results. In such systems, the critical locus
is interrupted by the triple-point curve with both the high- and low-temperature
branches terminating in singular end points having critical identity of the liquid and
vapor phases. Between these temperatures, there is a region of solid-fluid phase
equilibrium at all pressures. Since all the measurements were made in the solid-fluid
region, however, the solubility curves appear to be continuous through this region.
The effect of pressure on solubility is illustrated in Fig. 3. Several isotherms
are plotted against the corresponding methane density. The Francis [17] correlation
was used to obtain internally consistent methane densities for pressures up to 13.7
MPa. Although the methane contained nitrogen and ethane impurities as well as
varying solute concentrations, the accuracy of the data and the density correlation
justified the use of pure methane densities. The solute concentration plotted logarithmically against density produced straight lines with all isotherms for a given binary
having approximately the same slope. The average slope for the methane-benzene
system is 0.0053 m 3 /kg with a maximum deviation of 6.8 % and an rms deviation of
3.8%. The slopes of the two methane-toluene lines are 0.0077 and 0.0071 m 3 /kg
and the slope is 0.0022 m 3 /kg for the single methane-n-hexane isotherm.
The reproducibility in a given series of samples at the same temperature was
within ±0.5%. Temperatures are believed accurate to ±0.1 K and pressures to
within ± 0.04 MPa. The uncertainty in the pressure has a negligible effect on the
accuracy of the solubilities; however, the temperature uncertainty can produce
inaccuracies in the reported solubilities of 0.2 to 1.5% over most ofthe range covered.
.00 0
/
1.:.
200 0
E
a.
a.
o 1000
"0
E
/
199.8 K
..." ~ 121.8
/188.7
80 0
~/
60 0
'0 0
/
/'144.2
VISS.I
f!
,P
~.
20 0
BENZENE
0
Fig. 3. The effect of solvent density on
solubility.
10~
300
METHANE
340
DENSITY.
t;.
TOLUENE
"
n - HEXANE
380
kg /
m'
420
.60
326
G. P. Kuebler .... C. MeKiDIey
Gas chromatography reference standards are believed accurate to within ± 1.0%
and the standard and sample peak areas are believed accurate to within ±0.5%.
The total standard error, defined as the square root of the sum of the squares of the
individual errors, is approximately ±2%.
ACKNOWLEDGMENT
The authors acknowledge the Cryogenic Systems Division of Air Products and Chemicals, Inc. for
sponsorship ofthe work and permission to publish the results.
REFERENCES
I. M. v. Stackelberg, M. Heinrichs, and W. Schulte, Z. Physik. Chem., A170:262 (1934).
2.
3.
4.
5.
6.
7.
8.
9.
10.
II.
12.
13.
14.
15.
16.
17.
R. Heastie and C. Lefebvre, Proc. Phys. Soc. (London), 76: ISO (1960).
W. H. Yunker and G. D. Halsey, Jr., J. Phys. Chem., 64:484 (1960).
R. H. Jensen and F. Kurata, AIChE J., 17: 357 (1971).
A. Neumann and R. Mann, Kiiltetechnik-Klimatisierung, 22(6): 182 (1970).
G. T. Preston, E. W. Funk, and J. M. Prausnitz, J. Phys. Chem., 75 (15):2345 (1971).
M. F. Fedorova, Zh. Fiz. Khim., 14(3):422 (1940).
H. Cheung and E. H. Zander, Chem. Eng. Prog. Symp. Series, 64(88): 34 (1968).
M. J. Hiza and R. N. Herring, in: Advances in Cryogenic Engineering, Vol. 8, Plenum Press, New York
(1962), p. 158.
Air Products and Chemicals, Inc., unpublished work.
J. H. Hildebrand and R. L. Scott, Solubility of Nonelectrolytes, 3rd ed., Reinhold Publishing Corporation, New York (1950), p. 16.
A. Neumann, R. Mann, and W. D. Szaighary, Kiiltetechnik-Klimatisierung, 24(5): 143 (1972).
H. L. Chang, L. J. Hurt, and R. Kobayashi, AIChE J., 12(6): 1212 (1966).
J. P. Kohn, AIChE J., 7(3):514 (1961).
H. L. Chang and R. Kobayashi, J. Chem. Engr. Data, 12:517 (1967).
J. Shim and J. P. Kohn, J. Chem. Engr. Data, 7(1):3 (1962).
A. W. Francis, Ind. and Engr. Chem., 49: 1779 (1957).
H-j
PHASE BEHAVIOR OF THE METHANE-CARBON
DIOXIDE SYSTEM IN THE SOLID-VAPOR
REGION
G. M. Agrawal· and R. J. Laverman
Chicago Bridge and Iron Company
Oak Brook, Illinois
INTRODUCTION
Small amounts of carbon dioxide are usually present in natural gas mixtures.
The relatively high triple-point temperature of this contaminant necessitates its
removal from natural gas prior to cryogenic processing in order to avoid formation
of solidsin the process piping, valves, and the heat exchangers. It may also be necessary
to remove carbon dioxide from the natural gas stream which passes through turboexpanders in typical expander cycle liquefaction processes because, although commercial turboexpanders can operate into the region of partial condensation at the
turbine outlet, present design practice requires that the gas expansion process be
stopped short of the carbon dioxide frost point condition. It is therefore desirable
to have information on the phase behavior of carbon dioxide in light hydrocarbon
mixtures, especially in the range of carbon dioxide concentrations which are often
encountered in the liquefied natural gas (LNG) applications.
The phase behavior of methane-carbon dioxide mixtures has been studied over
a wide range of compositions, temperatures, and pressures by several investigators
[1-10]; however, little published information is available on the solid-vapor region
of the methane-carbon dioxide phase diagram shown schematically in Fig. 1.
Pikkar [11] made some measurements on the frost points of methane-carbon dioxide
mixtures, but to the authors' knowledge, these data have not been published in the
literature.
The present study has been undertaken in order to measure frost points of
methane-carbon dioxide mixtures in this pressure-temperature region of practical
importance to the LNG industry. Measurements were also made on two methanenitrogen-carbon dioxide mixtures whose composition was intentionally selected to
be similar to typical natural gas. These data have been used to verify the thermodynamic analysis employed here.
METHOD
Phase equilibrium studies often employ experimental techniques in which the
coexisting phases are sampled and analyzed after equilibrium conditions have been
established in a phase equilibrium cell. In the study of the solid-vapor equilibria of
EXPER~ENTAL
• Present address: Abbott Laboratories, Chicago, Illinois.
327
328
G. M. Agrawal and R. J. Laverman
I~O~----------~----------~--~
________ __________ __________
~
~
~
50\,. 10 I. l l QUIO
'000
llOulO e M.
THREE
VAPOR r"'URE
il"I'IA'S[
-------1-- / -
'--IOUIO CO 2 VAPOR PA(SSVR(
--
lOCU!.
I
O ~~==~----~----~--~~----~~--~------~----+-----~----~
' )00
-100
' 200
100
o
200
T(MPERATURE.
-r
Fig. I. Schematic phase diagram for the CH C C0 2 system.
the methane-carbon dioxide system, however, a much simpler method can be used
because the solid phase formed is pure carbon dioxide. In the present study, a nonsampling visual technique is used in which a known gas mixture of methane-carbon
dioxid~ is charged into the cell, and the pressure and temperature at which the solid
phase just begins to form are determined.
The method by which the point of inception of the formation of the solid phase
is determined can be based either on a purely visual observation or it can be determined from an apparent discontinuity which occurs at the phase boundary when
certain experimental variables are plotted in a suitable manner. The present study
utilized visual observation. A methane-carbon dioxide gas mixture of known composition was introduced into a phase equilibrium cell which was placed in a temperature-controlled cryostat. The cell pressure was maintained constant while the
cryostat temperature was gradually lowered. The temperature at which the first solid
crystal appeared in the cell was taken as the frost point temperature of the original
gas mixture at that pressure.
Experimental Apparatus
Figure 2 shows schematically the apparatus used for making the frost point
measurements. The equilibrium cell was suspended in an unsilvered glass cryostat.
This dewar contained a liquid-nitrogen-cooled, temperature-regulated liquid bath
(Freon 12) which surrounded the cell. The glass dewar was supported on a base plate
and surrounded with a stainless steel retainer vessel having plexiglass viewing windows.
The phase equilibrium cell was constructed from a modified Pemberthy sight
gage. Figure 3 shows a cross-sectional view of the cell. The rear gage glass was replaced
with a chrome-plated brass plate which could be cooled with liquid nitrogen to a
temperature slightly colder (1 to 2°F) than the bath. This produced a "cold spot" on
the inside surface of the brass plate. The cold spot where the solids first form could
Phase Behavior of the Methane-Carbon Dioxide System in the Solid-Vapor Region
G
329
2' (;01.. 0 So por COOL A Nt
Ou H. E 1
to N Z COL 0 SP OT COOL A N"T
rEs.
INL.[l
S. ... MPl. [
CONNE.C' I ON'S,
G AS. OUl L E T
l'
¢
C, N 2
COOL AN 1
e ....'
~
OUTL["
N2
BA h-l
COOl.. ANl HH.E T
IO'I'1AS( EOu tl. 16R' IUM
CELL
uNS,ILV[R[O
.... OCW. R
PLEx I CH• .t. S.S.
W I NDOw
S.Tt RRER
SVPPOR t
ROO':.o
Fig. 2. Schematic diagram of experimental apparatus.
be viewed easily through the front window by means of a focused light source. This
arrangement has been found to give consistent and reproducible results.
The desired gas mixtures were prepared from research-grade gases obtained
from the Matheson Gas Company. Stated purities were as follows:
Methane
Carbon dioxide
99.99 mole % minimum
99.995 mole %minimum
The gas mixture composition was determined using a Beckman GC-5 gas chromatograph. The accuracy of the composition measurements varied from ± O.Olmole % (for
carbon dioxide contents of 0.10 mole %) to ±0.1 mole % (for carbon dioxide contents
of 10 mole %).
Experimental Procedure
The mixture was charged into the equilibrium cell to the desired pressure, which
was measured by a 16-in. Heise pressure gage with a 0 to 1000 psig range to ± 0.5 psi.
The cryostat, which used Freon 12 as the bath fluid, was gradually cooled while the
cold spot in the equilibrium cell was continuously maintained about 1 to 2°F colder
than the bath temperature. Temperatures were measured by calibrated copperconstantan thermocouples to ± OSF. The temperature and pressure readings were
recorded when solid carbon dioxide first appeared on the cold spot of the brass plate,
and also when the solid carbon dioxide completely disappeared upon gradual warming of the cryostat.
The frost points for each mixture were determined at various pressures, starting
with the highest pressure and successively reducing the pressure.
G. M. Agrawal and R. J. Lavennan
330
L N2 COLO SPO T
COOl. ... N'
IN I.E 1
GN 2 CO LO SPOT
COO L A NT
COOL ANT
OUTlET
EN T RY
luBt
T[SoT GAS
GI.. A $S. W INOOW
L N2 COOl. E O
SRASS
COLD
SPOT
- -B -IA=--- 1Ii
PL A I E
'I"I[RMOCOU P L E S
COVER PLATE;
lEST GAS
Fig. 3. Cross-sectional view of
phase equilibrium cell.
Table I. Experimental Frost Point Data for the
CH ..-Nz-CO z System
Composition, mole
Point
No.
%
Pressure,
psia
Temperature,
OF
1
2
3
4
5
99.88
0.12
26.0
50.0
75.0
99.0
124.0
- 212.1
-201.9
-199.9
-193.6
-190.1
6
7
8
9
10
11
12
13
99.03
0.97
26.0
55.0
76.0
101.0
157.0
200.0
259.0
300.0
-180.4
-174.9
-169.4
-164.5
-160.0
-156.6
-152.6
- 150.0
Phase Behavior of the Metb~arbon Dioxide System in the Solid-Vapor Regioo
Table I. Continued
Point
No.
Composition, mole %
CH 4
Nz
COz
Pressure,
psia
Temperature,
of
14
15
16
17
18
19
20
21
22
23
98.20
1.80
25.0
27.0
50.0
51.0
76.0
101.0
125.0
176.0
228.0
328.0
-174.9
-172.9
-164.5
-163.5
-156.6
-153.1
-146.2
-146.0
-142.8
-141.5
24
25
26
27
28
29
30
31
32
33
34
96.93
3.07
25.0
26.0
51.0
51.5
76.0
101.0
203.0
299.0
302.0
311.0
404.0
-172.4
-170.9
-160.5
-160.1
-152.2
-144.7
-136.8
-132.8
-129.9
-131.8
-128.4
35
36
37
38
39
40
41
42
89.33
10.67
25.0
51.0
76.0
103.0
126.0
150.0
175.0
205.0
-139.8
-127.9
-121.9
-115.5
-110.5
-105.7
-103.8
-103.1
43
44
45
46
47
48
49
50
51
52
98.36
0.68
0.96
25.5
52.0
76.0
101.0
151.0
203.0
250.0
301.0
325.0
354.0
-180.8
-173.9
-169.0
-167.0
-161.1
-158.6
-155.1
-152.2
-150.7
-150.7
53
54
55
56
57
58
59
60
61
96.13
2.94
0.93
25.0
50.0
76.0
101.0
151.0
202.0
248.0
301.0
352.0
-182.2
-174.4
-169.9
-166.0
-161.0
-157.1
-154.3
-150.2
-148.7
331
332
G. M. Agrawal and R. J. Lavena..
1000,-------------,-------------,------------,
SOLID &
LIQUID
- - CALCULATED
w~ 1 0 0 + - - - - - -
'"
~
'""-
SOLID & VAPOR
TEMPERATURE,
IF
Fig. 4. Comparison of experimental and calculated frost points for the
CH 4 -C0 2 system.
EXPERIMENTAL RESULTS
Table I summarizes all of the experimental data. These data include frost point
measurements on five different methane-otrbon dioxide binary mixtures and two
different methane-nitrogen-carbon dioxide ternary mixtures. The binary data have
also been plotted on the pressure-temperature (P- T) diagram shown in Fig. 4, and
the data for one of the ternary mixtures are shown in Fig. 5.
THEORETICAL PREDICTIONS OF THE FROST POINTS
Thennodynamic Relationships
A theoretical analysis of the solid-vapor equilibria of methane-carbon dioxide
mixtures can be made using the following three assumptions [12.13]:
1. The condensed phase is essentially pure, consisting of the less volatile component of the mixture (e.g., pure solid carbon dioxide in the present study).
2. The condensed phase is incompressible.
3. The Benedict-Webb--Rubin (BWR) equation of state adequately describes
the thermodynamic behavior of the gas phases of each component.
For a binary system, by equating the fugacities of the less volatile component in the
coexisting phases and by applying the above assumptions, Kirk et al. P2] have derived
Pb_ Behavior of the Metb~arbon Dioxide System in the SoHd-Vapor Region
1000
-
0
I
CALCULATED
EXPERIMENTAL
}
~
333
JI
01
lOa
:>
:::...
.
0
0
0
II<
0
/
'0
-200
c!
COMPOSITION (MOLE -I. )
CH 4
-,eo
-'60
TEMPERATURE.
:
9'.13 .,.
fII12 :
:2.94-/.
CO2 :
0.93-1.
I
-120
-F
Fig_ 5_ Comparison of experimental"lUld calculated frost points for a
CHc N 2 --C0 2 system.
an expression for the calculation of the gas-phase composition in equilibrium with a
condensed solid phase at any pressure and temperature as follows:
3Clmm
+ ( RT3 ym -
2CmYlm
RT3 y
m
2
3C1mm
+ 2RT3V.m 2
2CmYlm
-
RT3 y V.
mm
CmYlm )
2 -
RT3V.
m
4
-Ym
exp~
(1)
m
Equation (1) involves only the eight BWR constants and the following relationships:
B ..
'J
= (B
) .. _ (AO)ij _ (Co)ij
0 'J
RT
RT3
(2)
(3)
334
G. M. AgrawallUld R. J. Lavermllll
The following mixing rules are used for the eight BWR constants:
Ao
=
b
=
(~YiAM2r,
Co
(t Yi l l3 t
b
c =
r
a
(t Yi l l3
c
where
=
=
(tYiCb;2r,
(t Yi l l3t
a
Bo =
y
tX ='
=
(tYiYfl2r
(t Yi l l3
[t ~ YiYJ{Bo)ij]
tX
r
(4)
(5)
Note that the use of (1) for the prediction of frost points of a known methanecarbon dioxide gas mixture at a given pressure requires an iterative procedure. Thus, a
temperature T is assumed, and the mole percent of carbon dioxide in the gas phase Y I
is calculated from (1). The iterations are continued until the calculated value of YI
matches the actual value. These calculations are most conveniently done on a
computer.
Equations (1) through (5) outlined above have been used in conjunction with the
following information:
1. BWR coefficients for methane have been taken from those published by Lee
et al. [14]. The BWR coefficients used for carbon dioxide are the same as those published by Orye [15], except for a change in the value of the Co coefficient. As has been
suggested by Orye, this coefficient has been made temperature dependent and expressed as a polynomial function of temperature. The constants of the polynomial
have been derived by simply fitting the BWR equation to the vapor pressure data for
carbon dioxide given by Canjar et al. [16]. The resulting relationship developed for
Co is shown in (6) or (7), depending upon the temperature:
Co
=
0.17218911
X lOll -
0.19943893 x 10 9 T
+ 0.97806287 x 106 T2 - 0.23207245 X 10 4 T3
+ 2.6861852T 4 - 0.OO12197792T 5
for T < 547.56
(6)
(7)
Co = 0.1698 X 10 10
for T ~ 547.56
where T is in oR.
2. The vapor pressure of solid carbon dioxide found in the literature [17,18] can
be expressed by
10giOP = 10.35607 - (2460.65/T)
(8)
where P is in lbrlfe and Tis in oR.
3. The specific volume data for solid carbon dioxide available in the literature [18]
has been expressed by
P. = 121.061 - 0.0676564T
(9)
where p. is in Ib"./ft 3 and T is in oR.
COMPARISON OF MEASURED AND PREDICTED FROST POINTS
Figure 4 shows that the comparison between the predicted frost points and the
measured values is fairly good. A point-by-point comparison for all the binary data
Phase Behavior of the Methan~arbon Dioxide System in the Solid-Vapor Region
335
Table II. Comparison of Measured and Calculated Frost Point Data for the
CH 4 -N 1 -C0 1 System*
Point
No.
Composition, mole %
CH 4
N2
CO 2
Pressure,
psia
Measured
frost point, TM ,
OF
Calculated
frost point. Te .
of
Te - TM ,
of
Ref.
1
2
3
4
99.563
99.710
99.795
99.814
0.437
0.290
0.205
0.186
48.1
78.3
125.5
144.0
-184.5
-184.5
-184.5
-184.5
-184.3
-184.4
-184.5
-184.9
0.2
0.1
0.0
-0.4
11
11
11
11
5
6
7
8
9
10
95.20
97.45
98.24
98.59
98.78
98.82
4.80
2.55
1.76
1.41
1.22
1.18
45.9
94.9
149.9
213.5
288.3
326.7
-147.9
-147.9
-148.0
-147.7
-147.5
-147.5
-147.5
-147.3
-148.3
-148.0
-148.3
-148.5
0.4
0.6
-0.3
-0.3
-0.8
-1.0
11
11
11
11
11
12
13
14
15
16
11
89.8
94.8
96.33
97.07
97.38
97.50
54.5
114.5
191.9
285.1
346.5
441.2
-130.1
-130.2
-130.2
-130.2
-130.2
-130.3
-130.7
-131.3
-130.3
-130.4
-131.3
-131.8
-0.6
-1.1
-0.1
-0.2
-1.1
-1.5
11
11
11
11
11
11
17
18
19
20
21
36.5
36.5
36.5
36.5
36.5
172.2
337.1
145.9
437.8
579.4
-187.0
-178.1
-171.7
-162.3
-166.0
-180.3
-176.3
-170.3
-162.7
-164.7
6.7
1.8
1.4
0.4
1.3
20
20
20
20
20
10.2
5.2
3.67
2.93
2.62
2.50
63.3
63.3
63.0
63.0
63.0
0.21
0.21
0.45
0.45
0.45
11
* The average absolute temperature difference between the measured and calculated frost points for the
data of Pikkar [II) is OSF.
reported in Table I has shown an average absolute difference of only 2.6°F between the
measured and calculated results for the forty-two measurements.
Although no published data have been found in the literature for frost points of
methane-carbon dioxide gas mixtures, some unpublished data have been reported
by Pikkar [11]. The latter has measured the frost points of several mixtures of this
binary by employing both a nonsampling technique (as used in the present study) as
well as a sampling technique in which a gas mixture in equilibrium with solid carbon
dioxide is sampled and analyzed. Some typical results obtained by the latter method
are shown in Table II along with the calculated values. An excellent comparison can
be seen between the measured and the calculated values with an average absolute
difference of only OSF between the calculated results and the measured values for the
sixteen data points.
EXTENSION OF THEORETICAL METHOD TO CERTAIN
MULTICOMPONENT SYSTEMS
Industrial applications, particularly in the LNG industry, rarely involve binary
mixtures, and hence the study of a mUlticomponent system is of practical importance.
Natural gas generally contains some small amount of nitrogen. It was felt desirable
to measure the frost points of methane-nitrogen-carbon dioxide mixtures whose
composition is similar to a typical natural gas, and to extend the frost point prediction
336
G. M. Ap'anl_ R. J. Layer..
method to handle such ternary mixtures where the amount of nitrogen is fairly small.
Noting that the basic assumption of a pure, incompressible solid phase which was
made in the derivation of the binary analysis also is valid for the case of a methanenitrogen-carbon dioxide mixtures, it is conceivable that one can consider the ternary
mixture as a pseudo-binary mixture consisting of carbon dioxide and a pseudoftuid, in
which the properties of the pseudoftuid are taken as those of the methane-nitrogen
mixture in question. If this is done, one can use the frost point prediction method
described earlier for the binary mixtures and extend it to the ternary system.
The frost point data for the mixture (2.94 % N 2 , 96.13 % CH 4 , and 0.93 %CO2 )
reported in Table I have been analyzed using this approach. The BWR constants for
the pseudoftuid (2.94 moles of nitrogen + 96.13 moles of methane) have been obtained
by combining the BWR constants of nitrogen as rePorted by Ellington et al. [19]
using the mixing rules of (4). Frost points are then calculated for a pseudo-binary
mixture containing 0.93 % carbon dioxide. The results are shown in Fig. 5, and it is
clear that the comparison is satisfactory.
More recently, limited frost f,Oint data for this same ternary gas mixture have
been published by Haufe et al. [2 ], utilizing a sampling technique. Table II shows a
comparison ofthese data with the predicted values, and again the comparison supports
the applicability of the calculational method suggested here.
'OOOT----------.,---------,-------,
.'":!
~ 100
::>
...'"'"0:
..
SOLIO
&,
VAPOR
'O~~~--~~~~~~~~~~--~~~~~
-250
-200
TEMPERATURE, -,.
-'50
Fig. 6. Calculated frost points for the CH.--C0 2 system.
-'00
Phase Behavior of the Methane-Carbon Dioxide System in the Solid-Vapor Regioo
337
CONCLUSIONS
Experimental data have been presented on the frost points of five gas mixtures
of methane-<:arbon dioxide and two gas mixtures of methane-nitrogen-<:arbon
dioxide. A wide range of compositions and pressures of industrial importance has
been covered.
Theoretical predictions of frost points have been made using the theory applicable to solid-vapor equilibria and by employing the BWR equation of state for the
calculations of the gas phase fugacities. The predicted values of the frost points for
methane-carbon dioxide mixtures have been found to compare closely with the
measured values as well as some unpublished data of Pikkar [11].
A method has been proposed for extending the theoretical procedure to the
calculation of the frost points of certain methane-nitrogen-carbon dioxide gas
mixtures. A limited amount of data available on the frost points of these mixtures
have been found to compare satisfactorily with the values calculated by the proposed
method.
ACKNOWLEDGMENT
The authors wish to thank Chicago Bridge and Iron Company for giving permission to publish the
results contained in this paper. The authors also wish to thank H. J. Kuklinski for his diligent and careful
work in both constructing the experimental apparatus and performing the tests reported here.
NOTATION
Ao, Bo, Co,
a, b,c,
ex, y
B
C'
P
=
constants in BWR equation
=
defined in equation (3)
= defined in equation (2)
= total system pressure
vapor pressure
p
=
R
T
V
= gas constant
= temperature
v
y
Z
p,
=
=
=
=
=
molal volume of the gas phase
molal volume of the condensed phase
mole fraction in gas phase
compressibility
density of the condensed component
Subscripts
= reference to condensed component
i, j, It
01
reference to more volatile component in a binary mixture
reference to components in mixtures
= gas mixture
= condensed component at its saturation pressure
=
=
REFERENCES
I. U. K. 1m and F. Kurata, J. Chern. Eng. Data, 17(1):68 (1972).
2. R. H. Jensen and F. Kurata, "Heterogeneous Phase Behavior of Solid Carbon Dioxide in Light
Hydrocarbons at Cryogenic Conditions," paper presented at the AIChE 66th National Meeting,
Portland, Oregon, (August 1969).
3. P. N. A. Neumann and W. Walch, Chernie. Ing. Techn., 40:241 (1968).
4. G. L Kaminishi and T. Toriumi, Rev. Phys. Chern. (Japan), 38(1):79 (1968).
5. G. L Kaminishi, Y. Arai, S. Saito, and S. Maeda, J. Chern. Eng. (Japan), 1(2): 109 (1968).
6. H. Cheung and E. H. Zander, Chern. Eng. Progr. Syrnp. Ser., 64(88): 34 (1968).
7. C. J. Sterner, in: Advances in Cryogenic Engineering, Vol. 6, Plenum Press, New York (1961), p. 467.
3J8
G. M. Agraw...... R. J. Lal'ermu
8.
9.
10.
11.
12.
13.
14.
IS.
16.
17.
18.
19.
20.
J. A. Davis, N. Rodewald, and F. Kurata, AIChE J., 8(4):537 (1962).
J. Brewer and F. Kurata, AIChE J., 4(3):317 (1958).
H. G. Donnelly and D. L. Katz, Ind. Eng. Chem., 46(3): 511 (1954).
M. J. Pikkar, Ph.D. Dissertation, University of London, London, England (1959).
B. S. Kirk. W. T. Ziegler, and J. C. Mullins, in: Advances in Cryogenic Engineering, Vol. 6, Plenum
Press, New York (1961), p. 413.
M. J. Hiza, Cryogenics, 10(2): 106 (1970).
T. W. Lee, S. B. Wyatt, S. H. Desai, and K. C. Chao, in: Advances in Cryogenic Engineering, Vol. /4,
Plenum Press, New York (1969), p. 49.
R. V. Orye, 1& EC Process Design and Development, 25(4):579 (1969).
L. N. Canjar and F. S. Manning, Thermodynamic Properties and Reduced Correlations for Gases,
Gulf Publishing Company, Houston (1967).
F. Din, Thermodynamic Functions of Gases, Vol. 1, Butterworths, London (1962), p. 123.
Matheson Gas Data Book, 4th ed., Matheson Company, Inc. (1966), p. 87.
R. T. Ellington, O. T. Bloomer, B. E. Eakin, and D. C. Garui, Thermodynamic and Transport Properties
of Gases, Liquids and Solids, McGraw-Hili Book Company, New York (1959), p. 102.
S. Haufe. H. D. Muller, and G. Tietze, Chem. Techn.l4{IO): 619 (1972).
H-6
CALCULATION OF LNG EXCESS VOLUMES BY
A MODIFIED HARD-SPHERE MODEL
J. B. Rodosevich and R. C. Miller
The University of Wyoming
Laramie, Wyoming
INTRODUCTION
In the past few years, there has been a considerable increase in interest related to
thermodynamic properties of liquefied natural gas (LNG). In particular, accurate
liquid densities are needed to provide equitable custody transfer of LNG.
Experimental results, empirical correlations, and semitheoretical models for
LNG have appeared in the recent literature. Shana'a and Canfield [1], Klosek and
McKinley [2], and the present authors [3] have reported experimental density measurements for a wide variety of pure components and mixtures. The results of the first two
studies were used by Klosek and McKinley [2] to develop an empirical correlation
applicable for temperatures from 90 to 130 K and average mixture molecular weights
from 16 to 33. The present authors [3] also reported an empirical correlation for
methane-rich LNG mixtures for the temperature range from 91 to 115 K.
The main disadvantage of the empirical correlations is that they are applicable
only to restricted ranges of composition and temperature. Calculations based on
more fundamental liquid models have an inherent advantage in this regard. Prausnitz
and co-workers [4.5] have developed semiempirical cell model partition functions
which may be used to calculate excess properties for LNG mixtures. As applied by
these authors, their models do not predict LNG excess volumes which are in quantitative agreement with experiment. This same conclusion can be made for all current
liquid mixture theories in which there is no use of excess volume data in the formulation.
By sacrifice of one excess volume data point for each binary system, the model of
hard spheres in a uniform potential field was shown [6.7] to account reasonably for the
temperature and composition dependence of the excess volume for some simple
binary liquid mixtures (A-CH4' N 2-CH 4, and N 2-A). This model was also used
successfully to predict excess volumes for the ternary mixture N2-A-CH4' without
utilization of any ternary data in the formulation. The primary purpose of the present
investigation is to attempt an extension of this model to LNG mixtures.
MODIFIED HARD-SPHERE MODEL
A model of the liquid state was proposed by Longuet-Higgins and Widom [8]
in which repulsive forces between molecules are represented by rigid spheres and
attractive forces by a uniform potential field. A close approximation to the equation
339
J. B. Rod-.icll ..... R. C. MUler
of state was presented as
1 + y + y2
a
--(1 _ y)3
RTV
PV
RT
where y
(1)
= b/4 V. Excess properties can be calculated from this equation of state if the
a and b parameters are determined for pure fluids and if suitable prescriptions are
written for application to mixtures [7.9].
This hard-sphere model is applicable to nonpolar species that are spherical in
shape. Prigogine et al. [10] proposed that for chain-type molecules, the number of
degrees offreedom that are density dependent depends upon the chain length. The net
effect on the equation of state would be to introduce a third parameter c
PV
RT = c
(1 (1+ _y +y)3y2) -
a
RTV
(2)
For spherical species, the parameter c must be unity. Actually, c is not a true
constant for nonspherical species since c -+ 1 in the ideal gas limit, but it will be
assumed that c is a constant for the liquid state. If the molecule is composed of r
monomer units, then c is related to r (for r > 1) by
3c = 3 + r
(3)
which is the total number of density-dependent degrees of freedom for the molecule.
