AN ABSTRACT OF THE THESIS OF
Florian Kapsenberg for the degree of Master of Science in Mechanical Engineering presented
on May 27, 2011.
Title: Lateral Fluid Motion in Nucleate Boiling Through Asymmetric Surface Structures
Abstract approved:
________________________________________________________________________
Vinod Narayanan
This thesis presents a feasibility study of a means to passively effect liquid motion parallel to
a heated surface though surface geometrical modifications. Such a passive system is
beneficial for electronics cooling applications as it reduces the pumping equipment normally
required in flow loops and is desired for space applications, where launch costs greatly
restrict the weight of onboard systems. The surface geometry considered was a repeated
array of asymmetric silicon ratchets with reentrant cavities located on the shallow face. A
serpentine thin film heater provided heat to the surface. A complete experimental facility in
which to perform experiments was designed and constructed as part of this work to comply
with requirements set by microgravity flight services in anticipation of future experiments in
microgravity.
Experiments were performed using deionized and degassed water at atmospheric pressure
at subcoolings of 5°C and 20°C. Applied area averaged heat flux was varied between 2 W/cm2
and 19 W/cm2 at these conditions. Magnified high-speed videos were used to resolve bubble
behavior near the surface. A preferential bubble growth and departure direction was
confirmed for both subcoolings and lateral fluid motion was confirmed in the high subcooling
condition. Repeatability was confirmed with separate experiment performed 58 days later.
Tracking of bubbles was accomplished using a custom bubble-tracking algorithm, designed to
resolve only bubbles within a two-dimensional plane normal to the viewing direction.
Instantaneous velocities of individual bubbles parallel to the surface were shown to be in
excess of 600 mm/s immediately following departure, and liquid flows with mean velocities
between 25 mm/s and 35 mm/s parallel to surface were observed in the plume farther from
the surface. A simplified semi-empirical model of bubble growth phase is proposed to explain
the observed mean liquid velocities.
©Copyright by Florian Kapsenberg
May 27, 2011
All Rights Reserved
Lateral Fluid Motion in Nucleate Boiling Through Asymmetric Surface Structures
by
Florian Kapsenberg
A THESIS
submitted to
Oregon State University
in partial fulfillment of
the requirements for the
degree of
Master of Science
Presented May 27, 2011
Commencement June 2011
Master of Science thesis of Florian Kapsenberg presented on May 27, 2011
APPROVED:
________________________________________________________________________
Major Professor, representing Mechanical Engineering
________________________________________________________________________
Head of the School of Mechanical, Industrial, and Manufacturing Engineering
________________________________________________________________________
Dean of the Graduate School
I understand that my thesis will become part of the permanent collection of Oregon
State University libraries. My signature below authorizes release of my thesis to any
reader upon request.
________________________________________________________________________
Florian Kapsenberg, Author
ACKNOWLEDGEMENTS
I express my sincere thanks to my advisor Dr. Vinod Narayanan for providing excellent
guidance throughout my research, for providing meaningful feedback, and for always
remaining calm when things seem to take a turn for the worst. I greatly appreciate the
freedom Dr. Narayanan has provided me in the design, construction, and experimental
phases of my research. His expertise and understanding was crucial to the successful
outcome of this project.
I would like to thank Dr. Sushil Bhavnani and Naveenan Thiagarajan at Auburn University for
their collaboration on this project and for their hard work in producing the silicon wafers that
made this project a success. Without their tireless efforts this project would not have been
possible.
I owe great thanks to Dr. James Liburdy and Dr. Nancy Squires for providing research
opportunities and the inspiration to study Thermal-Fluid Sciences during my undergraduate
career at Oregon State University. Dr. Squires’ Thermodynamics course was truly one of my
favorite at OSU, and I am grateful for having had Dr. Liburdy as my senior project advisor and
for having provided me with the opportunity to work at the wind tunnel facility.
I thank my family for supporting me all these years, my girlfriend Agata Szewczuk for her
support and, along with my friends David McIntosh, Kevin Harada, Jesse Grimes, and Joe
Hortnagl, for providing the periodically necessary distractions from academic work.
Lastly, I thank my lab mates Ruander Cardenas, Preeti Mani, David Haley, Mohammad
Tabatabaee, and Mohammad Ghazvini and all my student colleagues in the Thermal-Fluid
Sciences group for helping me along the way.
Financial assistance for the project was provided by NSF under grant number 0854503 and by
NASA under grant number NNX09AJ98G.
TABLE OF CONTENTS
Page
1. INTRODUCTION ............................................................................................................................................ 1
2. LITERATURE REVIEW ................................................................................................................................ 5
2.1. The Pool Boiling Curve ....................................................................................................................... 5
2.2. Forces on a Bubble............................................................................................................................... 8
2.3. Pool Boiling in Microgravity .......................................................................................................... 12
2.4. Pool boiling using augmented surface geometries............................................................... 14
2.5. Conclusions on Literature Review .............................................................................................. 20
3. SCOPE .............................................................................................................................................................. 21
3.1. Objectives and Tasks......................................................................................................................... 21
4. EXPERIMENTAL FACILITY ..................................................................................................................... 23
4.1. Existing Microgravity Simulation Methods and Facilities ................................................. 23
4.2. Test Section........................................................................................................................................... 28
4.3. Experimental Facility Requirements.......................................................................................... 30
4.4. Initial Chamber Design..................................................................................................................... 31
4.5. Final Chamber Design....................................................................................................................... 33
4.6. Experimental Facility........................................................................................................................ 56
5. EXPERIMENTAL PROCEDURES ............................................................................................................ 66
5.1. Startup Procedure .............................................................................................................................. 66
5.2. Continuous Operating Procedure ................................................................................................ 73
5.3. Shutdown Procedure ........................................................................................................................ 75
6. DATA ANALYSIS .......................................................................................................................................... 77
6.1. Determination of Bubble Kinematics ......................................................................................... 77
6.2. Calibration and Uncertainty Analysis ........................................................................................ 85
TABLE OF CONTENTS (Continued)
Page
7. RESULTS AND DISCUSSION.................................................................................................................... 98
7.1. Qualitative Visualization ................................................................................................................. 98
7.2. Bubble Kinematics .......................................................................................................................... 109
7.3. Repeatability ..................................................................................................................................... 121
7.4. Force Balance .................................................................................................................................... 130
8. CONCLUSIONS ........................................................................................................................................... 136
9. RECOMMENDATIONS FOR FUTURE INVESTIGATION............................................................. 138
REFERENCES .................................................................................................................................................. 140
APPENDICES................................................................................................................................................... 143
LIST OF FIGURES
Figure
Page
1.1
Cross-section of geometry proposed to impart a net momentum on the fluid
parallel to the heated surface. ...................................................................................... 3
2.1
Generic pool boiling curve. ............................................................................................ 6
2.2
Pool boiling curve for water under controlled heat flux conditions. ............................ 8
2.3
Forces on a vapor bubble attached to a heated surface. .............................................. 9
2.4
Force exerted on a bubble by the surrounding liquid as a result of bubble
growth. ........................................................................................................................ 12
2.5
The pool boiling curve in microgravity. ....................................................................... 14
2.6
Two-dimensional representation of the experimental setup used by Chyu
and Fei ......................................................................................................................... 15
2.7
Various slot geometries studied by Das et al. ............................................................. 16
2.8
Experimental results from Stroock et al. ..................................................................... 18
2.9
Leidenfrost effect on asymmetrically structures heated surface as
demonstrated by Linke et al. ....................................................................................... 19
4.1
Three orientations required to effectively demonstrate the flow's nongravitational dependence. ........................................................................................... 24
4.2
Configuration of the test section and thin film aluminum heater silicon wafer. ........ 29
4.3
Microscope image of a cavity on the test section with overlaid image of
calibrated optical grid. ................................................................................................. 30
4.4
Schematic of test section power and measurement circuit. ....................................... 30
4.5
Wafer holder and clamp designs. ................................................................................ 34
4.6
Transparent view of the wafer holder showing the infrared camera
microscope lens position in relation to the test section. ............................................ 35
4.7
Layout of the two flow loops implemented in the experimental facility. ................... 38
4.8
1D resistance network used to determine the minimum tube length required
for the condensing coils............................................................................................... 39
LIST OF FIGURES (Continued)
Figure
Page
4.9
Overview of condensing loop. ..................................................................................... 41
4.10
The chamber’s fill and recirculation loop with emergency overflow
protection. ................................................................................................................... 43
4.11
Overview of the components of the fill/drain flow loop. ............................................ 43
4.12
Required laser light sheet incidence angle on the test section as it depends
on the test section orientation. ................................................................................... 44
4.13
Selection of the laser light sheet prism. ...................................................................... 46
4.14
The laser light sheet as it is deflected by a prism in the upright orientation. ............. 47
4.15
The laser light sheet as it is deflected by a prism in the inverted orientation. ........... 47
4.16
The laser light sheet as it enters the chamber in the vertical orientation. ................. 48
4.17
Demonstration of the laser light sheet illuminating a bubble..................................... 49
4.18
Infrared camera mounted to custom support designed to interface it with
the chamber. ............................................................................................................... 50
4.19
Cutaway view of the infrared camera mounted to the chamber using the
custom support............................................................................................................ 51
4.20
Custom bracket designed to relieve the camera's lens mount adaptor from
the large cantilever weight of the K2/SC lens. ............................................................ 52
4.21
High-speed camera mounted to the chamber. ........................................................... 53
4.22
LED Array circuit. ......................................................................................................... 54
4.23
A cutaway view of the chamber showing the position of the LED array. .................... 55
4.24
The LED backlighting array. ......................................................................................... 55
4.25
Overview of the experimental facility. ........................................................................ 60
4.26
Static stress analysis performed one of two vertical frame members that
support the chamber. .................................................................................................. 62
LIST OF FIGURES (Continued)
Figure
Page
4.27
3-Axis Accelerometer used to monitor chamber accelerations and act as a
high-speed camera trigger........................................................................................... 63
4.28
High-speed camera triggering setup. .......................................................................... 63
4.29
The camera triggering transistor circuit functioning as a solid state relay. ................ 64
4.30
Schematic of electrical connections for the cartridge heater circuit and the
experimental facility systems. ..................................................................................... 65
4.31
Overview of various electrical systems in the experimental facility. .......................... 65
5.1
Installation of test section in chamber. ....................................................................... 68
5.2
Flowchart for startup procedures................................................................................ 72
5.3
Flowchart for operation procedures. .......................................................................... 74
5.4
Flowchart for shutdown procedures. .......................................................................... 76
6.1
Stretching the imported image histogram over the full grayscale spectrum.............. 78
6.2
Image processing steps to resolve bubbles. ................................................................ 79
6.3
Tagging of regions during image processing. .............................................................. 81
6.4
Bubble pair area ratio criterion applied. ..................................................................... 82
6.5
Velocity vector map obtained from bubble tracking velocimetry............................... 83
6.6
Estimated outline of bubbles as resolved by the filter. ............................................... 84
6.7
Use of thermocouple extension cables. ...................................................................... 87
6.8
Method of calculating the image resolution. .............................................................. 89
6.9
Vector distortion error in a uniform distortion field. .................................................. 90
6.10
Vector distortion error in a variable distortion field. .................................................. 91
6.11
Magnitude of pixel displacement magnitude due to lens distortion. ......................... 93
6.12
Gradient of pixel displacement magnitude due to lens distortion.............................. 94
LIST OF FIGURES (Continued)
Figure
Page
6.13
Results of the computation of the angular error in a vector (Eq. 6.11). ..................... 95
6.14
Optical grid placed over the test section to determine accuracy of ratchet
period spacing.............................................................................................................. 96
7.1
Sequence of 20 frames at low subcooling (Test 1). ................................................... 100
7.2
Sequence of 20 frames at high subcooling (Test 2). .................................................. 104
7.3
Microscope pictures of each cavity in the test section. ............................................ 108
7.4
Geometry for cavity 2a construed from a microscope image (Fig. 7.3). ................... 109
7.5
Bubble velocity field at low subcooling (Test 1). ....................................................... 111
7.6
Bubble velocity field at high subcooling (Test 2). ...................................................... 116
7.7
Sequence of 20 frames at low subcooling to demonstrate repeatability of the
experiment (Test 3) 58 days after Tests 1 and 2. ...................................................... 122
7.8
Sequence of 20 frames at high subcooling to demonstrate repeatability of
the experiment (Test 3) 58 days after Tests 1 and 2. ................................................ 126
7.9
Growth of the bubble normal to the surface imparts a net momentum on the
fluid. ........................................................................................................................... 131
7.10
Control volume for estimation of liquid plume velocity as it depends on the
momentum imparted on the fluid during one bubble growth cycle ........................ 132
7.11
Preliminary reasoning for observed behavior of bubbles in subcooled
conditions. ................................................................................................................. 134
LIST OF TABLES
Table
Page
4.1
Overview of various methods of creating a microgravity environment and
common associated parameters ................................................................................. 25
4.2
Advantages and disadvantages of achieving the three-orientation capability
by means of a rotatable chamber. .............................................................................. 32
4.3
Material compatibility with 3M FluorinertTM Electronic Liquid (FC-72) ...................... 36
4.4
List of instruments employed in the experimental facility. ......................................... 56
4.5
List of equipment employed in the experimental facility. .......................................... 57
6.1
Uncertainty in any dimensional computation derived from images due to
uncertainty in ratchet spacing, pixel count, and lens distortion (maximum as
well as RMS)................................................................................................................. 97
7.1
Experimental conditions for the two tests performed in January (Test 1 and
Test 2), and the test performed in March (Test 3). ..................................................... 98
7.2
Bubble diameter during growth before departure. .................................................. 133
7.3
Bubble accelerations calculated to determine if Marangoni flow plays a role
in the strong acceleration of bubbles immediately after departure. ........................ 135
LIST OF APPENDICES
Page
Appendix A. Derivation of Marangoni Flow Reaction Force....................................................... 144
Appendix B. Experimental Procedures Specific to FC-72 ............................................................ 145
B.1. Cleaning the Chamber ................................................................................................................... 145
B.2. Filling and Degassing ..................................................................................................................... 145
Appendix C. Bubble Tracking Algorithm ............................................................................................ 146
Appendix D. Data Analysis ........................................................................................................................ 147
D.1. Thermocouple Calibration and Uncertainty Analysis ..................................................... 147
D.2. Applied Heat Flux ........................................................................................................................... 151
Appendix E. Annotated Vector Plots for Tests 1, 2, 3 .................................................................... 152
Appendix F. Part Drawings ....................................................................................................................... 168
LIST OF APPENDIX FIGURES
Figure
Page
C.1
Flowchart for bubble tracking algorithm functions. .................................................. 146
D.1
Calibration curves for each thermocouple. ............................................................... 149
E.1
Labeled vector plots for Test 1. ................................................................................. 153
E.2
Labeled vector plots for Test 2. ................................................................................. 158
E.3
Labeled vector plots for Test 3. ................................................................................. 163
F.1
Assembly drawing of chamber. ................................................................................. 169
F.2
Drawing of chamber wall with laser port cutout and bolt flanges. ........................... 170
F.3
Drawing of chamber wall with wafer holder port cutout and bolt flanges. .............. 171
F.4
Drawing of bottom chamber wall with wafer holder cutout. ................................... 172
F.5
Drawing of top chamber wall with wafer holder cutout. .......................................... 173
F.6
Drawing of side chamber wall with window cutout and thru-holes for
condensing equipment and cartridge heaters. ......................................................... 174
F.7
Drawing of side chamber wall with window cutout. ................................................. 175
F.8
Drawing of window clamp. ........................................................................................ 176
F.9
Drawing of circular port cover. .................................................................................. 177
F.10
Drawing of laser port used for experiments in the vertical orientation. .................. 178
F.11
Assembly drawing of laser port used for experiments in the upright and
inverted orientation. ................................................................................................. 179
F.12
Drawing of base flange for laser port. ....................................................................... 180
F.13
Drawing of side wall of laser port. ............................................................................. 181
F.14
Drawing of bottom wall for laser port. ...................................................................... 182
F.15
Drawing of top wall for laser port. ............................................................................ 183
F.16
Drawing of window mounting plate for laser port. ................................................... 184
LIST OF APPENDIX FIGURES (Continued)
Figure
Page
F.17
Bracket for mounting the rotational platform for the prism to the laser port. ........ 185
F.18
Drawing of wafer holder. ........................................................................................... 186
F.19
Drawing of wafer clamp. ........................................................................................... 187
F.20
Drawing of arms used to support the high-speed camera. ....................................... 188
F.21
Drawing of high-speed camera mounting plate. ....................................................... 189
F.22
Drawing of high-speed camera lens support bracket. .............................................. 190
LIST OF APPENDIX TABLES
Table
Page
D.1
Thermocouple uncertainty measurements and calibration curve constants. .......... 148
D.2
Uncertainty of applied heat flux due to uncertainties in heated surface area
and applied power. .................................................................................................... 151
NOMENCLATURE
]
acceleration [ ⁄
area [ ]
drag coefficient
specific heat (constant pressure)
[ ⁄
]
bubble diameter [ ]
bubble contact diameter [ ]
distortion error [ ]
force [ ]
]
bubble departure frequency [
]
gravity (-9.81) [ ⁄
convective heat transfer coefficient
[ ⁄
]
latent heat of vaporization [ ⁄ ]
]
electrical current [
Jakob number
]
thermal conductivity [ ⁄
length [ ]
Marangoni number
⁄ ]
momentum [
vector length [ ]
heat flux [ ⁄ ]
heat transfer rate [ ]
bubble radius [ ]
Reynolds number
Stokes number
direction of distortion gradient
time [ ]
temperature [ ]
wall temperature [ ]
pool temperature [ ]
wall superheat [ ]
log mean temperature difference [ ]
pool subcooling [ ]
̇
uncertainty
velocity [ ⁄ ]
width [ ]
electrical potential [ ]
volumetric flow rate [ ⁄
]
Greek Symbols
ratchet angle [ ]
prism tilt angle [ ]
laser beam deflection angle [ ]
difference
contact angle [ ]
wavelength [ ]
⁄ ]
dynamic viscosity [
density [ ⁄ ]
]
surface tension [ ⁄
]
shear stress [ ⁄
prism angle [ ]
resolved vector angle [ ]
true vector angle [ ]
internal bubble angle [ ]
Subscripts
buoyancy
contact pressure
drag
bubble growth
liquid
Marangoni
maximum
root-mean-square
saturation
test section
vapor
LATERAL FLUID MOTION IN NUCLEATE BOILING THROUGH
ASYMMETRIC SURFACE STRUCTURES
1. INTRODUCTION
With advances in electronics towards smaller scales, the heat dissipation requirements per
unit area are increasing dramatically. This large increase in heat flux requires ever more
ingenious heat dissipation solutions. In the past this problem has been solved first by moving
from natural convection to forced convection through the use of fans, followed by heat sinks
and their associated development, then by single-phase liquid cooling in jet impingement,
spray impingement or microchannel cooling configurations.