Due to molecular size considerations for normal hydrocarbon chains, Prigogine and
co-workers chose the following definition for r in terms ofthe number of carbon atoms:
r
= (nc +
(4)
1)/2
That is, r = 1 for methane, r = 2 for propane, and r = 3 for n-pentane. Equations (3)
and (4) allow calculation of c for the normal hydrocarbons. Values of c for ethane and
higher alkanes correlate well with Pitzer's acentric factor
c
~
(1O/3)w + (7/6)
(5)
allowing calculation of c values for the branched hydrocarbons. This may not prove
to be the best approach for branched hydrocarbons; however, it will be used here
for lack of better information. Values of c for the species considered in this work are
given in Table I. Note should be made that these c values are considerably larger than
those determined by Prausnitz and co-workers [4.5] from their semiempirical cell
Table I. Pure Fluid Data and Corresponding Equation of State Parameters
species
T,K
CH 4
C2H6
C 3H S
i-C4 H lO
n-C4H lO
i-C,H12
C 6 H14*
111.63
111.63
111.63
111.63
111.63
111.63
111.63
77.35
N2
* 2-Methylpentane.
v,
cm 3 jmole
37.97
47.86
62.41
78.19
76.68
91.37
106.19
34.67
IlH.,
Jjmole
8,179
17,170
23,620
28,140
30,570
34,360
41,370
5,595
to-'a,
J-cm 3 jmole2
b,
cm 3 jmole
c
Refs.
2.755
7.773
14.165
21.28
22.73
30.55
42.95
1.718
62.65
90.66
126.3
163.0
162.5
197.4
234.4
56.23
1.00
1.50
1.67
1.79
1.83
1.91
2.11
1.03
19
3,20
3,20
3,20
2,20
14,20
14,20
21
Calculation of LNG Excess Volumes by a Modified Hard-Spbere Model
341
model approach, which utilized C as an empirical factor to be determined from pure
fluid data.
Following Snider and Herrington [II], the constants a and b can be determined
from experimental pure liquid molar volumes and heats of vaporization. Use must be
made of (2) and
I1Hv =
RT
c(1
+ y+
y2)
(1 - y)3
+
1 _ 2PV
RT
(6)
For LNG computations, the normal boiling temperature for methane (111.63 K) was
chosen for evaluation of a and b for all hydrocarbons. The data used and the resulting
constants are also listed in Table I.
Nitrogen is a nonhydrocarbon species which is often a dilute component in LNG.
From the work of Prigogine and co-workers, there is no good way to estimate its c
factor, although it may be anticipated that a value only slightly larger than unity
should work. Based on LNG mixture results, c = 1.03 has been arbitrarily taken for
nitrogen and the a and b evaluated from V and I1H v at the normal boiling point
177.35 K). The resulting values of a and b are presented in Table I.
The equation of state has been applied to mixtures by assuming a one-fluid
theory utilizing the van der Waals-type relations [7.9]
a = LLxixPij
i
j
i
j
i
j
(7)
(8)
(9)
The mixing rules chosen for the cross parameters were [7]
bij
=
[1{bf/3
+ bW)(1
- jij)F
aij = (aiiajj)I/2(bUbiibjj)I/2(1 - k i)
cij
= 1{C ii
+ cjj)
(10)
(11)
(12)
The deviation parameters jij and kij must be evaluated from binary mixture data or
estimated in some manner.
Since the equation-of-state constants a and b are proportional to the molecular
parameters [5.7] for size a, energy 1:, and coordination numbers s,
a oc sI:a 3
(13)
b oc a 3
(14)
Equations (10) and (11) are equivalent to the molecular mixing rules
a ij = t(a ii + ajj)(l - jij)
I:ij = (l:iil:jj)I/2( 1 - kij)
Sii =
Sjj =
1{Sij
+ Sji)
(15)
(16)
(17)
In (17), sij represents the total number of j-type molecules that can be placed as
nearest neighbors around a central i-type molecule.
342
J. B. Rodosevicb ad R. C. Miller
With the foregoing specifications the excess properties follow from
(18)
V E = V - LXiV;
GE = -PV - -a- - cln [VO - Y)]
-
RT
RT
RTV
- '"7 Xi {PV;
RT -
- Y)]
+ C [3Y(2
2( 1- y)2
ai
RTV; - ci In [V;(1 - Yi)]
Yi)]}
+ ci [3YJ2
2(1 _ - Yi)2
(9)
CALCULATIONS FOR LNG MIXTURES
The only adjustable parameters in the model outlined above are the k12 andj12
deviation parameters for each binary pair of species to be considered. They may be
readily evaluated if GE (or HE) and V E are known for the binary mixture at a single
composition and temperature, by forcing the model to reproduce the experimental
values. If GE (or HE) data are not available, the parameter k12 can be estimated from
gas mixture PVT data [12]; however, it appears that j12 must be fixed from liquid
mixture V E data
Data for GE and V E that were used in determining the binary deviation parameters are presented in Table II. Unfortunately, the data are far from complete for
the binary pairs of interest in LNG mixtures. Table III presents the k12 and j12
values used in the calculations to be presented here. Values in parentheses are estimates, based on the other values in the table. Many ofthe k12 values for LNG minorcomponent interactions were taken from the analysis of gas-phase data by Chueh and
Prausnitz [13].
Calculated excess volumes as a function of temperature and composition are
compared with experimental results [1,3,6,14] in Figs. 1 through 4 for binary systems
containing methane and the following second components: ethane, propane, ipentane, and nitrogen. Figure 1 is a plot of excess volume vs. composition at 108 K
for CH 4--C 2H 6 and CH 4-C 3H s ' The model describes the experimental composition
dependence reasonably well for the entire range, despite the fact that neither system
was forced to agree with data at this temperature (see Table II).
Figure 2 shows excess volumes vs. composition at 115 K and 150 K for methaneisopentane. Similar comparisons have been made for the methane-2-methylpentane
system. The temperature dependences for V E are well described by the model, while
n.
Table II. Experimental Mixture Data Used to Fit Deviation
Parameters i12 and k12
CH4-C 2 H 6
CH4-C 3 H 8
CH 4-i-C4H 1o
CH 4-n-C 4H 1o
CH4-i-C sHI2
CH4-C 6 H 14
CH 4-N 2
•
T,K
x(CH 4)
GB,J/mole
VB, cm 3 /mole
Refs.
115.00
90.68
108.00
108.00
115.00
115.00
110.09
0.7000
0.6500
0.9152
0.9206
0.7000
0.7000
0.4934
110t
187
-0.62
-0.53
-0.48
-0.54
-1.30
-1.25
-1.83
16,3
17
3
15
14
14
18,7
345
190
• 2-Methylpentane.
t Estimated from vapor-liquid equilibrium data above 130 K.
343
Calculation of LNG Excess Volumes by a Modified Hard-Sphere Model
Table III. Binary Deviation Parametersj12 and k12 Defined by Equations (10),
(11), (15), and (16)*
CH 4
C 2H 6
C 3HS
i-C 4 HIO
n-C 4H IO
i-C s H12
C 6 HI4t
N2
CH 4
C 2H 6
C3HS
i-C 4H lo
n-C 4H IO
i-C s H12
C 6 HI4t
-0.0017
0.0139
-0.0124
0.0583
-0.0227
(0.100)
-0.0241
(0.100)
-0.0326
(0.140)
-0.0458
0.1745
-0.0069
0.0336
( -0.(01)
(0.005)
(-0.002)
0.01;
(-0.002)
0.01:1:
( -0.(03)
0.02t
(-0.004)
0.03t
( -0.015)
0.05t
(0)
(0)
(0)
(0)
(0)
(0)
(0)
(0)
(-0.02)
0.09t
(0)
(0)
(0)
(0)
(0)
(0)
( -0.03)
(0.12)
(0)
(0)
(0)
(0)
( -0.03)
0.12t
(0)
(0)
(-0.04)
(0.15)
( -0.05)
(0.18)
N2
* Upper number iSjl2' lower number is k 12 .
t 2-Methylpentane.
t Taken from Chueh and Prausnitz [13].
the predicted composition dependences are slightly more sharply peaked functions
than the data would indicate. Nevertheless, the agreement for these systems is quite
gratifying when the great differences in molecular size and the large temperature
range are considered.
Figures 3 and 4 are plots of excess volume vs. temperature with composition as a
parameter. For illustrative purposes, three curves are presented for each of the binary
systems CH 4 -C 2 H 6 and CH c N 2 • In general, the experimental temperature dependence of V E is well followed by model predictions for all binary systems considered in
this work (see Table II).
Reasonable uncertainties in the k12 values do not significantly affect the predicted
temperature and composition dependences of V E • For example, a GE value of 150
J/mole was assumed for methane-ethane instead of the 110 J/mole listed in Table II.
..
'0
E
t
i
E
·0.4
-0.6
o> -D.'
:>
'....
~
::l
v
~
!
I~;
~
-, 0
o>=>
~
~ ·),0
-1.0
-1.~L.O-0:L.•-~0.'--.----=0"",-7
0.,:----:
MOLE FRACTION METHANE
Fig. 1. Calculated and experimental excess
volumes vs. composition at 108 K for CH C C 2H 6
and CHCC3HS' Data:' 0 and D, reference 3;
• and . , reference 1.
-<I
°o~'----;;'o.'o,---..cO.•.--.;,06----cO~.~
MOLE FRACTION ISOPENTANE
Fig. 2. Calculated and experimental excess
volumes vs. composition at 115 and 150 K for
CH 4-i-C s H 12 • Data: 0 and D, reference 14.
J. B. Rodosevich and R. C. MlUer
-OJ ... - -......
"'.
E ·0.2
......
~......
E.
i
~
......
• ·0.3
'"0.
>
...........
0 ...
.
:
...",
........ .. ~
'.'.
030.91 ~
0
=1.00 fOI
_
= 1.50 FOI C2H6
.
......
"
___ c
C2H6
~ ."
>
...... e:.
......
c
1 ·12
" ... ...
05.22 ~ C:zH6
A '414"
·0.6
1,0.•
~....
~
-20
C
"
... ·24
\,
...
.,.
o 512 ~ H2
A 15.51 '"
50.66
Q
.O.7.tco-=-"-""'00,,--7.'0'-;-'----,,,~o-,~,.
~
.3.2,:'-.----it'",---=,..
..----",.'".--.1:"."--"*'".
TEMPERAlURE. K
TEMPUATUIE. K
Fig. 4. Calculated and experimental excess
volumes as a function of temperature and
composition for CH 4 -N z . Data: ·0 and D.o.
reference 3; D. references 6 and 7.
Fig. 3. Calculated and experimental excess
volumes as a function of temperature and
composition for CH 4 -C zH 6 • Data: O. D.o. and
D. reference 3.
Thei12 and k12 were then recalculated as -0.0001 and 0.0215, which are significantly
changed from the values in Table III. However, the calculated V E values for Fig. 3
were changed by less than 0.01 cm 3/mole.
The dashed curves in Fig. 3 were calculated on the one-fluid theory hard-sphere
model using c = 1.00 for both methane and ethane. Again, the data quoted in Tables I
and II were used to reevaluate the pure ethane values for a and b, as well as for k12
and i12' With this approach, the temperature dependence of V E is described very
poorly. Repeating a third time these calculations of V E vs. T. but using c = 1.20 for
Table IV. Comparison of Calculated and Experimental
Excess Volumes for Some Ternary Mixtures*
Mixture
x(CH 4 )
x(C Z H 6 )
x(C 3 H s)
1
2
3
4
5
0.8466
0.7238
0.3884
0.8409
0.9055
0.1025
0.1668
0.3216
0.1086
0.0509
0.1094
0.2900
Mixture
T.K
Experimental
VB. cm 3 /mole
VB. cm 3 /mole
1
1
1
2
3
4
4
4
4
5
5
100
108
115
108
108
91
100
108
115
91
108
-0.39
-0.46
-0.55
-0.73
-0.60
-0.27
-0.39
-0.56
-0.85
-0.24
-0.50
-0.38
-0.45
-0.54
-0.65
-0.69
-0.27
-0:38
-0.55
-0.85
-0.22
-0.49
* Compositions are in mole fractions.
0.0498
x(N z)
Ref.
0.0505
0.0447
3
1
1
3
3
Calculated
Caicalatioo of LNG Excess Volumes by a Modified Hanl-Spbere Model
345
ethane (Renon et al. [4]), yielded results intermediate between the dashed and solid
curves. Thus, on the model proposed here, the value of c for ethane must be close to the
Prigogine value of 1.5 to account for the temperature variation of yE.
Calculations have been made for some ternary LNG mixtures for which experimental data are available. Comparisons between model prediction and experiment
are presented in Table IV. The agreement is quite satisfactory, with the maximum
discrepancy occurring for the nearly equimolar methane--ethane-propane mixture.
In this case, the deviation in yE would contribute an error of about 0.2% in calculating
the mixture molar volume.
SUMMARY
The liquid model of hard spheres in a uniform potential field has been modified
by introduction of a third parameter dependent on molecular shape. Mixing rules
have been developed so that liquid mixture excess properties can be calculated. A
preliminary set of mixing rule deviation parameters has been presented; some of these
parameters are based on binary mixture·data. This model has been shown to accurately
predict the temperature and composition dependences of the excess volume for
multicomponent LNG mixtures.
ACKNOWLEDGMENT
Partial financial support for this investigation was provided by The American Gas Association, Inc.,
Arlington, Virginia.
REFERENCES
I. M. Y. Shana'a and F. B. Canfield, Trans. Faraday Soc., 64:2281 (1968).
2. J. Klosek and C. McKinley, in: Proceedings of 1st Intern. Conference on LNG, Chicago, Illinois (1968),
paper No. 22.
3. J. B. Rodosevich and R. C. Miller, AIChE J., 19:729 (1973).
4. H. Renon, C. A. Eckert, and J. M. Prausnitz,lnd. Eng. Chem. Fundamentals, 6:52 (1967).
5. J. Winnick and J. M. Prausnitz, Chem. Eng. J., 2:233 (1971).
6. Y.-P. Liu and R. C. Miller, J. Chem. Thermodyn., 4:85 (1972).
7. D. R. Massengill and R. C. Miller, J. Chem. Thermodyn., 5:207 (1973).
8. H. C. Longuet-Higgins and B. Widom, Mol. Phys., 8:549 (1964).
9. K. N. Marsh, M. L. McGlashan, and C. Warr, Trans. Faraday Soc., 66:2453 (1970).
10. I. Prigogine, N. Trappeniers, and V. Mathot, Faraday Soc. Disc. No., 15:93 (1953).
II. N. S. Snider and T. M. Herrington, J. Chem. Phys., 47:2248 (1967).
12. R. C. Miller, J. Chem. Phys., 55: 1613 (1971).
13. P. L. Chueh and J. M. Prausnitz, Ind. Eng. Chem. Fundamentals, 6:492 (1967).
14. A. J. Davenport, J. B. Rowlinson, and G. Saville, Trans. Faraday Soc., 62:322 (1966).
15. J. B. Rodosevich, M. S. Thesis, University of Wyoming, Laramie, Wyoming (1973).
16. I. Wichterle and R. Kobayashi, J. Chem. Eng. Data, 17:9 (1972). .
17. H. F. Stoeckli and L. A. K. Staveley, Helv. Chim. Acta, 53: 1%1 (1970).
18. R. C. Miller, A. J. Kidnay, and M. J. Hiza, AIChE J., 19: 145 (1973).
19. R. D. Goodwin, "Tables of Provisional Values of Thermodynamic Functions for Methane," NBS
Rept. 10715, Boulder, Colorado (September 30, 1971).
20. R. D. Gunn and T. Yamada, unpublished correlation.
21. F. Din, Thermodynamic Functions of Gases, Butterworths, London (1961).
H-7
A NEW CORRELATION BETWEEN HEATING
VALUES FOR LNG CUSTODY TRANSFER
p. C. Johnson and J. P. Lewis*
Distrigas Corporation
Boston, Massachusetts
and
G. M. Wilson
Brigham Young University
Provo, Utah
INTRODUCTION
The day-to-day delivery of LNG to various customers requires an accurate
and reliable method for computing the actual amount delivered [1-3]. Delivery might
be from an LNG barge or tanker to an LNG storage tank; it might be from a storage
tank to a pipeline; or it might be from a tank-truck to an on-site storage vessel.
For contract purposes, several possibilities exist for calculating heating values
ofliquefied natural gas. If the volume of LNG transferred is measured, one would want
to know heating values on a Btu/unit volume basis; or, if the weight is measured,
then one would want to know heating values on a Btu/unit mass basis. At the present
time, ship and barge contracts for LNG custody transfer are being written by metering
volume transferred either by changes in tank liquid levels or by an appropriate flow
measurement device. The weight basis is used, however, for small deliveries by tanktruck.
Heating values on a weight basis depend only on the composition of the LNG,
while heating values on a liquid volume basis depend both on composition and
temperature. For this reason, weight measurement would be desirable if suitable
methods were available. However, since both methods are dependent on composition,
a method for determining the LNG composition is necessary in either case.
Since there are many opportunities for error in sampling and analysis of LNG,
alternative calculation methods would be desirable. One attractive method would
be to measure gas heating values on a Btu/sef basis and then relate this to heating
values on either a weight or liquid volume basis. Gas heating values can be measured
accurately and continuously with commercially accepted calorimeters which have
good calibration stability. By continuous flow measurement, sampling errors are also
minimized. By this method, the following alternatives exist for calculating total Btu
transferred.
* Present address: Transco Energy Company, Houston, Texas.
346
A New Correlation Between Heating Values for LNG Custody Transfer
347
1. Calculate heating values on a liquid volume basis from gas heating values
QT=L1V.QL;
QL=f(QV)
(1)
2. Calculate heating values on a weight basis from gas heating values
(2)
3. Calculate heating values on a weight basis from measurement of gas specific
gravity and gas heating value
QT
W
Q
=
L1W. QW
=
379.37ZQv
;
(sp. g.)Mair
(3)
where 379.37 is the molar volume, in ft3/lb-mole, of an ideal gas at 60°F and 1 atm.
4. Calculate heating values on a liquid volume basis from measurement ofliquid
density and gas heating value
QT
=
L1 V PLQW;
QW
=
f(QV)
(4)
5. Calculate heating values on a liquid volume basis from measurement of
liquid density, gas specific gravity, and gas heating value
QT
QL
=
L1V. QL
=
PL[379.37Z QV ]
(sp. g.)Mair
(5)
This presentation examines each of these possibilities and provides a comparative
analysis between these five cases.
RESULTS
Binary and multicomponent gas mixtures of two or more components were
examined. These include nitrogen, methane, ethane, propane, iso-butane, n-butane,
iso-pentane, and n-pentane. Compositions were chosen with heating values ranging
from 1000 to 1500 Btu/scf. For comparison purposes, heating values were first
calculated for assumed compositions on gas, liquid, and weight bases. Input data
were obtained using the component heating values from API-44 as summarized
by Mason and Biederman [4], liquid density from the correlation of Klosek and
McKinley [5], and the saturation temperatures from the P-V-T, Inc. Mark V computer program [6].
Assumed gas compositions and calculated heating values are summarized in
Tables I through IV. Based on these input data, the following results were obtained:
1. Liquid heating values on a volume basis can be predicted from gas heating
values if the equivalent liquid saturation pressure and nitrogen content are known.
The results are shown in Figs. 1 and 2. Figure 1 shows that deviations on binary
mixtures with heating values between 1000 and 1250 Btu/scf range from zero to 1%,
while mixtures with heating values between 1250 and 1500 Btu/scf range from 1 to
3%. In Fig. 2, deviations on typical natural gas mixtures are smaller, with errors on
the order of ± 0.2 % when a correction for nitrogen content is made. If no correction
for nitrogen content is made, a 1% concentration of nitrogen will cause an error of
about 0.7%. This would indicate that nitrogen content must be known. However.
it need not be measured during each transfer, since it changes slowly and predictably.
P. C.
J~,
J. P. Lewis, and G. M. Wi. .
Table I. Gas Compositions Used in Making Correlation
Bubble point, of
[6]
Mixture
No.
Analysis,· mole %
(balance methane)
Btu/scf
10 psia
12 psia
14 psia
16 psia
18 psia
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
l00%CH 4
7.5%C 2 H 6
15%C 2 H 6
30% C 2 H 6
45%C 2 H 6
15%C 3 H s
15%i-C4 H IO
7.5%n-C 4 H,o
15%n-C 4 H,o
15%i-C s HI2
5%n-C sH'2
10%n-C sH'2
15%n-C sH'2
1 %N2
I%N 2,7.5%C 2H 6
I%N 2,15%C 2H 6
I%N 2 ,15%C 3 H s
1 % N 2 , 15% i-C4 H,o
1 % N 2, 7.5 % n-C 4 H,o
1 % N 2, 15% n-C 4 H,o
1 % N 2, 15 % i-C s H'2
1 % N 2, 15 % n-C S H'2
1,011.6
1,069.0
1,126.4
1,241.6
1,357.0
1,239.8
1,351.4
\,l82.1
1,353.2
1,465.6
1,162.9
1,314.8
1,467.3
1,001.5
1,058.8
1,116.3
1,229.7
1,341.3
1,171.9
1,343.0
1,455.4
1,457.1
-266.6
-265.0
-263.6
-260.1
-255.4
-264.2
-264.6
-265.6
-265.1
-265.5
-266.1
-265.9
-265.9
-271.7
-270.9
-270.4
-271.7
-212.2
-273.3
-273.9
-275.7
-276.7
-262.9
-261.4
-259.7
-256.1
-251.2
-260.4
-261.0
-262.0
-261.4
-261.9
-262.4
-262.2
-262.2
-267.6
-267.0
-266.2
-268.2
-268.2
-269.2
-269.7
-271.5
-272.4
-259.8
-258.1
-256.5
-252.6
-247.6
-257.1
-257.7
-258.7
-258.1
-258.4
-259.1
-258.9
-258.8
-264.3
-263.4
-262.9
-264.0
-264.7
-265.6
-266.1
-267.8
-268.7
-256.1
-253.9
-252.1
-249.6
-244.5
-252.8
-254.9
-255.9
-255.2
-255.8
-256.2
-256.0
-256.1
-261.4
-260.4
-259.4
-260.9
-261.7
-262.4
-262.9
-264.6
-265.4
-254.4
-252.6
-250.7
-246.9
-241.7
-251.5
-252.2
-253.2
-252.6
-253.0
-253.6
-253.4
-253.4
-258.6
-257.5
-256.7
-258.0
-258.9
-259.6
-260.0
-261.6
-262.6
• Some of these mixtures are theoretical mixtures since solids would form in some cases at LNG temperatures.
2. Heating values on a weight basis can be predicted more accurately than on a
volume basis; the results are shown in Figs. 3 and 4. Figure 3 shows that deviations
for binary mixtures with heating values between 1000 and 1250 Btu/scf are within
0.15%, while mixtures with heating values between 1250 and 1500 Btu/scf show
deviations between 0.15 and 0.35%. This is nearly a factor of ten in improvement
over predicted heating values on a volume basis. The effect of ignoring nitrogen
content is more serious, however, since a 1% concentration, if uncorrected, can
produce an error of about 1.6% in the predicted heating value. Table IV shows similar
behavior for typical natural gas compositions, with errors of about 0.05% after
correction for nitrogen content. Since the effect of ignoring nitrogen content is large
for both weight basis and volume basis predictions, some method must be adopted
for estimating this effect. Possible methods would be to estimate it from boilofflosses
or from the measured heating value of the boiloff gas. The weight basis approach
does not require knowledge of liquid temperature or equivalent liquid saturation
pressure for prediction; thus, it would be easier to use.
3. The calculation of heating values on a weight basis from the measurement
of gas specific gravity and gas heating value is a direct calculation method in which
the accuracy is determined by the accuracy of the measurements. For a specified
gas heating value, the only error introduced would be the error in the measured gas
specific gravity. Commercially accepted instruments are available with good calibration stability and accuracies of ±0.1% or better. This method does not require a
knowledge of the nitrogen content as is required by the first two methods.
349
A New Correlation Between Heating Values for LNG Custody Transfer
Table II. "True" Btu Values per Pound and per Liquid Cubic Foot for Gas
Compositions Used in Making Correlation
QL, Btu/ft 3 (liquid).
Mixture
No.
QW
Btu/lb
10 psia
12 psia
14psia
16 psia
18 psia
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23,885
23,679
23,497
23,189
22,940
23,159
22,852
23,298
22,876
22,626
23,350
22,953
22,647
23,471
23,292
23,133
22,835
22,560
22,955
22,584
22,360
22,381
638,630
663,330
686,700
728,220
762,990
721,810
752,010
701,030
752,710
781,810
694,320
741,530
781,240
637,820
663,270
687,450
723,290
753,550
702,980
755,640
786,170
786,240
634,210
659,170
682,340
723,900
758,760
717,660
748,170
696,950
748,700
777,990
690,090
737,470
777,310
632,890
658,660
682,590
719,390
749,210
698,350
751,120
781,720
781,680
630,550
655,340
678,690
720,170
754,990
713,980
744,610
693,200
745,100
774,260
686,340
733,860
773,700
628,880
654,400
678,810
714,640
745,340
694,130
747,060
777,710
777,720
626,100
650,440
673,730
716,970
751,690
709,240
741,580
689,990
741,920
771,470
682,980
730,640
770,800
625,390
650,910
674,810
711,190
742,060
690,430
743,500
774,200
774,090
624,070
648,940
672,160
713,890
748,750
707,820
738,690
686,920
739,090
768,490
680,010
727,780
767,920
621,990
647,430
671,640
707,880
738,970
687,210
740,310
770,940
771,050
• Based on reference 5 for liquid density and reference 6 for bubble point temperatures.
Table III. Typical Natural Gas Compositions Compared in this Study
Mixture analysis, mole %
Mixture
No.
CH 4
C 2 H.
C 3HS
n-C 4 HIO
Q",
N2
Btu/scf
Bubble point at
14.7 psia, OF
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
97
96
96
91
91
91
86
86
86
96
95
95
90
90
90
85
85
85
3
4
2.5
8
6
5
10
8
7
3
4
2.5
8
6
5
10
8
7
0
0
1
1
2
2.5
3
4.5
5
0
0
1
1
2
2.5
3
4.5
5
0
0
0.5
0
1
1.5
1
1.5
2
0
0
0.5
0
1
1.5
1
1.5
2
1,034.5
1,042.2
1,057.3
1,088,0
1,110.6
1,121.9
1,156.5
1,175.4
1,186.7
1,024.4
1,032.0
1,047.1
1,077.9
1,100.5
1,111.8
1,146.4
1,165.2
1,176.6
-258.0
-258.0
-258.0
-256.7
-256.8
-256.8
-255.8
-255.8
-255.9
-263.0
-262.8
-262.9
-262.3
-262.5
-262.6
-262.0
-262.3
-262.4
350
P. C. Johnson, J. P. Lewis, and G. M. Wilson
Table IV. "True" Btu Values per Pound
and per Liquid Cubic Foot for Typical
Natural Gases Compared in this Study
QL at 14.7 psia,*
Mixture
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Btu/ft 3 (liquid)
639,320
642,900
648,330
661,470
669,240
672,990
686,750
692,630
696,210
638,160
641,480
647,030
660,850
668,730
672,600
686,670
692,880
696,460
23,800
23,772
23,713
23,614
23,534
23,495
23,397
23,338
23,304
23,397
23,373
23,321
23,234
23,164
23,130
23,044
22,993
22,963
* Based on reference 5 for liquid density and reference
6 for bubble point temperatures.
2 .-----~----,------r----~----~
-4 L-----~----L-----~----~--
1000
11 00
1200
1300
a V, BTU / ser
1400
__~
1500 Fig. 1. Heating value on a liquid volume basis;
deviation error for binary mixtures.
351
A New Correlation Between Heating Values for LNG Custody Transfer
3r-----,------r----~------;_--__,
2t-
o
o
JI
..0
It-
'"
1..0
a
a
__ {;>--t:r- - - - -6"/:,.- - -
--_
'--"'"IS
"
o 1----_-_-_O_o-_C5_2_-un-_-,..-;-O-"-"'o-_--<-~£o;_----'oL
"'O"-c-t-'-=-='----t
..0
a
.!:
_O-_~-----Q.o-----..().
0-"
c:
.~
.!!
~
-I
a
..
..
Typical Natural Gases
r-
o No N2
c
o I" N2 with correction
~
a..
'" I" N2 without correction
-2 t-
Fig_ 2. Heating value on a liquid
volume basis; deviation error for
typical natural gases.
-3~----~----~----~------~--~
1000
1100
1200
aV ,
BTU/SCF
4_ The calculation of heating values on a liquid volume basis from measurement
of liquid density and gas heating value has the advantage of good prediction accuracy
for the heating value on a weight basis_ Initial attempts at direct liquid density
measurement have not been very successful; so this method has not proved possible_
However, it is expected that new methods for liquid density measurement will be
developed which will make this method more attractive than using the first method_
3
2
0
With N2 Correctlon
0
1% No !\!2
Correct ion
0
~
·~I
It-
O.
..
0
0
•
0
c:
o lo
0
0
0
"-
vv
0
0
0
0
0
c:
0
.
0
>
-I
r-
-
a
Q
0
0
0
0
0
-
-2 r-
Fig. 3. Heating value on a weight basis;
deviation error for binary mixtures.
0
0
-3~----~----~----~--~~--~
1000
1100
1500
1200
1300
1400
aV,
BTU/SCF
P. C. Joa.-, J. P. Lewis, ad G. M. WiIIoB
352
3r---~-----~1----r---~-----'
o
2
r-
o
.i~!,l
IC
,;
1-
0
Typical
Natural
Ga •••
-----.
~-
_.6_
on..
(Wi ~ N3....Correction)
.5:
.g
;;
.:
o
-I
•e
-
I" NI
~
:.
Typical
Natural
Ga...
r-
___ ..a--oJl- --~o~~rrection)
.J:J--~
-2
-
-3~----~--~~--~----~----~
1000
1100
1200
aV ,
Fig. 4. Heating value on a weight basis;
deviation error for typical natural gases.
BTU/SCF
If this occurs, prediction accuracy would be comparable to accuracies of ±0.05%
obtainable by the second method outlined above. Knowledge of liquid temperature
would not be required. This may prove to be a very desirable method as new density
measurement equipment is developed.
5. The calculation of heating values on a liquid volume basis from measurement ofliquid density, gas specific gravity, and gas heating value is a direct calculation
method requiring no knowledge of nitrogen content or liquid temperature. Accuracy
obtainable would depend on the accuracy of the specific gravity and liquid density
measurements. Errors less than 0.1 % could probably be achieved.
HEATING VALUE ON LIQUID VOLUME BASIS FROM
HEATING VALUE ON GAS BASIS
The heating value on a liquid volume basis is related to the heating value on a
gas basis through
QL = 379.37ZQv/YL
If one could derive a relation between
YL
~
(6)
YL and QV such that
f(QV)
(7)
then one would have a method for predicting QL from QV.
At constant saturation pressure, YL can be assumed through first-order terms
to be given by
Yd379.37Z =
L xtBj
(8)
j
The gas heating value is also dependent on composition as given by
QV =
L xjhj;
j
hj = HJ379.37Z
(9)
A New Correlation Between Heating Values for LNG Custody Transfer
353
By appropriate rearrangement, one can combine these two equations. For example,
if one considers a natural gas system containing nitrogen, this rearrangement results
in
ilL
_
v _ h Bz
379.37Z - Bo + (Q
0) hz
-
-
BCH•
(10)
h
CH.
where
Bo
=
X N2 B N2
ho
=
XN2hN2
+ (1 + (1 -
X N2 )B cH •
(11 )
xN,)hcH.
(12)
(13 )
hz =
L (1
X·
-
I
X N2 -
X CH .)
( 14)
hi;
Correlation parameters Bi and hi are summarized in Table V.
The ratio (B z - BCH.)/(h z - hCH.) is found to be nearly constant and independent
of gas composition. In this case, one can use a generalized parameter kB for the ratio.
Equation (10) for ilL now becomes
ilJ379.37Z
=
Bo
+ (Qv - ho)k B
(15)
Deviations in Figs. 1 and 2 were calculated using a value of 5.5 x 10 -7 ft 3 /Btu for
k B . Since Bo and ho are both dependent on nitrogen content, ~ correction can be
made directly in (15). Figures 5 and 6 provide these parameters for the calculation.
The following example serves to illustrate this method. Assume a gas with a
heating value of 1130 Btu/scf, a tank holding pressure of 16 psia (1.3 psig + 14.7 psia
barometric pressure), and a nitrogen content of 0.5% . From Fig. 5, Bo = 16.05 X 10- 4
ft 3 (liquid)jscf; while from Fig. 6, ho = 1007.5 Btu/scf. From (15), ilL can be evaluated
as
V.
379.~7Z
+ (1130 - 1007.5)(5.5
=
16.05 x 10- 4
=
16.72 x 10- 4 ft3 (liquid)/scf
x 10- 7 )
Table V. Correlation Constants Developed from this Study
Component
h" Btu/scf
w;.lb/scf
B;. ft 3 (liq uid )/scf
Nitrogen
Methane
Ethane
Propane
i-Butane
n-Butane
i-Pentane
n-Pentane
n-Hexane
0
1011.59
1783.72
2563.28
3351.99
3376.40
4218.53
4249.17
5169.78
0.073869
0.042367
0.079930
0.118362
0.157888
0.158576
0.200563
0.201561
0.246921
- 5.5 x 10- 4 + 0.25 X 10- 4 (psia)
15.4 x 10- 4 + 4.6 X 10- 0 (psia)
BeH • + 3.82 x 10- 4
BeH • + 8.73 x 10- 4
BeH • + 13.56 x 10- 4
BeH • + 13.60 x 10- 4
BeH • + 18.03 x 10- 4
BeH • + 18.14 x 10- 4
BeH • + 22.66 x 10- 4
P. C. Jom.-, J. P. Lewis, and G. M. WU-
16.3 .....- - - - - - - - - - - - - - -....
I 6.0 x 10. 4
....U
15.7
II)
~
cD
15.4
15.3
15.2 Not.: Valid only for QV less
than 1250 BTU/SCF
15.1
15.0
Fig. 5. Graphical aid for computing QL
from QY; Bo vs. nitrogen composition at
varying pressures.
14.9.10-40~------2----3----4---..J5
Mole Percent Nitrogen
Finally, QL can be calculated from
QL
1130/(16.72 x 10- 4)
=
379.37ZQv/VL
=
675840 Btu/ft 3 (liquid)
=
1020 .....- - - - - - - - - - - - - - -....
.0450
1010
1000
_h_o_
....
u
.0440
990
ID
....
u
....
:::>
....
(/)
....
(/)
~
980
;
970
.0430
__
W_o_
960
950~----------------J~420
o
2
3
Mole Percent Nitrogen
4
5
Fig. 6. Graphical aid for computing QW
and QL from QY; ho and Wo vs. nitrogen
composition.
A New Correlation Between Heating Values for LNG Custody Transfer
355
HEATING VALUE ON WEIGHT BASIS FROM
HEATING VALUE ON GAS BASIS
The heating value on a weight basis is related to the heating value on a gas
basis by the relation
(16)
An analysis of the problem shows that if one could derive a relation between Mav and
QV such that
(17)
we would have a method for predicting QW from QV. The average molecular weight
of a mixture is given by
Ma.l379.37Z
=
L XiW i ;
Wi =
MJ379.37Z
(18)
i
Analogous to the relationship developed for the heating values on a liquid volume
basis, one can derive a similar expression for the heating values on a weight basis.
Thus
(19)
where
(20)
i #- N 2 , CH 4
(21 )
and other parameters are as previously defined. Correlation parameters Wi and hi
are summarized in Table V.
The ratio (w z - wcH.)/(hz - hcH .) in (19) is found to be nearly constant and
independent of gas composition; in this case, one can use a generalized parameter
kw for the ratio. Equation (19) for Mav now becomes
(22)
Deviations in Figs. 3 and 4 were calculated using a value of 4.89 x 10- SIb/Btu for
kw. Since Wo and ho are dependent on nitrogen content, a correction can be made
directly in (22). Figure 6 provides these parameters for the calculation.
A second example serves to illustrate this method. Assume in this case the same
gas as previously with a heating value of 1130 Btu/scf and a nitrogen content of 0.5% .
From Fig. 6, Wo = 0.04252 and ho = 1007.5 Btu/scf. Equations (22) and (16) then
permit the evaluation of M av and QW, respectively:
37~;;Z
=
0.04252
=
0.04852
+ (1130
- 1007.5)(4.89 x 10- 5)
QW = 379.37ZQv/M av = 1130/0.04852
=
23290 Btu/lb
P. C. JohMoa, J. P. Lewis, ud G. M. Wilson
The liquid density can be estimated from a combination of QL and QW since
QL = PLQW or PL = QL/Q W. Thus, for the 1130 Btu/scf containing 0.5% nitrogen
stored as LNG at 16 psia, we have from the calculation examples the following results:
QL = 675,840 Btu/ft 3(liquid)
QW = 23,290 Btu/lb
PL
=
675,840/23,290
=
29.018 Ib/ft 3
SUMMARY
A direct correlation between heating values for liquid natural gas has been found
which minimizes effects from other physical properties. Thus the heating value per
liquid cubic foot or per pound can be calculated from the gross heating value per
standard cubic foot of vaporized gas. These direct correlations are only affected by
the nitrogen content of the gas, so a correction must be made in the calculation procedure. These new correlations make it possible to calculate the heating value per
pound or per liquid cubic foot without knowing the relative amounts of the hydrocarbons present. The heating value of the gas is readily measured and the nitrogen
content can often be predicted from vaporization losses. Because of the simplicity
of the correlations, they are recommended for day-to-day conversions of volume or
weight to total Btu delivered between various buyers and sellers of liquid natural
gas.
This presentation is intended as an introductory paper to stimulate discussion
of alternate means for determining heating values of LNG. It seems that existing
procedures for determining gas heating values should be used for determining liquid
heating values because the procedures are well established between buyer and seller.
This presentation therefore has examined various methods by which this might be
done.
NOTATION
change in molar volume at constant saturation pressure due to addition of component i,
fe(liquid)/scf
Bo = defined by equation (11)
B.
= defined by equation (13)
Hj
= gross heat of combustion of component i on a mole basis
hi
= gross heat of combustion of component i per scf
ho = defined by equation (12)
h.
= defined by equation (14)
kB = constant taken as 5.5 x 10- 7 ftl/Btu
kw = constant taken as 4.89 x 10- Sib/Btu
M Ai. = molecular weight of air
M.v = average molecular weight of mixture
Mi = molecular weight of component i
QL = heating value on a liquid volume basis
QT = total heat transferred
QY = gas heating value on a volumetric basis, standard conditions
QW = gas heating value on a weight basis, standard conditions
VL = molar volume of LNG
~ V = volume transferred
Wi
= mass of component i per scf
Wo
= defined by equation (20)
w.
= defined by equation (21)
~w = weight transferred
Bi
=
A New Correlatioa Between Heating Values for LNG Custody TrlUlSfer
Xi
Z
= mole fraction of component i
= compressibility factor
PL = density ofliquid
sp.g. = specific gravity of gas
REFERENCES
I. G. J. Boyle and D. Reece, in: Proc. 2nd Intern. LNG Conference, Paris, France (1970).
2. J. A. Brennan, J. W. Dean, D. B. Mann, and C. H. Kneebone, "An Evaluation of Cryogenic Volu-
metric Flowmeters," NBS Tech. Note No. 605 (July 1971).
3. F. F. Shamp,lSA Trans. 10(3):219 (1971).
4. D. M. Mason and N. P. Biederman, "Natural Gas Physical Constants and Heating Value Calculation
Procedures," Institute of Gas Technology Rept., Project 8540 (December 1969).
5. J. Klosek and C. McKinley, in: Proc. 1st Intern. LNG Conference, Institute of Gas Technology, Chicago,
Illinois (1968), Session 5, paper 22.
6. The P-V-T, Inc. Mark 5 Computer Program, available from the Natural Gas Processors Association, 808
Home Federal Building, Tulsa, Oklahoma.
/-}
SCALED PARAMETRIC EQUATION OF STATE
FOR OXYGEN IN THE CRITICAL REGION
J. M. H. Levelt Sengers
National Bureau of Standards
Washington, D.C.
and
W. L. Greer* and J. V. Sengers
University of Maryland
College Park, Maryland
INTRODUCTION
The art of correlating thermodynamic properties of fluids has greatly evolved
in the last decade, due not only to the increased availability and capability of computers, but also to the refinement and sophistication of statistical tools. Thus, recently
several engineering correlations of thermodynamic properties of fluids have been
presented that fit precise data generally to within their accuracy. An exception has
to be made, however, for the data near the critical point since persistent systematic
errors, exceeding the experimental precision, show up in the critical region.
Thermodynamic properties show an anomalous nonanalytic behavior in the
critical region that cannot be represented by analytic equations used in engineering
correlations [1]. The shape of the coexistence curve illustrates this problem, i.e.,
analytic equations can only yield quadratic or quartic coexistence curves, while it
has been known since the beginning of the century that this curve is virtually cubic.
An even more immediate piece of evidence of nonanalyticity is the observed weak
divergence of the specific heat at constant volume. For analytic equations of state,
this quantity is bound to remain finite when the critical point is approached in the
one-phase region. A nonanalytic description is therefore mandatory.
SCALING LAWS
The first step toward a systematic description of the critical anomalies was the
formulation of power laws. Thus, the coexistence curve is described asymptotically
as
IIp = ± Bill Till
(1)
The critical isotherm is described asymptotically by the power law llP = D(llp) Illpi " - 1.
Along the critical isochore, the compressibility diverges as KT = I1llTI-Y and the
specific heat per unit volume as Cy/V = (A/a) III TI-IZ. The various paths along which
the power laws are defined are indicated schematically in Fig. 1.
• Present address: Department of Chemistry, George Mason University, Fairfax, Virginia.
3S8
Scaled Parametric Equation of State for Oxygen 10 the Critical Region
liT
liT
SCALING
a Mev
f3
a_a 1 _a ll
-COEX.
8-0
Y • KT
8 - CRiT.
ISOTH.
8
liP
8 - -lib
8- .l/b
liP
8--1
COEXISTENCE
BOUNDARY
COEXISTENCE
BOUNDARY
Fig. 1. Special curves in the I1T-l1p plane and the
power-law exponents defined along them.
Fig. 2. Values of the linear model contour parameter (J at the special curves in the I1T-l1p plane.
The next step was a formulation of the thermodynamic behavior that incorporates all these power laws. For this purpose, Widom postulated that the anomalous
part of the Helmholtz free energy density is a homogeneous function of its arguments.
It is a well-known property of homogeneous functions that the number of independent
variables can be reduced by one. Therefore, homogeneity of thermodynamic properties
in the critical region implies that, when properly scaled, they become functions of
only one variable which is a combination of the original two independent variables.
Here we shall need the scaling law for the chemical potential p.. Just as the
familiar P(V, T) equation of state follows by differentiating the Helmholtz free energy
A(V, T), so the p.(p, T) equation of state is obtained by differentiating the Helmholtz
free energy density A(V, T)/V. The reason for using the rather unfamiliar thermodynamic relation p.(p, T) is that it shows a striking antisymmetry with respect to the
critical isochore [1.2]. Along the critical isotherm, the asymptotic behavior of the
p.(p) relation is the same as that of the P( V) relation:
!J.p.(p, 7;,)
=
D(!J.p) I!J.pl<l - 1
(2)
If we now divide !J.p.(p, T) by (2), then the resulting scaled chemical potentials are
no longer functions of !J. T and !J.p separately, but rather of the single variable
x = !J.T/I!J.pll/fI. Hence, the scaling law for the chemical potential is
!J.p./(!J.p)l!J.pl<l-1
= h(x)
(3)
That thermodynamic properties do indeed scale in the critical region has been
demonstrated convincingly [2].
In order to use the scaling law (3) for actual data fitting, it is necessary to have
an expression for the function h(x). This has turned out to be a major problem.
The function h(x) is constrained by a number of conditions; some of these have to
do with thermodynamic stability (KT > 0, C y > 0); others, however, arise from the
fact that the chemical potential has to be completely free from nonanalyticities in
the entire one-phase region except at the critical point. It has not yet been possible
to devise a closed form for the function h(x) that fulfills all conditions.
J. M. H. Levelt Seagers, W. L. Greer, aDd J. V. Seagen
Schofield et al. [3] proposed an alternative way of formulating a scaled equation
of state. In their approach, the physical variables are written in terms of two parametric variables rand (), where the variable r describes "distance from the critical
point" and the variable () "location on a contour of constant r." The purpose is to
incorporate all anomalies represented by the power laws in the r dependence while
the () dependence is kept strictly analytic. In this way, nonanalyticities can only occur
at the critical point, while no singularities will appear anywhere else in the one-phase
region. Schofield et al. proposed simple forms for these transformations that incorporate the power laws, the scaling laws, and the antisymmetry of the JI.-P relation:
!!"JI. = arfJl>() (1 _ ()2)
!!..P
= rfJm«(),
!!..T
= r(1 -
m«() antisymmetric
(4)
b2 ()2)
By constructing the ratios !!..JI./(!!..p)l!!..pl<l-l and x = !!..p/I!!..TI 1 / 1l, it is seen that both
of these scaled variables depend only on (), so that the scaling law (3) is implied by
the parametric equation (4). The variable () runs from -1 to + 1, as indicated in Fig. 2.
The constants a and b are as yet unspecified, and so is the functional form of m«().
From an inspection of experimental data, Schofield et al. found, however, that m«()
is linear for a number of substances, magnets as well as fluids. Thus
m«()
= k()
(5)
With the assumption in (5) for m«(), the parametric equation is called the "linear
model." The latter, in addition to its simplicity, has the additional advantage of being
analytically integrable so that other thermodynamic properties, e.g., pressure and
free energy, can be obtained in closed (parametric) form [3.4].
PROCEDURE FOR DATA ANALYSIS
A statistical method for fitting the linear model to experimental data [5] has
been developed and a brief description of the procedure is given below.
The linear model, as formulated in (4) and (5), contains seven adjustable parameters, namely Pc' 1'." a, b, k, p, f>. Because the problem is nonlinear in most of the
parameters, it is impractical to try and determine all seven parameters in one fit to
experimental data. A reduction in the number of adjustable parameters is obtained
as follows. First, in many cases, reliable data are available for the coexistence curve,
so that the constant B and exponent p in (1) can be determined. From (4) and (5),
it is seen that at coexistence «() = ± 1, see Fig. 2)
l!!..pl/I!!..TlfJ = B = k/(b 2
-
I)P
(6)
Thus, with Band P known, the value of k is fixed once the value of b is chosen.
Second, the value of Pc is readily obtained from an analysis of the coexistence-curve
diameter. Third, it was suggested by Schofield et al. that a good value of b 2 is obtained
from
b~HL
= (f> - 3)/(f> - 1)(1 - 2P)
(7)
In this paper, however, b 2 has been left as a free parameter, in addition to Ye, f>, and
a. The parameter a has been calculated varying Ye, f>, and b2 stepwise on a threedimensional grid The choice of Ye, f>, and b2 that leads to the smallest standard
Scaled Parametric Equadoo of State for Oxygen iD the Critical Region
361
deviation of the fit is considered optimum. Error and weight assignments have been
made on the basis of the estimated experimental errors in Il, p, and T.
The calculation of a proceeds in two steps. First, the scaled variable x is constructed. From (4) .~nd (5), we have
x = AT/IApll/1l = (1 - b2 ( 2 )/k 1/1I 1011/1l
(8)
°
A choice of b 2 is made, k is calculated from (6), and then (8) can be solved for in
terms of the experimental quantity x. The errors in x, due to the estimated experimental errors in T and p, are calculated and propagated into using (8). Thus, to
each experimental p, T point, a value of along with its variance is assigned for
given choices of band 1'".
After a choice of b is made, the corresponding value of a and its error are calculated for each experimental Il(P, T) point using
°
a(1 - (2)/k"IOI"-1
°
= AIl/ApIApl"-1
(9)
For the set of values of a thus obtained, a weighted average a and its variance are
then calculated. The procedure is repeated for other choices of b 2 , b, and 1'". The
best fit is that with the smallest value of the e~timated variance X2 of the fit. If the
error assignments are correct and the model is adequate, the minimum X2 should
be close to unity. It is also checked whether the individual a values are independent
of to within one or two standard deviations.
The linear model has been tested for a number of gases with satisfactory results
[5.6]. It therefore seems reasonable to proceed with the assumption that the linear
model gives an adequate description of the critical region of fluids.
°
APPLICATION TO DENSITY PROFILES
In the vicinity of the critical point, the medium becomes so compressible that
gravity sets up considerable density gradients. Since the sum of the chemical potential
Il(P) per particle and the gravitational potential mgh is constant, we have, at two
levels hi and h2 in the cell with densities PI and P2' ll(h 2) - ll(h 1 ) = mg(hl - h 2).
Or, in terms of All and a reduced height h*,
(10)
Thus, the chemical potential differences in the cell are equal but opposed to the
reduced height differences. Therefore, if the density is measured as a function of
height, direct information about the Il(P) relation is obtained. Such measurements
were recently performed in oxygen by Weber C]. He used a stack of five capacitors
spaced at 2.5-cm intervals along the height of the cell. The 8-P relation for each
capacitor was calibrated by filling the cell away from the critical point at a pressure
and temperature where the density was uniform and could be calculated from the
known equation of state.
The results of Weber's experiments are p(h) data, usually in groups of five, with
most data concentrated along special curves; namely, the critical isochore, critical
isotherm, and coexistence curve. The density profile data span 0.5 K in temperature
and ±0.16 in Ap. Weber analyzed his own data for power-law behavior along the
special curves. To this end, he had to extrapolate to the position of the meniscus
and to make curvature corrections on the supercritical isotherms. On the other hand,
all his data can be analyzed directly by means of the linear model equation of state.
J61
J. M. H. LeYelt Seagers, W. L. Greer, IUld J. V. Seagers
Data not located on the special curves will be included, no extrapolations will be
required, and no curvature corrections will have to be made.
Consider two levels i and j in the cell. As discussed before, the values 0i and OJ
are calculated with their variances from the scaled variable x, using (8) and trial
values for 1'" and b 2 • For these two points, using (9) and (10), a value of a can then be
calculated (designated as aij) from
_.
aij - (hi
•
-PlI[ OJ{1 - 0/)
O~1 - 0/)
- hj )IATI
11 _ b20/IPlI - 11 _ b20/IPlI
]-1
(11)
The variance in aij is estimated from the errors in 0i and OJ, and the estimated experimental errors in ATand the levels hi and hj. A weighted average a is computed from
the set of aij. The process is repeated for other choices of 1'", ~, and b 2 . The set of
parameters 1'", ~, and b2 that minimizes the standard deviation of a is considered
optimum. Since both i andj run from 1 to 5, not all choices of(i]) pairs are.independent.
For those sets of data reported by Weber in which five densities were recorded along
the length of the cell, the (i]) pairings (1,31 (1,4), (2,51 and (3,5) have been used. If
fewer than five points were reported, the density pairs (i]) were again chosen to give
symmetric separations. In addition to the data presented by Weber [7] new, unpublished data kindly provided by Weber have also been included.
RESULTS
In fitting the profile data, errors of 0.3 mdeg in temperature, 0.05% in density,
and 0.01 em in the location of each capacitor have been assumed; an error of 0.038 em
in hi - hj ' where i and j are usually two to three spacings apart, has been assumed.
In accordance with Weber's own analysis of his extensive set of coexistence curve
data obtained in the same cell, values for f3 = 0.353 and B = 1.8190 have been selected.
Weber obtained T.: = 154.576 ± 0.001 K as a best value. Since the fit here is quite
insensitive to the choice of b 2 , the latter has been fixed at the value from (7) recommended by Schofield et al. Weber's values of Pc = 436.2 kg/m 3 and p. = 50.43 bar
have been used as reduction factors.
Varying ~ and 1'" stepwise, values of a have been calculated for approximately
160 (hi' hj) pairs. The lowest values obtained for were about 2.6; they were obtained
for a number of pairings (~, 1',,), such as (4.29; 154.577); (4.30; 154.576); and (4.35;
154.575). The best fit, for (4.29; 154.577), is listed in Table I, along with the optimum
fit at Weber's choice of 1'". Table I also lists the fit for the parameter pair (4.533;
154.576), which was selected by Hohenberg and Barmatz on the basis of Weber's
'r
Table I. Linear Model Parameters for Oxygen·
1'.,K
6
a
b~L
x2
*B =
Best fit
Best fit for
1'. = 154.576 K
Fit for
6, 1'. from Ref. 4
154.577
4.29
12.41 ± 0.52
1.3337
2.59
154.576
4.30
12.55 ± 0.52
1.3399
2.66
154.576
4.533
25.82 ± 1.10
1.4759
3.67
1.8190,p = 0.353, Pc
= 436.2 kg/m 3 , and Pc = 50.43 bar from reference 7.
Scaled Parametric Equation of State for Oxygen in the Critical Region
363
Table II. Critical Exponents and Coefficients for Oxygen
Best fit
tX
f3
Y
(5
A
A'
A"
B
r
r'
D
0.133
0.353
1.161
4.29
1.077
0.931
3.089
1.819
0.09949
0.02533
2.018
7;
Best fit for
= 154.576 K
0.129
0.353
1.165
4.30
1.118
0.973
3.092
1.819
0.09907
0.02513
2.025
(5,
Fit for
7; from Ref. 4
0.047
0.353
1.247
4.533
3.773
3.644
5.157
1.819
0.05421
0.01277
3.608
power-law analysis. Their estimate of 26.60 for a is confirmed rather well, but X2 for
this choice of b is 3.67, considerably higher than for lower values of b. A choice of
T.: = 154.572 would bring X2 down to 2.8, but this value of T.: is uncomfortably low.
The various critical exponents and coefficients for the power laws implied by
these fits are presented in Table II.
Considering the individual data pairs, one finds that the fit of the linear model
is indeed quite good along the coexistence curve and the critical isotherm. However,
it turns out that data points at the critical isochore in a small temperature range
above T.: are responsible for the fact that X2 is larger than one. The solution for e
from (8) fails to converge for points near Pc that are within 0.02° from the critical
temperature, so that these points were eliminated automatically. The points near
Pc at temperatures between 0.02 and 0.04° above T.: appear to deviate systematically
by several standard deviations irrespective of the choice of b. This poor fit of the
linear model to the slightly supercritical data must cause the rather large discrepancy
between exponents from Weber's power-law fits and those from linear model fits.
Specifically, Weber's)' = 1.25 and the present}' = 1.16 are at the extremes of the
range of known}' values. It is planned to investigate the behavior of the linear model
near e = 0 more thoroughly before proceeding with thermodynamic calculations
for oxygen.
ACKNOWLEDGMENTS
The part of the research conducted at the University of Maryland was supported by the National
Aeronautics and Space Administration, Grant NGR-21-002-344. Computer time.for this project was
provided by the Computer' Science Center of the University of Maryland. The work done at the National
Bureau of Standards was supported by the Office of Standard Reference Data.
NOTATION
A, A', A"
a
B
b2
CvlV
D
g
= coefficients of power laws
parameter in linear model equation of state
coefficient of power law
= parameter in linear model equation of state
= specific heat per unit volume
= coefficient of power laws
= gravitation constant
=
=
J. M. H. Levelt Seagers, W. L. Greer, ud J. V. Seagen
h
h*
KT
k
m
P
Pc
!!P
r
T
7;
!!T
V
x
= height
= (mgpjPc)h
= isothermal compressibility
= parameter in linear model equation of state
= molecular weight
= pressure
= critical pressure
= (P - Pc)/Pc
= parametric variable
= temperature
= critical temperature
= (T - 7;)/7;
= volume
= !!T/I!!pI1IJl
Greek letters
CX,P,I',lj = exponents of power laws
r,r'
8
/l
!!/l
P
Pc
!!p
x2
= coefficients of power laws
= parametric variable
= chemical potential
= {P(p, T) - /l(p" T)}p)P c
= density
= critical density
= (p - Pc)/Pc
= variance of fit
REFERENCES
I. J. M. H. Levelt Sengers, Ind. Eng. Fund., 9:470 (1970).
M. Vicentini-Missoni, J. M. H. Levelt Sengers, and M. S. Green, J. Res. NBS, 73A:563 (1969).
P. Schofield, J. D. Litster, and T. Ho, Phys. Rev. Lett., 23: 1098 (1969).
P. C. Hohenberg and M. Barmatz, Phys. Rev. A, 6:289 (1972).
W. L. Greer, J. M. H. Levelt Sengers, and J. V. Sengers, Bull. Am. Phys. Soc., 17:277 (1972), and J.
Phys. Chern. Ref Data, to be published.
6. T. A. Murphy, J. V. Sengers, and J. M. H. Levelt Sengers, in: Proceedings 6th Symposium on Thermophysical Properties (P.E. Liley, ed.), ASME, New York (1973), p. 180.
7. L. A. Weber, Phys. Rev. A, 2:2379 (1970).
2.
3.
4.
5.
/-2
SUPERFLUID THERMODYNAMIC TRANSPORT
LIMITS FOR LIQUID HELIUM n
C. Linnet, R. C. Amar, Y. G. Wang, and T. H. K. Frederking
University 0/ California at Los Angeles
Los Angeles, California
INTRODUCTION
Recent advances in liquid He II technology pose questions concerning the
utilization of "superphenomena." In particular, it is useful to know the transport
conditions that separate classical flow (with large flow resistances) from nonclassical
phenomena (with small or entirely negligible resistances). The purpose of the present
investigation, therefore, is a thermodynamic evaluation of the critical flow rates at
the initiation of observable flow resistance (upper stability limit at which superfluidity
becomes thermodynamically unstable). The critical flow rate is evaluated on the
basis of the modified (Gorter-Casimir) two-fluid model [1] and the extended theory
of second-order phase transitions [2.3]. The comparison with experiment shows that
in general the thermodynamic critical rate is an upper bound to experimental data
which approach the limit closely when T -+ TA • Good agreement also is obtained
for specific geometries in a limited temperature range below the A.-point.
THERMODYNAMIC FUNCTIONS
Two-Fluid Thermodynamics
The two-fluid approximation is based on the observation [4] that the He I
entropy above the A.-point is quite well represented by a linear function of temperature
[with a free energy density f.. = -(y/2)T2, wherey is the volumetric entropy coefficient].
The evaluation in this study has been based on a constant power-law coefficient
n = dOog S)/d(log T) = (1 + 6)/(1 - 6). The free energy density is considered the sum
of the two fluid contributions
/ =
f. + f..(1
- y)"
(1)
where f. = - yU 0 (U 0 is the volumetric condensation energy), and In is obtained
from the He I result. Minimization of/with respect to y = p./ p yields the equilibrium
values of y and / subject to the conditions that at T = TA , we have y = 0, and as
T -+ 0, we have y -+ 1. Thus, the thermodynamic derivatives are evaluated readily.
These results are given in Table I for a general power law with an approximate
exponent n of 5.
The volumetric entropy coefficient of saturated liquid He II is equal to 0.115
J/cm 3 -K2. The condensation energy for an exponent of 5 then has the value of
0.18 J/cm 3 • Further details of the present two-fluid properties may be found elsewhere [5].
c. LiImet, R. C. A....... Y. G. WIIIlI. aad T. H. K. Frederking
Table I. Thermodynamic Two-Fluid Functions*
Eq.No.
Supertluid fraction
y=l_t 2/(1-')
at equilibrium
Equilibrium free
I = tyT2[(e - i)t 2 /(1-,) - e]
energy density
Entropy S = - dl/dT
S = l'T.t(1 +.)/(1-.)
(volumetric)
Specific heat
C = T(dS/dT)
C = l'T.t(1 +.)/(1-')(1 + e)/(I - e)
(volumetric)
Free energy density
difference
INI = tyT/[(e - 1)t2/(1-.) - e + t 2]
INI =1.-1
Condensation energy
Vo = tyT/e = INol
Vo = 1,1/0 1
*t =
T/T•. Exponents: n
= d(log S)/d(log T); e = (n
- I)/(n
Result for n
=
5 and e = 2/3
(2)
(3)
1= -1YT/(1
(4)
S = l'T.t S
+ !t 6 )
(5)
(6)
INI = -tyT/[t 2 -
~(t·
+ 2)]
(7)
+
I).
Phase Transition Functions
The theory of Landau phase transitions includes quantum mechanical constraints which have been left out of the previous two-fluid discussion. The Landau
free energy density is expressed in terms of the wave function 1/1 for He II with
Y = 1I/I1 2m/p. The thermodynamic Ginzburg-Pitaevskii-Mamaladze version [3] of
the theory considers the first two terms of a series in ascending powers of 11/112,
(T.. - T) « 1';.,
(8)
Then we have at thermodynamic equilibrium [o(af)/oll/l12 ~ 0], an equilibrium number density
11/1 01 2 = a./P
(9)
and an equilibrium free energy difference
lafl
= a. 2 /2P = lI/IoI 2a./2
(10)
Note that on the right-hand side of (10), N appears as the product of the particle
energy of condensation a. and the number density 11/1012. Further, quantum mechanical
uncertainty is reflected in the coherence distance [3] (the de Broglie wavelength for a.)
e
When afand 11/1012 (and
and (10) by
e = h/(2ma.)1/2
(11)
a(T) = 2 N/Il/loI 2
(12)
= 2 N/Il/loI4
(13)
e, respectively) are known, a. and P can be evaluated from (9)
and
P(T)
Superfluid Thermodynamic Transport Limits for Liquid Helium II
367
Once IX and f3 are known, other thermodynamic functions can be deduced from the
phase transition theory. Of particular concern is the upper thermodynamic stability
limit for transport [6.7]. For instance, the superfluid critical velocity is given by
v.c =
(~)1/2(IX/m)1/2 =
r
(14)
1/21i/m~
while the limiting mass flux is obtained from
(15)
jc = ~mll/loI2v.c = (~)3/211/1012m(lX/m)1/2
Mamaladze has evaluated V.c ' ~, and other quantities in his version of the phase
transition theory for the vicinity of the A-point. For completeness, his results are
listed in Table II.