For larger heat fluxes, there exists the possibility of implementing phase-change heat
transfer, such as evaporation or boiling, in the cooling application. Depending on the
operating surface temperatures, fluids with appropriate boiling point can be selected. The
advantage of phase-change heat transfer is that rather than removing heat from the system
by sensible heating of a fluid, the phase of the working fluid is converted from liquid to vapor
thermodynamically, at a constant saturation temperature. In most cases, the amount of
energy required to convert liquid to vapor is very high (for water at atmospheric pressure the
heat of vaporization is 2257
⁄
compared to a specific heat of 4.217
⁄
[1]),
thereby resulting in reduced inventory of fluid and smaller pumps for the same heat transfer
rate compared with single-phase liquid cooling methods.
Phase change heat transfer systems that include boiling can either consist of pool boiling or
convective boiling systems. Pool boiling systems rely on a stagnant pool to dissipate the
thermal energy; bubbles are generated at the liquid-solid interface and are transported to
the liquid-vapor interface through buoyancy forces. Secondary motion in the liquid phase is
induced by natural convection and the movement of bubbles. Convective boiling systems aim
to increase the intensity of fluid motion by directing an inflow of the liquid phase across or
onto the heated surface [2]. There are a variety of convective boiling systems, ranging from
jet-impingement boiling where a liquid jet is directed onto the heated surface, to spray
impingement boiling where a liquid spray is directed onto the heated surface, and
microchannel boiling, where flow is directed through a heat sink with microscopic channels
in the heated surface.
2
The underlying trend in these ever more intricate heat removal techniques is that although
they become more effective, they also become more bulky, heavy, and complex to
implement. Any sort of forced liquid convection system requires pumps, additional plumbing,
and the associated instrumentation and control. The cost of these systems is often also a
limiting factor. From a reliability and cost point of view, it is therefore desirable for there to
exist heat removal solutions for high flux cooling that would contain no moving parts and
would ideally be self-controlling.
A significant advance in heat transfer was made when such a passive phase-change system
was first discovered in the form of heat pipes [3]. A heat pipe consists of a sealed tube that is
heated at one end and cooled at the other. Evaporation takes place at the hot end and the
vapor travels to the cool end where it is condensed. The condensed liquid is then transported
back to the hot end by capillary forces induced by a porous wicking liner in the inside walls of
the tube. This system is highly effective as it requires no external power and does not require
any moving parts, but is limited by the pressure drop in the wick for high heat fluxes. Other
types of heat pipes are the looped heat pipe, and pulsating heat pipe [4].
The weight and complexity associated with convective heat dissipation solutions is even
more problematic for space applications, where a simple and lightweight system is highly
desirable in order to keep costs down. For this reason forced convection solutions are not
ideal. For example, the pathfinder spacecraft featured a single-phase mechanically pumped
cooling loop; the pump assembly alone weighed about 8 kg [5]. At the estimated launch cost
per pound at that time this equates to an estimated launch cost of between $70,593 and
$238,427 for the pump alone, depending on the configuration of the Delta II launch vehicle
used [6]. Nevertheless, the high cost of forced convective systems has not limited their use.
In terrestrial conditions, natural convection is supplied by gravity-induced buoyancy forces,
created by differences in the medium density caused by local heating. The microgravity
environment does not permit liquid cooling systems to rely on natural convection for
adequate heat transfer as buoyancy forces are no longer significant. Thermal management
systems that use boiling heat transfer are not currently used in microgravity systems either
since bubble departure from the surface by buoyancy is greatly reduced. A puzzling problem
3
now exists as well-studied means of high heat flux heat transfer and simple natural
convection are no longer practical in the microgravity environment. There consequently
exists a need for a new technology that has the simplicity of passive systems such as heat
pipes, and the heat flux capacity of two-phase forced convection systems.
The work discussed in this thesis aims to generate a non-gravity induced net fluid
momentum parallel to the heated surface solely by the application of heat during nucleate
boiling. Figure 1.1 shows a schematic of the surface geometry studied.
Figure 1.1: Cross-section of geometry proposed to impart a net momentum on the fluid
parallel to the heated surface. Reentrant cavities serve to lower the nucleation temperature
by maintaining a nucleation site and generate bubbles preferentially on one side of the
ratchet pattern.
The geometry studied is a 1 mm pitch ratchet pattern, with faces at angles of 30° and 60°,
and a pyramidal reentrant cavity in each 30° angled face. Heat is applied from below by a
thin-film heater and initiates nucleation inside the reentrant cavities, upon which the bubble
grows, protrudes from the cavity, detaches from the surface and leaves the surface at an
angle to the vertical. Small cavity sizes (on the order of 100μm) ensure small bubble
departure diameters, subsequently resulting in high departure velocities. The use of
reentrant cavities ensures consistent nucleation sites, reduced temperature overshoot for
initiation of boiling when compared to polished surfaces, and an increase in the heat transfer
coefficient by up to an order of magnitude as vapor remains trapped in the cavity after
bubble departure, precluding the need to reach nucleation temperature for subsequent
bubbles [1]. Additionally, when orienting these cavities preferentially in one direction, a net
4
flow is expected to develop during nucleate boiling. This type of thermally-driven flow would
have great advantages for space applications as it would not require a pump to move the
fluid.
The goal of this thesis work is to perform a proof-of-concept study of this design. A
significant portion of the effort entailed design and fabrication of an experimental facility
that could be used to test the concept under different orientations and in a microgravity
environment afforded by parabolic flight. The hypothesis that lateral motion could be
induced was tested using high-speed image analysis of bubble motion including
determination of bubble velocity, acceleration, and size of the individual bubbles departing
from the surface. Additionally, a model is proposed that describes the driving mechanism of
the induced flow over the structured surface.
5
2. LITERATURE REVIEW
The research presented in this thesis pertains to pool boiling from a novel design of a heated
surface with augmented surface geometry intended to generate a net fluid motion parallel to
the surface. Such fluid motion could result in development of pumpless flow loops for
terrestrial as well as space applications. Microgravity environments were of particular
interest since the reduction of buoyancy-induced natural convection would imply higher
component of fluid motion parallel to the surface. The literature review presented in this
section is consistent with the scope of the thesis and examines the basics of pool boiling,
microgravity pool boiling, pool boiling with augmented surfaces, and thermally actuated
pumping.
2.1. The Pool Boiling Curve
Boiling heat transfer is not only used for the purpose of removing heat from a surface, but
also to convert liquid into vapor in power generation cycles. The widespread implementation
of boiling heat transfer has generated a significant amount of research on the subject,
leading to among other things the development of the pool boiling curve. A boiling curve is
normally plotted as the variation of heat flux
with wall superheat (
) and
provides a convenient illustration of the performance of any particular boiling surface. Good
performance is considered to be high levels of heat flux at low wall superheat conditions (i.e.
as the curve moves upward and to the left, the heat transfer performance is increased).
Figure 2.1 shows a generic pool boiling curve as it applies to controlled wall superheat
conditions.
6
Figure 2.1: Generic pool boiling curve. All major boiling regimes are shown (plot adapted
from Incropera et al. [1]).
First and foremost, it is important to distinguish between the different pool boiling regimes
(characterized by the different ranges of wall superheat), as each is dominated by distinctly
different mechanisms. The pool boiling curve is commonly broken down into four distinct
regimes with characteristic behavior: free convection (up to point A in Fig. 2.1), nucleate
boiling (line AC), transition boiling (line CD), and film boiling (point D and beyond). Incropera
et al. [1] and Carey [8] provide a detailed description of each of these regimes. In the free
convection regime, few or no bubbles form on the surface and depart at a relatively low
frequency such that the flow over the surface, and thereby the heat transfer coefficient, is
dominated by natural convection. The nucleate boiling regime is characterized by the
continuous nucleation, growth, and departure of vapor bubbles from the heated surface.
This regime can be further broken into two subregimes: the isolated bubble regime (line AB
in Fig. 2.1), and the regime of columns and jets, known as the fully developed nucleate
boiling regime (line BC in Fig. 2.1). In the isolated bubble regime, the departure of bubbles
induces significant fluid motion near the heated surface, increasing the convection
coefficient. This increase in convection coefficient allows for a greater heat transfer rate for
7
the same wall superheat
. In the fully developed nucleate boiling regime, the widespread
coalescence of bubbles near the surface inhibits fluid flow, subsequently reducing the
convection coefficient (point B in Fig. 2.1). However, the increase in wall superheat
maintains the trend of the increasing heat flux until the reduction in convective effects is
balanced by the increasing wall superheat. The heat flux at this point is the maximum
obtainable heat flux in the nucleate boiling regime. Because of the high heat flux and low
required wall superheat associated with nucleate boiling, this is a desirable operating regime
for most engineering applications. Transition boiling occurs when
is increased beyond
this maximum value. When this occurs, nucleation happens at such a high rate that bubble
coalescence on the surface covers large regions of the heated surface with a vapor layer.
Because of the relatively low thermal conductivity of vapor, this further reduces the heat
flux. Increasing
even further results in more of the surface being covered in vapor until
the liquid no longer touches the surface and a stable vapor film forms. The temperature at
which this occurs is called the Leidenfrost point (point D in Fig. 2.1), beyond which film
boiling takes place. The dominant heat transfer mechanisms in film boiling are conduction
through the vapor layer, and radiation from the heated surface to the liquid-vapor interface.
When heat flux is the independent variable in the system (common for most experiments
and characteristic of heat generation in electrical systems), the maximum heat transfer rate
that occurs at the onset of transition boiling is called the critical heat flux (CHF), as any
further increase in heat flux would result in a rapid increase in surface temperature, rapidly
moving to the film boiling regime; this discontinuity can be shown by line EF in Fig. 2.2a – the
pool boiling curve for water subject to increasing controlled heat flux.
8
(a)
(b)
Figure 2.2: Pool boiling curve for water under controlled heat flux conditions. Plots are
shown for both during increasing heat flux (a) and decreasing heat flux (b) (plots adapted
from Carey [8]).
Likewise, reducing the heat flux during film boiling maintains the vapor film until reaching the
Leidenfrost point, upon which the flow abruptly transitions to nucleate boiling (line GH in Fig.
2.2b). This transition is similarly associated with an abrupt reduction in wall superheat. Such
a scenario could occur in quenching of metals for example. This behavior indicates that when
controlling the heat flux, the surface temperature exhibits hysteresis.
Hysteresis can also occur at the onset of nucleate boiling and is especially evident when the
fluid is highly wetting on the surface and/or the surface is very smooth with no cavities or
trapped vapor. For water on a smooth surface, a wall superheat as high as 17°C can be
necessary to initiate nucleation [9]. When controlling heat flux, once nucleation occurs, the
surface temperature abruptly decreases due to the increased heat transfer coefficient (line
BC in Fig. 2.2a). At this point, if the heat flux were to be reduced, the temperature would
follow a different path as nucleation has already taken place (line HA in Fig. 2.2b).
2.2. Forces on a Bubble
Before discussing the dynamic behavior of bubbles in pool boiling, it is important to review
the forces on them in a heated pool. It should be noted that three distinct phases exist for a
bubble: growth, departure, and post-departure. Figure 2.3 shows the forces acting during
9
bubble growth on a surface. The forces acting are the buoyancy force
between the bubble and the solid surface
force induced by Marangoni flow
, the surface tension
, the contact pressure force
, and the drag force due to bubble growth
, the reaction
. The forces
acting on a bubble after departure are the buoyancy force, the reaction force induced by
Marangoni flow, and the drag force caused by the motion of the bubble relative to the
surrounding fluid
. Each of these forces is explained in detail below.
Figure 2.3: Forces on a vapor bubble attached to a heated surface. The sum of the forces
dictates whether the bubble will depart from the surface or remain (diagram adapted from
Zhang and Chao [10]).