For purposes of the present investigation, the Ginzburg-Landau parameters
need to be determined ftom the two-fluid power laws (Table I, n = 5). From these
parameters, ratios such as 1X/1X0 = (p/P.)(L\f/L\fo) are obtained as a function oft = T/T;,.
for an extended temperature range. In particular, (6) and (7) yield
L\f/L\fo = 1 - tt 2 +
tt6
(16)
Equation (11) permits evaluation of the coherence distance ratio
~/~o
= (lXo/IX)1/2
(17)
and (14) yields the critical velocity ratio
v.c/v.co = (1X/1X0)1/2 = (L\f/1110)1/2/yl/2
(18)
The ratios (16) to (18) are shown in Fig. 1 vs. t. It is noted that the changes with
temperature are small as long as t « 1. However, in the A-point vicinity, the temperature effect is rather drastic. Once T -+ T;,., the two-fluid functions (17) and (18)
are superseded by the Mamaladze functions [3], which differ qualitatively from the
present set. Only the free energy density difference (16) is consistent with the
Mamaladze scaling at the A-point: As T -+ T;,., L\f '" (T;,. - T)2 in both cases [from
(16), L\f/L\fo -+ 6(1 - t)2 when t -+ 1; Fig. 1].
Figure 2 shows details of the free energy difference for several reduced temperatures t as a function of 11/11 at and near the equilibrium value 11/101. Ordering is complete
as the equilibrium order parameter reaches its largest value 11/1 <XlI at t = O.
A determination of absolute values of the thermodynamic quantities requires a
knowledge of the number density of particles participating in the condensation
process (which at most will involve p/m particles per unit volume of 4He). The
condensation energy 1110 as well as the order of magnitude of the coherence length
Table II. Mamaladze Scaling Functions*
Function
Superftuid density ratio
Coherence length
Parameteroc
Parameter P
Critical velocity
*(TA - T)« TA •
Reference quantity
= 2.73
= 1.11
10- 8 cm K2/3
10- 16 ergjK4 / 3
PR = 3.52 x 1O-3gergcm3/K2/3
vR = 5.76 X 103 cm/sec-K 2/3
~R
OCR
X
X
C. Linnet, R. C. Amar, Y. G. Wang, and T. H. K. Frederking
/
flf,
!lr.
/
1,5
/
/
/
../'
- .........
....
\
6f1 - t 2 ,
/
. . . ~ <, Y~ Ivtlc-o
\
\
,
....
....
....
,,
\
Fig.!. Thermodynamic ratios vs. t
=
Fig. 2. Dimensionless free energy density difference vs. dimensionless order parameter.
TIT;..
eo are fairly well known. Therefore, 11/1001
can be obtained from (10) and (11). It
turns out that the number density is smaller than p/m. Accordingly, the ordering
in He II can be characterized by a condensate fraction Peon! P < 1 with values predicted between 0.024 and 0.25 [8- 10]. Since exact values of the condensate fraction
are not known, numerical calculations have been carried out for 0.2 and 1.0.
Figure 3 provides a plot of the coherence length vs. T for the two values of the
condensate fraction. Though values of bulk He II-bulk solid systems are not easily
accessible in experiments, the temperature effect e(T} can be compared with results
of He II film investigations [11-14] and studies of vortex core parameters 5]. The
temperature dependence of the data appears to agree reasonably well with the present
power-law approximations. (For details of substrate effects on film properties, the
references may be consulted.)
2
e
e
COMPARISON OF TRANSPORT DATA WITH THE
THERMODYNAMIC LIMIT
The predicted thermodynamic limits are compared with transport results for
finite mass flow and zero net mass flow. The latter yields the largest cooling rates
when the thermohydrodynamics of the liquid He II is the governing factor (e.g.,
critical conditions of cylinder cooling in saturated He II at the peak flux).
1.4
r
1
Fig. 3.Coherence length vs. temperature.
369
Superftuid Thermodynamic Transport Limits for Liquid Helium II
P"OAh
Fig. 4. Data of resistive zero-net-massflow cooling (horizontal cylinders near
TJ.
10 6,.
•
.
em
<>
0
"'"
•
I
T K
2.11
2. IS
2.''-
....
1I
t
,
I
I
I
•
,
I·
•
. l .-
I
<>
,.~
I
..
<>I
Q
?
,J
~
,
<>
I
I
7
",o
Jp
c,
,"
d
..
..I
P
Q
2.13
2.12
2.10
6
,I
I'
I
/'
0 ,.0" .- ; /
OL-~__~~__~~D~~__~~____~____L - -
o
Figure 4 is an example of zero-net-mass-flow data for cooling of horizontal
cylinders (Pt-lO Rh: 1.78 x 10- 3 cm diameter; Ga: 0.12 cm diameter). The bulk
liquid is near saturation; cooling takes place at the peak flux. Details of these experiments have been described elsewhere 6 ,17]. The heat flux
e
q
=
(19)
STjeff
is the consequence of an applied thermo mechanical pressure difference
fl.PT = pS fl. T
(20)
Using the set of fl.P~j) data, the critical values (jerr = jc) are obtained by an extrapolation to vanishing driving pressures. Similar curves are known for other configurations of single heaters and for mass flow.
102
Peon
If
-
___
-
. ---
1
- --
---
~ ~' -
+
",
.......
""'- "\
.\
em -a
"
\
'0
,JO.em
0
"<>
Fig. 5. Critical fluxes of zero-net-mass-flow
cooling.
,.7
2,31
I"
'.'
'..
T.
K
,
T
C. Linnet, R. C. Amar, Y. G. Wang, and T. H. K. Frederking
370
Figure 5 shows critical zero-net-mass-flow data j.iT) for cooling of horizontal
cylinders over an extended temperature range [17]. Use has been made ofthe BhagatWiner equation
(21)
When T ~ Tl , Pn ~ p, resulting in the relation given by (19). Figure 5 indicates that
the thermodynamic limit is a meaningful bound only near the A-point. The smaller
the diameter, the better these and other data [16] approach the limit asymptotically.
There is a distinct size effect, as is known also for the resistive data at supercritical
flow. At low T, the data are one order or more below the limit.
Figure 6 plots the critical mass flux data for finite He II mass flow driven either
mechanically or thermomechanically through various geometries [19-22]. For this
kind of flow, the temperature dependence of the data agrees quite well qualitatively
with the thermodynamic jiT) prediction. Again, numerically the data approach the
stability limit asymptotically, as T ~ Tl . The agreement again is closer, the smaller
the diameter. For large duct geometries, there is again a size dependence. At
T« Tl , the small-diameter data tend to be about one order below the limit.
An exception is the orifice geometry [21], whose data fall surprisingly close to
the thermodynamic stability limit. It is noted that super leaks had been arranged
upstream and downstream of the orifice to prevent disturbances in the liquid. The
agreement between prediction and data is quite good in view of the uncertainties
usually resulting from surface effects during channel flow. The orifice results indicate
that the condensate fraction of He II is below 20%.
In summary, the experimental evidence suggests that the upper thermodynamic
stability limit for transport constitutes the maximum rate of superflow (at negligible
resistance); however, its utilization has to rely on very specific geometries.
,
.0
..
Ceo
o
.
stzr
a
5 _lD- f
+
1.2 - 10' "
"
, - 1D- "
0
em
3.1 · '0- ~
2 - '0
0
.I
,
i .4 - 10· 4
•
0 .
_I
0
c
9
!ilifF.
"
2.
"
"
'_2
'.6
T,
T. K
Fig. 6. Critical mass fluxes of finite mass transport.
Reference 19-slit width; reference 20--pore size;
reference 21-orifice diameter; reference 22nominal pore size.
SuperOuid Thermodynamic Transport Limits for Liquid Helium II
371
ACKNOWLEDGMENT
This work has been supported in part by the National Science Foundation under Grant GK-23307.
NOTATION
C
=
specific heat (volumetric)
f
= free energy density
h
= Planck's constant divided by 2n
(!!J.f difference between disordered and ordered 4He)
= effective mass flux
m = mass of 4He atom
n
= exponent
!!J.PT = thermomechanical pressure difference
q
= heat flux
S
= entropy (volumetric)
T
= temperature
T. = temperature at A-point
t
= TIT.
UC = condensation energy (U o = !!J.fo)
v
= velocity
y
= superfluid density ratio, pJp
jeff
Greek letters
= parameter in equation (8)
P = parameter in equation (8)
C(
y
= volumetric entropy coefficient
= exponent in equation (1)
= coherence length
= density
= wave function, "'00 at T = 0
11/1012 = equilibrium order parameter
Subscripts
c
= critical
con = condensate
n
= normal component
R
= reference quantity
s
= superftuid component
o =T-+O
REFERENCES
I. C. J. Gorter, in: Progress in Low Temperature Physics, Vol. I, North-Holland Press, Amsterdam,
Netherlands (1955), p. I.
2. L. D. Landau and E. M. Lifshitz, Statistical Physics, Pergamon Press, London (1958), p. 430.
3. V. L. Ginzburg and L. P. Pitaevskii, Soviet Phys.-JETP, 7:858 (1958); Yu. G. Mamaladze, Soviet
Phys.-JETP, 25:479 (1967).
4. G. Borelius, Cryogenics, 3:96 (1963).
5. C. Linnet, Ph.D. Dissertation, University of California, Los Angeles, California (1971); UCLA Engr.
Rept. 7109 (1971).
6. H. J. Mikeska, Phys. Rev., 179: 166 (1969).
7. W. F. Vinen, Statistical Physics, Phase Transitions, and Superfluidity, Vol. 2, (M. Chretien, E. P. Gross,
and S. Deser eds.), Gordon and Breach, New York (1966), p. 150.
8. R. D. Puff and J. S. Tenn, Phys. Rev., AI: 125 (1970).
9. H. A. Mook, R. Scherm, and M. K. Wilkinson, Phys. Rev., A6:2268 (1972).
10. J. F. Fernandez and H. A. Gersch, Phys. Rev., A7:239 (1973).
II. R. P. Henkel, E. N. Smith, and J. D. Reppy, Phys. Rev. Letters, 23: 1276 (1969).
12. E. S. Sabisky and C. H. Anderson, Phys. Rev. Letters, 30: 1122 (1973).
13. I. Rudnick and J. C. Fraser, J. Low Temp. Phys., 3:225 (1970).
Y71
C. LiDDet, R. C. Amar, Y. G. Wang, and T. R. K. FrederkiBg
14. M. Chester, L. C. Yang, and J. B. Stephens, Phys. Rev. Letters, 19:211 (1972); L. C. Yang, Ph.D.
Dissertation, University of California, Los Angeles, California (1973).
15. M. Steingartand W. I. Glaberson,J. Low Temp. Phys.,8:61 (1972); G. W. Rayfield and F. Reif, Phys.
Rev., I36A: 1194 (1964).
16. C. Linnet, T. R. K. Frederking, and R. C. Amar, in: Proceedings of 13th Intern. Conference on Low
Temperature Physics, Vol. 1, Plenum Press, New York (1974), p. 393; R. L. Raben, R. A. Madsen,
A. C. Leonard, and T. R. K. Frederking, in: Advances in Cryogenic Engineering, Vol. 17, Plenum Press,
New York (1972), p. 323.
17. R. L. Raben, M.S. Thesis, University of California, Los Angeles, California (1967).
18. D. W. B. Matthews and A. C. Leonard, in: Advances in Cryogenic Engineering. Vol. 19. Plenum Press.
New York (1974), p. 417.
19. W. E. Keller and E. F. Hammel, Physics, 2:221 (1966).
20. J. R. Clow and J. D. Reppy, Phys. Rev. Letters, 19:291 (1967); J. S. Langer and J. D. Reppy, in:
Progress in Low Temperature Physics, Vol. 16 (C. J. Gorter ed.) North-Holland Press, Amsterdam,
Netherlands (1970), p. 1.
21. G. B. Hess, Phys. Rev. Letters, 27:977 (1971).
22. A. Elsner, in: Advances in Cryogenic Engineering, Vol. 18, Plenum Press, New York (1973), p. 141;
T. H. K. Frederking, A. Elsner, and G. Klipping, in: Advances in Cryogenic Engineering, Vol. 18,
Plenum Press, New York (1973), p. 132.
23. S. M. Bhagat and B. M. Winer, Phys. Letters, 27A: 537 (1968).
DISCUSSION
Question by R. C. Hendricks, NASA Lewis Research Center: Please locate the V spacing of 1 x 10- 4
em on Fig. 6. What are the general trends of these points? Are they the same as the + and • points for
3 to 5 X 10- 4 em spacing, or are these significantly close to verify the theoretical prediction?
Answer by author: The triangles in Fig. 6 (with uncertainty range indicated by vertical lines) represent
critical mass fluxes extrapolated from resistive flow data [22]. These fluxes are one order of magnitude
lower than the orifice data (+ and • symbols) and fall in line with size-dependent and apparently roughness- and impurity-dependent channel flow results. In addition, the data appear to be affected by the pore
size distribution of the porous medium (for further details, Elsner [22] may be consulted). No uniform
flow conditions appear to exist, and the present thermodynamic limit is an upper bound.
Question by R. C. Hendricks, NASA Lewis Research Center: Do you feel the vortex formation is more
nearly approached by orifices rather than channels where secondary flows and vortices become more
prevalent?
Answer by author: The experimental evidence may be interpreted as follows: The orifice geometry
of Hess [21] favors localized vortex formation. In contrast, channels may generate vorticity at many locations
along the flow path. In line with this interpretation, the superleaks upstream and downstream [21] may
be considered as "vorticity filters." It is surprising to see the critical orifice data so close to the thermodynamic uniform flow prediction. This may indicate that uniform flow is approached closely at the orifice
(as it is in classical flow at high Reynolds numbers).
Questions of the details of vortex kinematics have been raised also in other discussions and in the
context of vortex models. They usually require adjustable parameters because of complicated conditions.
For the orifice geometry, however, recent studies ofGamota [Phys. Rev. Lett., 31 :517 (1973)] show rather
simple critical phenomena The experimental evidence indicates that a vortex ring is created of quantum
number unity with a diameter equal to the orifice diameter. This result is consistent with the form of the
thermodynamic limit for uniform flow. (Further details of the calculations are contained in UCLA Rept.
7433, 1974.)
}-1
A SIMPLE METHOD FOR CHARPY IMPACT
TESTING BELOW 6K
s. Jin, w. A. Horwood, J. W. Morris, Jr., and V. F. Zackay
University of California
Berkeley, California
INTRODUCTION
Impact testing at temperatures down to 77 K is routinely accomplished according
to ASTM specifications. The specimen is transferred in open air from a liquid bath
to the test machine anvil and the pendulum is released within 5 sec. However, this
method is impractical at lower temperatures since the low heat capacity of the metal
leads to a rapid increase in specimen temperature during transfer.
To obtain valid low-temperature data, DeSisto [1] designed an apparatus consisting of an evacuated enclosure which houses both an impact machine and an automatic feed mechanism which rapidly inserts specimens cooled in liquid helium into
the test machine anvil. He was able to obtain data at instantaneous sample temperatures near 8 K. In other work, Kiefer et al. [2] employed a glove-box arrangement to
manually transfer specimens from an open-mouth dewar of liquid hydrogen to the
test machine under inert atmosphere. They report test temperatures near 25 K when
tested within 2 sec.
While these techniques have permitted impact testing at low temperature, the
elaborate apparatus required makes them impractical for use in many laboratories.
Therefore, a new and simple method was sought which could be employed in current
research on the design of tough alloys for cryogenic use. Since this research concentrates on alloys of body-centered cubic structure whose mechanical properties are
typically sharp functions of temperature in the cryogenic region, it was particularly
desirable that the test technique allow not only close temperature control but also
ensure the absence of appreciable thermal gradients in the specimen at the moment
of impact.
The approach selected was to encase the test samples within individual insulating
boxes, which permits maintaining a controlled sample temperature near liquid
helium temperature during the impact test. The requirements for a suitable boxing
arrangement are that it must be an efficient insulator which withstands cooling to
liquid helium temperature but contributes negligibly to the measured Charpy
energies. These requirements seem well satisfied if one uses a thin-walled Lucite box,
insulated with open-cell styrofoam which is continuously bathed with liquid helium.
The test procedure and its calibration are discussed in the following sections.
EXPERIMENTAL METHOD
A Charpy V-notch specimen is wrapped with open-cell styrofoam which is
grooved along the specimen length direction to allow easy passage of liquid helium
373
374
S. Jin, W. A. Horwood, J. W. Morris, Jr., and V. F. Zackay
2_,7 1!J~""ml
- ---
Luc."
Fig. 1. Grooves made on each piece of the
styrofoam layer. (Grooved side is to face
the metal specimen.)
(~~-~}
- --
.,,,--r;==:======:;;==:;,
Styrofoam
ProlectlOfl rube
Outllll!f DIp!
GIo..e
(lucile )
Fig. 2. Specimen packing inside a Lucite box (top view).
(Fig. 1). The styrofoam surface is then wrapped with Scotch tape. The insulated
specimen is then inserted into a rectangular Lucite box (with one or two faces open)
of in. (0.8 mm) wall thickness having two holes of 156 in. (7.9 mm) diameter (Fig. 2).
The cover for the open face is glued to the box system with ethylene dichloride
solution. The Scotch tape and the styrofoam layer underneath the holes are removed.
A polyethylene inlet pipe of iin. (9.5 mm) ID and! in. (12.7 mm) in length is glued
to the box by Duco cement. A Lucite outlet pipe of i in. (9.5 mm) ID and 1 in. (25.4 mm)
in length is attached with ethylene dichloride solution.
This box system is placed on the anvil of a Charpy test machine. Liquid helium
is transferred to the system through the inlet pipe from a pressurized dewar using
an ordinary vacuum-insulated transfer line. The pressure inside the dewar is maintained close to 2 to 3 psig (1.4-2.1 x 10 4 N/m2) by an external supply of helium gas.
Figure 3 is a schematic view of the test arrangement. Figure 4 shows the liquid helium
being transferred to the box system located on the test machine anvil. A specimen
temperature of 5 to 6 K is quickly attained and maintained as long as the liquid
helium is transferred. Given proper assembly, the Lucite box does not crack during
liquid helium transfer. The liquid helium flows easily through the system and can
be observed through the box wall. The released pendulum breaks the entire assembly.
tr
Fig. 3. Schematic illustration of the test arrangement (front view).
A Simple Method for Charpy Impact Testing Below 6 K
375
Fig. 4. Transfer of liquid helium to the box system.
(Liquid helium is about to emerge from the vent)
When the system is struck, the weak bond at the base of the polyethylene inlet pipe
fractures, thus ensuring the safety of the helium transfer line.
The open-cell styrofoam layer has two functions. It absorbs the difference in
thermal contraction between the Lucite box and the specimen on cooling to liquid
helium temperature, thereby preventing cracking of the box and the escape of liquid
helium. It also allows the interstitial flow of liquid helium around the specimen for
maximum cooling efficiency. After steady state is reached, a layer of liquid helium
surrounds the metal specimen. Moreover, the insulated Lucite box is extremely
brittle at liquid helium temperature and absorbs very little energy during testing.
Because of the restricted space in the anvil of the test machine (which, like most
older machines, does not offer the specimen clearance recommended in ASTM
E-23-72), the specimen length was reduced to 2.01 in. (51 mm) from the ASTM
standard 2.17 in. (55 mm) length. However, the shortened length does not seem to
introduce appreciable error. If the anvil were modified to allow larger specimens, a
full-size specimen could be used.
TEMPERATURE CALIBRATION
The specimen temperature was measured as a function of time by means of the
following arrangement. A hole of 0.1 in. (2.5 mm) diameter by 0.2 in. (5.1 mm) in
depth was drilled in the calibration specimen, and a section of brass tubing of 0.125 in.
(3.2 mm) diameter by 0.5 in. (12.7 mm) in length was silver-soldered over the hole
to act as a collar to keep liquid helium away from the monitoring thermocouple (as
shown in Fig. 2). For specimens that were actually impacted, plastic tubings of the
same size were glued to the specimen. The insulating spacer and the box were then
assembled around the specimen. A thermocouple (Au + 0.07 %Fe, Chromel) was
inserted into the drilled hole so that its tip was in good contact with the specimen.
S. Jill, W. A. Horwood, J. W. Morris, Jr., ..... V. F. Zackay
376
+1I88+--_......:!C4.2"--"-K-
- - - - -
- - - - - - - - -
!!.~ ~ -
-,.----r--~-6
o
K
--~--------------77 K
(Ref.renc. Temperature)
-300 K
(Room Temperature)
-46000
20
40 60 80 100
k-O-;2b:O---;!'H~!:'-;'6<-:~~~-;:!;;~
Time (seconds)
Fig. 5. Cooling curve for specimen.
The thermocouple was then glued to the brass tube and the brass tube sealed to the
Lucite box to prevent escape of liquid helium. This configuration ensures that the
thermocouple measures the temperature at the centerline of the specimen rather
than that of the bath of liquid helium in which the specimen is immersed.
A typical temperature-time curve obtained using a dewar pressure of 2 to 3 psi
during transfer is shown in Fig. 5. Liquid helium transfer is initiated at t = O. Between
t = 0 and t l ' liquid helium is consumed in cooling the transfer line and does not
significantly affect the specimen temperature. At t l ' liquid helium begins to cool
the box system. The specimen temperature drops to below 6 K in about 60 to 80 sec.
It then remains constant for as long as liquid helium is transferred. At t 2' the noise
associated with the rapid evaporation ofliquid helium abates and liquid helium begins
to emerge from the outlet pipe. This noticeable change in noise level may be used
as a sign that the test temperature has been reached. The pendulum may then be
released. The specimen remains at 5 to 6 K until impact.
After testing, the helium transfer is halted and the next specimen set up for
impacting. The transfer line cooldown consumes very little helium after the first test.
Liquid helium consumption for the first test is approximately five to six liters. In
continuous operation, each subsequent test consumes two to three liters of liquid
helium.
Specimen temperatures at points A and B in Fig. 2 (each point is approximately
one-fourth of the specimen length apart from the notch) were measured. A negligible
temperature gradient was noted in the specimen when the liquid helium transfer
is maintained by a dewar pressure of 2 to 3 psig. However, below a dewar pressure
of 1 psig. a temperature gradient of several degrees was noted between points A and
B in the specimen. Therefore, it is necessary to keep the dewar pressure above a
certain level to maintain a uniform specimen temperature.
IMPACT TEST CALmRATION
There are two principal sources of error in the proposed impact test technique:
a positive error is introduced by the Lucite box and styrofoam layer, and a negative
error is due to the shortened specimen length. Both errors appear to be small.
A Simple Method for Cbarpy Impact Testiq Below 6 K
m
Table I. Comparison of Charpy Impact Testing
Test temperature
Method of cooling
Specimen length
Number of specimens tested
Average impact energy
Standard deviation
ASTM standard method
Proposed method
77K
Open-air transfer
2.17 in. (55 mm)
17
124.5 ft-lb
(168.8 N-m)
8.7 ft-lb
(11.5 N-m)
77K
Box system
2.01 in. (51 mm)
17
120.9 ft-lb
(163.9N-m)
13.8 ft-lb
(18.7N-m)
Plastics such as Lucite are usually very brittle at low temperatures. Several
impact tests were made at liquid helium temperature with the packing system minus
the metal specimen. The impact energy absorbed by the Lucite box and the styrofoam
spacer was less than 1 ft-lb (1.36 N-m). When a box system with a specimen is impacted,
the face of the Lucite box behind the specimen is crushed. This contributes an additional small, positive error.
Since the anvil of the test machine was not modified, the specimen length had
to be shortened from 2.17 in. to 2.01 in. For very tough alloys, this shortened length
could introduce a negative error in the measured impact energy because the maximum
possible angle of bend is decreased.
For purposes of calibration, seventeen specimens of commercial 304 stainless
steel of 2.17 in. length were tested at liquid nitrogen temperature according to the
ASTM specification (open-mouth transfer). Another seventeen specimens of the
same material of 2.01 in. length were tested inside the Lucite box system by transferring liquid nitrogen to the box. The results are compared in Table I. The difference
in the average impact energy obtained by these two methods was 3.6 ft-lb (4.9 N-m)
or approximately 3% of the average impact energy. This difference is much less than
the average scattering of the test data [The standard deviation of data obtained using
the ASTM standard method was 8.7ft-lb (11.5 N-m).] Less extensive comparison
tests with other cryogenic steels also showed good agreement between the standard
and the proposed methods.
The small deviation in average impact energy between the standard and the
proposed method seems essentially to be due to the difference in specimen length.
This error may decrease if the anvil of the impact machine is adjusted to accommodate
a 2.17-in. boxed specimen. The increased scatter of data obtained with the proposed
method was anticipated because the boxing arrangement introduces additional
variables and it is difficult to ensure that the notch in the boxed specimen is well
centered in the anvil. Given that the Charpy impact test is at best a semiquantitative
measure of the toughness of a material, the mechanical accuracy of the proposed
test method seems reasonable.
DISCUSSION
The suggested test method has been used routinely in research on cryogenic
alloys at Berkeley. Some experimental results are reported elsewhere [3]. It offers
two principal advantages. First, the tests are inexpensive and may be performed
on any available impact machine whose specimen holder will accommodate the
boxed specimen. Second, the method allows maintenance of a steady, uniform sample
S. Jill, W. A. Horwood, J. W. Morris, Jr., .... V. F. ZKluty
temperature of 5 to 6 K and avoids thermal gradients induced by sample transfer
to the anvil.
The method should be useful for estimating the shelf energy and transition
temperature of materials of reasonable impact toughness. It will be of less value in
testing very brittle materials, since the impact fracture of the boxed system may
then introduce significant error.
ACKNOWLmGMENTS
The authors are grateful to G. A. Gachis and V. D. Santos ofthe Cryogenic Dewar Coordination Board
of the Lawrence Berkeley Laboratory for advice and discussion. This work was supported by the Atomic
Energy Commission through the Inorganic Materials Research Division of the Lawrence Berkeley Laboratory, and by the Office of Naval Research under Contract NOOO14-69-A-0200-1062, NR 031-762.
REFERENCES
I. T. S. DeSisto, "Automatic Impact Testing to 8 K," Tech. Rept. 112/93, Watertown Arsenal Lab.
(July 1958).
2. T. F. Kiefer, R. D. Keys, and F. R. Schwartzberg, in: Advances in Cryogenic Engineering, Vol. 10,
Plenum Press, New York (l965), p. 56.
3. S.Jin,J. W. Morris, Jr., and V. F. Zackay, in: Advances in Cryogenic Engineering, Vol. 19, Plenum Press,
New York (1974), p. 379.
DISCUSSION
Comment by H. M. Long, Oak Ridge National Laboratory: Your method is very interesting and you
should be commended for publishing it. We at the Linde Division, Union Carbide Corporation, had the
same problem in 1958 while developing a liquid helium transport trailer for the U. S. Air Force. We
obtained Charpy impact tests at 5 to 6 K by encasing the specimen in a paper boat which was immersed
in liquid hydrogen first and then in liquid helium to freeze the hydrogen around the specimen; finally,
the frozen hydrogen-encased specimen was laid on the impact anvil and the hammer released. We instrumented several specimens with thermocouples and timed the temperature rise. We then accepted only
those tests that occurred within the time to the desired temperature. Our scatter was acceptable.
Answer by author: We were aware of the liquid hydrogen method, though I am not sure of the source
of our information. We certainly agree that it is an imaginative and useful technique. However, the method
requires hydrogen facilities, and, as you point out, involves temperature control difficulties which are
avoided with the more elaborate boxing procedure.
J-2
AN IRON-NICKEL-TITANIUM ALLOY WITH
OUTSTANDING TOUGHNESS AT
CRYOGENIC TEMPERATURE
S. Jin, J. W. Morris, Jr., and V. F. Zackay
University of California
Berkeley, California
INTRODUCTION
Recent research in this laboratory has led to the identification of a group of
ferritic alloys from the iron-nickel-titanium system which show an unusual combination of strength and ductility at cryogenic temperatures. The details of this
research are reported elsewhere [1]. This presentation provides some of the best
results obtained to date with an Fe-12 Ni-O.25 Ti alloy and compares these to the
best available properties of two commonly used cryogenic steels: type 304 stainless
steel and Fe--9 Ni-1 Mn-O.1 C (9% nickel steel).
ALLOY COMPOSITION AND PROCESSING
Low-carbon alloys of nominal composition Fe--12 Ni-O.25 Ti were obtained by
vacuum melting from the prime components. The actual compositions of two ingots
are shown in Table I. Two ingots of 21 in. (7.0 cm) diameter were homogenized at
1050°C for 80 hr, forged to plates of 1 in. (1.9 cm) thickness at 1100°C and air-cooled
to room temperature. The alloys were heat-treated in the ')I-phase region (above
about 7()()oq and the two-phase (oc + ')I) region (below about 7()()oq in sequence
with cold-working or isothermal decomposition at 550°C.
The four processing sequences which will be specifically mentioned below are:
(1) 900°C (2 hr), air quench + 700°C (2hr), air quench + 550°C (2 hr), water quench;
(2) 700°C (2 hr), air quench + 30% cold work + 700°C (2 hr), air quench + 550°C
(2 hr), water quench; (3) 700°C (2 hr), air quench (repeated four times) + 550°C (2 hr)
water quench; (4) 900°C (2 hr), air quench + [7()()OC (2 hr), air quench + 660°C (2 hr),
air quench] (repeated four to six times) + 550°C (2 hr), water quench. Treatment (1)
imparts outstanding impact toughness at liquid nitrogen temperature. Treatment (2)
yields an alloy having outstanding impact toughness to temperatures below 6 K.