The most familiar force is the buoyancy force, a force induced by differences in hydrostatic
pressure around the bubble, which normally causes a bubble to rise to the surface when
subject to gravity. This force can be computed as the weight of the bubble less the weight of
the displaced fluid:
(
)
(2.1)
Whenever a bubble is in contact with a solid surface, it is subject to surface tension forces
that act on the bubble at the contact line, tangent to the bubble’s liquid-vapor interface
10
( )
(2.2)
Surface tension also causes the bubble’s internal pressure to be higher than the external
pressure. This higher internal pressure is imparted on the un-wetted portion where the
bubble is in contact with the solid surface, and is therefore called the contact pressure. The
force due to the contact pressure is modeled using the Young-Laplace equation:
(2.3)
In a pool with a non-uniform temperature, Marangoni convection, an interfacial convective
current set up through surface tension gradients, acts to push the bubble toward areas of
lower surface tension (higher temperature) [10]. Briefly, at a liquid vapor interface, any
surface temperature gradients cause a flow of liquid from regions of low surface tension to
regions of high surface tension. In any subcooled pool boiling condition, the fluid near the
heated surface is hotter than the pool and exhibits a lower surface tension than the fluid in
the surrounding pool. In the presence of vapor bubbles these surface tension gradients
generate a concentrated flow of liquid from the heated surface to the surrounding pool (Fig.
2.3). These surface tension-driven currents travel around the bubble, and create a force on
the bubble towards the surface, effectively trapping the bubble on the surface. The local
shear stresses at any angular location
due to Marangoni flow on a spherical bubble can be
modeled by the following equation as described by Wang et al. [11]:
(
)(
)
(2.4)
When assuming a linear vertical temperature gradient, integrating this equation over the
total surface area of the bubble gives
(
)(
)
(2.5)
Derivation of Eq. 2.5 is provided in Appendix A. In terrestrial conditions, Marangoni
convection has little effect on bubble departure for large bubbles as the large buoyancy
11
forces greatly mask such small effects; this can be shown by modeling the ratio of the force
due to Marangoni convection to the buoyancy force:
(
)
(
)(
)
Assuming properties of water at saturation temperature (
[
⁄
],
[
⁄
(2.6)
⁄
[ ⁄
],
]) and negligible vapor density, this ratio becomes
(
)
(2.7)
It becomes clear that low gravity, steep temperature gradients, or small bubble diameters
are necessary for the Marangoni convection forces to become important. For example, in
terrestrial conditions, with a bubble diameter of 10 mm and a temperature gradient of
2°C/mm, the ratio of Marangoni to buoyancy forces is -0.015, indicating the force due to
Marangoni flow is two orders of magnitude lower than the buoyancy force. Wang et al. [11]
observed that on a thin (25 or 100 µm) heated wire with bubble sizes smaller than 50 µm and
subcooling greater than 30°C, strong liquid jets with velocities on the order of 100 mm/s
were observed to emanate from the wire, occasionally reaching 600 mm/s. These jets were
observed following the collapse of a larger bubble and the associated forces induced by
these jets were strong enough to drive the bubble against the buoyancy force toward the
wire. For this case, using Eq. 2.4, and the conditions presented by Wang et al. [11], the force
due to Marangoni flow was computed to be 31 times greater than the buoyancy force.
There is also a drag force present whenever the bubble is in motion relative to the
surrounding liquid. This occurs both when moving through a pool, and growing on the
surface. To determine this force, the inviscid model is used as the vapor viscosity is much
lower than the liquid viscosity, rendering the no-slip boundary condition invalid. For a bubble
moving through a pool, the drag force is:
(2.8)
12
which is the equation for drag on a sphere. For a bubble growing on a surface, Zeng et al [12]
described a net force that was imparted on the fluid (Fig. 2.4). The reaction is a downward
force on the bubble that can be modeled as the drag the bubble experiences moving at a
velocity equal to the growth rate of the diameter
Figure 2.4: Force exerted on a bubble by the surrounding liquid as a result of bubble growth.
This force can be characterized as a drag force, referencing the diameter growth rate as the
bubble velocity.
(
)
(2.9)
which is merely Eq. 2.8 with the exception that the velocity has been replaced with the
growth rate of the diameter. This is a very basic form of the model; a number of other
correlations exist for this growth rate that account for the presence of a wall and the
hemispherical shape a bubble traditionally has during the initial (inertially controlled) growth
phase on a smooth surface.
2.3. Pool Boiling in Microgravity
The pool boiling curve described in Section 2.1 applies only to terrestrial conditions where
gravity serves to remove vapor from the surface primarily though buoyancy forces. The
complication that arises when performing pool boiling in microgravity is that once a bubble
nucleates and grows in size, the near absence of a buoyancy force impedes the bubble’s
13
ability to leave the surface. Fortunately, there have been numerous experiments in the field
of microgravity pool boiling, both under saturated and subcooled conditions. Most
commonly experiments were aimed at characterizing the behavior and dominant
mechanisms of two-phase flows in a microgravity environment. Such studies include
determination of the heat transfer coefficient, critical heat flux, bubble formation and
growth, and bubble coalescence.
Lee et al. [13] performed experiments with R-113 on a number of space shuttle flights and
observed that since buoyancy was no longer the driving force for bubble behavior in
microgravity, bubbles tended to remain on the surface, growing ever larger in size by
coalescing with nearby bubbles, leading in turn to greater surface dry out and a faster onset
of critical heat flux. Once bubbles grew sufficiently in size and coalesced, a single dominant
vapor bubble remained and was lifted just off the surface. This behavior was attributed to
the rapid evaporation of the thin liquid layer remaining after the coalescence of a large
number of small bubbles; the added momentum away from the surface was strong enough
to drive it off the surface. This hypothesis was supported by an observed rapid decrease in
surface temperature, consistent with rapid evaporation. The large bubble was subsequently
kept from the surface by the continued addition of momentum from the small bubbles as
they grew towards and coalesced with the large bubble. This continued coalescence was
observed at steady state, and confirmed that there is a clear nucleate boiling regime in
microgravity conditions.
What was even more surprising was that despite the greatly reduced bubble departure
frequency, during low subcooling and low heat fluxes, several studies observed increase in
the heat transfer coefficient in microgravity over terrestrial gravity [13-18]. In fact,
experiments performed with R-113 refrigerant showed a 32% increase in the heat transfer
coefficient. This increase in the heat transfer coefficient was attributed to the large vapor
bubble just off the surface, which acted as a large reservoir and absorbed the nucleating
bubbles through coalescence before they would depart naturally, inhibiting dryout.
Additionally, Lee et al. [13] postulated that the continuous process of nucleation and
coalescence induced turbulent eddies that enhanced the heat transfer coefficient. These
14
effects have since been generally accepted as the reason for the enhanced heat transfer
coefficient in low subcooling conditions in microgravity over terrestrial gravity. Under high
subcooling conditions (greater than 22.2°C for R-113), no such large vapor bubbles were
observed, and the heat transfer rates were attributed solely to the turbulence caused by the
erratic motion, coalescence, and departure of bubbles. The high heat transfer coefficient in
high subcooling conditions in microgravity compared with terrestrial conditions has also
been suggested to be caused by Marangoni convection around the vapor bubbles [18]. Based
on the experiments performed with R-113, Lee at al. [13] developed a microgravity pool
boiling curve and compared it to a terrestrial gravity pool boiling curve (Fig. 2.5). The
increase in heat transfer coefficient for low heat fluxes in the nucleate boiling regime can be
seen by the microgravity boiling curve’s shift to the left compared to the equivalent pool
boiling curve for R-113 in terrestrial gravity. Additionally, for the microgravity condition, a
significant reduction in the critical heat flux is clearly visible.
Figure 2.5: The pool boiling curve in microgravity. Microgravity pool boiling studies using R113 were performed on a number of space shuttle flights in order to generate this curve (plot
was adapted from data presented by Lee at al. [13]).
2.4. Pool boiling using augmented surface geometries
Past research on passive heat transfer enhancement has been performed predominantly by
means of augmented surface geometry. Such geometric features include continuous slots
15
and reentrant tunnels, cavities, angular/wedge-shaped geometries, and porous surfaces [1,
19-22]. The improved heat transfer characteristics of these augmented surfaces have been
well documented, and in some cases result in heat transfer coefficients more than double
that which are observed on smooth planar surfaces. Modifications to surface geometry on
the order of 1 mm such as slots, wedges, and reentrant tunnels are created to act as
favorable sites for nucleation and vapor entrainment, reducing the required wall superheat
temperatures required to maintain nucleate boiling.
Chyu and Fei [21] studied the effect of wedge angles on the heat flux during nucleate boiling
by varying the contact angle of a solid wall with the heated surface (Fig. 2.6). It was observed
that decreasing the contact angle allowed for the formation of a stable vapor phase at the
contact line, and subsequently allowed for higher heat transfer rates for a given wall
superheat over that of a plain surface. The greatest enhancement was observed for a wall
contact angle of 15°, where heat transfer rate was over twice as high as for a plain surface
for a wall superheat of about 2.5°C.
Figure 2.6: Two-dimensional representation of the experimental setup used by Chyu and Fei
[21]. This setup was used to study the effect of wedge angle on heat flux (diagram adapted
from Chyu and Fei [21]).
Das et al. [19,20] studied the effects of continuous slots in heated surfaces on the boiling
heat transfer coefficient. The types of slots studied were rectangular slots with either a
circular reentrant geometry, rectangular reentrant geometry, or no reentrant geometry at
their base (Fig. 2.7). The effect of slot tilt and orthogonal patterning (a grid of slots) was also
studied.
16
Figure 2.7: Various slot geometries studied by Das et al. [19,20]. Reentrant, non-reentrant,
and angled slots were studies. Slot grids were also created using these geometries, creating
rectangular pillars on the surface (diagram adapted from Das et al. [19,20]).
All slots were created using a wire EDM technique and were 0.4 mm wide, 2 mm deep and
spaced 3 mm apart. The results from this study reveal that significant enhancement can be
achieved by use of such geometry. Enhancement factors as high as 12 over a planar surface
were observed for the grid-style slot arrangement with circular reentrant geometry at the
base, keeping in mind that when accounting for the increased surface area of the slots, the
wetted surface area was increased by a factor of 3.65. Based on these results, follow-on
studies with rectangular slots, and rectangular slots with circular reentrant geometry at the
base were performed. The slot angle to the surface was varied, revealing that this too
increased the heat transfer coefficient. Again the slots with circular reentrant geometry
greatly surpassed the plain slots in performance. For both geometries, a maximum heat
transfer coefficient was observed for all heat fluxes when the slot was tilted 45° to the
surface.
2.4.1. Surface enhancements for fluid motion
While the surface modifications in the previous section were aimed at heat transfer
enhancement, this section discusses surface enhancements for fluid motion. The first of
these enhancements discussed are wicking structures, designed to facilitate the advection of
fluid from the pool to the surface via capillary action. Liter and Kaviany [22] studied the
effects of porous coatings with a periodic variation in thickness. The aim of this study was to
periodically facilitate the advection of liquid to the heated surface by means of capillary
wicking, and vapor away from the heated surface along predefined paths, reducing the flow
17
resistance. Such systems are referred to as capillary-artery evaporator systems. The structure
studied was comprised of copper beads 200 µm in diameter, sintered to form conical
structures arranged in a close-packed hexagonal pattern. This arrangement greatly increased
the heat transfer rate over a plain surface for the same degree of wall superheat. In fact, the
critical heat flux achieved with this spatially modulated structured surface was over three
times higher than that of a plain surface, and 58% higher than a uniform thickness porous
layer coating.
A particular case of fluid motion by surface enhancements has been recently demonstrated
in single phase and film boiling regimes [23-26]. All three studies employed a similarly
structured ratchet-like surface as discussed in this thesis, with the exception of the absence
of any cavities. Additionally, the mechanisms by which the fluid motion was hypothesized to
occur differ greatly between these studies and the work discussed in this thesis.
The single-phase study was performed by Stroock et al. [23] and aimed at characterizing the
net flow in a thin layer of oil with a free surface exposed to air by taking advantage of the
natural presence of Marangoni-Bérnard convection cells. These convection cells are formed
due to non-uniform temperature gradients at the liquid-air interface. Hot regions exhibit
lower surface tension that cooler regions, driving flow near the surface towards the cooler
regions. Where possible, this effect is further amplified by the up-welling of hot liquid under
the hot surface regions, and down-welling of cold liquid under the cold surface regions. This
thermo-capillary positive feedback effect leads to the formation of stable convective cells. It
was observed that in the presence of a surface of asymmetrical topography (grooves with a
steep slope on one side and shallow slope on the other) the primary convective cells did not
remain entirely intact; a component of the flow moved to the next cell at either the bottom
or top of the cell. This movement was not symmetrical as it occurred preferentially in one
direction depending on the depth of the liquid layer and the surface temperature gradients.
At a fluid depth half that of the topographical period
, the convective cells matched the
topographical period; the flow was in the direction of the steeply-sloped face at low
temperature gradients, and in the direction of the shallow-sloped face at high temperature
gradients (Fig. 2.8). At a fluid depth that corresponded to the topographical period, the
18
convective cell period was double that of the topographical period; the net flow at this depth
was in the direction of the steeply-sloped face at all temperature gradients. In both cases,
the amplitude of the net flow increased rapidly at conditions above the critical Marangoni
number (a condition which dictates at which vertical temperature gradient the dominant
form of heat transfer moves from conduction to convection).
(a)
(b)
Figure 2.8: Experimental results from Stroock et al. [23]. Diagram shows hypothetical
Marangoni-Bénard convective cells and a secondary flow pattern over a surface with
asymmetric topography at high temperature gradients (a) and associated experimental
results (b) (diagram and plot adapted from Stroock et al. [23]).
These results were later replicated in a similar, but entirely separate study performed by Jo
at Oregon State University using a closed loop experimental facility [24]. Both studies
demonstrated small flow rates; the maximum liquid-air interface velocity observed by Jo
during experiments was 0.85 mm/min. Even so, the ability to generate transverse flow
merely by applying a heat flux was clearly demonstrated.
Linke et al. [25] conducted a study on the thermally actuated pumping of liquids on surfaces
above the Leidenfrost point in the film boiling regime (Fig. 2.1). Water does not contact the
heated surface in the film boiling regime, and often moves around erratically depending on
the smoothness of the surface. These characteristics were take advantage of by creating an
asymmetrically structured surface (period of 1.5 mm) and heating it above the Leidenfrost
19
point. What was observed was that when liquid droplets were deposited on the heated
surface they would consistently accelerate in the direction of the shallow facing slopes.
Figure 2.9 shows the structured surface with a droplet of film-boiling water. At point A the
droplet contacts the crest of the ratchet and vapor escapes to the left and the right of this
point. The vapor escaping to the left was expected to escape to the sides of the ratchet (into
the page and out of the page), whereas the vapor escaping to the right was expected to
enact some viscous drag on the droplet to the right before escaping to the sides in the trough
of the ratchet pattern. This net viscous drag was hypothesized to be the driving force for the
droplet acceleration, and even propelled droplets up inclines.
Figure 2.9: Leidenfrost effect on asymmetrically structures heated surface as demonstrated
by Linke et al. [25] The direction of vapor transport between droplet and ratchets is shown
(diagram adapted from Linke et al. [25]).