Table I. Chemical Analysis of Two Ingots of Nominal Composition
Fe--12 Ni-O.25 Ti
Ingot
Ni
Ti
I
II
12.14
10.90
0.24
0.23
eNS
0.01
0.008
0.003
0.004
<0.005
<0.005
379
P
Mn
0
Fe
<0.005
<0.005
<0.005
<0.005
0.014
0.021
balance
balance
380
S. JiB, J. W. Morris, Jr., and V. F. Zackay
Treatment (3) also imparts very good impact toughness at 6 K, while avoiding the
cold-working step. Treatment (4) yields an anoy of extremely fine grain size ( < 1 Jlm)
which shows exceptional ductility in fracture toughness tests at liquid nitrogen
temperature.
In succeeding sections, the mechanical properties of this alloy will be compared
to those of type 304 stainless steel and 9% nickel steel. The comparison alloys were
procured commercially and treated according to recommended procedures [2] to
optimize toughness at cryogenic temperature. The type 304 stainless steel was austenitized at 1020°C for 1 hr, then ice-brine-quenched. The 9% nickel steel was processed
in the sequence: 900°C (2 hr), air quench + 790°C (2 hr), air quench + 550°C (2 hr),
water quench.
CRYOGENIC TENSILE PROPERTIES
The tensile properties of the Fe-12 Ni-O.25 Ti alloys [treatments (1), (2), (3), and
(4)] were tested at both 77 and 6 K. The tests were conducted in an Instron machine
equipped with either a liquid nitrogen cryostat (77 K tests) or a liquid helium cryostat
(6 K tests). Specimens of 0.5 in. (1.27 cm) gage length and 0.125 in. (0.31 cm) gage
diameter were tested (strain rate:;: O.04jmin). The tensile properties of the two
comparison alloys, 9% nickel steel and 304 stainless steel, were measured under
identical conditions.
The results of the cryogenic tensile tests at 77 and 6 K are given in Table II.
At liquid nitrogen temperature, the Fe-12 Ni-O.25 Ti alloys, treatments (1), (2), (3),
Table II. Results of Cryogenic Tensile Tests
Alloy
77K
304
9 % nickel steel
Fe-12 Ni-{).25 Ti
Treatment
(1)
(2)
(3)
(4)
6K
304
9 % nickel steel
Fe--12 Ni-{).25 Ti
Treatment
(1)
(2)
(3)
(4)
Yield strength,·
ksi (N/m2)
Tensile strength, t
ksi (N/m2)
Elongation,
R.A.,
81 (5.58 x 108 )
146 (10.07 x 108 )
235 (16.21 x 108 )
172 (11.86 x 108 )
47.3
29.6
65.6
66.8
142 (9.80
141 (9.74
138 (9.53
141 (9.74
153 (10.55
155 (10.69
151 (10.42
163 (11.24
108 )
108 )
108 )
108 )
31.0
32.3
33.8
33.1
71.2
75.8
78.5
76.2
105 (7.24 x 10 8 )
208 (14.34 x 108 )
270 (18.63 x 108 )
231 (15.94 x 108 )
41.6
21.2
.52.7
59.1
184 (12.69
183 (12.63
180 (12.42
181 (12.48
197 (13.58 x 108 )
199 (13.73 x 108 )
196(13.51 X 108 )
205 (14.14 x 108 )
21.7
22.3
24.1
22$
68.0
70.2
73.6
70.4
x
x
x
x
108 )
108 )
108 )
108 )
x
x
x
x
108 )
108 )
108 )
108 )
x
x
x
x
%
%
Ingot
used
I
I
I
II
I
I
I
II
• For the 77 K tests, 0.2 % offset yield strength was taken. However, at 6 K, a serrated yielding behavior
was observed in all the specimens tested. This is a rather common phenomenon in the tensile tests of ironbase alloys near liquid helium temperature. The yield strength in this case was taken from the first discontinuous yielding point.
t The tensile strength at 6 K was obtained from the upper locus of discontinuous peaks in the stress-strain
curve.
An Iron-Nickel-Titanium Alloy with Outstanding Toughness at Cryogenic Temperature
gil. Ni STEEL
Fe-12Ni-0.25 Ti ALLOY
381
304 STAlPl.ESS STE EL
OUENCH Aft) TEMPERED THERMAl.. CYCLING TREATNfNT
ANNEALED
75 FT-La
205 FT-LB
130 FT-LB
y. S.:208,OOOPSI
Y. S.: 185,000 PSI
Y.S.: 105,000 PSI
Fig. I. Charpy specimens tested at 6 K. The Fe-12 Ni-O.25 Ti alloy was processed according
to treatment (3).
and (4) showed virtually identical tensile properties. The yield strength of this alloy
is about the same as that of the 9% nickel steel but the ductility is better, as can be
seen in Table II. At liquid helium temperature, however, the yield strength of Fe12 Ni--O.25 Ti alloys is slightly below that of the 9% nickel steel. This is probably
due to the smaller amount of carbon in Fe-12 Ni--O.25 Ti alloys (Table I) than in the
9% nickel steel, which contains approximately 0.1 % carbon. The Fe-12 Ni--O.25 Ti
alloys are distinguished by their excellent ductility (66 to 77% reduction in area at 6 K).
CRYOGENIC IMPACT TOUGHNESS
Charpy impact tests were conducted at liquid nitrogen temperature using ASTM
standard techniques and at 6 K using the "boxing" technique described elsewhere [3].
The Fe-12 Ni--O.25 Ti alloy was tested in each of the four processing conditions
described above. The 304 and 9% nickel comparison alloys were also tested.
The results of the Charpy impact tests are shown in Table III. At 77 K, the Fe12 Ni--O.25 Ti alloy showed outstanding impact toughness in each of the four treatments. The Charpy energy ofthe alloy processed in sequence (2) exceeded the capacity
Table III. Charpy V-Notch Impact Tests at Cryogenic
Temperatures
Energy absorbed, ft-Ib (or N-m)
Alloy
304
9% nickel steel
Fe-12 Ni--{).25 Ti Treatment
(I)
(2)
(3)
(4)
17K
6K
128 (174)
92 (125)
130 (176)
75 (102)
204 (276)
>225* (305)
205 (278)
140 (190)
26 (35)
>225* (305)
203 (275)
134 (182)
Ingot
used
I
I
I
II
* The capacity of the impact test machine was exceeded and the pendulum was
stopped by the specimens.
382
S. JIB, J. W. Morris, Jr., and V. F. Zackay
Fe - 12 Ni - 0.25 Ti
30 %
COLD WORKED
Fig. 2. An Fe-12 Ni-D.25 Ti Charpy bar processed
according to treatment (2) and tested at 6 K. The
impact energy exceeded the capacity of the test
machine (225 ft-Ib).
ofthe test machine. When the test temperature was lowered to 6 K, the alloy processed
through sequence (1) became relatively brittle, having a Charpy energy below those
of the 304 and 9% nickel steels. Alloys processed through sequences (2), (3), and (4),
however, retained high impact toughness. The Charpy energy obtained through
treatment (2) again exceeded the capacity of the machine.
The results of these tests are illustrated in Figs. 1 and 2. In Fig. 1, the impact
fracture surface of the Fe-12 Ni-O.25 Ti alloy, processed through sequence (3), is
compared to the fracture surfaces of the 304 and 9% nickel steels. The higher impact
toughness of the Fe-12 Ni-O.25 Ti alloy is apparent in the severe deformation of its
fracture surface. Figure 2 is a photograph of a sample of the Fe-12 Ni-O.25 Ti alloy
processed through sequence (2) and tested at 6 K .The impact fracture propagated
through roughly 90% of the thickness of the bar before the hammer was stopped.
FRACTURE TOUGHNESS
While only preliminary fracture toughness testing has been completed to date,
the results confirm the unusual ductility of the processed Fe-12 Ni-O.25 Ti alloy.
Fracture toughness tests were conducted at 77 K on an MTS machine equipped with a
liquid nitrofen cryostat. The specimens were ASTM standard "compact tension"
specimens [ ] of 0.70 in. (1.78 cm) thickness. Alloys tested included Fe-12 Ni-O.25 Ti,
treatments (1) and (4) (specimens were prepared from ingot II), and 9% nickel steel.
The results of these tests are shown in Fig. 3, which includes the load~isplace­
ment curves obtained. The shape ofthe curves shows that none ofthe alloys tested was
actually in plane strain loading condition at "pop-in" (unstable crack propagation as
indicated by the sudden drop in the load-displacement curve). However, the 9%
Fe - 12 Ni -
025 Ti
Fig. 3. Load-<iisplacement curves
obtained in fracture toughness tests
comparing the 9 % nickel steel and
two treatments of the Fe- 12 Ni0.25 Ti alloy at 77 K.
An Iron-Nickel-Titanium Alloy with Outstanding Toughness at Cryogenic Temperature
383
Fig. 4. Post-test photographs offracture toughness specimens: 9% nickel steel (left), Fe-12 Ni
-D.25 Ti, treatment (l)(middle), and Fe-12 Ni-D.25 Ti, treatment (4)(right). The fracture in the
last specimen propagated very slowly and the test was halted before the final fracture. The
specimen has been cut open to reveal the fracture surface.
9 - Ni Steel
1---1
5o,LL
Fe -12 Ni - 0.25 Ti
Fe -12 Ni - 0. 2 5 Ti
Treatment 1
Treatment 4
1---1
50f'-
Fig. 5. Scanning electron fractographs of the samples shown in Fig. 4: Fractographs of 9 %nickel steel and
Fe-12 Ni-D.25 Ti, treatment(l), show the faceting characteristic of quasicleavage. Fractograph of Fe-12 Ni
--{).25 Ti, treatment (4), shows the "dimples" characteristic of ductile fracture.
384
S. Jill, J. W. Morris, Jr., aDd V. F. Zackay
nickel steel was in nearly plane strain loading (its load-displacement curve approximates to a straight line) and exhibits a fracture toughness value KQ in the range 130 to
150 ksi
[144--166 x 106 (N/m 2)fo]. The Fe-12 Ni-O.25 Ti alloy, processed
through sequence (4), was virtually immune to unstable crack growth in this test. The
induced crack in this sample grew in a stable manner with marked concomitant
plastic deformation until the test was finally terminated.
The three fracture toughness specimens are compared in Fig. 4. The enhanced
ductility ofthe Fe-12 Ni-O.25 Ti alloys is visually apparent. Scanning electron fractographs of the fracture surfaces of the three samples are presented in Fig. 5. These
fractographs were taken along the center line of the samples slightly ahead of the
front of th~ preinduced fatigue crack. They show that crack propagation through the
center of the 9% nickel and the Fe-12 Ni-O.25 Ti, process (1), alloys occurred in a
semibrittle manner through quasicleavage. Since crack propagation in the Fe-12 Ni0.25 Ti alloys were well away from plane strain conditions during these tests, no fracture
toughness measure was taken. Tests using thicker samples are currently in progress.
.JiD.
DISCUSSION
The results presented in the previous sections indicate that the Fe-12 Ni-O.25 Ti
alloy can be processed to have an excellent combination of strength and toughness at
very low temperatures. The impact toughness values in particular, as far as is known,
are well above any which have ever been obtained at cryogenic temperatures.
It is not yet clear whether the properties reported here are peculiar to Fe-Ni-Ti
alloys of high-nickel, low-impurity content with titanium's scavening effect on interstitials, or whether they may be reproduced in less highly alloyed systems through
proper processing. For example, reports of recent Japanese work on the 9% nickel
alloy [5] indicate that Charpy impact values as high as approximately 160 ft-lb (217
N-m) can be obtained at liquid nitrogen temperature through careful alloy preparation.
Nor is it clear whether the rather complex processing sequences that has been used to
obtain these properties can be simplified to processes of more commercial appeal.
These questions are now being investigated.
ACKNOWLEDGMENTS
The work reported here was supported by the U. S. Atomic Energy Commission through the Inorganic
Materials Research Division of the Lawrence Berkeley Laboratory and by the Office of Naval Research
under Contract NOOOl4-69-A-0200-1062, NR 031-762.
REFERENCES
I. S. Jin, J. W. Morris, Jr., and V. F. Zackay, to be published.
2. ASTM Standards Designation A353-70A.
3. S. Jin, W. A. Horwood, J. W. Morris, Jr., and V. F. Zackay, in: Advances in Cryogenic Engineering,
Vol. 19, Plenum Press, New York (1974), p. 373.
4. Annual Book of ASTM Stamklrds, Vol. 31, E399-70T, ASTM, New York (1970).
5. T. Ooka, H. Mimura, S. Yano, K. Sugino, and T. Toizumi, J. Japan Inst. Metals, 30:442 (1966).
J-3
COMPRESSIVE LOAD-DEFLECTION
CHARACTERISTICS OF SEVERAL FOAM
MATERIALS AT ROOM TEMPERATURE,
77 K AND 4.2 K*
w. F. Stewart, D. T. Eash, and W. A. May
Los Alamos Scientific Laboratory, University of California
Los Alamos, New Mexico
INTRODUCTION
A 300-kJ superconducting inductive energy storage experiment at the Los
Alamos Scientific Laboratory involves supporting a superconducting coil within a
fiberglass cryostat with a secondary coil on the outside of the cryostat as shown in
Fig. 1. The energy storage experiment is a model for a pulsed plasma thermonuclear
fusion energy source. The experiment involves charging the primary superconducting
coil with a current of 10,000 A, driving a switch section of the circuit from its superconducting state to its normal resistive state, producing a large voltage across the
switch. The transient current decay in the primary loop induces a large current into
the secondary coil and a resistive load connected in series.
At the levels of current and voltage involved, large transient forces can be produced by misalignment of the primary superconducting coil and the secondary coil.
A rigid support for the primary coil is required to prevent vibration or motion of
the primary coil when subjected to these transient forces. The primary coil and
support hardware have a weight of approximately 272 kg and the force transient
Cryostat Inne,
Shell (fibe,v1ols)
Cryostat Out.,
Sholl (fibergla .. l
SupportinQ Arm
Load Coli
Flexible Foam Pod
Primary
Superconducf ing Coil
Secondory Coil
Resistive Load
Fig. I. Layout of the LASL 300-kJ superconducting
inductive energy storage experiment.
• Work performed under the auspices of the U. S. Atomic Energy Commission.
38S
FiberQloss- Polyurethane
Support System
w. F. Stewart, D. T. EasII, aad W. A. May
has a rise time of about 0.001 sec. Magnetic misalignment is expected to result in
transient forces of 340 kg/cm for vertical misalignment, 170 kg/cm for concentricity
misalignment, and 98 m kg/deg torque for angular misalignment.
The superconducting coil was supported from the cryostat lid since the fiberglass
cryostat does not have any mounting features on the inner shell. The supporting
post from the cryostat lid to the coil cannot be made strong enough to support the
lateral loads without excessive heat leak. The post has sufficient strength to support
the weight of the coil and the tensile load caused by vertical misalignment of the
primary and secondary coils. In fact, the primary and secondary coils are deliberately
misaligned vertically to ensure that the post loading is always tensile and never
compressive. The lateral loads and moments caused by concentricity and angular
misalignments are supported both by a fiberglass-polyurethane support section at
the bottom of the coil and by four adjustable supporting arms at the top of the coil.
A foam pad (flexible at room temperature but rigid at 4.2 K) was considered to be a
better means of adapting the coil-supporting arms to the inner shell of the cryostat
than springs or other mechanical or pneumatic (gaseous helium) damping devices.
These were determined to be unsatisfactory for a variety of reasons (difficulty of
adapting to irregularities of the cryostat inner shell; difficulty of adjusting at room
temperature and maintaining proper adjustment during cooldown, operation, and
warmup; difficulty of operating in liquid helium and in high electrical and magnetic
fields). Highly localized loading of the cryostat inner shell is avoided by using a
flexible material pad on the ends of the adjustable supporting arms since the flexible
material conforms to the irregularities of the inner shell when pressed against it at
room temperature. An initial compressive load of 0.7 kg/cm 2 is used. When the
temperature is lowered, the flexible material hardens and a rigid (not springy) area
loading of the cryostat inner shell is achieved. With four upper lateral supporting
arms with 323 em 2 foam each and under maximum loading conditions, the compressive load in the flexible foam is not expected to exceed 3.5 kg/cm 2 • The flexible
foam is not affected by electrical or magnetic fields. Several flexible material candidates
for this application were tested to determine which would be best suited for this
purpose since no published data suitable for evaluating these materials were found.
Arvidson et al. P] reported some limited compressive loading data on a flexible
polyurethane foam (0.03 g/cm 3 ) at the 1972 Cryogenic Engineering Conference.
MATERIALS AND SPECIMENS
The materials tested included various densities of a gas-blown flexible polysiloxane foam (Silastic). a polyether-based flexible polyurethane foam. a proprietary
flexible cellular silicone [2], and a low-density rigid polystyrene foam, as detailed
in Table I.
The specimens were all of the same configuration: flat, rectangular, 5.1 cm wide,
15.2 em long, and approximately 0.5 em thick. The specimens were cut from pieces
that were molded 30.5 em square by approximately 0.5 em thick. The test specimens
were cut from the interior of the square and away from the edges of the molded pieces
and were loaded in the same orientation as they were formed.
TEST PROCEDURE
All tests were conducted with a 4536-kg-capacity Instron testing machine with
a crosshead speed of 0.76 em/hr and under essentially static conditions. Material
response is not expected to be altered significantly by the dynamic conditions of
actual operation. The test arrangement is shown in Fig. 2.
387
Compressive Load-Deflection Characteristics
Table I. Description of Materials Tested
Density,
g/cm J
Material
General description
Polysiloxane foam
(Silastic)
0.42
0.45
0.47
0.51
0.55
Blend of Dow Corning Corporation product.<>; hydrogen-blown,
foam-in-place polysiloxane system; 72.1 wt. % S-5370 resin,
24.0 wt. % RTV-3110 resin, 3.9 wt. % S-5370 catalyst; roomtemperature cure in mold and post cure of 3 hr at 120°C
Polyurethane foam
(CPR-X-2-141-A)
0.32
0.44
Polyether-based, CO 2-blown, foam-in-place polyurethane system;
CPR-X-2-141-A available from CPR Division of Upjohn Company ; 41.4 wt. %resin, 57.6 wt. %isocyanate, 1.0 wt. %catalyst;
cured I hr in mold at 120°C and post cure of 6 hr at 120°C
Cellular silicone
0.55
Union Carbide Y-3260 cellular silicone, urea used as a temporary
filler. Commercially fabricated by Union Carbide.
Polystyrene foam
(Styrofoam)
0.032
Dow Chemical Styrofoam FR-33. Commercially fabricated by
Dow Chemical.
The test specimen was mounted between two stainless steel compression plates,
one of which was rigidly attached to and moved with the crosshead. The other was
attached to the load cell via a pull rod through a yoke and pivot pin assembly. The
amount of compression of the flexible material was measured by two linear variable
differential transformers (L VDT), one on each end of the two stainless steel blocks.
The displacements measured by the two LVDT's and the average of these two displacements were recorded as a function of the applied load. The L VDT's were precalibrated at the various testing temperatures. If a foam specimen was nonuniform
and compressed more at one end than the other, this could be detected and recorded
by the output signals from the LVDT's. This was the case with one of the 0.42 g/cm 3
density polysiloxane specimens and was verified by rotating the specimen 180
0 •
Pull rod (connected to load cell )
t
Dewar
Specimen
Fig. 2. Test arrangement.
Compression Plate (connected to
pull rod vio yoke ond pins)
388
w. F. Stewart, D. T. Easb, aud W. A. May
The testing conditions were determined mostly by the operating conditions
described earlier (0.7 kg/em 2 preload during cooldown and 3.5 kg/cm 2 maximum).
Compressive load-deflection data were obtained at room temperature. The specimen
was then submerged in liquid nitrogen with either a compressive load of less than
0.07 kg/em 2 or 0.7 kg/em 2 maintained during cooldown. Compressive load-deflection
data were again obtained. Then, the specimen was either allowed to return to room
temperature and the compressive load-deflection data obtained once more or it was
submerged in liquid helium, load-deflection data obtained, and then returned to
room temperature and tested again.
Several of the specimens were subjected to a vacuum while compressed to
0.7 kg/cm 2 to observe the effect.
Several cycles of loading and unloading at room temperature, at 77 K, and at
4.2 K were made to simulate the thermal and loading conditions that exist during
cryostat and support system warmup, cooldown, and operation.
No attempt was made to determine the glass transition temperature of these
materials.
RESULTS
The compressive load-deflection data obtained for those materials tested are
presented in Figs. 3 through 8. In these figures, the deflection is presented as a
percentage of the original thickness and the compressive load is presented as the
load cell force divided by the surface area of the specimen. These results are indicative
of the general behavior of the materials since the number of samples tested was
insufficient for a statistical prediction of their behavior. Where possible, the hysteresis
effects of loading and unloading cycles are shown, but in many cases the hysteresis
effects were too small to be resolved in these figures. Data scatter is not shown in
the figures but, in general, the compression did not exceed a scatter of 2 % of the
original thickness at a given load.
The flexible materials that were tested became extremely rigid at the lower
temperatures but returned to their original flexible condition upon warming to room
temperature. When compressed to 0.7 kg/cm 2 during cooldown, these flexible
materials seemed to be slightly more rigid than when compressed to less than 0.07
kg/cm 2 during cooldown.
There was very little, if any, difference observed in the compressive load-deflection
data obtained at liquid nitrogen and liquid helium temperatures.
The relatively large deflection for the first few kg/em 2 of load at liquid nitrogen
and liquid helium temperatures may be a low-temperature property of the material
or it could be a result of bowing of the specimen (caused by the edges cooling more
rapidly than the central portion of the specimen), friction in the pullrod or in the
LVDT mounting fixtures, a shift in the lower compression plate when unloaded, or
a combination of these possibilities. A more elaborate test fixture would have been
required to enable this phenomenon to be examined in greater detail.
At room temperature, the flexible polyurethane foam is a tougher, more tearresistant material than the cellular silicone, which is in tum tougher than the polysiloxane Silastic, which is very easily torn or damaged.
The flexible polyurethane foam, a mostly closed-cell material, exhibited about
a 0.56 kg/em 2 increase in load with a slight decrease in deflection when subjected to
a vacuum with an initial 0.7 kg/cm 2 compression (see Fig. 6). This increase dropped
off only slightly over a 35 min period. When the vacuum was broken with gaseous
389
Compressive Load-Deflection Characteristics
A · PoIY5I1oJ.one. 0 .55 Q/c m'
8· POIY'Sllol.one. 0 . 51 q/cm'
Cellula, s ilico ne, 0 .55 Q/em'
O· Poly ur e thane. 0 .44 Q:/(m3
5.0
c-
4.5
E · POlys. lo.on e , 0.47 Q/c ml
F - Polys ilo lCo f'l e, 0.45 ;/(m3
G - Polyurelha ne, 0 .32 q / cm l
4.0
H -POly sllo llon e
I
0.42 Q/c m l
A
N
BCD
E
...."
~
""o
...J
Fig. 3. Load-deflection characteristics at room
temperature.
De fl ectio n
t3
Ce-llulo r sl hcone, 0 .55 Q'tm3
lN 2 . 0 .7 qk/c.m 2 compren'lt e
load dur ing (ooldo w",
15
t~~___
15
-
load
I~~
I~f~r-=
__ _ _ _ _ _ _ _ _ _ __
c
2
"
~
~
0
_
o" ' lnq cooldo wn
~
POI !furetha"e, O.44q/ cm J
L N 2 , 0 .7 k'9 ' (.m 2 comp,en lve
_
load d Uring eaoldo .... n
- - - - - - - - - - - - --
;J.
%
-
Polyu r et hane, 0 .4 4 Q/cm.l
LHe, 07 k.g/cm2 compress ive
-
load
dunnQ cool do wn.
load dU ' II'u;J coolOO\ll' n
~t-~f-
-
=
Pol yurelhone. 0 ,.3.2 q/cm l
comg , ess lye
l 2' 0.7 \cg/ cm2
l oad dUring cooldo\llH'I
- - - -_ _ __ _ __ _ _ _ POlyst lOJC one, 0 ,55 c,/cm'
k~
:~...:;;;..!~~~~:;;::;;:;;:::::::===== ~:,:,,~::~::. :O::d:::
LN2"
O. 7
/em 2 eomp t USI YC!
IO<JcI dutlng Cooldown
=-==-"==-=-=~~-------
:~-
2
2 ' 01 1I.9/en'l2
m,
com pr ess ' \le
load dur In g cool down
- - - Lood ,nQ
Un lood, nQ
=I
~
PoI YSllo a<Jne . 0 ,5 5 Qlcm 3
LN2.<0,01I1og/Cm2
eomp rUS I... e
L
0
4
~
Polyul elhone. 0.44 Qlem'
LN Z ' <0.07 II,g I cm 2 compr esSI 'I e
j
1..
' .c:_==-==_-=_=-=_=-=_=""_=-"===-_____ __ L
POlysl l o ;.:o ne, 0 . 51 g lem}
2,< 0,07 Q/ cm2 comprUSI ... e
lood dU'lnO
6
cooldo \ll n
8 ~~~~--~~~~~~-f~~--~-7~----~
o
0 .5
10
I 5
2 0 25
3.0
3 5
Load, kg I cm2
4.0
4.5
5 .0
Fig. 4. Load-deflection characteristics at 77 and 4.2 K.
EF
GH
w. F. Stewart, D. T. Eash, and W. A. May
390
0
- - - Loodin g
Unlood i ng
2 " ..... _
;:::-:::=-=::-::~-=::-::::-==:-::=-==~__
POly slfo:a;Qne. 0.147 QI em l
LN:2" <O.07kg/cm Z compressive
:t=
;fl
c
~
u
'"
';;
0
1000 dU f l n9 c~dO 'wlll n
lood dtJf Ing
1--
cool down
Polys llollone. 0.45 q/cm}
- --
-
-
-
-
-
-
- --
-
-- l
2'
0 .7 ktit/C mZ
looel 4 Uf l n4jl
compre5!5ll1e
j
j
CQ(lldOw n
0
2
----
PO I ~ Sllo .. one. 0.42 q / cm3
L
2' <0.0 7 kq / cm2
lood
(luring
comp r e"S$I\/@
coo iClo w n
32
~
Pol ),'S llo xOM. 0. 4 2 'O/e",3
- - -- - - - - - - - -- - l N2 . _0.49 'gIC m2
10011 a u, InC) coolda,.,.n
36 L-~L-~~~~~~~~~~~__~__~__~______~
o
0.5
1.0
I 5 20
2.5
3.0 35 4 .0
Load. kg / cm 2
4.5
5 .0
Fig. 5. Load-deftection characteristics at 77 K.
50
B
A· Room fempero ,ure, In,l lol lesl
0 . 7 k;/cm 2 cOfTlpres5IVe Iood dUring cooldOwn
a c·LN2 ,
D· Room tempe ralu re oher C
LN2 ~O.07k9/cm2 eompre'$$lve load during cooldown
E-
45
~ - Vac uum lU i 01 room lemperolun: of Ie, E
G- Room lempe: ral ur e ofl er F
H-lN Z '
.
~.O
B.C •
H,',J
0 .7 kCtJ/cm2:
Com pr U!I ... e
D
GKA
lood dUring coot do wn
t LHe, 07 Q/cm2 comp fe5 $I Ve
load dU f lnq coot down
J - LHe afle f I
K - Room Ie.mpefolute olle r J
A
2.0
15
1.0
--Lood In 9
- -
'5
20
25
Deflec tIon . %
-
Unlood l 'u~
40
45
Fig. 6. Typical test cycle.
391
Compressive Load-Deflection Characteristics
30
- - -
Lood l n9
Unloa d In;
25
N
E
....u
~
20
15
."
0
0
...J
Cellulor SIlicone
POlys, lo_one
(0.55 qICm 3 I
(0.45 9/om 3 I
L 2 ,cOO7 'kq /c m 2
LN2 .
compress ive lood
duunQ cool do wn
com preuIVe
o 7 ~9/cm2
load
du ring cool dOwn
I
I
10
5
0
0
Fig. 7. Load-defiection characteristics for overload condition.
2
3
"
I
18
20
Del lee 1100. 'M
helium, the load dropped below 0.7 kg/cm 2 (the initial load) and did not return to
0.7 kg/cm 2 within several hours.
The flexible cellular silicone, a more open-cell material than the polyurethane,
exhibited about a 0.28 kg/cm 2 increase in load with a slight decrease in deflection
when subjected to a vacuum with an initial 0.7 kg/cm 2 compression. The 0.28 kg/cm 2
increase dropped to zero (i.e., returned to the initial 0.7 kg/cm 2 ) in approximately
1 min, indicating that the gas escaped from this specimen. The flexible cellular silicone
exhibited more of a hysteresis effect than the other materials upon load removal at
room temperature.
None of the specimens, examined visually, was observed to crack or break
during the low-temperature testing. They were not examined microscopically to
determine if there was any individual cell breakdown. This was true for all but the
rigid Styrofoam, which did take a permanent deformation (see Fig. 8).
One polysiloxane specimen (0.45 g/cm 3 ) was compressively loaded to over
28.1 kg/cm 2 at 77 K to observe the material's behavior under higher compressive
loading (an overload condition for the experiment). At 12.9 kg/cm 2 compressive
load, the deflection increased 0.6% and then the material became very rigid, as
shown in Fig. 7. A cellular silicone specimen (0.55 g/cm 3 ) was also loaded to over
28.1 kg/cm 2 at 77 K and did not exhibit any sudden shifts in deflection, as shown in
Fig. 7. Following this test, the room-temperature load-deflection data were essentially
the same as before the test for the cellular silicone specimen.
4 .0
--LOOdlng
- - -
Unloadln Q
I
I
4 ,2
I(
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Fig. 8. Load-defiection characteristics for polystyrene foam (0.032 g/cm 3 ).
I
I
IC)
20
lO
40
0
Del leet Ion . %
10
lO
I
I
40
W. F. Stewart, D. T. Easb, and W. A. May
392
ACKNOWLEDGMENT
The assistance and advice of J. D. Rogers during the testing program and the preparation of this
report are greatly appreciated and acknowledged.
REFERENCES
I. J. M. Arvidson. R. L. Durcholz. and R. P. Reed. in: Advances in Cryogenic Engineering. Vol. 18,
Plenum Press. New York (1973). p. 194.
2. F. A. Smith (Union Carbide Corporation), U. S. Patent 3,238,157 (March I, 1966).
DISCUSSION
Question by R. P. Reed, National Bureau of Standards: How did you calibrate the strain at 77 and
4.2 K?
Answer by author: The LVDT's used to measure the amount of compression of the test specimens
were precalibrated by immersion in liquid nitrogen and liquid helium. The motion of the compression plate
was measured by a dial indicator attached to the pullrod of the Instron and this was compared to the
output of the LVDT's for a calibration.
Question by R. P. Reed, National Bureau of Standards: The "negative work hardening" shown in the
4.2 K stress-strain curve of Fig. 8 is interesting. Do you have any comments on this phenomenon?
Answer by author: At this time, we do not have an explanation for this phenomenon. This curve
was obtained from the testing of a single specimen. Thus, it is not known if this phenomenon is peculiar
to this single specimen, peculiar to this single test, or if it is actually characteristic of this material at this
temperature. Additional testing would be necessary to verify this. This result may not be a work hardening
process but a foam cell collapse process.
J-4
A STRUCTURAL PLASTIC FOAM THERMAL
INSULATION FOR CRYOGENIC SERVICE
R. B. Bennett
Amspec Inc.
Columbus, Ohio
INTRODUCTION
Today, with the impending energy crisis and the rush to construct LNG facilities
as one method to alleviate the energy shortage, it is not surprising that we find plastic
foams and other thermal insulating materials playing a leading and necessary role
in cryogenic applications. Obviously, the cost of thermal insulating materials is a
minor part of an LNG contract when compared to the total project cost. However,
the proper selection ofthermal insulations in their various forms can be a determining
factor in the overall success and performance of an LNG facility. Plastic foam insulation can playa significant and vital part in the performance and economics of each
LNG installation.
Most cryogenic thermal/structural applications can be satisfied by the use of
plastic foams, provided the specifier and designer has a thorough understanding and
knowledge of each manufacturer's insulation products. One major application for
a plastic foam requiring significant load-carrying requirements is the base or bottom
insulation for land LNG tanks. This is a specific end use where plastic foams can
offer economical long-term performance. An example of such a plastic foam is
Styrofoam* HD-1623. This product was developed specifically for cryogenic tank
base insulation and after years of testing and research was only recently made available as a sales product.
This presentation will be confined to the specific topic of a single foam product
and its physical characteristics as they relate to the structural/thermal application of
the LNG tank base.
PHYSICAL CHARACTERISTICS
The structural plastic foam considered here is a high-density [52.8 kg/m 3
(3.3 pcf)] extruded polystyrene foam. Basically, it is not a new product; high-density
extruded polystyrene foams have been used in many other applications and have
been in continuous use as a load-bearing insulation in excess of twenty years.
It is important to note that most plastic foams are anisotropic, and this structural
plastic foam is no exception. Cell size, cell shape, and cell orientation will influence
the mechanical responses obtained when tested along the three orthogonal axes.
Quality control is of vital importance in this case, and during manufacturing, cell
• Trademark of The Dow Chemical Company.
393
394
R. B. Bennett
250
1700
1600
225
I
200
v/
175
125
100
75
25
(-260'F)
II
/V
II
1400
1300
1200
I V r--r--II
150
50
A111K
1500
r---...
"-.
/(+~S:'F~
1100
1000
900
~
800
700
600
500
400
V
300
V
200
100
STRAIN, %
Fig. 1. Compressive stress-strain plot for polystyrene foam developed for
low-temperature insulation.
size and orientation are controlled to optimize the mechanical behavior in all three
directions giving the plastic foam high load-carrying capabilities in the rise or vertical
direction. This product has a minimum compressive strength of 861 kPa (125 psi)
in the rise direction at its yield point and at 297 K ( + 75°F). As the temperature is
reduced, the compressive strength is increased to an average compressive strength
of 1550 kPa (225 psi) at the LNG temperature (see Fig. 1). It is interesting that the
strain level at yield is about the same for either temperature stress-strain plot. The
compressive modulus increases as temperature is reduced, but the yield point remains
at the same strain level, indicating that the product has the capability to withstand
impact and shock loadings without brittle fracture. This is extremely important
when the designer must consider the cyclical loading differentials that occur through
product loading and unloading.
An important design consideration for load-bearing thermal insulation is its
long-term compressive creep characteristics. In order to predict long-term creep
behavior without having to wait twenty to thirty years for creep test results, it was
necessary to conduct accelerated testing that would yield predictable long-term data.
This was accomplished by using the time-temperature superposition technique for
extrapolating long-term results from relatively short-term testing, where the temperature variable was used to accelerate time for the thermal plastic product. Three
A Structural Plastic Foam Thermal Insulation for Cryogenic Service
load levels were selected: 138, 276, and 414 kPa (20, 40, and 60 psi). Normally,
276 kPa (40 psi) would be the maximum design load based upon a safety factor of
three. The 138- and 414-kPa loads provide data plots on either side of the 276-kPa
load to help confirm and correlate data reference points.
Figure 2 indicates the long-term compressive strain of the plastic foam under
three stress levels at 296 K ( + 73°F). At lower temperatures, the creep rate would of
course be reduced; however, the 296 K data will give the designer a conservative
margin of safety. [Recent testing at 275 K ( + 35°F) showed the creep rate to be three
to four times less than the room-temperature creep rate shown.] Figure 2 shows the
three stress levels beginning above 0% strain, because the elastic deformation due
to initial loading is included. Creep must be determined as the total strain at a selected
time less the initial elastic deformation. A paper by Hart et al. [1] provides additional
information on the creep behavior of cellular polystyrene.
Because it has an elastic deformation under stress, the plastic foam under
discussion shows virtually no change under repeated cyclic loading; a slight compaction of the cut surface cells was noted where foam blocks came in contact with
one another. After 60,000 cycles at 310 kPa (45 psi), there was less than 1% reduction
in overall foam thickness. Cyclic loading is an important consideration when designing tank bases for LNG receiving facilities where product loading frequently varies.
Freeze-thaw cycling on submerged specimens showed no evidence of deterioration after 40 cycles.
With respect to the thermal properties of the plastic foam (Fig. 3), we find that
it has a thermal conductivity of 0.0224 W/m-K (0.155 Btu-in./hr-ft2-0F) at a mean
temperature of 200 K (- lOO°F). At a mean temperature of 297 K, it has a thermal
conductivity factor of 0.0303 W/m-K (0.21 Btu-in./hr-ft2-oF~ It is apparent that the
plastic foam exhibits thermal conductivity characteristics similar to those observed
for urethane. The condensation hump is due to the increase in the conductivity factor
.z"
~kP'
E
(10 pol)
.. 3
a::
>
a
8
2
L----
0.1
10
1110
--
~
1Il00
~ ~kP'
(CO pol)
~
10.Il00
131 kP.
VpolI
f,...
1110,Il00
nME. HOURS
Fig. 2. Predicted compressive creep at 296 K for a structural plastic foam insulation
for cryogenic service.
R. B. Bennett
.-
.21
.1'
.1'
~
"II:
.12
IE"
.at
i
~
/
/
V
/
/
/
/
/"
~,'
.,.,,',.
~'
/
~ (r'#
/
.011'
K~
.0173
.
I
.00ao .!!
•
.01
.0017
.113
.1IIM3
D
.1
(-a)
II
(-3DD)
117
(-lID)
144
(-IDD)
172
(-lID)
IDD
(-100)
221
(-50)
215
(D)
213
(+50)
311
(+100)
lSI
(+150)
MEAN TIMP.:K ('F)
Fig. 3. Thermal conductivity as a function oftemperature for a structural plastic foam insulation
for cryogenic service.
when the Freon gas in each cell condenses into a liquid. The dashed line represents
the predicted thermal conductivity value after twenty-five years of aging at 297 K.
This is due to air migration into the cells to replace part of the mixed blowing agent
that has diffused out of the cells. Since air has a higher thermal conductivity value
than the escaped blowing agent, it results in a slightly higher value for this property.
All published thermal conductivity values from this laboratory are based upon the
predicted twenty-five year plot. Computations were made to determine the rate
methane gas would diffuse into the cells of the foam and its effect upon the thermal
conductivity ofthe foam. The assumed environment was a 100% methane atmosphere
at room temperature and 1 atm pressure. Calculations based upon permeability,
solubility, and other pertinent data regarding methane and polystyrene show the
inflow of methane into a 101.60-mm (4-in.) block of this polystyrene foam to take
about ten years to reach 1 atm pressure within the cells. The thermal conductivity
value at this time would be 0.0253 W/m-K (0.175 BTU-in./hr-ft 2 -OF) at a mean
temperature of 297 K, indicating that if methane were to replace the foaming agent,
the thermal conductivity value would remain unchanged from the values as shown
in Fig. 3. At lower temperatures, the methane diffusion rate into the foam would be
substantially reduced and would take several hundred years at the cryogenic temperature of LNG to reach equilibrium.
Additional testing by independent laboratories of this polystyrene foam in an
LNG environment was carried out to determine the foam's compatibility to such
exposure. (See Table I showing physical properties.) The Yarsley Laboratories in
England ran 500- and 1000-hr immersion tests on the foam in liquid methane, ethane,
and propane. The foam specimens were checked for compressive strength before
and after immersion. After immersion, half the specimens were conditioned at room
temperature for 24 hr before testing compressive strength, and the other half were
A Structural Plastic Foam TbermallDsulatiOll for Cryogenic Service
397
Table I. Physical Property Data of a Structural Plastic Foam Insulation
for Cryogenic Service
Property
Density, kg/m 3 (pel)
Thermal conductivity,
W/m-K
(Btu-in./hr-ft 2 - OF)
at mean 297 K ( + 75°F)
at mean 200 K (-100°F)
Compressive strength at
yield-vertical direction,
kPa (psi)
at 297 K ( + 75°F)
at III K ( - 260°F)
Water absorption, % by volume
Water vapor transmission,
metric perm cm (perm in.)
Linear coeft'. of thermal
expansion, m/m-K (in./in.-OF)
LNG compatability
Test method
ASTM C-303
ASTM C-518
ASTM C-177
ASTM
ASTM
ASTM
ASTM
D-1621
D-1621
D-272
C-355
Value
52.8 (3.3)
0.0303 (0.21)
0.0224 (0.155)
861 (125)
1550 (225)
<0.1
1.0 (0.6)
ASTM C-696
6.3 x 10- 5 (3.5 x 10- 5)
Excellent
Freeze/thaw resistance
Dynatech/R & D,
Yarsley Lab
Yarsley Lab
Yarsley Lab
Dow---60,OOO cycles
at 310 kPa (45 psi)
Dow
Corrosion
Dow
Compressive creep
Flame spread
Dow
ASTM E-84
Ethane compatability
Propane compatability
Cyclic loading
Excellent
Excellent
< 1 % reduction
in thickness
No evidence of
deterioration after
40 cycles
Does not cause
corrosion
Fig. 2
<25
tested at the liquefaction temperature of the gas. Those specimens conditioned after
exposure showed no apparent change in compressive strength compared to the specimens that were unexposed Also. those exposed specimens tested at the liquefaction
temperature showed no apparent change in compressive strength from specimens
that were unexposed and tested at the same liquefaction temperature. The thermal
conductivity values were checked at mean temperatures of 297 and 200 K on both
exposed and unexposed specimens and showed no change due to immersion in the
liquefied gases. Also, visual and microscopic examination revealed no physical
deformation, warping, or cellular changes in the foam because of the exposure.
The Dynatech RjD Laboratories in Cambridge, Massachusetts, repeated this
testing using LNG obtained from the Boston Gas Company. Again the SOO- and
1000-hr immersion testing had no apparent effect upon the thermal or mechanical
properties of the foam.
The polystyrene foam under consideration contains a fire-retardant additive
which allows the product to achieve a flame spread rate ofless than 2S-in. thicknesses
up to 101.60 mm (4 in.) according to ASTM E-84 Tunnel Test and is listed by Underwriters' Laboratories Guide BRYX, Foamed Plastics. However, this should not be
construed to mean that the product is incombustible. The polystyrene foam is an
organic product and under the proper circumstances, it will burn. Therefore. reasonable precautions should be taken during the installation of the foam to reduce the
R.B.ae.ett
possibility of accidental fire. For example, it is advisable to restrict welding during
installation and to cover expansive areas of exposed foam with a fireproof tarpaulin
to minimize accidental fire before the sand or concrete finish is applied over the foam.
Further, it is recommended that chemical or carbon dioxide extinguishers or a fire
hose be strategically placed in the area where foam is being installed.
The average linear coefficient of thermal expansion for the structural plastic
foam is 6.3 x 10- 5 m/m-K (3.5 x 10- 5 in./in.-OF) from room temperature to 111 K
(-260°F). This is relatively high for a structural material. However, knowing this
characteristic, the material can be installed in such a manner as to prevent any direct
openings between layers that may cause convection heat flow. Offsetting joints
between each layer of boards and installing a polyester film or a glass-reinforced
Kraft aluminum foil laminate between layers near the cold side will isolate openings
and restrict convection flow.
The fact that foamed polystyrene has a low rate of water absorption is significant
to the designer and assures him that the product will not pick up moisture while in
storage or during installation. By ASTM C-272, the percent by volume water pickup
in this polystyrene foam is less than 0.1 % and this low value is essentially due to
moisture clinging to the surface of the foam specimen. Also, the foam has a water
vapor transmission rate of 1.0 metric perm cm (0.6 perm in.) according to ASTM
C-355, further indicating the product's resistance against moisture migration and
infiltration.
ADVANTAGES
Performance will reflect economics. The LNG tanks that have been installed
using this plastic foam as base insulation are performing well within design expectations. The fact that less insulation thickness is needed because of the low thermal
conductivity value of the foam can mean savings in shallower foundations, lower
tank shell, or perhaps additional tank capacity. Installation time is reduced, because
the insulation boards are shipped to the job site in trucks unpackaged or at the most
bundled and held together with glass tape. Therefore no unnecessary handling or
unpackaging is required. The large boards, 2.74 or 1.37 m (9 or 4! ft) long x 406.4 mm
(16 in) wide x 76.2 or 101.6 mm (3 or 4 in) thick, are lightweight, easy to handle,
fabricate with ordinary woodworking tools, can take handling abuse and foot traffic
without breaking, and will not dust or irritate the skin. These advantages add up to
less installation time and smaller work crews, which again mean greater savings.
INSTALLATION
The insulation boards are installed dry over a leveling bed of sand or concrete
(see Fig. 4). The boards are laid side by side with all joints staggered and tightly
butted. A fiberglass batt insulation is installed and compressed between the insulation
boards and ring foundation wall for the inner containment vessel. The compressed
batt will act as a spring keeping the joint between the ring foundation and floor
insulation closed, thus preventing or minimizing the effect of convection heat flow.
Subsequent layers of boards are installed with all board joints broken between
layers to reduce through openings. Subsequent boards can be secured in place by
using two or three treated wood skewers driven through the board at an angle into
the board beneath. To further assure against convection flow between boards, a
film may be installed above and between the two top layers offoam where the greatest
board contraction will be realized. The film layer on top is especially necessary when
399
A Structural Plastic Foam Thermal Insulation for Cryogenic Service
INNE R TANK SHELL
(
OUTER
TANK
SHELL
POLYESTER FlU/! A80ve AND
BETwEEN TOP TWO L.AYERS
OF INSULATION
CRANUl.AR
FILL
:\
)
... ~~:::O:'"':' .
CONCRETE FOOT ING
"" ,::
im~11{i~~;~~
STYROFOAM
HD. 1!&i23
INNER TANK SHElL
SAND OR CONCRETE
Fig. 4. Typical insulation and construction of
cryogenic storage ta nk shell and floor.
~.
•
SAND .. ··
HEATING e-l.EMENTS
(;)
~
dry sand is used to cover the foam to prevent sand from filtering between the board
joints. The top layer offoam boards must be protected by either sand, poured concrete,
precast concrete planks, or other material that will prevent exposure of the foam to
sparks and heat from welding of the floor plate. The foam must be protected from
solvents and fuels, which often can cause serious attack on the foam. The film layer
will offer protection to the foam, as will the sand cover or concrete topping. Also
safety cans should be used to store and dispense solvents. A solvent-dampened rag
is recommended for cleaning metal plates. A preferred solvent for cleaning and degreasing plates is denatured alcohol, which is an effective solvent for degreasing and
is also compatible with the polystyrene foam.
SUMMARY
The preceding is an example of how a single, rigid, plastic foam was designed to
fulfill a specific application. However, a variety of plastic foams on the market today
are being successfully used and specified in various applications throughout the
construction field. It is important that the designer is aware of these foam products
and their proper application.
The engineer knowledgeable in the proper use of plastic foams often can attain
greater design flexibility and better performance with increased savings in today's
ever-spiraling economy. Through cooperating foam manufacturers and engineering
design firms, new uses and applications for plastic foams are being realized.
REFERENCE
I. G. M. Hart, C. F. Balazs, and R. B. Clipper, J. Cellular Plastics, 9(3):3 (1973).
J-5
A DIFFERENTIAL THERMAL ANALYSIS
APPARATUS FOR USE AT CRYOGENIC
TEMPERATURES*
E. Catalano, J. A. Rinde, and J. C. English
Lawrence Livermore Laboratory, University of California
Livermore, California
INTRODUCTION
A differential thermal analysis (DT A) apparatus for use in the temperature
range from 20 to 300 K has been constructed and put into operation. This apparatus
has multiple sample cells for simultaneous testing of up to twelve samples. Both
temperature control and data sampling are handled by a minicomputer using a
modified version of an interpretive language.
The DT A apparatus is used to measure and record any difference in temperature
between a sample material and a reference cell. In this way, values for melting points,
glass transition temperatures, solid-solid transitions, and approximate values for
heats of fusion can be determined. A cryogenic DT A apparatus has been previously
reported e]. The multiple sample feature enhances the use of this instrument for
screening properties of materials.
DESCRIPTION OF APPARATUS
A cross section of the DT A apparatus is shown in Fig. 1. The key items of this
apparatus are the twelve sample cells arranged around the central reference cell.
The reference and differential thermocouples are placed in small holes that are drilled
into the cells. Good heat conduction is obtained between the sample and the sample
cells by adding a drop or two of a heat transfer fluid, octadecane or diethylphthalate,
to each sample cell. The sample cells are located in a sample compartment consisting
of a base plate and a cover that are held together by high-strength aluminum screws
and an indium wire gasket to make a vacuum-tight seal. The sample compartment
rests upon a miniature Joule-Thomson refrigerator that produces up to 4 W of
cooling at 23 K and has a stainless steel gas fill tube connected to the base plate. A
copper strap thermally connects this stainless steel tube to the radiation shield and
also connects both of these to the liquid nitrogen reservoir.
The outside vacuum cylinder contains the liquid nitrogen reservoir, which can
be moved up and down by turning the large nut located on the top of the vacuum can.
In this way, the liquid nitrogen reservoir can be lowered into contact with the sample
compartment to lower the temperature rapidly to approximately 80 K. To reduce
the sample compartment temperature to 20 K, high-pressure nitrogen and hydrogen
• Work performed under the auspices of the U. S. Atomic Energy Commission.
400
A Differential Thermal Analysis Apparatus for Use at Cryogenic Temperatures
Heal!!r
401
liI:.ftlrenct!l Gel!
Somplw c:el l
TkerrntX:oup le!.
Th.rmoooupi e
leoc:l w ir ••
Cryo-t ip
Nylon iUfJport
~~~E~~tft---A'1c.teWS
Indiutn Qlulce r
Somple c;ompartml!'n r
FiII.i.d>e
COPp!!lf !..rap
Fig. 1. Cross section schematic of DT A
apparatus.
gas are started flowing through the refrigerator and the liquid nitrogen reservoir is
backed away from the sample compartment. For these low temperatures to be
achieved, the space within the apparatus must be pumped to a good vacuum:
1 x 10- 6 mm of Hg or better. This is accomplished using a 4-in. oil diffusion pump
located directly below the apparatus, but not shown in Fig. 1.
The hardware associated with the minicomputer data acquisition and control
system are described elsewhere [2]. Both differential and reference temperatures are
measured with copper-constantan thermocouples, calibrated against an NBScalibrated platinum resistance thermometer. The signals from these thermocouples
are carried through shielded cables to a Cunningham* Model 309 crossbar scanner,
a Vidar Model 520 digital voltmeter, and the Digital Equipment Company (DEC)
PDP-8/1 computer. The program used is a modified version of the DEC FOCAL 69
Program.
OPERATION
Samples are placed in the sample cells and the various connections and vacuum
seals are made. The liquid nitrogen reservoir is lowered into contact with the sample
compartment and liquid nitrogen is added if any liquid samples are being run. If
not, the total system, including the sample compartment, is evacuated to a pressure
of 1 Torr or lower. The sample compartment is then valved off from the vacuum
*Reference to a company or product name does not imply approval or recommendation of the product
by the University of California or the U. S. Atomic Energy Commission to the exclusion of others that
may be suitable.
402
E. Catalano, J. A. Rinde, and J . C. English
pump and back-filled with 300 to 500 Torr of helium gas. The helium gas acts as
the heat transfer agent to conduct heat from the sample compartment walls, where
the heater is located, to the sample cells. After the system has reached temperature
equilibrium at the lowest desired temperature, generally 77 or 20 K, a computer
program that measures the DT A reference temperature and compares it to a programmed, linearly increasing temperature is started. The difference between these
temperatures then controls the amount of power to be fed to a heater wrapped
around the DTA sample compartment. At various time intervals, the reference and
differential thermocouples are read by means of a precision voltmeter and the data
are stored on the computer memory for later processing. The need for the computer
is obvious when one realizes that 7280 separate readings are required to obtain data
at 0.5 K intervals from 20 to 300 K from thirteen thermocouples. Typical data
sampling rates for a heating rate of 1 K /min are at intervals of 0.5 or 1.0 K.
EXPERIMENTAL RESULTS
A DT A scan from liquid nitrogen temperature to room temperature for eleven
polymer samples was made at a temperature rate of 1 K /min. The individual recordings of ~ T vs. T data can be shown on a single plot, but since the data were collected
and stored in the disk memory of the computer, it can be plotted separately or
collectively, and at various scale factors. Figure 2 shows six curves from the above
run. The polystyrene curve is an example of a material showing no thermal transitions
in this temperature range, while the polyethylene terephthalate (PET) + diethylphthalate (DEP) curve is a good example of first-order transitions. The PET, Mylar
film heated to 173°C and cooled at l.2°Cjmin, shows no thermal changes in this
region and the two peaks are due to the small amount of DEP present, as heat transfer
fluid, changing from a glassy solid to a crystalline solid at about 248 K and then the
crystals melting at 270 K. Since the plot shows an exothermic peak that is the result
106
1.\0
186
Te ~~Oh.fI. I(
273
Fig. 2. Thermograms showing first- and second-order
transitions.
A Differential Thermal Analysis Apparatus for Use at Cryogenic Temperatures
403
of crystallization, it should also show the glass transition (Yg) for DEP at a lower
temperature. This is probably the cause of the slight dip in the curve at 186 K where
the DEP glass transition is located.
The four other curves in this figure show the glass transitions of Adiprene
(L-lOO)/moca samples cured under different conditions. The glass transition values
observed for Adiprene agree with the reported value [3] of 223 ± 5 K. It is seen that
glass transitions appear as step changes in the baseline and therefore are not as well
defined as first-order transitions.
The DT A apparatus presently is being operated at temperatures from 20 to
77 K; however, at this time no thermograms have been obtained in this region
showing either first- or second-order transitions.
REFERENCES
I. R. F. Robbins, "Behavior of Polymeric Materials at Cryogenic Temperatures," Cryogenic Division,
NBS, Boulder, Colorado, Contract No. NASA H-92120 (1968).
2. J. C. English, "Calorimetry Data Acquisition and Control," Lawrence Livermore Laboratory, Livermore, California, Rept. UCRL-73046 (1971); presented at Decus Spring 1971 Symposium, Atlanta,
Georgia, May 13-15, 1971.
3. E. I. du Pont de Nemours and Co., "Adiprene Product Bulletin."
K-]
FORCED CONVECTION HEAT TRANSFER
TO SUBCRITICAL HELIUM I
P. J. Giarratano, R. C. Hess, and M. C. Jones
Cryogenics Division
NBS Institute for Basic Standards
Boulder, Colorado
INTRODUCTION
The intensive development of superconducting technology for electrical power
equipment calls for a good knowledge of heat transfer to helium. Traditionally,
simple bath cooling is used to provide the necessary rates of heat transfer to stabilize
the conductor, and this appears still to be quite adequate in dc applications. However,
in ac applications and various applications where large field sweep rates are experienced, either in normal operation or under fault conditions, losses occur and there
may be design benefits from circulating the helium.
In a previous study [1], heat transfer coefficients for forced flow of supercritical
helium were measured and a correlation was developed to predict the heat transfer
in this region. As an extension of that study, heat transfer coefficients have been
measured under conditions of forced flow of subcritical helium. Particular emphasis
has been placed on the determination of conditions under which a transition from
nucleate boiling to film boiling occurs (critical heat flux) since an unacceptable rise
in wall temperature may occur at this point (from the point of view of cooling superconductors).
It is hoped that the information developed in this program will assist in broadening the designer's choice and permitting closer design. The data presented here are
a summary of over 200 experimental runs, including 74 transition observations.
DESCRIPTION OF EXPERIMENTAL APPARATUS AND MEASUREMENT
A schematic of the boiling heat transfer flow loop is shown in Fig. 1. * A centrifugal pump, previously described [2], circulates the liquid around the flow loop.
To maintain a constant inlet temperature to the test section, heat is removed from
the liquid as it passes through approximately 13 m of 0.32-cm-ID copper tubing
located in a heat exchanger reservoir of approximately 2.5-liter capacity. The heat
exchanger reservoir is continuously supplied with liquid helium from a storage
dewar.
Test Section
The test section is a 0.213 cm ID x 20 cm long stainless steel tube with a wall
thickness of 0.016 cm. It is resistance-heated along 10 cm of its length (LID = 50).
• Identification of any materials or their manufacturer by the National Bureau of Standards in no way
implies a recommendation or endorsement by the Bureau. Furthermore, use of other trade names is
for the sake of clarity and does not in any way imply a recommendation or endorsement by the Bureau.
404
405
Forced Convection Heat Transfer to Subcritical Helium I
, - C::!--~,-~""""---::r1
'-1_I_I,
Mio~
I_
."'........,1
fin '"""
.... -.
_I-
... [111111.......
, ... 1..
,-.Mi.iooSWIiI
III .. h,llII_
lI ..
I.
iIIt,lIist..
~. llI1mlioo .iII
e• • raid. .
....... 11 • • •
Fig. I. Schematic of experimental apparatus.
A stainless steel-Pyrex seal located in the bottom of the loop provides electrical
isolation of the test section. A wire-wound preheater upstream of the test section
allows variation of quality at the inlet of the tube. Accurate measurement of the
current and potential drop across the two heaters is used in the calculation of heat
flux. At the lowest power levels, the uncertainty is less than 2%.
Temperature Measurement
Outside tube wall temperatures are measured at ten points along the heated
test section length with calibrated carbon resistance thermometers. One lead of the
carbon resistor is soldered to a small copper block which also provides a tempering
post for the resistor's electrical leads. The block is thermally clamped to the test
section but is electrically isolated from it by a thin (0.0127 cm) Mylar film. The thermometers are positioned 1 cm apart, with the first thermometer located 0.6 em from the
inlet. Inside wall temperatures are obtained by calculation of the temperature drop
through the tube wall using the thermal conductivity of stainless steel given by (AI)
in the appendix of the paper by Giarratano et al. [1]. Bulk fluid temperatures are
measured at four points along the flow loop (upstream of preheater, upstream of
test section, and two measurements downstream of the test section) with germanium
resistance thermometers. These thermometers are potted with vacuum grease in
copper wells which are soldered to the tube.
The carbon resistance thermometers and the germanium resistance thermometers used in this study were calibrated in a separate apparatus, over the range
4 to 20 K, against three germanium resistance thermometers (GT 1011, 1027, and
1024). These three germanium thermometers had been previously calibrated against
NBS secondary standard germanium thermometers (GT 722 and 734).
In this calibration, the agreement between GT 1011, 1027, and 1024 over the
temperature range 4 to 9 K was within 0.01 K for the worst case and was generally
within 0.005 K. Above 9 K, the agreement was within 0.05 K. An average value of
the three germanium thermometer readings was taken to be the true temperature.
The calibration data for each thermometer were fit with a curve of the form
log T =
N:7
L
N:O
A(N)(log R)N
406
P. J. Giarratano, R. C. Hess, and M. C. Jones
with an rms percent error of 0.06 in temperature over the range 4 to 19 K. This
corresponds to approximately 0.003 Kat 4.2 K and 0,01 K at 19 K.
However, in the heat transfer apparatus during preliminary runs, it became
apparent that there were carbon thermometry errors that were bulk temperature
dependent and which partially decayed in time (time constant of the order of hours).
Extensive tests indicated that the time dependence of the error was probably due to
residual spurious radiation heat leaks, not evident in the calibration apparatus but
impossible to eliminate totally in the heat transfer apparatus, and the bulk temperature dependence was probably due to shifts in calibration. Therefore, an in situ
recalibration procedure was adopted which allowed correction of the carbon thermometers on the basis of the germanium thermometer readings taking into account
both time and bulk temperature dependence.
The estimated uncertainty in outside wall temperature after correction is at
most 0.02 K (the maximum shift in calibration of the germanium thermometers
upon which corrections for carbon thermometers were based). However, the error
in temperature difference between the inside wall and the bulk temperature, due to
uncertainty in the thermal conductivity of the stainless steel wall, the wall thickness,
and bulk temperature, is of order 0.05 K at heat fluxes of 0.02 W/cm 2 and is of order
0.15 K at heat fluxes of 0.2 W/cm 2 .
In the two-phase region, since the pressure drop across the test section was of
the order of a few mm Hg for the data presented, the intermediate bulk temperatures
were taken to be the saturation temperatures corresponding to the inlet static
pressures. This approximation was within experimental error.
Pressure and Flow Measurement
Provision was made to measure pressure drop across the flow orifice, preheater,
and test section and the static test section inlet pressure using calibrated pressure
transducers and a static Bourdon tube pressure gage accurate to 0.01 atm located
outside the cryostat at room temperature.
The pressure transducers were located at room temperature, which resulted in
large temperature gradients along the pressure tap lines. Consequently, the pressure
drop readings across the test section were unsteady (noise of the same order as the
pressure drop itself). However, generally, when there was no surging, the pressure
drop across the test section was less than 7 mm Hg.
Since the fluid at the outlet of the preheater was slightly subcooled for much
of the data presented, a calorimetric method determined the flow rate under these
conditions. The calorimetric flow rate from the preliminary runs was also used to
determine a discharge coefficient for the flow orifice so that the orifice may be used
for flow measurement when calorimetric flow determination is not possible, i.e.,
two-phase out of the preheater. The discharge coefficient so obtained agrees with
that of an identical orifice section which was calibrated in a separate apparatus.
For the calorimetric method, applicable for 75% of the runs, accuracy of the
flow ranges from 2 to 9% (more uncertainty at the lower operating pressures due to
less subcooling available). This is due to uncertainty in measuring the temperature
rise of the bulk fluid through the preheater section. For 25% of the runs where the
orifice was used, there is an additional maximum uncertainty of 35% (3/T limits)
due to fluctuations in the readings of pressure drop across the orifice.