This study revealed that significantly high droplet velocities could be achieved with this
geometry. In fact, speeds on the order of 5 cm/sec were achieved during experiments with R134a refrigerant.
A Follow-on study on this phenomenon was performed by Ok et al. [26] with water, and was
primarily concerned with the effect of ratchet size on droplet velocity. The smallest ratchet
period studied was 800 nm, and resulted in a maximum measured mean velocity of 40
cm/sec at a surface temperature of roughly 240°C. Maximum mean droplet velocities
decreased with increase in ratchet period. As the wall superheat was increased the droplet
speed decreased, and the all period-specific curves collapsed onto one. This perioddependent behavior at lower wall superheat was explained by the contact the droplet made
with the top of the ratchets, increasing the heat transfer rate and splitting the vapor layer
over each ratchet in two. The shorter the ratchet period, the more such contact lines existed
20
under the droplet to generate vapor and the greater the net viscous drag on the droplet. The
collapse of the individual curves at high wall superheat was explained as follows: as the wall
superheat increases, the droplet is levitated off the surface, and the flow between the two
sides is less distinct. Larger ratchet periods provide a larger gap below the droplet for the
vapor to escape, whereas smaller ratchet periods create a larger flow resistance for vapor
escape, as the cross sectional area of the triangular channel is much less. The greater viscous
drag on the vapor creates a thicker equilibrium vapor layer, offsetting the performance gain
due to a small ratchet period, and making the behavior similar for all ratchet periods.
2.5. Conclusions on Literature Review
The above literature review covers a broad range of previous work in the field of pool boiling
using structured surfaces both in terrestrial and microgravity conditions, and serves to put
the work discussed in this thesis into context.
This review has revealed that studies relating to pool boiling heat transfer are mostly
concerned with heat transfer enhancement without regard for lateral fluid motion or passive
pumping effects. Some studies have been conducted on the passive pumping effects of
structured surfaces by application of heat, specifically in the single phase natural convection
regime, and the two-phase film boiling regime, but a gap exists in the nucleate boiling regime
of the boiling curve where such fluid motion has been demonstrated, highlighting a need for
thermally actuated pumping research in in the nucleate boiling regime.
It is also important to note that much of the research surrounding pool boiling in
microgravity was centered around characterizing the performance of a flat heated surface in
the nucleate boiling regime, and that no studies were found in which the advantages of
structured surfaces were discussed as they pertain to microgravity applications, nor were
such experiments conducted. Additionally, some of the dominant pool boiling effects in
microgravity such as Marangoni convection in subcooled conditions, and the presence of
large vapor bubbles in saturated conditions, highlight the need for effective bubble removal
from the heated surface to delay the onset of critical heat flux, allowing for greater heat
fluxes at low surface temperatures.
21
3. SCOPE AND OBJECTIVES
A review of the literature in Chapter 2 has revealed thermally actuated pumping in the
nucleate boiling regime has not been studied, nor has there been research into the effects of
such surfaces in microgravity. Because of the high heat fluxes that are possible with phase
change heat transfer, thermally actuated pumping of liquids in the nucleate boiling regime
has the potential to offer great benefits in the form of a passive high flux thermal
management device. Thermally actuated pumping has been demonstrated with an imposed
vertical temperature gradient in a single-phase liquid regime in the presence of a liquidvapor interface by Stroock et al. [23] and in the Leidenfrost film boiling regime by Linke et al.
[25]; It is therefore of great benefit to research possible methods of doing so in the nucleate
boiling regime. Both experiments demonstrated a net pumping effect using asymmetric
ratchet-type surface structures, and both were able to drive fluid in the same direction (i.e.
in the direction of the shallow-sloped face). The work discussed in this thesis will focus on
the same type of repeated asymmetric ratchet surface geometry, such that a single type of
surface structure can potentially provide thermally actuated pumping in the same direction
in the single-phase, nucleate boiling and Leidenfrost regimes.
3.1. Objectives and Tasks
A multi-year collaborative study between Oregon State University and Auburn University is
investigating the potential for thermally activated pumping of fluids in the nucleate boiling
regime using repeated asymmetric surface structures. The experimental and modeling effort
at Oregon State University will be aimed at documenting the bubble dynamics and flow
around a small array of such ratchets in a pool of stagnant fluid and quantify forces that
cause fluid motion in a direction parallel to the heated surface. The effort at Auburn is
devoted to silicon fabrication methods for the surfaces as well as to characterize the net flow
of liquid in a closed loop with an evaporator comprised of the ratcheted surfaces.
As a part of the larger effort described above, the objective of this thesis work is to carry out
proof-of-concept experiments to test the hypothesis that bubbles emanating from cavities
22
located on an asymmetrically structured surface have a component of velocity parallel to the
heated surface as described in Chapter 1. A positive indication that a net lateral flow is
generated by the structured surface would justify future work in this area and would be the
first step in evaluating the potential for thermally actuated pumping using asymmetrically
structured surface geometries similar to those employed by Stroock et al. [23] and Linke et
al. [25]. While asymmetry is not necessary for the purpose of proving the hypothesis, it is
nevertheless implemented to maintain a surface geometry consistent with that employed by
Stroock et al. [23] and Linke et al. [25] such that a single surface geometry can generate
lateral fluid motion in both the nucleate and film boiling regimes. In order to validate the
hypothesis, information on bubble velocity, acceleration, departure diameter, and volume
needs to be obtained. High-speed visualization with backlighting provides for both qualitative
and quantitative visualization of the bubble motion and is hence the method of choice for
the experiments presented in this thesis.
In order to carry out the above proof-of-concept experiments, a major task entails design
and construction of an experimental facility. The facility design also needs to accommodate a
future study of the potential of the passive geometry for thermally-actuated pumping in a
microgravity environment. These future experiments on gravitational dependence of the
flow are expected to take place with three different orientations of the heated surface to
quantify the bubble and liquid motion with respect to the buoyancy force, as any flow
perpendicular to or in the opposite direction of the buoyancy force would indicate that
buoyancy is not the driving mechanism of the flow. For this reason, the surface must be
capable of being installed in the pool in three orientations as indicated in Fig. 4.1. Additional
requirements were that the temperature of the liquid-surface interface could be mapped
using infrared thermography, and that Particle Image Velocimetry (PIV) and Laser Induced
Fluorescence (LIF) measurements could be made to study the two-dimensional flow and
temperature variations in the liquid near the heated surface.
23
4. EXPERIMENTAL FACILITY
This section covers the design of the experimental facility. First a review covering various
methods of creating a microgravity environment is presented, followed by a brief review of
such previously designed facilities. A description of the test section, the experimental facility
requirements, the various designs considered for this thesis, and ultimately a detailed
description of the final design, including all associated instrumentation and hardware is
presented next. Although the full functionality of this experimental facility was not used to
obtain the results presented in this thesis, it was nevertheless implemented to set the stage
for follow-up work. Part drawings for all manufactured parts are included in Appendix F.
4.1. Existing Microgravity Simulation Methods and Facilities
Researching new technologies for space applications ideally requires validation in space.
However, since the cost of transporting experiments to space in prohibitively expensive, it is
desirable to simulate or create an artificial microgravity environment where experiments can
be conducted as a reasonable simulation of the space environment [28]. For this reason a
review of such existing methods was carried out, along with test facilities used for such
experiments.
4.1.1. Simulating a Microgravity Environment
The primary difference between gravity and microgravity conditions is the near absence of
buoyancy forces in microgravity. Because the buoyancy force acts in the vertical direction
alone, it is still possible to simulate microgravity conditions by studying only the convective
behavior normal to or opposing the direction of the buoyancy force (Fig. 4.1). For example, in
terrestrial conditions, if naturally occurring convective motion was observed to transport hot
liquid laterally or downward in a cold pool, such an observation would indicate buoyancy was
not the driving mechanism for convection. For this reason, among others, researchers
sometimes employ vertical or downward-facing heated surfaces to study the flow effects
when buoyancy does not aid bubble departure [29,30]. Results from such experiments may
then justify further experimentation in microgravity conditions.
24
Figure 4.1: Three orientations required to effectively demonstrate the flow's nongravitational dependence.
4.1.2. Creating a Microgravity Environment
The need for a non-space-based microgravity environment has led to the development of
several types of microgravity simulation facilities. There are three primary means in which a
microgravity environment can temporarily be created, these are: drop towers, sounding
rockets, and parabolic flights. There are advantages to each of these services in the form of
acceleration expected, duration of the microgravity conditions, and various others. Table 4.1
outlines the various parameters associated with each service.
25
Table 4.1 Overview of various methods of creating a microgravity environment and common
associated parameters [31-35].
Service Type
Drop Tower
Sounding
Rocket
Parabolic Flight
Min g
1x10-5
1x10-5
Max g
30-65
N/A
2x10-2
9
Duration
Advantages
2-5 sec
Very repeatable
Several min Experiment is actually brought into
space
10-20 sec
Operator can accompany experiment
As each facility/method has its own advantages, the desired facility/method should be
chosen based on the experimental requirements.
Drop Tower
Drop tower facilities usually consist of a large tower, at the top of which a drop vehicle is
suspended. The experiment is housed inside this vehicle and experiences very little
acceleration when released from its suspended position. In order to minimize the
accelerations associated with aerodynamic drag, some facilities implement either a
secondary aerodynamic fairing within which the experiment floats freely, or create a near
complete (0.005 Torr) vacuum in the chamber in which the vehicle falls [31,32]. Facilities that
implement these features are capable of delivering accelerations as little as 1x10 -3g for the
aerodynamic fairing type and 1x10-5g for the vacuum chamber type. The primary
disadvantage of a drop tower is the high acceleration upon impact with the stopping device;
it is not uncommon for drop towers to subject payloads to impact accelerations of 30-65g
[31,32].
Sounding Rocket
For longer and more accurate microgravity simulations, the sounding rocket is often used.
Sounding rockets can house numerous experiments and have a vertical launch and return
trajectory. The microgravity conditions at the apex of the trajectory are often of very high
quality. The primary advantage to this service is that the payload is actually brought into
space, which allows experiments to be conducted that require a true space environment in
addition to microgravity. In some cases, sounding rockets achieve altitudes over 1,500 km
(higher than the international space station) before returning to ground level [33]. Because
26
sounding rockets can travel to such high altitudes, they are free of aerodynamic drag and are
thereby capable of providing microgravity conditions with accelerations commonly as low as
1x10-5g for several minutes; at least one study reported continuous accelerations as low as
10-6g [16]. The disadvantages of sounding rockets are the payload size restrictions,
inaccessibility to the experiment, and the possibility that the experiment is not recovered
after the flight.
Flying Parabolic Arcs
The most familiar means of simulating microgravity is flying parabolic arcs, where an aircraft
flies a parabolic trajectory to match the acceleration of gravity. There are numerous federal
agencies such as NASA and ESA, and commercial organizations such as Zero Gravity
Corporation that provide these services. The primary advantage of these aircraft-based
microgravity simulations is the number of repeated parabolas that can be performed (up to
40 per flight), the allowance for people to accompany the experiment, the absence of any
large accelerations, and the availability of external power sources [37]. Experiments are
normally subject to aircraft vibrations, but can often be free-floated to physically disconnect
the experiment from the airframe (a technique that makes the airframe analogous to the
aerodynamic fairing employed in a drop tower) to reduce accelerations below those
experienced by the airframe.
4.1.3. Microgravity Experimental Facilities
There have been a number of pool boiling studies performed in microgravity. Perhaps the
most successful and widely cited microgravity pool boiling experiments are by Lee et al. [13]
and are therefore the focus of this review. Because the aim of these experiments was largely
the observation of bubble behavior as it pertains to the different regimes in the pool boiling
curve, obvious similarities exist compared to the needs for the experimental facility
developed for this thesis work. The experimental facility employed by Lee et al. [13] was a
pressurized chamber, with two compartments; one compartment housed the boiling surface
and the R-113 fluid, the other served as a reservoir to accommodate expansion associated
with the generation of vapor bubbles in the pool and was filled with nitrogen. The two
27
chambers interfaced by means of a bellows, allowing for expansion into the nitrogen
chamber without increase in pressure. A pressure transducer in the nitrogen chamber
monitored the pressure, and regulated the flow of nitrogen from a storage tank into the
chamber when the pressure was too low, and vented nitrogen if the pressure was too high.
Such a pressure regulating configuration was also used by Straub [38]. The pool was heated
through the chamber walls, and a stirrer in the pool served to create a uniform pool
temperature at the start of each experiment. Perhaps the most unique feature of this
experimental facility was the ability to image the pool from the side and from below the
heater surface at the same time using a single camera. The bottom view was used to count
the number of bubbles generated as well as to determine their size, whereas the side-view
was used to observe events occurring in the pool above the heated surface. The ability to
view the pool from below was accomplished by sputter depositing a thin (400 Å) gold layer
that served as a semitransparent heater as well as resistive thermometer onto a polished
transparent quartz substrate. A number of thermistors were placed at various distances from
the heater surface as well as throughout the pool. R-113 was used for its low boiling
temperature, thus minimizing heat losses to the environment, and its electrical nonconductivity, allowing it to be used in direct contact with the gold film heater. Because this
experiment was used in a microgravity environment it did not require the heated surface to
be installed in more than one orientation. For the purposes of the present experimental
facility design, the primary considerations taken from the experimental facility employed by
Lee et al. was to make accommodations for future pressure regulation equipment.
Experiments in both microgravity and simulated microgravity can be performed using a
variety of resources and techniques. This review has revealed that for the research discussed
in this thesis, it would be most appropriate to initially research the effects of various surface
orientations, followed by flying parabolic flight in an aircraft, primarily because an aircraft
allows the operator to accompany the experiment, and does not subject the experiment to
high accelerations.
28
4.2. Test Section
The structured surface was fabricated by Dr. Sushil Bhavnani and Naveenan Thiagarajan at
Auburn University in Auburn, AL and consisted of two wafers bonded to one another – the
structured surface and the heater wafer. The heater wafer consisted of a 500 µm thick, 76.2
mm (3.00 in) diameter silicon wafer, etched from a larger 100 mm diameter wafer. A small
section (11.8 x 8.0 mm) of the structured surface with seven ratchets and four cavities per
face was bonded to one side of the large wafer, and a thin film aluminum heater was
deposited on the other. From here on in this thesis the side of the silicon wafer with the
heater is referred to as the backside, the side with the structured surface is referred to as the
topside, and the asymmetrically structured surface is referred to as the test section. The
aluminum heater was deposited through e-beam evaporation, after which thermal isolation
trenches were etched 350 µm deep around the heater to reduce conduction losses in the
wafer. The test section was made from a separate silicon wafer and the ratchet pattern was
created using Deep Reactive Ion Etching (DRIE). The angles of the pattern were controlled by
varying the plasma gas ratios during this process, and were nominally etched to 24° for the
test section used in this thesis work. The cavities were etched from the opposite side of the
test section using potassium hydroxide. This resulted in an isotropic etch of the silicon at an
angle of 54.6°, creating pyramidal cavities. Once the test section was completed, it was
bonded to the topside of the wafer using a gold and tin preform as solder (Fig. 4.2).
29
Figure 4.2: Configuration of the test section and thin film aluminum heater silicon wafer. The
test setion was etched from a separate silicon wafer and bonded with gold and tin. The thin
film heater was deposited through e-beam evaporation.
The cavity mouths in the sloped ratchet face were trapezoidal in shape due to the pyramidal
etch from the back. Using a 25 mm x 25 mm fixed frequency grid distortion target with dots
of 0.0625 mm diameter and 0.125 mm spacing (Edmund Optics stock no. NT58-509), the
cavity mouth was estimated to be 250 µm at the base, 120 µm at the top, and 205 µm tall
after correcting for the 24° slope of the face (Fig. 4.3). An image of the optical grid taken at
the same working distance and magnification as the cavity image was overlaid to provide a
scale, as in-situ placement of the optical grid did not permit both the cavity and grid to be in
focus simultaneously due to the topography of the test section and the narrow focal depth of
the microscope lens.
30
Figure 4.3: Microscope image of a cavity on the test section with overlaid image of calibrated
optical grid.
Power was supplied to the test section by a variable power supply (Tenma 72-8350). Two
multimeters (Tektronix DMM912) were used to measure the supplied current and the
voltage across the heater. Figure 4.4 shows a schematic of the electrical circuit used to
power the test section and measure the voltage and current.
Figure 4.4: Schematic of test section power and measurement circuit. Multimeters were used
to measure the supplied current and the voltage across the measurement leads.
4.3. Experimental Facility Requirements
The experimental facility was designed to study the effects of two-phase microgravity pool
boiling through minimally-intrusive whole-field measurements. The silicon wafer was to be
mounted in the three aforementioned orientations in such a way as to allow the
31
simultaneous imaging of the ratchet-fluid interface through infrared thermography from the
backside of the wafer, high-speed imaging of the bubbles departure dynamics from a side-on
view, and facilitate future implementation of Particle Image Velocimetry (PIV) and Laser
Induced Fluorescence (LIF) measurements (also from a side-on view). Additionally, the pool
temperature would need to be controlled to permit saturated or subcooled experiments, and
there would need to be provisions for condensation of vapor. The chamber was to be
constructed out of 316 Stainless Steel as this would insure compatibility with most fluids,
prevent oxidation, and insure a long operational life. Because the desired method of
simulating microgravity was flying parabolic arcs, the entire experimental facility was to be
made compatible with NASA Document AOD-33897 [39].
4.4. Initial Chamber Design
The primary design challenge for the chamber was the triple orientation requirement for the
test section. The first design considered was to mount the wafer in the chamber and allow
the chamber and all associated cameras and instruments to be rotated as one assembly to
provide the desired orientations. Table 4.2 lists the advantages and disadvantages of this
design.
32
Table 4.2: Advantages and disadvantages of achieving the three-orientation capability by
means of a rotatable chamber.
Advantages of Rotatable Chamber
Disadvantages of Rotatable Chamber
 Pool temperature and power input
 The ability to rotate the entire
conditions could remain more consistent
experimental facility requires a lot of
between the various orientations.
additional hardware and design effort
 Little time is required to change the
wafer’s orientation as the chamber can be
leading to increased cost.
 The degree of freedom about the axis of
rotated, precluding a need to remove and
rotation introduces the possibility of errors
reinstall the wafer for the various
in orientation repeatability and makes the
orientations, greatly simplifying this part of
chamber more susceptible to external
the experimental procedure.
vibrations and forces.
 Allows for off-90° changes in the wafer’s
orientation if this ever becomes necessary.
 The externally-mounted laser will require a
complicated mirror system to maintain
accurate and repeatable alignment with the
target for all orientations, increasing the
cost and complexity.
 Size envelope of experimental facility is
large
Though a rotatable chamber has great benefits from an experimental procedure point of
view, the cost of manufacturing and complexity of the optics ultimately resulted in the
decision not to follow through with this design. Instead, it was decided to keep the chamber
in a fixed orientation, and change the wafer’s orientation alone, greatly simplifying the
overall design.
4.4.1. Initial Pool Temperature Control
The pool was initially heated using a hot oil loop; it consisted of a 3 m (10 ft.) long 316
stainless steel tube through which hot oil was pumped. The hot oil was pumped from a
33
remotely-located circulation heater (Cole Parmer Polystat Series) which was capable of
controlling the oil temperature. Unfortunately this configuration not only lacked
temperature feedback for to control pool temperature, it was not capable of raising the pool
temperature above 40°C, largely due to the number of flow restrictions in the flow loop
which greatly reduced the flow rate.
4.5. Final Chamber Design
Because the fixed chamber design was adopted, the focus was on how best to provide the
test section with the capability of being mounted in the three orientations of interest with
minimal cost and effort. Considering the wafer’s fragility and high manufacturing cost and
procurement time, it was not favorable to require dismounting of the wafer itself when
changing orientations, but rather to create a wafer holder upon which the wafer could be
affixed, which itself could then be reoriented. Such a wafer holder would have to support the
wafer at its perimeter, keeping the central area on the backside of the wafer exposed to the
outside environment to permit IR thermography. By creating three identical circular ports in
the chamber walls (bottom, side, top), the wafer holder could be placed in any one of the
ports, providing the three-orientation capability.
4.5.1. Wafer Holder
In addition to interfacing with the chamber, the wafer holder was designed to hold the
silicon wafer while providing for IR access from the backside and to provide electrical
connections for the power supply and for voltage measurements. Four electrical connections
were needed in total – two for supplying power and two for measuring voltage. Since the
backside of the wafer is exposed, the holder must also provide a seal around the wafer. The
O-ring was of round cross section, 1.59 mm (0.063 in) thick, and made of FC-72 compatible
EPDM (Ethylene Propylene Diene Monomer).The initial design featured multiple clamps that
clamped the wafer to the O-ring (Fig. 4.5a). The purpose of this design was to provide a clear
view of the test section between the clamps when viewed from the side. A number of clamp
tests were performed with sample wafers, but the individual clamps did not provide uniform
34
clamping force and caused numerous failures of sample wafers. It was discovered that in
order to prevent failures, the requisite clamping force on the wafer needed to be uniformly
distributed around the circumference of the O-ring.
However, supporting the wafer continuously around its circumference over the O-ring would
obscure the side-on view of the test section. To solve this problem, a 3-piece clamp was
designed (Fig. 4.5b). The sides of the clamp were made of 12.70 mm (0.500 in) thick optically
clear polycarbonate (the inner face of which was located 25.4 mm (1.00 in) from the test
section centerline), while the rest of the clamp was made of Polyetheremide (a translucent
brown thermoplastic also known as Ultem1000). Six screws secured the clamp to the wafer
holder, sandwiching the wafer and the O-ring in between. This design provided a continuous
flat surface that pressed down on the wafer, and also served to support the wafer where the
pogo pins made contact, precluding stress concentrations at these locations. Lastly, a
removable top wall was added 25.4 mm (1.00 in) above the wafer to reduce the momentum
effects of any currents in the surrounding pool. This clamp design proved highly effective,
providing a circumferentially uniform clamping surface while providing an unobstructed sideon view of the ratchets.
(a)
(b)
Figure 4.5: Wafer holder and clamp designs. The low-profile individual clamps on the initial
design caused stress concentrations that lead to failures in sample wafers (a). The new
design (b) was a continuously supporting clamp where part of the clamp was made of
optically clear polycarbonate to provide an unobstructed side view of the test section.
35
The holder was designed to bring the wafer into view through the chamber windows, and
provide for an unobstructed view of the backside of the wafer for infrared thermography
(Fig. 4.6). The working distance for the IR microscope lens was 30 mm so the pogo pins and
lead wires were kept short enough to maintain clearance in this small gap. The bottom of the
holder featured a flange with a hole pattern to mate with the round chamber ports in any of
the three aforementioned orientations.
Figure 4.6: Transparent view of the wafer holder showing the infrared camera microscope
lens position in relation to the test section. A cutout in the polycarbonate wafer mount
allows the microscope lens to be spaced 30 mm to focus from the backside of the wafer,
permitting infrared thermography of the heated surface.
4.5.2. Chamber Design
The 316 stainless steel chamber featured a wall thickness of 12.7 mm (0.500 in) in order to
allow enough depth for screw threads and O-ring grooves, and inside dimensions 22.8 cm x
22.8 cm x 15.2 cm (9.00 in x 9.00 in x 6.00 in). All chamber O-rings were 3.18 mm (0.125 in)
thick with a round cross-section. The material chosen for all O-rings was EPDM as it offered
good chemical compatibility with 3M FluorinertTM Electronic Liquid FC-72 (Table 4.3). All Oring grooves were 3.56 mm (0.140 in) wide and 2.84 mm (0.112 in) deep, allowing for 10%
36
compression. The circumference of each O-ring groove was sized to induce 2-4% stretch
when installed.
Table 4.3: Material compatibility with 3M FluorinertTM Electronic Liquid (FC-72)
Material
Buna-N
PTFE
Viton
EPDM
Silicone
Thermoplastics (PVC, Ultem,
Polycarbonate, Acrylic, etc)
PFA
Compatibility with FC-72
Poor
Good
Poor
Good
Not Recommended
Good
Good
Because the test section was to be imaged in three different orientations, it needed to be
mounted in such a way as to keep it in view for external visible spectrum cameras (for PIV,
LIF, and High-Speed Imaging) that were to image the flow near the ratchets, for this reason
large windows were used. Additionally, the windows were chosen to be rectangular,
precluding any cropping of the image by the edge of the window for a large field of view. The
windows were made of 6.35 mm (0.250 in) thick borosilicate glass from GrayGlass and sealed
against the chamber with a 3.18 mm (0.125 in) thick EPDM O-ring, clamped with a stainless
steel flange. The flange was polished smooth to prevent the glass from cracking. The wafer
mount ports were designed as circular cutouts because the wafer itself was round. This
allowed the wafer holder to be an axis-symmetric part that could be machined on a lathe.
Two 6.35 mm (0.250 in) thick 316 stainless steel covers with mating geometry identical to
that of the wafer holder were designed to cover the circular ports not filled by the wafer
holder. One of the covers featured four individual thermocouple ports, while the other was
blank. Installing the thermocouples in one of the port covers allowed them to always remain
in the same orientation relative to the wafer; when the wafer was moved from the upright
orientation to the vertical orientation, the port cover with the thermocouples was moved
from the vertical port to the bottom port. Lastly, in order to make the chamber compatible
with NASA Document AOD-33897, four flanges with 9.53 mm (0.375 in) diameter holes in a
vertical 5.80 cm (2.00 in) spaced pattern were added to the chamber that would serve as an
37
interface to mount to a rack. These flanges also served to support the high-speed camera
and the backlighting assembly.
4.5.3. Pool Temperature Control
Maintaining control over the chamber pool temperature was critical to performing
experiments at subcooled and saturation conditions. In general, precise temperature control
of any kind can be accomplished by means of a PID control loop supplied with temperature
measurements from thermocouples, controlling some type of heat source. This configuration
allows the pool temperature control to be performed automatically and autonomously. As
discussed in Section 4.4.1, pool temperature control was initially performed using a hot oil
loop. However, because of its ineffectiveness, the hot oil loop was replaced with two
electrically powered 600W cartridge heaters (Watlow part # G6A56-BG12) that were
immersed in the pool and controlled by a PID controller (Omega part # CNi3253-DC) through
a separate high-current relay (Omega part # SSR330DC50). The cartridge heaters greatly
simplified the heating system and were far more effective at heating the pool than the hot oil
loop. There were no active cooling systems; cooling of the pool relied on heat losses through
the chamber walls since future flight experiments would not permit the use of the chiller.
4.5.4. Chamber Flow Loop
The entirety of the fluid flow loops consisted of a condensing equipment flow loop and a
fill/drain flow loop. Figure 4.7 shows the layout of the two flow loops; the following sections
will describe each in detail.
38
Figure 4.7: Layout of the two flow loops implemented in the experimental facility. A
condensing equipment flow loop was responsible for condensation of the vapor in the
chamber, and a fill and drain loop was responsible for filling, draining, and filtering the pool.
In order to condense vapor in the chamber, condensation coils were installed. These coils
were designed to be used with either the Thermo Scientific NESLAB ThermoFlex5000
Circulation Chiller with positive displacement pump, capable of removing up to 5kW at 20°C
at a flow rate of up to 15.5 L/min, or the Thermo Scientific ThermoHaake C30P Heating
Refrigerated Water Bath Circulator, capable of removing 800W at temperatures as low as 3°C
at a flow rate of up to 24 L/min. The length of tubing needed for condensation of vapor was
estimated as a part of design of the coils. The condensing coils were designed to fit inside the
chamber while keeping clearance with the circular port around which it was installed so that
there would be no interference with the wafer holder when installed in the inverted
orientation. For this reason the tubes were coiled to maintain clearance around the circular
port (Figs. 4.9a, 4.15, 4.19b). An additional restriction was that the condensate dripping onto
the liquid-air interface would not affect the flow near the test section. This requirement was
39
met by default as the coiled tubing configuration used to provide clearance for the holder in
the inverted orientation left the area above the test section free of condensing equipment in
all orientations. Double shut-off quick disconnect tube couplers from Swagelok allowed the
chiller to easily be disconnected from the condensation coils.
The length of tubing required to condense at a minimum rate of 1200W (the maximum
heating capacity of the cartridge heaters) was determined using Matlab®; all fluid properties
used were for water. A 1D resistance network, shown in Fig. 4.8, was used to model the heat
transfer from the saturated vapor to the coolant in the condensing coils.
Figure 4.8: 1D resistance network used to determine the minimum tube length required for
the condensing coils.
The heat transfer coefficient for the flow in the tube was calculated using single-phase
convection:
̅
⁄
[
]
[
]
(4.1)
The total heat transfer rate was calculated using the resistance network shown in Fig. 4.8.
Since no explicit solution for the tube length as a function of total heat transfer rate could be
found, this value was determined iteratively by first guessing a length and then solving for
the total heat transfer rate by minimizing the residuals of the inner and outer log mean wall
temperatures. The length was increased until the total heat transfer rate was greater than
1200W. Initial guesses for an average inner and outer log mean wall temperatures as well as
coolant exit temperature were set. The following equations were solved iteratively in order;
first the condensation heat transfer coefficient was calculated:
40
̅
where
*
(
⁄
)
(
)
+
(4.2)
was determined by
(
)
(4.3)
and the Jakob number was determined by
(
)
(4.4)
To determine the total heat transfer rate, the log mean temperature of the system was
calculated:
(
((
)
For the first iteration, the value for
(
)⁄(
)
))
(4.5)
was determined by the initial guess, whereas all
subsequent iterations used the value of
from the previous iteration. The value of
for the next iteration was calculated as follows:
̅
(
(
))
(4.6)
̇
The total heat transfer rate was then evaluated by dividing the log mean temperature
difference by the sum of the thermal resistances, specifically the inner convection, tube wall
conduction, outer condensation resistances:
(
( ⁄
̅
)
̅
)
(4.7)
Using Eq. 4.1-4.7, a total length of at least 1.2 m of 6.35 mm (0.250 in) diameter 316 stainless
steel tubing was deemed necessary to remove the heat generated in the chamber. Because
of the high vapor pressure of FC-72, a precautionary reflux condenser was installed on top of
41
the chamber. The reflux condenser used was 300 mm jacket length borosilicate glass Graham
condenser. The condenser was cooled using the same source as the stainless steel
condensing coils, and was installed such that the two flow loops were in parallel. This parallel
arrangement reduced the overall pressure drop in the flow loop, allowing for a greater flow
rate than in a series arrangement. Figure 4.9a shows an overview of the complete
condensing loop; Figure 4.9b shows film condensation of FC-72 on the chamber’s condensing
coils during experiments
(a)
(b)
Figure 4.9: Overview of condensing loop. Stainless steel coiled tubing was installed inside the
chamber and an externally mounted Graham reflux condenser installed in parallel to
minimize pressure drop; double shut-off quick disconnects connect the condensing loop to
the chiller (a); condensing coils inside the chamber condensing FC-72 during a test (b).