Extraneous Heat Exchange
To minimize heat leak from room temperature to the test section by conduction,
all electrical leads and the thin-walled, 0.317 -cm-diameter stainless steel pressure
Forced Convection Heat Transfer to Subcritical Helium I
407
transmission lines are thermally anchored to the outside copper surface of the liquid
helium heat exchanger. A length of multifilament niobium-titanium superconducting
wire is used for power leads between the heat exchanger and the preheater. The
small diameter and low thermal conductivity of this wire further minimize heat leak
due to conduction, and louIe heating in the leads is eliminated. Error in heat flux
due to axial conduction from the ends is less than 2% for the worst condition. Extraneous heat to and from the test section is therefore considered negligible. Boilofffrom
the liquid helium heat exchanger is routed through coils soldered to the pump housing
and the radiation shield. This arrangement minimizes heat leak from room temperature via conduction along the pump housing and provides a low-temperature
radiation shield around the flow loop. The test section portion is further protected
from radiation by a copper shield thermally anchored to the heat exchanger. The
evacuated copper enclosure (vacuum less than 10- 7 mm Hg), which is submerged in
a bath of liquid nitrogen, provides the first-stage radiation shielding from room
temperature.
Experimental Measurement
For a fixed system pressure, pump speed, and quality at the inlet of the test
section, the power to the test section was increased in steps from zero. At a certain
power level, a discontinuous rise in the wall temperature occurred at the outlet end
of the test section (critical heat flux). Further increase in power caused the discontinuity to move up the test section toward the inlet. The upper limit of power was
usually determined by excessive wall temperatures at the outlet. After a change in
test section power, thermometer voltages stabilized generally within a few seconds
and were recorded by a digital voltmeter and automatically punched on paper tape
together with all other pertinent voltages and run information.
Since the flow loop is a closed system, as power was applied, it was necessary
to vent the system to maintain a constant test pressure and conversely a decrease in
power applied required adding and condensing gas in the loop to maintain the
pressure. This procedure was repeated for different pump speeds, inlet quality, and
pressure.
During a measurement, the temperatures and pressure were stable to within
their accuracy prior to a transition in the heat transfer mechanism (critical heat flux
exceeded). When a transition occurred, it was accompanied by fluctuations in
temperature (for the wall stations in the transition region) of a rather random nature,
and an amplitude of the order of 0.5 K. Oscillations in pressure were not noticeable
but this was at least partly due to the pressure transmission line being hea vily damped.
Attempts have been made to establish the variability of the data from day to
day by limited repetition of a measurement under the same set of conditions (e.g.,
same pressure, pump speed, and inlet quality). For heat fluxes below the critical heat
flux, typical deviations in wall temperatures, at a given position on the test section,
were 0.02 K. For heat fluxes above the critical heat flux, the wall temperature deviations were of the order of 0.05 K and the position of the transition was repeatable
to the nearest thermometer, i.e., to within 1 cm.
HEAT TRANSFER RESULTS
Temperature profiles along the wall for subcritical helium heat transfer are
shown in Figs. 2 and 3 for pressures from 1.1 to 2.0 atm* and mass velocities from
4.5 to 63 gjsec-cm 2 . A typical profile is extremely flat up to some point at which a
* 1 atm
=
0.1013 MN/m2,
P. J. Giarratano, R. C. Hess, and M. C. Jones
408
32
"
32
PRESSuRE • 1.1 aIm
.... ss VELOCITY' 4.50 / , - em f
QUALITY @ INLET . - 0 .03 10 -0.02
28
28
24
24
20
20
::l
;
I-
:l
J
16
12
12
•• 0 .209 Of
,,,,,1,,---
8
-----,,-----
\ T: X
32
28
TOU\)(
0
8
14
20
10
"
•
c
....
12
24
LID
32
PRESSuRE' 1.50tm
MASS VELOClTY • 14 q' ,-cm'
12
oJ 16
oJ
16
16
PRESSURE • 1.1 atm
MASS VELOCITY' 28 - 32. qtl-em l
OUALITY @ INLET ' 0.0
24
'"
0
"
16
0
PRESSURE • U otm
"'ASS VELOC ITY· 13 q / l-em l
QUALITY @ I NLET '-0.07
:::l
}
QUALITY@INLET • -0.07
8
6
8
Fig. 2.Typical wall temperature profiles for various inlet conditions (pressure : 1.1 and 1.5 atm).
409
Forced Convection Heat Transfer to Subcritical Helium I
7 ,5
16
PRESSURE · 1.4.'m
MASS VELOC ITY' 63 Q I. - em'
QUALITY @INLET • 0 .0
14
6.5
Q •
0 .602 W/em'
c
:.: 6 ,0
10
.J
.J
e
OJ
7.0
12
"
PRESSURE' 2 .'m
MASS VELOCITY· 13 Q/. · em'
QUALITY@ INLET · - 0.05 ,. - 0 .03
.J
.J
•
~
8
l-
5.5
~
I-
5.0
6
TIUlK
4
4 .5
q' 0 ,171
T IUlK
4 ,0
0
0
7.5
16
32
40
8
0
48
16
24
16
PRESSURE • 2 •• "'
MASS VELOCITY· 27· 28 Q/ . - em'
QUA L ITY
INLET .-0.11 '0 - 0,10
7.0
14
6.5
12
2 a.m
PRESSURE •
q " .0 13 W/e m\
10
j 8
I-
6
4.0
2
16
24
LID
32
40
OJ
~
~
OJ
,"
~~
.
•. ""o~
-
4
OJ
OJ
,q'0' 755~
.'
~
4.5
OJ
-
.
/
;
48
MASS VELOC I TY ' 47 Q I.- em'
QUALl TY@INLET • -0.3
®
o
40
32
LID
,,- -
:::1:;:=
q '
.
.
.
O~~I
. -
-. - - -
T BUL K
48
Fig. 3. Typical wall temperature profiles for various inlet conditions (pressure : 1.4 and 2 atm).
410
P. J. Giarratano, R. C. Hess, and M. C. Jones
KUTATELADZE COARELATION FOR
NUCLEATE POOL BOILING 8;
CRITICAL HEAT FLUX [ 3 ]
o
MASS VELOCITY
9 /1 -c:m l
o
•
6
..
3 - 15
26- 37
12-14
28-42
PRESS.
aim
1.1
2.0
Fig. 4. Heat flux q vs. temperature difference
Twan - Tbu1k in the nucleate boiling region
under forced flow conditions.
sharp rise is observed for the higher heat fluxes. This has been identified, with reference
to a vast body of literature on boiling heat transfer, as indicating a hydrodynamic
transition from wetted-wall, e.g., nucleate boiling, to dry-wall, e.g., film boiling, heat
transfer. Thus, at a particular point along the test section, there is a critical heat flux
above which the wall temperature rises quite steeply with heat flux.
Below the critical heat flux, heat transfer is very high, as evidenced by the wall
temperatures being nowhere more than about 0.3 K above the bulk fluid temperature.
Indeed, this behavior is strongly reminiscent of boiling heat transfer without forced
convection, i.e., pool boiling. Figure 4 relates the heat flux q vs. the difference between
the wall and bulk fluid temperatures I!J. T. The I!J. T plotted is an average of the I!J. T's
for all positions on the tube having an LID> 20, to preclude possible entrance
effects. Furthermore, since the wall temperature profile is so flat in this nucleate
boiling region and since the bulk temperature is not changing, such a plot gives an
indication of the average heat transfer in the nucleate boiling region. The nucleate
pool boiling correlation of Kutateladze [3] has been plotted as solid lines for 1.1
and 2.0 atm. At the upper extremities of the solid line, the critical heat flux for pool
boiling, given by a second correlation due to Kutateladze [3], has also been indicated.
As shown, for example, by Brentari et al. [4], these correlations give a good average
representation of nucleate pool boiling data for cryogenic fluids. The present data
for forced convection heat transfer are represented by the first of these correlations
below the critical heat flux as well as any given set of pool boiling data, which are
notorious for their sensitivity to the precise nature and preparation of the boiling
surface, as well as its heating history (a strong hysteresis effect is often observed).
The lack of reproducibility in the present data in this region is quite in character
with nucleate pool boiling data and has apparently masked any trend with mass
velocity. It is concluded that forced convection has had no significant effect on the
rate of heat transfer below the critical heat flux and that the boiling mechanism itself
is the primary determinant.
Above the critical heat flux, wall temperatures are quite high and are now a
strong function of the mass velocity in addition to heat flux and pressure. At 1.1 atm,
the wall temperatures rise to unacceptable values from the point of view of cooling
of superconductors even at the highest mass velocities. At 2 atm, however, it is clear
Forced Convection Heat Transfer to Subcritical Helium I
411
that the temperature excursion above the critical heat flux has reached a limit which
is of some practical value and at the highest mass velocity (47 g/sec-cm 2 ), it is even
possible for the whole test section to be in film boiling below 8 K, with a heat flux
of 0.755 W/cm 2 . Even at 1.4 atm and 63 g/sec-cm 2 , a very useful heat flux can be
supported while wall temperatures remain in a useful range. It is noticed, too, that
when film boiling is well established throughout a good portion of the test section,
wall temperatures tend to fall again toward the outlet. This may be due to an increase
in velocity as more vapor is generated as the fluid moves down the section.
It has not been possible to find a suitable correlation for providing a unique
representation of the film boiling heat transfer coefficient even when attention has
been confined strictly to data for fully developed film boiling. The latter was defined
as data for which the given thermometer was downstream of the maximum wall
temperature. Correlations which were investigated for subcritical hydrogen [4] could
not represent the available data to better than an rms deviation of about 100%. It
has therefore been necessary to rely for the present on the data alone to describe the
film boiling region for helium.
The primary interest in the present work has been in the nucleate-to-film boiling
transition itself. The point of transition on the test section is quite clearly indicated
on the temperature profiles to the nearest wall thermometer station. For this purpose,
a point is taken midway between two adjacent thermometers where one shows no
significant temperature rise, and therefore presumably is still in nucleate boiling,
while the next thermometer does indeed show a significant rise. In this way, seventyfour data points have been recorded for the point of transition. The goal has been
to provide a means of predicting the transition, that is: Given the state of the fluid
at the entry to the heated section, i.e., enthalpy and pressure, the flow rate, and the
heat flux, it becomes desirable to determine at what position downstream the
transition will occur. Alternatively, given the state and flow rate, one may wish to
determine the heat flux-the critical heat flux-at which the transition will occur
at a certain position. The observed trends are presented below, while in the next
section the rationale for a correlation is discussed and the correlations that represent
the data are described.
The first and most obvious trend is that, for a given pressure and mass velocity,
the critical heat flux is a function of distance from the inlet to the heated section. In
experiments on heat transfer to subcritical helium under conditions of natural
convection, Johannes and Mollard [5] found that most of their data could be represented by a single curve when the critical heat flux was plotted against the distance
from the inlet to the heated section in diameters (the equivalent diameter De was
used for rectangular cross section, where De is equal to four times the ratio of the
cross-sectional area to the heated perimeter). The equation which represented their
data was
(1)
where qcr is the critical heat flux, L is the tube length at which transition occurred,
and De is the equivalent diameter based on the heater perimeter. Under conditions
of natural circulation, the mass velocities obtained by Johannes and Mollard were
from 2 to 3 g/sec-cm 2 . For the lowest mass velocities (4.5 g/sec-cm 2 ) at 1.1 atm pressure,
the critical heat fluxes in this study are only about 10% higher than given by (1).
412
P. J. GIarratIuIo. R. C. Hea, ... M. C. J _
However, a cursory examination of the data showed that there are considerable
departures which may be summarized as follows:
1. Critical heat fluxes may be as much as a factor two or three times higher
when mass velocities are in the range 30 to 60 glsec-cm 2 •
2. Critical heat fluxes may be as much as a factor oftwo lower at 2 atm pressure.
3. Critical heat fluxes are enhanced by subcooling the liquid helium at the inlet
to the test section.
There are thus at least five interdependent variables which characterize the transition:
the heat flux, the thermodynamic state (two variables), the mass velocity, and the
position in the tube.
Critical Heat Flux Correlation
Two factors have influenced the approach to the correlation of the transition
data in this study. The first is that considerable simplification of the problem is
possible if a local hypothesis is adopted which assumes that the transition occurs at
a point in the tube under conditions of local hydrodynamic similarity. The second
is the success which has been achieved for a very wide range of liquids (from helium
to water) in correlating the pool boiling transition on the basis ofthe similarity criterion
derived first by Kutateladze [3]. This criterion,
(2)
was derived from an analysis of the equations of motion of the two phases and
equations for the dynamic interaction at the vapor-liquid interface when turbulent
transport predominates. In (2), A is the latent heat of vaporization, (1 is the surface
tension, Pv and p, are vapor and liquid densities, respectively, and g is the gravitational
acceleration. The essential point in the derivation of (2) is that in pool boiling, the
only scale of the velocity is the quantity q/PvA and there is no scale for the pressure
difference between phases. The point of view has been taken that in forced convection
boiling, a well-defined scale of velocity also exists in the direction of the tube axis,
as does a pressure drop. Further, an added parameter is required to complete the
hydrodynamic description, namely the void fraction, or, equivalently, the quality
x and the ratio pJpv. It is then straightforward to show that similarity in forced
convection boiling can be specified by the following groups:
Here G is the mass velocity or mass flow per unit area. Now the ratio p, AP/G 2 for a
given L/D is determined by a Reynolds number defined as Re = GD/Jl." where Jl.,
is the liquid viscosity. Noting also that the first and second groups are the Froude
number and the Weber number, respectively, we arrive at the following equivalent
set of dimensionless groups:
{Fr, We, Re,
x, pJpv, KU}
This is as far as the local hypothesis takes us and the above groups should be
adequate provided the point under consideration is remote from any flow-perturbing
geometry, e.g., the entrance to this heated section. This would be indeed a nonlocal
effect. Now because some of the transitions were observed as close as three diameters
413
Forced Convection Heat Transfer to Subcritical Helium I
distance from the entrance, the group 1 + (D/L) was included, where L is again, as
in (1), the length up to the transition point. Since in the absence of forced convection
we must again recover the criterion (2), the following form suggests itself:
KU
=
C1
+ C2[FrC3WeC4(l + D/Lf'(l -
x)C6(PI/pvf7ReC8 ]
(3)
We have replaced x by 1 - x to avoid a negative term in the subcooled region.
Using nonlinear least squares fitting techniques [6] and the helium properties data
given by McCarty [7], it was possible to determine the constants in (3). The experimental data of Johannes and Mollard [5], with the equivalent diameter De based
on the wetted perimeter instead of the heated perimeter, were used to supplement
the data in this study. Those data provided low mass velocity input (natural circulation
with mass flow rates of order 2g/sec-cm 2). The standard deviation for the fit was
0.017 in units of KU, which was approximately 19% for an average value of KD.
The constants obtained are listed in Table I.
Table I. Constants for Equation (3)
C1
= 0.04
c2 =
C,
C6
0.36
C 3 = 0.40
C 4 = essentially zero
C7
C8
=
=
=
=
-0.73
2.72
0.11
-0.30
Due to the complexity of (3) and because some of the exponents were quite
small, suggesting some redundancy, several variations of (3) were tried in which terms
were removed. A good compromise between simplicity and goodness of fit was
achieved with the expression
(4)
where C 1 = 0.031, C 2 = 0.078, and C 3 = 3.92, and the standard deviation was 0.02
(approximately 22% for an average of KU).
In arriving at a final selection of a correlation, three data points of Jergel and
Stevenson [8] and two of Keilin et al. [9] were included in the evaluation. The latter
are particularly significant, because in their work, these authors observed transition
under rather different circumstances; their L/D was 278 and the transition occurred
at a quality of about 0.4 (e.g., in this study, the L/D < 48 and x < 0.2 for similar
pressures and mass velocities). It was found that some variants of (3) fitted to the
IJ.
I.l4I
1.111
0
0
. ,....u, _.u._ m
II
0
.
IW.Sllft
v rOld I SI""," II]
Fig. 5. Dimensionless critical heat flux
KU = q../Jep// 2 [Ug(PI - Pv)]l/4 vs.
correlating parllmeter (1 - X)3.92.
u.
l.ih .1 ' . I!]
1.1. 0 tall
I.IH
I.ltl
(I _I)'"
1.111
1 .•
414
P. J. Giarratano, R. C. Hess, and M. C. Jones
data in this study and those of Johannes and Mollard were quite unable to predict
the Keilin et al. data to better than an order of magnitude, whereas (4) came within
5%. Equation f4) and all the data discussed are plotted in Fig. 5
To illustrate more pointedly the effects of mass velocity pressure and inlet
conditions on critical heat flux, Table II has been prepared for a 10 cm long x 0.2 cm
ID tube using (4) in the calculations. It is apparent that, for a given pressure, increasing
the mass velocity from natural convection rates (2 g/sec-cm 2 ) to reasonable forced
convection rates (50 g/sec-cm 2 ) results in a threefold increase in the critical heat flux.
Furthermore, for a given pressure, the effect of increased subcooling at the inlet also
results in an increased critical heat flux. The maximum benefit from subcooling is
obtained at 2 atm, where it is seen that for an inlet temperature of 4 K and mass
velocity of 50 g/sec-cm 2 , the critical heat flux is approximately three times that at
saturated inlet conditions. Note that at 2 atm, (4) predicts that transition can occur
at negative values of x, representing subcooled helium. Such a circumstance has been
observed in these studies; it is possible to have subcooled film boiling under forced
convection conditions. Table II can be taken as a reasonable representation of the
experimental results on transition of this and other work.
Comparison of Modes of Heat Transfer
In Fig. 6, the present results of heat flux q as a function of wall temperature rise
~ T for L/D = 48 are represented by shaded areas for comparison with other modes
of heat transfer. The calculated values of the nucleate pool boiling correlations of
Kutateladze and the calculated values of q vs. ~ Tfor supercritical helium heat transfer
at 3 atm and at mass velocities of 10,30, and 50 g/sec-cm 2 have been included. These
were computed from
(5)
where Nu is the Nusselt number, Re is the Reynolds number, Pr is the Prandtl
number, Tw the inside wall temperature, and TB is the bulk fluid temperature. The
Table II. Example of Variation of Critical Heat Flux and Critical Quality, at LID of 50,
with Pressure, Mass Velocity, and Inlet Conditions [Using Equation (4)J*
P
G=2
= 1 atm
P
=
P
1.6 atm
= 2atm
50
G=2
5
30
50
G=2
5
30
50
Saturated inlet conditions
0.15
0.20 0.37
q.,
0.73 0.38
0.12
x.,
0.41
0.08
0.12
0.73
0.15
0.38
0.29
0.12
0.32
0.08
0.07
0.64
0.09
0.34
0.17
0.1 1
0.19
0.07
Inlet temperature of 4 K
0.15
0.20
0.40
q"
0.69 0.35
0.08
x"
0.46
0.04
0.13
0.53
0.19
0.46
0.55
0.21 -0.07 -0.13
0.09
0.15
0.47
0.62
0.19 -0.09 -0.35 -0.42
Inlet subcooling of 0.224 K
0.15
0.20
0.40
q.,
0.69 0.35
0.08
x"
0.46
0.04
0.12
0.65
0.17
0.31
0.08
0.50
* L = 10 cm,
quality.
5
30
0.34
0.04
D = 0.2 cm; G = mass velocity, g/sec-cm 2 ; q"
0.40
0.00
= critical
0.11
0.25
0.30
0.21 -0.05 -0.10
heat flux, W/cm 2 ; x.,
= critical
415
Forced Convection Heat Transfer to Subcritical Helium I
10
/ G' 10 gis -em'
..,J -"
I.lotm,
G . 27 - 37q/!. o(;m 1
N
...E
.....
~
atm .
G :012 · 15;lsocm 1
0 ,1
~
FORCED CONVECTION NUCLEATE BOILING
DATA OF THIS STUDY,
FORCED CONVECTION FILM BO ILING
DATA OF TH IS STUDY.
- - KUTATELADZE CORRELATION [3) FOR
NUCLE ATE POOL BOILING 8 CR ITIC AL
HEAT FLUX.
- - - - - FORCED FLOW OF SUPERCRITICAL
HELIUM@ 4 ,2 K. PRESS. 3 aim USING
CORRELATION OF REF, [ I ) ,
~
.01
0 ,1
I
t.T. K
10
100
Fig, 6, Comparison of various modes of helium heat transfer.
above correlation was developed in an earlier study [1] and rerresented the data of
that reference (LjD = 20 and 40) with an rms deviation of 8.5%.
The authors conclude from this comparison that for heat fluxes between 0.01
and 0.3 Wjcm 2 , subcritical pressures are preferable to supercritical pressures for
equivalent mass velocities in forced convection. Above these heat fluxes, up to about
1 Wjcm 2 and at pressures near 1 atm, super critical heat transfer is superior to subcritical, which is now in a dry-wall regime and shows a very steep temperature rise
for a small heat flux increment However, at subcritical pressures approaching the
critical pressure (e.g., 2 atm) and the highest flow rates observed, the heat transfer
rates are better or equivalent for the whole range of heat flux (0.01 to 1 Wjcm 2 ).
Note that even though critical heat fluxes at 2 atm are relatively low, compared to
1 atm, the temperature excursion after transition is moderate if a reasonable flow
rate is used (see last wall temperature profile in Fig. 3).
Finally, judging by the slopes of the curves in Fig. 6, at heat fluxes beyond the
range of the present data (> 1 W jcm 2 ), supercritical heat transfer should be superior.
CONCLUSIONS
1. The usual transition from wetted-wall to dry-wall heat transfer is observed
in heat transfer to subcritical helium I under forced convection.
2. Below the transition, forced convection has little effect on heat transfer and
the pool boiling correlation of Kutateladze for the heat flux vs. temperature
rise of the wall is sufficient.
416
P. J. Giarratano, R. C. Hess, and M. C. Jones
3. Above the transition, the heat transfer coefficient falls off drastically. For
superconductivity applications and low operating pressure (e.g., 1 atm) this
will usually have to be avoided except at mass velocities of the order of
60 glsec-cm 2 or higher, but may be tolerated if higher operating pressures
are used (e.g., 2 atm).
4. Transition is a function of many variables (see discussion of correlation) but
may be predicted by a relatively simple function of quality [equation (4)]
which implicitly contains those variables.
5. At heat fluxes between 0.01 and 0.3 W/cm 2 , subcritical heat transfer is
superior to supercritical for the same mass velocity. Above about 0.3 W/cm 2 ,
the reverse is true, for an operating pressure of 1 atm, and for an operating
pressure of 2 atm, the heat transfer rates are comparable.
ACKNOWLEDGMENT
This work was begun under sponsorship of the U. S. Atomic Energy Commission (Contract Agreement
AT(49-2)-1165) and completed with support from the Department of the Air Force, Wright Patterson Air
Force Base (Contract No. MIPR FY 14557200411).
REFERENCES
I. P. J. Giarratano, V. D. Arp, and R. V. Smith, Cryogenics, 11 :385 (1971).
2. H. Sixsmith and P. J. Giarratano, Rev. Sci. Instr., 41: 1570 (1967).
3. S. S. Kutateladze, Heat Transfer in Condensation and Boiling, State Sci. and Tech. Pub. of lit. on Machinery, Moscow (1952); (AEC translation 3770, Tech. Info. Service, Oak Ridge, Tennessee).
4. E. G. Brentari, P. J. Giarratano, and R. V. Smith, "Boiling Heat Transfer for Oxygen, Nitrogen, Hydrogen, and Helium," NBS Tech. Note No. 317 (September 1965).
5. C. Johannes and J. Mollard, in: Advances in Cryogenic Engineering, Vol. 17, Plenum Press, New York
(1972), p. 333.
6. C. Daniel and F. S. Wood, Fitting Equations to Data, John Wiley and Sons, New York (1971).
7. R. D. McCarty, "Thermophysical Properties of Helium-4 from 2 to 1500 K with Pressures to 1000
Atmospheres," NBS Tech. Note No. 631 (November 1972).
8. M. Jergel and R. Stevenson, Appl. Phys. Letters, 17(3): 125 (1970).
9. V. E. Keilin, E. D. Klimenko, and I. A. Kovalev, "Apparatus for Measuring Hydraulic Resistance and
Heat Transfer in the Two-Phase Flow of Helium," RTS 5062, National lending Library for Science
and Technology, Boston Spa, Yorkshire (1968).
DISCUSSION
Question by P. Thullen, Massachusetts Institute of Technology: What is the definition of quality x
used?
Answer by the author: The following definition of quality x was chosen so that the state of the fluid
in the subcooled region, as well as the two-phase region, could be designated by quality:
x% =
__ h.=----_h-='a=I..:.::liq""U=.id_
hlal . vapor - hsat.liquid
where h. is the enthalpy at the local pressure. From the above definition:
o< x <
x=
x >
x =
x <
I in the two-phase region and has the conventional physical meaning of the ratio of mass
vapor to mass mixture;
I for saturated vapor;
I for superheated vapor;
0 for saturated liquid; and
0 for subcooled liquid.
K-2
VAPORIZATION ONSET HEAT FLUX FOR
FLAT PLATES IN SATURATED LIQUID
HELIUM n
D. W. B. Mathews
National Defence Headquarters
Ottawa, OntariQ, Canada
and
A. C. Leonard
Royal Military College of Canada
Kingston, Ontario, Canada
INTRODUCTION
It is well known that liquid helium can exist in two different phases, He I and
He II. Helium I, the normal liquid phase, exhibits properties similar to other liquefied
gases. Helium II, the superfluid phase, possesses some very unusual characteristics,
not the least of which is the extremely high thermal conductivity of primary concern
here.
This unusually high heat transport capacity of He II has led to its more recent
probable use as a coolant for superconducting magnets used in linear accelerators
and magnetic levitation trains where the stability characteristics of the superconductor depend directly on the ability of the coolant to transfer heat. For this
reason, it is highly desirable to maintain the liquid in the vapor-free thermal counterflow regime of small temperature differences and not allow the film boiling regime
to be reached, because of the drastic increase in heater surface temperature. Specific
knowledge of the magnitude of the heat flux q"",x required to initiate film boiling is
therefore essential.
Experimental qmax data in He II to date have been restricted to electrically
heated cylinders or small tubes and channels. In the present work, it has been found
possible to extend this investigation to flat plates by depositing a very thin layer of
pure gold onto one face of a flat Pyrex plate and then heating the gold electrically.
THEORETICAL CONSIDERATIONS
Heat flow phenomena in He II have been explained by means of the two-fluid
theory. In this theory, two independent velocity fields are proposed to exist within
the bulk liquid, which permits heat to be transported equally to all areas of the liquid
when a temperature gradient is established. Using this concept to obtain theoretical
expressions for heat flow phenomena, numerous researchers have measured the
417
D. W. B. Mathews and A. C • .:-ant
418
maximum heat flux qmax for cylinders and channels immersed in liquid He II [1-10]
and compared the results with the two-fluid model expressions for these quantities.
The maximum heat flux data obtained using horizontal cylinders have consistently
proven to be dependent on the depth of immersion, diameter, and fluid temperature
I;, [9-12]. For channels, Chase found that channel length and orientation were
independent of the critical heat flux [5] and Vinen, in developing his mutual friction
concept [8], assumed that the channel walls themselves played no part in the generation,
maintenance, or decay of superftuid turbulence.
Several correlations have also been made by recent researchers for qmax data
obtained with channels and tubes in He II, based on the hydrodynamic principle of
dynamic similarity proposed by Staas et al. [3]. Their basic equations involve a
dimensionless Reynolds number
(1)
and a dimensionless pressure gradient parameter
Rep = pr3 VP/4'1/
(2)
n,
Frederking and Schweikle
Clement and Frederking [1], and Haben [12] all
proposed correlations using modifications of these groups but could obtain valid
relationships only within limited temperature ranges and heater sizes.
Haben et al. [9] further modified the dimensionless groups by utilizing thermodynamic scaling laws to relate heat flow in insulated channels to that from cylinders.
They reasoned that temperature and pressure gradients occurred primarily within
a coherence distance e(T) of the disordered layer around the heater. They thus
obtained a dimensionless driving force
N vp = (grad P)p 2 D 3 /'1n 3
(3)
and a dimensionless transport rate
(4)
where
j
= q/SI;,v n
(heat flux density)
(5)
(thermodynamic scaling laws)
(6)
and
pJp
=
eo/e
Using a reference distance eo of the order of
cm in the equations, they
obtained three general power-law approximations of the form
10- 8
(7)
based on the dependence of q* on size. In these cases, 0 < m < ! and the superscript
asterisk on q represents value at qmax.
This provided an excellent correlation of the qmax data in the temperature range
from 1.75 to 2.15 K but did not account for the temperature dependence of the
coherence length in the temperature range below 1.75 K or above 2.15 K.
Leonard and Clermont [14] recently extended these correlations by rearranging
the scaling laws as
e
(8)
Vaporization Onset Heat Flux for Flat Plates in Saturated Liquid Helium II
419
and introducing a further dimensionless term Do/D to account more completely for
changes in diameter, where Do is a normalizing diameter set equal to unity, to give
(9)
where
(10)
and
(11)
Equation (9) provided a correlation for a wide variety of cylinder sizes and
depths of immersion for the temperature range 1.1 to 2.17 K.
The success of the two-fluid hydrodynamics in predicting and correlating qmax
results for cylindrical heaters leads logically to the assumption that it will also provide
a successful model for other heater shapes. For a plate heater, the coherence distance
~ is very small and the flow can be assumed to resemble channel flow. It should,
therefore, be possible to obtain a correlation of qmax results for this heater configuration using the analysis and dimensionless groups of Leonard and Clermont [14] with
minor modifications. For this configuration, the groups are exactly similar to (10) and
(11), except that
D = A/2(W
+ L)
(12)
where D is now the limiting plate dimension perpendicular to the direction of thermal
counterflow-in this case, the hydraulic diameter. Thus, once again a correlation of
the form
(13)
should be obtainable.
APPARATUS AND EXPERIMENTAL PROCEDURE
The test specimens, which are the same as those designed and used by Merte [18],
were made of Pyrex approximately 0.438 cm thick coated on one side with a 50-nm
layer of gold. Over each end of the gold, a layer of silver was evaporated to form a
surface more suitable for electrical contacts. This left a gold surface 2.231 cm wide by
2.515 cm long.
A detailed description of the plate heaters with a heat transfer analysis is given by
Mathews [17]. This analysis predicts that better than 99.9 % of the heat generated, at
the time the qmax data are taken, passes from the gold surface to the helium bath.