42
Because the images acquired by the Phantom V310 high-speed camera are used for
qualitative evaluation of the test section performance as well as image processing, it was
critical that there was no visible particulate in the fluid that would otherwise affect the image
processing algorithm’s results. For this reason, a small gear pump (Cole Parmer part # WU00132-YQ) was used to pump fluid directly from the fluid reservoir through a filter into the
chamber. For water, a 15-micron sintered stainless steel filter (Swagelok part # SS-4F-15) was
used, and for FC-72, a 7-micron sintered 316 stainless steel filter (Swagelok part # SS-4F-7)
was used. The filters required regular cleaning to keep the pressure drop within the limits of
the pump. The wetted materials of the pump were chosen for compatibility with both water
and FC-72. In particular, EPDM O-rings were substituted for the stock Buna-N O-rings. The
tubing used was 6.35 mm (0.250 in) diameter 316 stainless steel tubing with 316 stainless
steel Swagelok fittings and valves. Perfluoroalkoxy (PFA) tubing was used to connect the hose
barb fittings on both inlet and exit of the pump to the flow loop. This tubing was transparent,
permitting flow in the loop to be confirmed by the operator, and Polypropylene nuts and
ferrules (McMaster part # 5016K222) were used to mate this tubing to the Swagelok
components. Once the chamber was filled, turning a two-way valve allowed the pump to
draw water from within chamber as opposed to the external reservoir, enabling the fluid in
the chamber to be continuously filtered, removing any particulate that may have been
present inside the chamber before filling. The pump also served to thoroughly mix the pool
before pool temperature data was acquired. Once an experiment was completed, the
chamber could be drained directly back into the reservoir through the drain tube after
turning the drain valve. An overflow tube was installed at the top of the chamber and
connected to the drain tube downstream of the drain valve. In the event the pump was
inadvertently left on, the excess fluid in the chamber would overflow through this tube into
the drain line, precluding any spills or a chamber over-pressure that could break the silicon
wafer. Figure 4.10 shows the complete pump fill and recirculation loop, along with the
overflow tube. Pictures of the various components in the pump fill and recirculation loop are
shown in Fig. 4.11.
43
Figure 4.10: The chamber’s fill and recirculation loop with emergency overflow protection. A
two-way valve selects either an external reservoir or the chamber as the source for the gear
pump. A 15-micron sintered stainless steel filter insures that particulate in the liquid being
pumped in from the reservoir or being recirculated is removed before being pumped into the
chamber.
(a)
(b)
(c)
(d)
Figure 4.11: Overview of the components of the fill/drain flow loop. Shown are the pump and
pump selector valve in the recirculation position (a), the 15-micron filter (b), the chamber
overflow (c), and the drain valve in the closed position (d).
44
4.5.5. Imaging
To enable future PIV and LIF measurements, provisions were necessary for a laser light sheet
to enter the chamber and be incident on the wafer. Because the surface of interest on the
wafer consists of a ratchet pattern with a 30 degree ratchet angle, it is necessary for the laser
light sheet to be incident to the surface at a minimum angle of 30° to illuminate the trough
regions of the ratchets. However, providing such laser sheet illumination in all three
orientations (30° down for the upright orientation, straight in for the vertical orientation, and
30° up for the inverted orientation) as shown in Fig. 4.12 without changing the position or
orientation of the laser, required an ingenious design.
Figure 4.12: Required laser light sheet incidence angle on the test section as it depends on
the test section orientation. The laser port was designed so that a fixed source could provide
all three incidence angles.
To permit illumination of the wafer in any of the three orientations, a special laser port was
designed. The laser port featured a prism that would deflect the light-sheet down by 30°
upon which it would enter the chamber through a window that was oriented perpendicular
to the incident light sheet. The entire laser port assembly could also be installed upside down
to deflect the laser light sheet upwards at 30°.
45
In order to deflect the laser light sheet by 30°, the correct prism dimensions and properties
had to be determined. The deflection angle of the laser light sheet is a function of the prism’s
index of refraction, the prism angle, and the orientation in which it is installed. A Matlab®
script was written to determine the prism angle and orientation that would permit deflection
of the light sheet by at least 30°. The prism material chosen was BK7 as this is a common
material for prisms; the refractive index for BK7 is N = 1.5195 for λ = 532 nm (PIV laser), N =
1.5234 for λ=473 nm (LIF laser) [40]. Prism angles
The prism tilt angle
ranging from 10° to 60° were studied.
was also studied.
Figure 4.13a shows a diagram of the prism with prism angle , the prism tilt angle , and the
incident and refraction angles
. The beam refraction due to the prism was computed
using the standard refraction formula:
( )
( )
(4.8)
When written in terms of the prism angles and indexes of refraction, this equation becomes
Where
incident angle
and
(
( ))
(4.9)
(
( ))
(4.10)
. The beam deflection
, the refracted angle
was computed using the
, and the prism angle :
(4.11)
Combining the above equations, the beam deflection becomes
*
(
*
(
)+)+
(4.12)
46
(a)
(b)
(c)
Figure 4.13: Selection of the laser light sheet prism. A diagram (a) showing the prism angle ,
prism tilt , and the total beam deflection angle shows the variables considered. Plots
show the various beam deflection angles as they correspond to laser wavelengths of the PIV
laser (b) and the LIF laser (c).
Beam deflection was calculated as a function of prism tilt, prism angle and index of refraction
(a function of the wavelength of the light sheet) for both the PIV laser with λ = 532 nm (Fig.
4.13b) and the LIF laser with λ = 473 nm (Fig. 4.13c). Each curve in Fig. 4.13b,c represents a
particular beam deflection. Both plots show that a 45° prism (indicated by the dashed
horizontal line) can deflect a 532 nm or 473 nm wavelength laser light sheet by at least 30°,
provided that the prism is titled more than 15°. Because 45° prisms are very common, this
was an effective means of deflecting the light sheet. The prism (Thorlabs part # PS908H-A)
was mounted on a rotation platform (Thorlabs part # RP01) so that the tilt could be adjusted.
This allowed the beam deflection to vary between roughly 27° and 38° if so desired (for
example if the ratchet angles were to be varied). Figure 4.14-4.16 show the implementation
of the prism and two laser ports in the chamber used to illuminate the test section in the
upright and vertical orientations. The laser port’s mating geometry to the chamber was
rectangular to allow only the intended upright or inverted installation of the laser port
assembly.
47
Figure 4.14: The laser light sheet as it is deflected by a prism in the upright orientation.
Figure 4.15: The laser light sheet as it is deflected by a prism in the inverted orientation.
48
A second laser port was designed for use in the vertical orientation that featured a simple
window mounted perpendicular to the incoming sheet, thus permitting the light sheet to
pass into the chamber undeflected. The laser port assembly was designed for the light sheet
to enter at its center, allowing the laser to remain fixed for this configuration as well.
Figure 4.16: The laser light sheet as it enters the chamber in the vertical orientation. A simple
laser port was used in this configuration as the laser was already aligned with the test section
to illuminate the troughs between the ratchets.
Figure 4.17 shows a demonstration of the deflected beam; the beam is incident from the
top-left of the image, illuminating a bubble in the pool. Fluorescent dye (“Sulforhodamine B”
from Sigma Aldrich) was dissolved in the pool to make the laser light sheet visible. An optical
filter was placed in front of the lens to block the 532 nm wavelength light, but allow the
fluorescence wavelength (607 nm) to pass through so that only light emitted by the dye was
visible.
49
Figure 4.17: Demonstration of the laser light sheet illuminating a bubble. The light sheet
(λ=532 nm) was refracted 30° by a prism to allow the fluid in the troughs of the test section
(not included during this test) to be illuminated.
Infrared imaging was to be performed using a Nippon Avionics TVS-8500 Series Camera in
conjunction with a TVM-8510U Microscope Lens. The camera featured 256 x 236 pixels
resolution and provided up to 120 fps recording output. The microscope lens provided a 2.56
mm x 2.36 mm field of view at a fixed working distance of 30 mm. Since the microscope lens
must be fixed at 30 mm from the backside of the wafer, an IR camera support (Fig. 4.18) was
designed and constructed to support the IR camera in such a way that the lens would remain
concentric within the wafer holder, thereby keeping the center of the wafer centered in the
image. The support interfaced with the chamber using a rectangular four-hole pattern
around the circular port. This mounting interface was common to all circular ports, allowing
the IR camera to image the backside of the wafer in any of the three orientations (Fig. 4.19).
An adjustable screw in the back of the support allowed the camera to be moved closer or
farther away to bring the wafer into focus.
50
Figure 4.18: Infrared camera mounted to custom support designed to interface it with the
chamber. The camera position was fixed so that the microscope lens was aligned with the
test section. An adjustable screw allowed the camera to be positioned closer or farther from
the backside of the wafer to bring the image into focus.
51
(a)
(b)
(c)
Figure 4.19: Cutaway view of the infrared camera mounted to the chamber using the custom
support. The holder mated to the chamber with the same interface geometry for the upright
orientation (a), inverted orientation (b), and vertical orientation (c), making the positions
completely interchangeable.
52
Imaging of the bubble dynamics over the ratchets was accomplished using a Phantom Vision
V310 high-speed camera, capable of 3200 fps at 1280 x 800 px in conjunction with an Infinity
Optics K2/SC microscope lens with CF-3 objective, providing a field of view between 8.39 mm
x 5.25 mm and 12.43 mm x 7.77 mm depending on the working distance. This setup allowed
for the observation of the boiling activity on the surface as well as in the plume of the test
section. The K2/SC microscope lens with CF-3 objective was chosen for its combination of
high magnification (2.06X to 3.05X), long working distance (125-92 mm respectively), and
narrow depth of field (0.02 mm at full aperture). The narrow depth of field allowed only
bubbles within 0.02 mm of the working distance to be in focus, making it easier to identify
only bubbles located in a two-dimensional plane (desirable in particle image velocimetry and
in bubble tracking from cavities located on a specific plane). Due to the weight of the K2/SC
lens, a custom bracket was designed and machined to relieve the lens mounting interface on
the camera from the cantilever forces excreted on it by the lens (Fig. 4.20).
Figure 4.20: Custom bracket designed to relieve the camera's lens mount adaptor from the
large cantilever weight of the K2/SC lens. The lens was hereby connected directly to the
structure of the camera.
The high-speed camera was mounted to two aluminum arms extending from the chamber
flanges using an ABS base plate as the adaptor. Slots were machined in the plate to allow it
to slide on the arms, permitting the camera to be moved closer or farther away from the
chamber to keep the ratchets within the working distance range. The 5.80 cm (2.00 in) hole
spacing in the flanges permitted the entire high-speed camera assembly to be moved up or
53
down in 5.80 cm increments, allowing the ratchets to remain in view for any of the three
orientations required; the bottom three holes were used in the upright orientation, the
middle three in the vertical orientation, and the top three in the inverted orientation. A
second hole pattern was created in the ABS base plate that also allowed the camera to be
moved 5.80 cm (2.00 in) to the side when imaging the wafer in the vertical orientation (Fig.
F.21 in Appendix F).
(a)
(b)
(c)
Figure 4.21: High-speed camera mounted to the chamber. Camera position is shown for the
upright orientation (a), inverted orientation (b), and vertical orientation (c).
Because high speed imaging required focused images (minimal motion-blur) at very high
frame rates, the exposure time was limited to just tens of microseconds. This short exposure
time introduced a secondary requirement for high-intensity illumination of the subject, and
in the case of this experimental facility, high-intensity backlighting, as backlighting allowed
the bubble to appear as a dark silhouette against a bright background. Initial experiments
were carried out using both incandescent bulbs and halogen bulbs behind a paper diffuser,
but these methods did not provide a uniformly illuminated background, did not ensure
repeatability in bubble tracking, and were eventually abandoned in favor of a densely-packed
array of white LEDs.
A custom 12.7 cm x 12.7 cm circuit board array of 625 white LEDs each of 17,000 mcd
intensity was designed to distribute the backlighting over the entire window area. To allow
for the densest possible packing of 2.54 mm (0.100 in) center-to-center, the base flange on
54
each LED was removed using a sanding wheel. To maximize reliability of the array, each LED
was fitted with its own current limiting resistor (100Ω) on the back of the PCB (Fig. 4.22).
100Ω
100Ω
LED
(20 mA)
LED
(20 mA)
5V
(a)
(b)
Figure 4.22: LED Array circuit. Shown is a picture of the current limiting resistors on the back
of the array (a) and schematic of the LED circuits arranged in parallel (b). Each LED had its
own 100Ω current limiting resistor on the opposite side of the board to maximize reliability
of the array.
The total power dissipation of the array was 62.5W, so two small fans were added to keep it
cool. The LED array was powered using a 5V source and the fans were powered using a 12V
source. The power supply chosen for this task was a standard ATX form factor computer
power supply as it supplied all the necessary voltages in a single unit. For this reason, a
standard 24-pin through-hole ATX power connector was installed in the LED array’s circuit
board to interface the LED array with the power supply. In order to power the fans, the 12V
power was routed from the 24-pin connector to two small 2-pin connectors on the board,
providing a 12V source for both fans. An aluminum sheet metal cowling served as a duct for
the fan flow, the mounting interface to the chamber flanges, and a structure to mount the
LED array and fans. Figure 4.23 shows a cutaway view of the chamber and LED array; a
diffuse polycarbonate plate was placed between the array and the window to provide
uniform backlighting. Figure 4.24 shows the actual array, its installed configuration, as well as
its backlighting performance.
55
Figure 4.23: A cutaway view of the chamber showing the position of the LED array.
(a)
(b)
(c)
Figure 4.24: The LED backlighting array. The array is shown before installation showing the
dense packing of 625 17,000 mcd white LEDs (a), in the installed position between the
chamber flanges showing the placement of the cooling fans and the aluminum ducting (b)
and in operation with the diffuser installed, providing intense diffuse backlighting for the
high-speed camera (c).
56
4.6. Experimental Facility
Because the future intent for this experiment was to fly it in a parabolic flight aboard aircraft
operated by organizations such as Zero Gravity Corporation, ESA, or NASA, the chamber and
all of the supporting equipment was to be contained within a single unit to ensure it meets
the standards set for such experiments in parabolic flight [34]. As such, all equipment needed
to be securely mounted to a rack, including data acquisition devices, cameras, power
supplies, PID controllers, switches, etc. Table 4.4 and Table 4.5 show the instruments and
equipment required to perform experiments.
Table 4.4: List of instruments employed in the experimental facility.
Instrument
Model #
Measured
Specifications
Variable
KMQSS-125U-6 Temperature -200°C – 1250°C
K-Type
Thermocouples
Accelerometer ADXL335
Thermocouple National
Data Acquisition Instruments
USB-9211
Accelerometer National
Data Acquisition Instruments
USB-6009
Acceleration +/-3g
Voltage
Voltage
Uncertainty Mounting
± 0.3°C
Fixed
Not
Calibrated
N/A
Fixed
4 channels
Built-in cold junction
12 samples/sec total
8 Analog input
N/A
channels (14-bit, 48
kS/sec)
2 Analog output
channels (12-bit 150
S/sec)
12 digital I/O
channels
32-bit counter
Fixed
Fixed
57
Table 4.5: List of equipment employed in the experimental facility.
58
Table 4.5 (continued)
Equipment Model #
High-Speed Phantom V310
Camera
High-Speed
Camera
Microscope
Lens
Infrared
Camera
Infrared
Camera
Microscope
Lens
Gear Pump
Infinity K2/SC
body
CF-3 objective
lens
Nippon Avionics
TVS-8500
Nippon Avionics
TVM-8510U
LED Array
Custom
PID
Controller
Omega
CNi3253-DC
Cartridge
Heaters
Watlow
G6A56-BG12
High Current
Relay
Voltage
Multimeter
Current
Multimeter
Omega
SSR330DC50
Tektronix
DMM912
Tektronix
DMM912
Cole Parmer
WU-00132-YQ
Purpose
Specifications
Mounting
High-speed imaging of flow 3287 frames per
Fixed
bubble dynamics
second
1280x800 px
resolution
26.5 x 11.9 mm CMOS
sensor
Magnification for High2.06X – 3.05X mag
Fixed
Speed Camera
125-92 mm working
distance
IR Thermography of test
section
Magnification for Infrared
Camera
-40°C – 900°C range
0.025°C resolution
2.56x2.36 mm Field of
View
30 mm working
distance
Pumping/recirculation of Max pressure 21 psi
pool fluid
max flow rate 0.67
gpm
Backlighting for high-speed 625 LEDs: 17,000 mcd
imaging
each
2 fans: 12 V, 60 mm
Maintaining pool
± 0.05°C/°C
temperature at desired
thermocouple
setting
temperature stability
Providing the control to
2x 600 W, 120 VAC
maintain the pool at a given
temperature
Switching the power to the 24 – 340 VAC, 50 A
cartridge heaters
4 – 32 VDC control
Measuring voltage across Estimated ± 0.005 V
the thin film heater
Measuring current through Estimated ± 0.0005 A
the thin film heater
Fixed
Fixed
Fixed
Fixed
Fixed
Fixed
Fixed
Fixed
Fixed
59
Table 4.5 (continued)
Equipment Model #
DC Power
PowerMan
Supply for IP-P350AJ2-0
Experimental
Facility
DC Power
Tenma
Supply for 72-8350
Test Section
Coolant
Thermo Scientific
Circulator
ThermoHaake
C30P
Purpose
Specifications
Mounting
Supplying power to various 3.3V, 5V, 12V, 350 W Fixed
experimental facility
systems
Powering the test section
thin-film heater
1-20 VDC, 5 A
External
Providing the cooling
800 W Heat Removal External
capacity for the condensing at 3°C
equipment
24 L/min flow rate
In order to mount the instruments and equipment effectively, a rack was designed to fully
contain the experimental facility within its largest dimensions in the configuration(s)
expected during parabolic flight. All three configurations (upright, vertical, inverted)
including the infrared camera were still to be possible, but the infrared camera and Graham
condenser would not need to be contained within the rack’s largest dimensions as these
would not be used in the flight configuration; the condenser would be ineffective in
microgravity, and the IR camera was not expected to contribute significant data because the
images recorded by the high-speed camera would be far more meaningful. Eliminating the
Graham condenser and infrared camera from this configuration allowed to rack to be more
compact, and would reduce the operator’s workload in the aircraft. Figure 4.25 shows the
experimental facility with all attached instruments and equipment as it was most frequently
configured for the work discussed in this thesis (i.e. without the infrared camera). All
instrumentation and switches were securely mounted in one area for efficient use.
60
(a)
(b)
Figure 4.25: Overview of the experimental facility. The entire experimental facility is pictured
in (a); overall dimensions and placement of modules is shown in (b).
61
The rack was built using a standard T-slotted aluminum frame system. This not only allowed
for the easy mounting of all modules, but also would allow the frame members to be
recycled for use in future experiments if so desired. All frame members were machined to
size on a milling machine to ensure a good fit. The ends of each primary frame section were
threaded with M8x1.25 thread to accept 3-way corner couplers. This ensured easy mating
and guaranteed perpendicular alignment of the frame members. The only modifications
made to the frame members were the thirteen 9.53 mm (0.375 in) holes spaced 2.54 mm
(1.000 in) apart in two vertical members to which the chamber was bolted. The chamber
together with the attached high-speed camera weighed an estimated 52.3 kg (115 lbs), which
prompted a stress analysis to insure that mounting it to the rack as a cantilever would not
result in material failure in the lab or in future flight experiments for which the facility must
withstand a maximum loading of 9g [34].
A simple single-part static stress analysis was set up to determine if the intended design was
strong enough to meet flight-worthy standards. The weight and location of the center of
gravity of the chamber and high-speed camera during 9g acceleration were simulated by a
single remote load of 2308 N (517.5 lbs), half of the total load at this acceleration as the
chamber is mounted to two such frame members symmetrically (Fig. 4.26a). The worst case
scenario for this analysis is that the frame member is a simply supported beam, meaning the
ends do not resist any internal moments. The results of the analysis, performed in Solidworks
COSMOS, are shown in a stress plot with greatly exaggerated deformation to show the
deformation mode (Fig. 4.26b). The analysis confirms that the maximum stress is not
expected to exceed 275 MPa, the yield strength for 6061-T6 Aluminum; the chamber was
subsequently installed in the rack in this manner.
62
(a)
(b)
Figure 4.26: Static stress analysis performed one of two vertical frame members that support
the chamber. The simulated load was applied at the center of gravity and was 4.5 times the
weight of the chamber with the high-speed camera attached, simulating 9g acceleration (a).
Results show that the maximum stress is not expected to exceed the yield strength of the
material (b).
4.6.1. Accelerometer
Because possible microgravity flight experiments would require the recording of
accelerations, an accelerometer was mounted as close as possible to the test section to one
of the vertical chamber support members discussed above (Fig. 4.25). The accelerometer
used was an ADXL335 3-axis accelerometer, purchased on an existing breakout PCB from
Sparkfun Electronics. The PCB was mounted to a custom non-conducting polycarbonate base
and clamped down since the PCB did not feature mounting holes (Fig. 4.27). A six-pin rightangle socket (Molex part # 70553-0109) was soldered in the board to provide the necessary
power, ground, and signal connections for reliable data acquisition. In addition to providing
acceleration data acquisition for documentation purposes, the accelerometer was also used
as a means of triggering the high-speed camera to initiate recording. Figure 4.28 shows the
components of the high-speed camera triggering setup. The Phantom V310 was purchased
with a breakout box for a number of the functions available in the user interface. Two of
63
these functions were of particular interest, namely “Trigger,” and “Event” (Fig. 4.28c). The
“Trigger” function initiates recording, and the “Event” function adds an event marker to the
timeline for the duration it is activate. Both of these functions are useful for autonomous
control of the camera using the accelerometer. A LabVIEW program was created that
compared the acceleration data to a user-defined threshold. When the magnitude of the
accelerations in each axis fell below their respective thresholds, the camera was triggered
through custom cables (Fig. 4.28a).
Figure 4.27: 3-Axis Accelerometer used to monitor chamber accelerations and act as a highspeed camera trigger. The accelerometer was mounted to a T-slotted frame member by the
chamber.
(a)
(b)
(c)
Figure 4.28: High-speed camera triggering setup. The camera’s breakout box controlled via
custom cables (a) by the accelerometer mounted to the experimental facility rack (b).
LabVIEW software monitored the accelerometer data and could control the camera’s trigger
function, event function, and if a third circuit is added, the pre-trigger function (c).
64
Figure 4.29 shows how the triggering was accomplished. A +5V signal was applied to the gate
of a transistor by the USB-6009 Module (triggered through LabVIEW), creating a conducting
path from the BNC connector’s conductor to ground, which was kept at the same potential
as the BNC connector shield. When the conductor was shorted to the shield, the camera
function corresponding to the connector to which the circuit was coupled was activated. This
configuration allows the computer to monitor accelerations and trigger the camera only
when the experimental facility experiences low gravity, as well as activating the timeline
marker when microgravity conditions are met, for example, providing the necessary
capability to operate the experimental facility autonomously during parabolic flight.
(a)
(b)
Figure 4.29: The camera triggering transistor circuit functioning as a solid state relay. Shown
is a picture (a) and schematic (b) of the circuit. The 5V source was applied by the USB-6009
Analog Input/Output Module and would short the BNC conductor to ground, triggering the
corresponding function on the high-speed camera’s breakout box.
4.6.2. Electrical Connections
Since there are a number of electrical systems in the experimental facility operating on a
variety of voltages, the power supply chosen to power the DC systems was a standard ATX
form factor computer power supply (PowerMan IP-P350AJ2-0) as it provided 12V for the
heater relay, the PID controller, the fans on the LED array, and the gear pump, 5V for the
LEDs on the LED array, and 3.3V for the accelerometer, providing all of the necessary DC
sources with a single power supply. Standard electrical connectors were used to connect the
power supply wiring harness to the various systems so that any one of them could be
disconnected if so desired. The cartridge heaters were powered directly from a wall outlet
120VAC source, switched by a relay which was controlled by the PID controller. Figure 4.30
65
shows a schematic of all of the electrical connections in the experimental facility, and Fig.
4.31 shows images of various systems and connections in the experimental facility.
23.3Ω Cartridge
Heaters
120 VAC
Control Panel
Chamber
Heater Relay
K-Type
Thermocouples
PID
Controller
Power Supply
12V DC
Gear Pump
LED Array Fans
5V DC
USB-9211
DAQ
Computer
LED Array
120 VAC
DAQ
3.3V DC
Accelerometer
USB-6009
Transistor
Triggering
Circuit
High-Speed
Camera
Figure 4.30: Schematic of electrical connections for the cartridge heater circuit and the
experimental facility systems. All electrical connections were made using properly rated
connectors so that systems could be easily disconnected from the circuit. A separate 120VAC
source was used to power the cartridge heaters to avoid excessively high current.
(a)
(b)
(c)
(d)
Figure 4.31: Overview of various electrical systems in the experimental facility. Shown is the
primary power supply (a), the control panel for all of the facility’s electrical systems (b), the
wiring for the PID controller, relay, and power switches (c), and the wiring harness for the
LED array (d).
66
5. EXPERIMENTAL PROCEDURES
Because a new phenomenon was being quantified and because of the fragility of the test
section and use of sophisticated hardware, it was of the utmost importance that a strict
procedure was followed. This chapter outlines the procedures followed during startup,
operation, and shutdown of the experimental facility. A detailed description of each step is
provided first, followed by a summary in the form of a flowchart that can be used as a
reference for future experiments. It should be noted that although accommodations were
made for future use of Particle Image Velocimetry (PIV), Laser-Induced Fluorescence (LIF),
and the Infrared camera, these systems were not used in any of the experiments discussed in
this thesis, and are therefore not discussed in these procedures. The startup procedures
differed slightly depending whether water or FC-72 was used as the working fluid. In general,
more care was required when handling FC-72 due to its volatility and cost. Procedures
specific to FC-72 are provide in Appendix B since this fluid was not used for any of the results
discussed in the thesis.
5.1. Startup Procedure
Significant preparations were made before experiments could take place. Primary concerns
for experiments were cleanliness of the wetted surfaces, purity of fluids, and fragility of the
test section. Two hours was usually required to fully prepare the facility and the test section
for experiments.
5.1.1. Cleaning the Chamber
Before powering any systems, the first step was cleaning of the chamber to remove any
particulate matter. This insured that there would be no visible particulate in the images
captured by the high-speed camera, and that the test section would remain uncontaminated.
In order to clean the chamber effectively, the top and bottom circular ports were opened.
The chamber was cleaned first by thoroughly blowing it out using a standard filtered
compressed air source, this removed any loose particulate. A tray was then placed below the
67
chamber and at least two liters of deionized water were poured inside along the walls and
windows of the chamber using a funnel and tube, and drained out the open circular port in
the bottom of the chamber. This rinse removed particulates adhering to the walls that were
not dislodged by the compressed air.
5.1.2. Cleaning and Mounting the Test Section
The silicon wafer with the test section was kept stored in a sealed container, and the topside
was cleaned before being mounted in the chamber. It was first rinsed with acetone followed
by deionized water and then left to dry. In order to ensure an optically clear path between
the test section and the camera, the polycarbonate clamp walls were cleaned using isopropyl
alcohol. Once cleaned, the clamp assembly was rinsed with deionized water and left to dry.
The presence of the O-ring in the groove on the polycarbonate wafer holder was checked
and the silicon wafer was placed on top of the holder with the ratcheted surface of the test
section facing up. The four contact pads on the thin film heater on the backside of the wafer
were visually aligned with the four copper pogo pins in the holder to ensure there would be
good electrical contact. Alignment of the test section was verified to be orthogonal with
clamp, so that the camera would capture events occurring between the crests of the ratchets
such as bubble growth and departure. Once the wafer was adequately aligned on the holder,
the Polyetheremide wafer clamp with optically clear polycarbonate sidewalls was carefully
placed over the wafer in such an orientation that the clear sidewalls were parallel to the
ratchet pattern direction.
68
(a)
(b)
(c)
Figure 5.1: Installation of test section in chamber. After the wafer is clamped to the holder
(a,b) the assembly is installed in the desired orientation in the chamber (c).
Clamping the wafer was a tedious process, as the pogo pins kept the wafer from resting in its
designated recess in the top face of the holder. For this reason, the clamp was very carefully
screwed to the holder while alignment was periodically checked (Fig. 5.1a). Only the two
#10-32 x 19.05 mm (0.750 in) stainless steel button-head cap screws were used for the initial
clamping. Once the wafer rested in its recess, the remaining four #10-32 x 38.10 mm (1.500
in) stainless steel cap screws were inserted. In order to prevent the wafer from fracturing
due to non-uniform compression against the O-ring, all six screws were tightened iteratively
in a star pattern until the clamp mated to the holder (Fig. 5.1b). When mated, the O-ring was
compressed by the wafer to make the seal. This method proved successful as it did not result
in a single failure of any of the silicon wafers.
After installation of the test section, all four electrical connections were checked with a
multimeter to ensure proper contact and the entire test section assembly was installed in the
desired chamber port using twelve #10-32 x 15.88 mm (0.625 in) stainless steel cap screws
with stainless steel washers (Fig. 5.1c). The O-ring seal was verified by checking that the
wafer holder mated flush to the chamber.
69
5.1.3. Filling and Degassing
Before filling of the chamber, the condensing loop was coupled to the Thermo Scientific
Thermo Haake C30P Heating Refrigerated Water Bath Circulator. The chiller was set to 4°C
and the pump speed was dialed to the high setting and allowed to reach steady state.
Operation of the cooling loop was checked by confirming the reflux condenser was filled with
cold water.
The chamber drain valve was set to the closed position, and the pump selector valve was set
to the fill position, allowing the pump to draw fluid from an external source. Because the
gear pump’s dry performance was always very poor, the liquid reservoir it drew from was
placed at a level higher than the pump, and in the case of water, some was syphoned
through the clear PFA tubing before connecting it to the pump’s feeder line, insuring there
was a positive pressure feeding the water to the pump. A second length of PFA tubing was
connected to the chamber drain line and inserted into the reservoir, preventing spills during
drainage or due to an overflow.