Considerable difficulty was encountered over a period of several months in
maintaining electrical contact with the silver and gold film. Several types of metal
clamps were constructed to keep lead-in wires in contact with the silver layer. However, the difference in thermal expansion coefficients of the materials caused a break
in contact when the plate was subjected to large temperature changes. In the process
of trying different soldering techniques, it was found that, principally due to its low
melting point, indium could quite readily be used to solder the electrical leads
directly to the silver film without loss of contact. The joint so formed proved capable
of withstanding repeated cycling of power without any appreciable breakdown. A
detailed description of the apparatus and the experimental procedures are presented
elsewhere [17].
D. W. B. Mathews aad A. C. Leoaanl
4110
RESULTS AND DISCUSSION
Data have been obtained using a flat Pyrex plate coated with a 50-nm layer of
pure gold (variation in the gold thickness was less than 0.04%) mounted in three
orientations in a He II bath-O° or horizontal facing up, 90° or vertical, and 180° or
horizontal facing down. These results are shown in Figs. 1 through 3 as a function of
bath temperature.
It is clearly evident that, in keeping with the results of Chase [5] for enclosed
channels, there has been no appreciable change in the qmax data for the different
orientations as there is in conventional fluids such as He I [15]. The slightly lower
values of qmax for the lower depths in the vertical orientation resulted from the fact
that the vertical centerline of the plate was used to determine the liquid level in this
orientation, leaving the upper half of the plate closer to the He II surface. In all other
orientations, the gold surface was the marker.
As expected, there is also a noticeable influence of the heater immersion depth,
similar to the results found with horizontal cylinders. The plate data presented for the
three orientations can be extrapolated to obtain a value for the maximum peak heat
flux density q:'ax of approximately 1 W cm
/ 2 at the hypothetical limit of zero depth of
immersion. This is consistent with the results of other researchers, presented in a
review by Frederking [16], where D (diameter) can be considered approaching infinity.
The effect of surface width was observed by splitting the plate in half, lengthwise.
The results are plotted in Fig. 4 and show an average increase in q:'ax of approximately
11 %. This is once again consistent with the data plotted by Frederking 6 ] in which
q:'ax approaches the hypothetical limit of 1 W/cm 2 asymptotically as D increases.
In Fig. 5, q:'ax for the flat plate is compared with cylindrical heaters for depths
of immersion of up to 70 cm. The plate data appear to be compatible with what would
be expected for a cylinder of infinite diameter, as mentioned above.
A correlation which included all the qmax data for plate heaters taken at this
time was made using the modified dimensionless groups of Leonard and Clermont
e
+
4
..
o
"
. .
60 eM
.
30 eM
20 CM
10
5
eM
CM
..
.
o
0
".
0
0
A'1
f!>.
"tP-
+
"0"
00
'Ij.
+
o
t.
HORIZONTAL POS ITION - FACING UP
THICKNESS
;;:r;
50
nm
O+---~----~--~----r----r----.----.--~----,----,----~
1.1
l2
1.3
1.4
1.5
1.6
1.1
T FLUID • I(
1.8
1.9
Fig. I. Maximum heat flux vs. bath temperature.
2.0
2.1
Vaporization Onset Heat Flux for Flat Plates in Saturated Liquid Helium II
...
.
60 CM
0
20 CM
Ii>.
10 eM
+
4
N
30 CM
2
5
v
3
"-
II)
I-
o
l
x
~
0 000
0
o
+....
* .. *
. .
.
eM
+
0
421
.
*
0
0
2
rr
Ii>.
o
o
0
0
VERTICAL
POS ITION
o+----r--_,----~--_r--_,r_--~--_r----._--._---.----,
1.1
12
1.3
1.'1
1.5
1.6
T FLU ID .
1.7
1.8
K
1.9
2 .0
2.1
2..2
Fig. 2. Maximum heat flux vs. bath temperature.
[14]. The data from cylinders and plates from this study listed in Table I were plotted
in the form In(Re v) - (Dol D)0.2 vs. Rep and a straight line relationship was determined
by inspection, as indicated in Fig. 7. The relationship so formed was best given by
(14)
where once again Do is a normalizing diameter and is set at 1 cm. In this case, D is
the limiting plate dimension perpendicular to the direction of counterflow and represents the hydraulic diameter.
5
.
60 eM
0
20 CM
Ii>.
10 eM
+
4
N
2
v
....
~
0
30 CM
5
+
It
CM
.
0
0
Ii>.
0
66
0
2
rr
o
Ii>.
0
6
Ii>.
.
. .. .
0
~
3
i
i
.
0
HOR IZONTAL POSITION - FACING DOWN
o+---~----~---r----.----r----._---r---.r---._---.----.
1.1
1.2
1.3
1.'1
1.5
1.6
1.1
T FLUID. K
1.8
1.9
Fig. 3. Maximum heat flux vs. bath temperature.
2.0
2.1
2.2
_ _ . D.W " B M athews and A .C. Leonard
422
.. + + ...
...
...
...
60 CM
30 CM
0
20 e M
A
10 eM
Q
5 Clol
o
.
..
0
..
o
.. "
0 0
..
...
...
* .. *
... "'" ..
-- ..... + + ...
...
o .. +
o ..
o
~o
0+
0 0000
A
A
0*'+
6
A AA
A
o
2
o
o
0
0
"""..
01.1;-----,--1.2
POSITION -
1.3
CYLINDER
o
A<4...
o o
l>Q
o
o
1.4-:----.---
r,.C;
0
"b
"\
HALF
0
\>
It
SIZE
PLATE
1.9--;;:--~-
T F'16
1.7
Fig 4 '1. 5
LUIO
K
I.B
. . Maximum h
eat flux vs. bath t emperature.
2.1
2.0
DIAIoiET ER -e M
•
l!!.
o
+
10
20
30;0--'-'--'- DEPTH OF IMMERSION.
40
eM
F'
•
. on q:;'ax
Ig. 5. Effect of d epth ofimm erslOn
'
*
- 2.2
0
423
Vaporization Onset Heat Flux for Flat Plates in Saturated Liquid Helium II
15
DEPTH OF IMMERSION - H
I TO 70C M
LIMITING HEATER DIMENSION - D
( PLATE DATA ONLY)
. 38 TO .~. CM
12
He!!
BATH
TEMPERATURE - Tb
1.2" TO
2.17"
9
CORRELATION
3
O+--r--.-~-.--.--r--.-~-.--.-~--r--r~--.-~--r--r--'-~
o
4
9
12
16
20
24
29
32
36
40
In (Rep'
Fig. 6.
qmax
data correlation.
Figure 6 illustrates the above correlation for the plate data of the present work
and includes approximately 600 data points. It should be noted that while giving
excellent correlation, the hydraulic diameter does not give a fair indication of the
actual plate size. The plates used had the dimensions 1.091 cm by 2.515 cm, to give
a hydraulic diameter of 0.38 cm, and 2.231 cm by 2.515 cm, to give a hydraulic
diameter of 0.59 cm.
15
DEPT H OF IMMERSION. - H
I TO 70CM
12
LI MITING HEATER DIMENSION - 0
.0017 TO . ~9 CM
HoI
BATH TEMPERATURE- Tb
1.1 K TO 2. 17K
9
"!
81 0
I
>.
tl! 6
CORRELATION
(!12).2
InIRev.' ~ In [ e 0
3
Rep .35
J
0~0~'--'4'--'--~9--'--"2---'--~~--'--'20---'--2~4--'--'28---'--3r2--'-~36~-'--'
40
In (R ep )
Fig. 7.
qmax
data correlation.
424
D. W. B. Mathews and A. C. Leonard
Table I. List of Researchers
Investigator
T,K
Haben [12]
Holdredge and McFadden [19]
Madsen and McFadden [20]
Goodling and Irey [21]
Vinsen, Agee, Manning,
and Hereford [22]
Lemieux and Leonard [II];
Leonard and Lady [13]
Present workt
D,*cm
0.OOI7-{).OO81
1.16-2.15
0.145,0.242
1.80-2.10
1.96,2.10
0.179
2.142, 1.802
0.567
1.4-2.12
0.0090
Immersion
depth, cm
0.5-20.9
4.2,11
13,15.2
4-32
19-19.95
1.1-2.17
0.OO25-{).0635
1-70
1.2-2.17
0.38,0.59
5-70
• D represents diameter in all cases except the present work, where it is the hydraulic
diameter.
t The present work represents qm•• data obtained with plates of two different sizes.
In Fig. 7, the correlation of (14) has been superimposed on that of Leonard and
Clermont given by (9), for a total of approximately 1800 data points. Clearly, the
plate data complete the correlation with an addition to the upper end of the curve
and provide a more accurate fit for a greater variety of heater sizes, depths ofimmersion, and bath temperatures with a new slope of 0.35. Table I lists the researchers
contributing data to this correlation and includes the range of pertinent variables.
CONCLUSIONS
Data collection of over 600 points taken with flat gold plates of two different
sizes has proven qmax to be dependent upon bath temperature, depth of immersion,
and heater surface area and independent of orientation, so that
qmax =
f(T", H, D)
(15)
A correlation has been obtained using the dimensionless groups of Leonard
and Clermont [14] for cylinders and including all the data taken with flat gold plates.
The correlation obtained is
(14)
where
Rev" = qmaxD/ST,,"n
(10)
Rep = (pD 3 /"n 2)(pST,JPvhf,)(pgH)(p.lPn)2
(11)
and D is the limiting heater dimension which, for plates, is the hydraulic diameter.
ACKNOWLEDGEMENTS
The research for this paper was supported (in part) by the Defence Research Board of Canada, grant
number 9510-37. Thanks are extended to H. Merte, Jr. from the University of Michigan, for providing the
specifications and source required for the production of the gold-faced heater plates.
Vaporization Onset Heat Flux for Flat Plates in Saturated Liquid Helium II
425
NOTATION
A
D
D
g
H
hI.
L
P
qrnax
q:;'ax
Rep
Re""
r
S
T"
Vn
W
= surface area of heater
= limiting heater dimension, diameter for cylinders and channels, hydraulic diameter for rectangular
plates
normalizing diameter, 1 cm
acceleration due to gravity
= immersion depth
= latent heat of vaporization
= length of rectangular heater
= pressure
= maximum heat flux (required to start film boiling)
= highest value of qrnax at a given depth
= dimensionless pressure gradient parameter
= dimensionless Reynolds number
= radius of cylinder heater
= entropy
= temperature of He II bath
= velocity of normal fluid component
= width of rectangular heater
=
=
Greek Letters
I7n
=
~
~o
=
=
P
Pn
Ps
p,
=
=
=
=
normal fluid component viscosity
coherence length required to establish thermodynamic equilibrium at a given bath temperature
coherence length required to establish thermodynamic equilibrium at absolute zero
total density of bulk fluid
density of normal fluid component
density of superfluid component
vapor density
REFERENCES
I. B. W. Clement and T. H. K. Frederking, in: Proc. of llR Commission I, Intern. Inst. Refrig., Annexe
1966, p. 49.
2. R. C. Chapman, Y. W. Chang, and T. H. K. Frederking, in: Advances in Cryogenic Engineering, Vol. 15,
Plenum Press, New York (1969), p. 290.
3. F. A. Staas, K. W. Taconis, and W. M. van Alphen, Physica, 27:893 (1961).
4. E. C. Alcaraz and H. H. Madden, Phys. Rev. A, 1(1):49 (1970).
5. C. E. Chase, Phys. Rev., 131(5): 1898 (1963).
6. J. T. Tough, W. D. McCormick, and J. G. Dash, in: Advances in Cryogenic Engineering, Vol. 15,
Plenum Press, New York (1970), p. 287.
7. T. H. K. Frederking and J. D. Schweikle, in: Proceedings 9th Intern. Low Temperature Physics Conference, Part A, Plenum Press, New York (1965), p. 251.
8. W. F. Vinen, Proc. Roy. Soc., A242:493 (1957).
9. R. L. Haben, R. A. Madsen, A. C. Leonard, and T. H. K. Frederking, in: Advances in Cryogenic
Engineering, Vol.ll, Plenum Press, New York (1972),p. 323.
10. M. A. Clermont, M. Eng. Thesis, Royal Military College of Canada, Kingston, Ontario (1971).
II. G. P. Lemieux and A. C. Leonard, in: Advances in Cryogenic Engineering, Vol. 13, Plenum Press, New
York (1968), p. 624.
12. R. L. Haben, M.S. Thesis, University of California, Los Angeles, California (1967).
13. A. C. Leonard and E. R. Lady, in: Advances in Cryogenic Engineering, Vol. 16, Plenum Press, New
York (1971), p. 378.
14. A. C. Leonard and M. A. Clermont, in: Proceedings 4th Intern. Cryogenic Engineering Conference,
Iliffe Sci. and Tech. Pub!., London (1973), p. 301.
15. M. Jergel and R. Stevenson, Intern. J. Heat and Mass Transfer, 14: 2099 (1971).
16. T. H. K. Frederking, "Thermal Transport Phenomena at Liquid Helium II Temperatures," Chem.
Eng. Progr. Symp. Series, 64(87):21 (1968).
17. D. W. B. Mathews, M. Eng. Thesis, Royal Military College of Canada, Kingston, Ontario (1973).
18. H. Merte, Jr., University of Michigan, Ann Arbor, private communication.
19. R. M. Holdredge and P. W. McFadden, in: Advances in Cryogenic Engineering, Vol. II, Plenum Press,
New York (1966), p. 507.
426
D. W. B. Mathews and A. C. Leoaard
20. R. A. Madsen and P. W. McFadden, in: Advances in Cryogenic Engineering, Vol. 13, Plenum Press,
New York (1968), p. 617.
21. J. S. Goodling and R. K. frey, in: Advamces in Cryogenic Engineering, Vol. /4, Plenum Press, New York
(1969), p. 159.
22. J. S. Vinson, F. J. Agee, R. J. Manning, and F. L. Hereford, Phys. Rev., 168: 180 (\968).
K-3
HEAT TRANSFER TO SLUSH HYDROGEN
c. F. Sindt
Cryogenics Division
NBS Institutefor Basic Standards
Boulder, Colorado
INTRODUCTION
The Cryogenics Division of the National Bureau of Standards has been involved
in a study of the characterization of slush hydrogen for several years. This analytical
and experimental program has been directed toward the investigation of characteristics that apply to the use of slush hydrogen as a propellant for rockets. Characteristics previously investigated include those of preparation, flow, pumping, aging,
solid particle configuration, and instrumentation. Characteristics common to most
applications that have not been investigated or have been studied very little are
those of mixing and heat transfer. This presentation covers heat transfer experiments
in a small laboratory apparatus.
EXPERIMENTAL APPARATUS
The experimental apparatus for the heat transfer experiments consisted of three
glass dewars nested together. The outer dewar was filled with liquid nitrogen at its
normal boiling temperature. The second dewar was filled with liquid hydrogen which
was maintained at the test liquid temperature. The inner dewar was the experimental
space and was filled with liquid or slush hydrogen, depending on the desired test
conditions.
Heat Transfer Unit
The heat transfer unit selected was similar to one used by Coeling and Merte [1].
It consisted of a 2.54-cm-diameter cylindrical block of electrolytic tough pitch copper.
This cylinder was 1.9 cm long and was drilled in six equally spaced places to accept
six carbon resistors which were used as the heat source. The carbon resistors were
embedded in a high-thermal-conductivity epoxy and were connected in parallel
configuration. The electrical resistance of the unit was 13.86 n at 20 K. A schematic
of the heater unit is shown in Fig. 1.
The surface of the heater exposed to the hydrogen was formed by a 0.05-mmthick, stainless steel sheet which was soldered to the face of the copper block opposite
that drilled for resistors. The stainless steel sheet and the copper block were tinned,
then mechanically pressed together as they were heated above the melting temperature
of the solder. The stainless steel sheet extended radially 3.1 mm beyond the copper
cylinder, where it was soldered to a hollow, stainless steel cylinder which was used
as a vacuum jacket. The vacuum was pumped continuously during the test with a
427
c. F. Sindt
428
,,,,.
luttl
n.......
lI.,tU'11II
hftum
I[Kl
Fig. I.Heat transfer unit.
diffusion pump. The vacuum jacket provided thermal insulation for the heating unit
except at the heat transfer surface.
Instmmentation
At 0.25 mm below the heat transfer surface, two thermocouples were embedded
in a low-melting-temperature metallic eutectic. One thermocouple was located at
the center ofthe heated surface; the other was at a radius of 0.96 cm and was centered
between two of the resistance heaters. An array of twelve thermocouples extended
outward from the external heated surface, spaced at approximately l.1-mm intervals.
Four of the thermocouples were located at the centerline, four at about one-half of
the heater radius, and the other four at the heater edge. The thermocouple array is
shown in the heater schematic in Fig. 1. The thermocouples were all type KP vs.
gold (0.07 at. % iron).
All thermocouples in the array and in the heater were referenced to a thermocouple located at the bottom of the test vessel. In this location, the absolute temperature was known for each test.
The power to the heater was provided by a dc power supply which maintained
voltage constant within ± 0.2%. The voltage at the heater and the voltage drop
across a precision resistor in series with the heater were measured to calculate the
power supplied to the heater. Pressure was measured with an absolute pressure
mercury manometer and was maintained constant within ± 135 N/m2 using a barostat.
TEST PROCEDURE
Four types of tests were conducted using three orientations of the heater surface.
The four test types were: (1) heat transfer at 1 atm pressure in liquid at the normalboiling-point temperature, (2) heat transfer at the triple-point pressure in liquid
hydrogen, (3) heat transfer at the triple-point pressure in settled slush hydrogen
(estimated solid fraction of 0.45), and (4) heat transfer in settled slush hydrogen at
1 atm pressure using helium as the pressurizing gas. The three orientations of the
surface were horizontal facing up, vertical, and horizontal facing down. Test procedure did not vary for the three orientations of the heater.
The test procedure with the normal-boiling-point liquid was initiated by bringing
all of the liquid to equilibrium at 1 atm pressure. The liquid was mixed until bubbles
formed at the lower mixer blade and did not collapse while rising to the surface. At
this time, all of the thermocouples were read to establish a zero-point offset. Power
was then supplied to the heater in increasing increments. The thermocouple data
and the power were recorded while the pressure in the vessel was maintained constant.
Heat Transfer to Slush Hydrogen
429
Heat was increased from 0.002 W/cm 2 to the point where the boiling regime changed
from nucleate to film (burnout). The heating rate was increased by increments to
the maximum, then decreased to determine hysteresis effects.
The procedure for tests with triple-point liquid was to pump the dewar to near
triple-point pressure and maintain this pressure without forming solid. During the
remainder of the test, triple-point pressure was maintained; otherwise, the test was
the same as for the normal-boiling-point liquid.
The procedure for heat transfer to slush at the triple-point pressure was to prepare
slush in the experimental dewar using the freeze-thaw method; the dewar was filled
with settled slush. Triple-point pressure was then maintained with the barostat.
Heat was increased and decreased as for the liquid tests, but burnout was not defined,
since the solid in the slush could not be maintained long enough to determine the
actual burnout heat rate with any certainty. Therefore, due to this apparatus limitation, the maximum heat rate used in the slush was arbitrarily selected as that used
in the liquid.
For the slush pressurized to 1 atm pressure, the procedure was similar, except
that after slush preparation and prior to adding heat, the pressure was raised to
1 atm by introducing helium gas precooled to the triple-point temperature of hydrogen. The pressure was maintained during the test. This test was always run last in
the series so that liquid, used in the other tests, was not saturated with helium gas.
The slush was not mixed after pressurization with the helium gas, so as to keep the
amount of helium desolved in the hydrogen to a minimum.
For both slush tests, the slush had to be replenished frequently, thus interrupting
the increasing or decreasing heat rate. To maintain consistency after slush preparation,
the heat rate was always increased from a lesser value for increasing heat flux tests
and decreased from an arbitrary larger value for decreasing heat flux tests. During
the higher heat flux tests, slush had to be prepared prior to each test point to assure
adequate slush depth over the heater during the period required to take data. Although
visibility in the slush is poor, prohibiting good observation, the slush appeared to
settle around the heating unit as melting occurred.
TEST RESULTS AND DISCUSSION
At least two separate runs were made for each type of test and heater position
to check for repeatability of tests and consistency of data. The temperature difference
between the two thermocouples in the heater was monitored to assure that the heater
surface was at a uniform temperature.
All of the test data were reduced to the parameters of power per unit of area
vs. the temperature difference between the bulk liquid and the heater surface. The
power per unit of area was calculated from the power supplied to the heater and the
diameter of the heater block. A correction was made for the heat loss through the
stainless steel fin formed by the extension of the sheet from the copper block to the
vacuum jacket. A correction was also made to the temperature difference for the
temperature drop from the copper block to the stainless steel surface exposed to the
liquid. Both corrections were relatively small. Curves fit to the data for each type of
test and test configuration are shown in Figs. 2 through 4. Actual test data points
are shown for some of the test conditions to illustrate typical data scatter. The L\ T
on the figures is the difference between the bulk fluid temperature and the corrected
heater wall temperature. The NBP (normal boiling point) and TP (triple point)
C. F. Sindt
430
'0
LII~ .... d
001 H',
SI",,,tJ Clot TP
~ 0 I
~
..
t-
LoJ
I
00'
o
ltqulod 01 NBP , Runl
LIQUIQ 01
BP,Run 2
o
CO@I"",q . Mer It!
SOIIO symbOl" IndiCa Ie deCteOS'"9
MO' flux
t.
Fig. 2. Heat transfer of hydrogen,
horizontal surface facing up.
T, K
'0
N~
~
>< 0'
~
...
.J
..
t-
W
I
001
o
6
0.
o
Slus h 01 TI> Prusure, Ryn ~
Slu!50h 01 TP Pressure. Aun 2
Liquid 01 Nep, Run'
Liquid at Bt:' ,Run 2
$01113 symbOlS Indicate deCfeO$I"q
heal flux
o 0OOL,__.l--'-...J.-L-'--.l-'-':I-\;o:---'-...J....J.-'-..LJ~~'o;---'---'-'...J.--'--~,~oo
Fig. 3.Heat transfer of hydrogen,
vertical surface.
431
Heat Transfer to Slush Hydrogen
10
/
/
/
/
/
/
/
I
I
"Ls~Liquid 01 TP,
I.
>-
<I
'"
:J:
SIUlh 01 TP
/
~ 01
...J
u.
I
/
/
/~
0
o Slush at TP Pressure (Run 1
b.
Slush at TP Pressure (Run 2)
o Liquid at TP Pressure
Fig. 4. Heat transfer of hydrogen,
horizontal surface facing down.
to T, K
conditions referred to on the figures are the pressures in the ullage during the test.
The data of Coeling and Merte [1] for a pressure of 117 kN/m2 (878 Torr) are
compared in Fig. 2 with data obtained in this study when the heater was in the upward
facing position. The agreement is good. Coeling and Merte did not show a hysteresis
effect for decreasing heat flux for a polished surface; their data for a rough surface
did indicate a hysteresis similar to that shown for normal boiling liquid. No hysteresis
was found in the other three types of tests for this heater position.
The point where the curves make a sharp break coincides with the time that
vapor bubbles were first observed at the heater surface, except for the slush pressurized
to 1 atm. In this case, vapor bubbles were not visible; however, they may have formed
and collapsed after leaving the surface and not have been observed, since visibility
in slush is poor. Vapor sites were apparent during the decreasing heat flux portion
of the normal boiling liquid tests. Three sites were still apparent at the lowest point.
Data for triple-point liquid and slush at triple-point pressure were not significantly
different when the heater was in the facing-up position.
When the boiling became vigorous in slush at triple-point pressure, mixing of
the slush resulted. The mixing began at a heat flux of 0.5 W/cm 2 and became vigorous
at 0.8' W/cm 2. Mixing resulted from the vigorous bubble action in the boiling liquid.
Since no bubbles persisted in the pressurized slush case, no apparent mixing occurred.
A significant difference exists between the heat transfer data for the heater
facing up and the heater oriented in the vertical position; hysteresis was present in
both triple-point liquid and slush at triple-point pressure with the heater in the vertical
position, while the hysteresis was absent with the heater facing up. The data for
22
0
0
,
8 '
0
20
,
E
0
0
I
.. ?
I
H .. .
0 12 (Ro 14/ 1
""l>I
~
=
0
~
HU'O 55 1Ro )
~
I
""f
18
z
'"
'I'
z
I
V
'C
..,.
,
00 I
00 0
I
i
,
I
I
2
'"
0
...J
16
I.
o
"
"
"
----
LH, 01 NBP
Slush 01 NBP
LH, 0 1 T P
S lush at T P
Ja .ob ( 1949) for air,
water, and al coho l
12.~----~----~----~~----±-----~9~----~'O~--~
Fig. 5. Natural convection heat transfer
to hydrogen, horizontal surface facing
up.
LOG ,O ' Rayl ei gh NumbN
22 ,-----,------r-----.,-----,------r-----,----- .
I
I
I
I
20
I
..
~z
.'"
I
I
";1
"
I
,"
I
~ /~
o
18
N
H.
. O I2 IRo )
I
.. 1
tI. /v
~
z'"
~
I
\I
'"o
I
o
16
...J
,
I
I.
I
I
I
I
I
I
I
,Ii
- 0551 Ro l
r/
I
",
o LH z 01 NBP
o
..
v
-- - -
Slush 01 NBP
L Hz 01 T P
Slush 0 1 TP
Jakob 11949) Inr oor ,
waf er , and o:cobo l
12.!------t------!-6----__!------±-- -----!c-----l,0'c-------111
LOG,o ' Ray leigh Number
Fig. 6. Natural convection heat transfer
to hydrogen, vertical surface.
433
Heat Transfer to Slush Hydrogen
22 r-- --r----,----r---,r---.----,----,
I
I
I
1
1
1
20
1
1
1
o
<;
&l
~
16
z
I
~
II.
:>
N••
Z
o
0
<5
0
-'
16
I
I
14
I
"055(Rol - I
I
I
I
I
I
1
I
1
" '"
I
I
I
I
I
I
1
"
4
v
v v
o
LH z 01 NBP
Sl ush 01 NBP
v Slush al TP
---- Jakob (1949) for air.
wa l er. and alcohol
'"
Fig. 7. Natural convection heat transfer
to hydrogen. horizontal surface facing
down.
LOG,o • Rayle iqh Number
slush of 1 atm pressure were nearly the same. Visual observation of the vapor formation with the heater vertical revealed that the vapor sites always formed at the top
of the heated surface first and progressed down across the surface as the heat flux
was increased.
Another characteristic that is apparent in the data for the vertical surface and
for one set of data for the horizontal surface facing up is the second break in the
curve for normal boiling liquid. This break occurred at the time vapor sites covered
the entire surface. In the vertical position, this occurrence was repeatable and the
data were consistent. In the horizontal facing-up position, the vapor site formation
was not consistent. On several occasions, sites appeared all across the surface simultaneously and vapor bubbles were very small and appeared as a cloud (data shown
as circles in Fig. 2). At other times, the vapor sites produced large bubbles and
increased in number gradually as heat flux was increased (data shown as diamonds).
This difference in bubble formation probably accounts for some difference in heat
transfer rate, although no large differences were observed.
The heat transfer characteristics in the heater facing-down position were significantly different than for the other two positions. As shown in Fig. 4, the heat transfer
to liquid at normal boiling temperatures is much larger at lower temperature differences. Vapor sites formed at the lowest heating rates, and a single large bubble grew
until it escaped over the edge of the heater unit. Vapor sites did not form at the lower
heating rates in liquid at triple point and slush at triple-point pressure. The large
discontinuity in these curves coincided with the first visible vapor formation. The
434
C. F. Sindt
discontinuity in the data occurred because bubbles formed and covered the
surface before escaping instead of escaping from vapor sites as in the other heater
orientations.
The characteristics of pressurized slush for the facing-down heater position
were similar to characteristics of other orientations, but the heat flux was less for
the same temperature difference. However, the break in the curve occurs at a lower
temperature difference, and the temperature difference for a given heat flux is less
from this point on.
The heat transfer in the free convection regime was examined further by comparing it to heat transfer in other fluids. The comparison was made using the dimensionless parameters of Rayleigh and Nusselt numbers, with the diameter of the heated
surface used for the characteristic dimension. The data were compared to data and
to equations given by Jakob [2] for heat transfer to cylinders, vertical planes, spheres,
and a block in the fluids, air, water, alcohol, and oil. The results are given in Figs. 5
through 7. The curve by Jakob is a curve fit to the data presented in Jakob for air,
water, alcohol, and oil. The Nusselt numbers from the data for the horizontal surface
facing up are larger than those predicted by the curve from Jakob [2]; however,
Jakob suggests an increase in Nusselt numbers from 28 to 100% for horizontal
facing-up surfaces as compared to the surfaces represented by the curve. The triplepoint liquid and slush data fit the curve quite well for the vertical orientation. In the
horizontal facing-down position, none of the data fit the curve or the equations. The
data for normal-boiling-point temperature liquid were not expected to fit the corre-
TP Pr ..... Uf.
10
"'5
0
....
0
3:
X
:::>
-'
"-
.....
c
0
0
01
0
00
<oJ
I
001
o
o
LI QU i d 01 Nap, Run 1
LiqUid 01 N8P, Run. 2
L iqU Id at T P Pressure
Slus.h al T P Pr'"@';Uure
- - - - - - Kul Qlelodze
- ~ - BOr"lshanslllY end Michenko
- - Brenlor l el 01 (1965 )
OOO,L-~~~~~~~~-LLL~~~~-LLLUD
0'
, 0
l> T. K
'0
'00
Fig. 8. Correlations for nucleate boiling
heat transfer to hydrogen.
Heat Transfer to Slush Hydrogen
435
lations because vapor was forming during all of the data points; therefore, the fluid
was not in a free convection state. For horizontal surfaces facing down, Jakob suggests Nusselt numbers as small as half of those for horizontal surfaces facing up.
Nusselt numbers for the data shown in Fig. 7 are approximately one-half of those
for the facing-up position shown in Fig. 5.
Heat transfer in the nucleate boiling regime was also compared to heat transfer
in other fluids. The comparison was made using several different equations for predicting boiling heat transfer to water. These equations are given by Kutateladze [3.4].
The comparison with heater facing-up data is given in Fig. 8. A similar comparison
by Brentari et al. [5] was made for hydrogen heat transfer data. The correlations of
Kutateladze [4] and Kichigin (from Kutateladze [4]) agree with the normal boiling
liquid data; however, these correlations do not fit the triple-point liquid data. All
of the correlating equations predict too much effect on the temperature difference
for the pressures of 1 atm and triple point.
The only tests in which more than 1-deg temperature gradient developed in the
hydrogen (at the first thermocouple in the array over the heater) was in slush pressurized to 1 atm. The fourth thermocouple in the array never did indicate a significant
difference from the bulk temperature.
CONCLUSIONS
The natural convection heat transfer to liquid hydrogen at normal-boiling-point
temperature can be predicted quite accurately for horizontal surfaces fac