The LED array was turned on first, allowing for a clear view of the inside of the chamber. This
allowed allow any problems such as a low fill rate (usually caused by a plugged filter) inside
the chamber to be spotted more easily. The pump was then switched on, drawing fluid
through clear PFA tubing from the reservoir, and pumping it through the inline 15-micron
sintered stainless steel filter into the chamber. Additional clear PFA tubing on the pump inlet
and outlet served as a means of checking if the pump was pumping liquid or if it remained
dry. While the chamber was filling, all O-ring and compression fitting seals were checked for
leaks. However, because all tests were conducted at atmospheric pressure, and the pump
pressure was always low, leaks were rare. Any leaks that were detected were the result of
improper compression of an O-ring or of the failure to tighten the compression fitting to the
specifications. The pump was left on until the chamber was filled to a depth of at least 18 cm
for the upright orientation. Once filled to the desired level, the pump selector valve was set
to the recirculating position, temporarily switching the pump off in the process. This allowed
70
the pool to be filtered continuously during the degassing phase. The reservoir was set on the
ground, allowing any potential overflow to drain back into the reservoir.
After filling the chamber, the power cable for the cartridge heaters was plugged in, the PID
controller set temperature (“SP2” on the controller) was set above the saturation
temperature of the liquid (100°C for water) to ensure continuous heating, and the cartridge
heaters were switched on. Prior to this point, the power switch to the cartridge heaters was
set to the “off” position and the power cable for cartridge heaters was unplugged. This was a
safety precaution that insured the heaters would not be powered while there was no liquid
in the chamber, precluding dangerously high temperatures and burnout of the heaters.
In addition to the rising pool temperature displayed by the PID controller, the presence of
convective currents over both cartridge heaters served as visual confirmation that the
heaters were functioning properly. The pool was left at saturation temperature for at least
one hour to properly degas the fluid. While degassing, the test section was powered to
initiate nucleation in the cavities. As described in Section 2.1, hysteresis can occur at the
onset of nucleate boiling. The test section was therefore supplied with up to 4.00 A of
current for at least one minute to initiate nucleation from all cavities, ensuring wallsuperheat conditions were uniform everywhere on the test section. The current was
gradually increased from zero to prevent any sharp temperature gradients from cracking the
silicon wafer. The current and voltage were each monitored by a Tektronix DMM912 digital
multimeter. After degassing and initiating nucleation on the cavities, the pump was switched
off.
5.1.4. Additional Hardware and Software
The high-speed camera was mounted in a position corresponding to the test section
orientation as described in Section 4.5.5. Before powering the Phantom Vision V310 HighSpeed Camera, the K2/SC Microscope Lens with CF-3 objective was mounted to it, and the
lens support bracket was installed. The camera’s power supply was then plugged in (instantly
powering up the camera) and the Ethernet cable was plugged into the computer. After
waiting until the LEDs on the back of the camera labeled “Ethernet” stopped blinking, the
71
Phantom Software was started, and the image was displayed by selecting “Setup and
Recording…” from the “Acquisition” pull-down menu. This order of operations was critical as
starting the Phantom Software too early could corrupt the camera’s firmware and make it
unusable.
To permit temperature and acceleration data monitoring/acquisition, as well as external
camera triggering, the temperature data acquisition module (National Instruments USB9211) and accelerometer data acquisition/camera triggering module (National Instruments
USB-6009) were plugged into the computer’s USB ports. Once the modules’ indicator LEDs
turned on, the LabVIEW data acquisition program was started, permitting the chamber
accelerations in three directions and internal temperatures to be monitored, and the highspeed camera to be triggered through the LabVIEW software as described in Section 4.6.1.
72
Clean chamber
 Blow out chamber with filtered compressed air
 Rinse inside of chamber with deionized water
Will FC-72
be used?
Yes
No
Install Test Section
 Clean test section with Acetone and rinse with deionized water
 Clean wafer clamp with Isopropyl Alcohol
 Check for presence of O-ring
 Align wafer with pogo pins in holder
 Check for good electrical connection to thin-film heater
 Clamp wafer to holder by gradually tightening screws in a star
pattern, continuously checking alignment
 Check electrical connections with multimeter
Fill Chamber
 Connect coolant hoses to condensing loop
 Turn on chiller pump and set temperature to 4°C
 Remove and flush chamber filter and reinstall
 Close drain valve, set pump selector valve to fill position
 Connect feeder tube and insert into liquid reservoir
 Turn on LED array for visibility
 Turn pump on and fill to desired level
 Turn pump off, set pump selector valve to recirculation
position and turn pump on again
Initiate Test Section Nucleation
 Power test section to initiate nucleation on all cavities
 Set chamber temperature to above saturation
temperature on PID Controller (“SP2”)
 Turn pump off.
 Blow out fill and
drain loop with
filtered compressed
air
 Allow 24 hours for
chamber to dry
thoroughly
Heat Pool
 Connect power cable for
cartridge heaters
 Set chamber temperature
above saturation
temperature on PID
Controller (“SP2”)
 Switch cartridge heaters on
Yes
Saturation
temperature
Reached?
No
Additional Hardware and Software
 Turn on high-speed camera and start software
 Plug in DAQ devices and start LabVIEW program
Wait 1 hour
Figure 5.2: Flowchart for startup procedures.
73
5.2. Continuous Operating Procedure
Before acquiring high-speed videos, the camera lens was set to focus on the desired row of
cavities. By selecting the “Edge Highpass 5x5” edge detection filter in the image processing
dialog box in the camera user interface, the image could be focused more accurately,
adjusting the lens focus to where bubble edges appeared brightest. By default the camera’s
aperture was set to 100% open to let the maximum amount of light in, permitting the
shortest exposure times. Short exposure times ensured the image was subjected to very little
motion blur. Additionally, a full open aperture maintained a narrow depth of field (as narrow
as 0.02mm), greatly blurring any bubbles outside the plane of interest, and thus allowed the
edge detection filters in the post processing to resolve only bubbles within the plane of
interest.
The test section was then supplied with the desired current, and the temperature controller
set-point temperature “SP2” was set to achieve the desire subcooling and left to reach
steady state, at which point the current and voltage displayed on the multimeters was
entered in the LabVIEW interface and the camera mode we set to “Capture,” awaiting the
external trigger signal. Before triggering the camera, the cartridge heaters were switched off
to prevent large scale convective flows from possibly affecting the flow near the test section.
At least 30 seconds were allowed to pass. To record data, the “Trigger Camera” button on
the interface was clicked, simultaneously recording temperature, current, and voltage data
and sending the trigger command to the USB-6009 module, shorting the camera’s “Trigger”
BNC connector on the breakout box. The accelerometer could also be used as a trigger if so
desired: a threshold could be set for each axis, and when the magnitude of the recorded
accelerations fell below the threshold of each axis, the data acquisition and camera would be
triggered. After data acquisition, the cartridge heaters were switched on again, and either a
new pool temperature was set in the PID controller, or a different current was supplied to
the test section, upon which the same procedures were followed to record the next set of
data.
74
Configure High-Speed Camera
 Set lens to focus on desired row
of cavities
 Set aperture to 100% open
 Set exposure time as short as
possible without losing bit depth
Set Experimental Conditions
 Supply desired power to test section
 Set desired pool temperature on PID
Controller, insure cartridge heaters are on
Prepare for data acquisition
 Enter Current and Voltage readings in LabVIEW
 Activate high-speed camera “Capture” mode
 Turn off cartridge heaters
Wait 30 seconds
for convective
currents to settle
Data acquisition
Trigger camera through LabVIEW interface to
simultaneously acquire temperature data
Yes
Perform
another
experiment?
No
Proceed to shutdown
Figure 5.3: Flowchart for operation procedures.
75
5.3. Shutdown Procedure
Shutdown began by gradually reducing the current supplied to the test section until it
reached zero. Shutdown of the test section was confirmed using the high-speed camera’s
real-time display, confirming the absence of convective activity. Once confirmed visually and
by instruments, the power supply was disconnected from the test section. The cartridge
heaters were also switched off and the cartridge heater power cable was unplugged from the
outlet and the experimental facility. Disconnecting heater power sources first ensured that it
the test section and/or cartridge heaters would not be powered while the chamber was
drained, precluding burnout of the heaters. Next the chamber drain valve was set to open,
allowing the pool to completely drain back to the reservoir after which it was set to closed.
The chiller was turned off and the cooling fluid in the reflux condenser was allowed to drain,
after which the condensing loop inlet and outlet hoses were disconnected. Next the LED
array was switched off and the experimental facility’s main power supply was switched off.
The LabVIEW program was closed and the USB-9211 and USB-6009 modules were unplugged
from the computer.
In order to prevent corruption of the camera’s firmware, the high-speed camera shutdown
procedure was performed as follows: first the Phantom program was closed, then the
Ethernet cable was unplugged from the computer and then the power supply was
disconnected from the camera.
76
Reduce power to test section
 Ramp down the test section
power, avoid abrupt power cutoff
to prevent cracking of the wafer
Confirm power cutoff
 Check that the current and voltage measurements read zero
 Confirm absence of convective activity over test section
Turn off cartridge heaters
 Put the PID controller in standby mode
 Turn off cartridge heaters
 Unplug cartridge heater power cable from
experimental facility
Drain chamber
Open drain valve
Turn off all equipment
 Unplug cartridge heater power cable from
experimental facility
 Turn off chiller and disconnect hoses
 Turn off LED array
 Turn off experimental facility power supply
 Close LabVIEW
 Disconnect DAQ modules from computer
 Close high-speed camera software
 Unplug high-speed camera Ethernet cable
 Disconnect high-speed camera power
Figure 5.4: Flowchart for shutdown procedures.
77
6. DATA ANALYSIS
This section covers all work performed to generate the data reported in Chapter 7. A detailed
description of the image processing algorithm written to resolve bubbles from the recorded
images and compute velocity, acceleration, diameter, and volume information of bubbles
that have departed from the surface is presented. Additionally, the calibration of instruments
and determination of uncertainties for all measured variables is discussed.
6.1. Determination of Bubble Kinematics
In order to obtain quantitative results for parameters like velocity, acceleration, and bubble
departure diameter, an image-processing algorithm was developed in Matlab®. The primary
focus of image processing described in this section was to resolve the individual bubble
velocities through bubble tracking. Briefly, bubble-tracking velocimetry entails finding the
displacement of each bubble from one frame (A) to the next (B) and then dividing by the
elapsed time to resolve the velocity. The bubble was first located within an image and
positively identified. Once positively identified, each bubble in the first frame was paired
with the corresponding bubble in the second frame. Any unmatched pairs were deleted and
resulting bubble pairs were converted to individual vectors and displayed on a full field of
view image of the resolved bubbles. The following paragraphs describe the image processing
steps in more detail.
After loading the image, the image histogram was stretched linearly over the entire grayscale
spectrum such that the maximum amount of contrast was achieved without washing out any
pixels (Fig. 6.1). Additionally, this ensured the images used in analysis always had a similar
histogram, improving repeatability.
78
(a)
(b)
Figure 6.1: Stretching the imported image histogram over the full grayscale spectrum. This
ensures contrast remains similar for all imported images, improving repeatability.
The first step in resolving the bubbles was to apply edge detection filters. Edge detection
filters are used to restrict the algorithm to tracking only bubbles within a two-dimensional
plane (i.e. the plane of focus). Since the K2/SC lens has a very narrow depth of field (as low as
0.02 mm at full aperture), the bubbles located outside the depth of field appear greatly
blurred, and are not resolved by the edge detection filters (depending on the sensitivity set
by the user). Three different filters – gradient, Laplace, and Sobel – were applied to the
image and the results of all filters were merged in the end. Using multiple filters allowed
certain bubbles to be resolved that individual filters may have missed. The perimeters of
bubbles still attached to the surface could not be resolved as there was no distinct edge
between the bubble and the surface to resolve; only departed bubbles could therefore be
resolved.
Figure 6.2 illustrates the steps of processing an image using the Sobel filter. Similar
processing steps were performed for the two other filters and the results were combined
prior to tagging the bubbles for tracking. Upon applying the edge detection filter (Fig. 6.2b),
the next step was to binarize the image (Fig. 6.2c). All pixels with a value above the userdefined binarization threshold were set to a value “1” and the remaining pixels were set to a
value of “0”. The optimal threshold value was determined iteratively for each of the three
filters and kept constant for all experiments to allow for meaningful comparisons.
Henceforth, the term “region” will be used to refer to any continuous group of binary pixels
with a value “1”. Following binarization, the image was further processed to resolve as many
79
regions as possible to represent the location and shape of bubbles in the image. Since the
edge detection filters only resolved bubble perimeters, the shape of the resulting region is in
the binarized image was usually a ring. For this reason all holes in the regions are filled to
obtain regions representative of a complete bubble (Fig. 6.2d).
(a)
(b)
(c)
(d)
(e)
(f)
Figure 6.2: Image processing steps to resolve bubbles. Shown is the original images (a), the
Sobel filter applied to original images (b), the binarized image (c), the filled hollow regions
(d), the applied convex hull criterion (e), and the deletion of noise and regions contacting the
image border (f).
It is possible that some bubble edges were not fully resolved (i.e. there is a gap in the edgedetected ring). To retain these regions, a convex hull process was applied as shown in Fig.
6.2e. The convex hull refers to the smallest envelope containing the region with continuous
convex geometry, and can also be thought of as the outline that a rubber band would trace if
it were pulled around the region. If the set of points representing the convex hull bridges a
gap that exceeds a threshold (set as a fraction of the region’s major diameter), the original
bubble edge is retained. However, if the threshold is not exceeded, the region is replaced
80
with its convex hull, effectively enclosing the incomplete edge of the bubble. Figure 6.2e
shows that the smaller bubbles are resolved using the convex hull whereas the larger bubble
to the lower left hand corner remains unresolved. The next image-processing step was
deletion of regions that were artifacts resulting from the background or from noise. These
regions were filtered simply based on an area threshold, specified by the user (Fig. 6.2f). Any
regions connected to the image border were also removed at this stage, as there would be
no way of resolving partial bubbles that move in or out of the frame in subsequent images.
After the regions were resolved, they were processed to retain only those regions that
represented the location and shape of bubbles in the image. Two criteria had to be met for a
region to “pass” as a bubble: (1) region solidity (area/convex area) > threshold – This criterion
was applied to eliminate regions resolved without continuous convex edges, and (2) region
eccentricity < threshold – This criterion was applied to eliminate regions with high
eccentricity. Figure 6.3 shows which regions were eliminated for not meeting the solidity and
eccentricity criteria, labeled red and green respectively. Such regions are unlikely to
represent the outline of a bubble, which is usually spherical in shape. All threshold values
were set by the user and could be adjusted to resolve bubbles whose shape would otherwise
not permit classification as a bubble.
81
Figure 6.3: Tagging of regions during image processing. Regions that did not meet the solidity
criterion (red) or eccentricity criterion (green) were removed, preserving the remaining
regions. Lighter shades represent regions from image A, darker shades represent regions
from image B.
Following identification of bubbles in each image, bubble tracking was performed. The first
step in bubble tracking was determining the locations of each region’s centroid. For any
region with equivalent diameter equal to or greater than 0.1 mm, the centroid was
computed as the centroid of the pixels in the region. For small regions with equivalent
diameters less than 0.1 mm) the centroid location was weighted by the intensity of the
corresponding pixels in the original image. Intensity weighting the pixels within the region
ensured that any misrepresentation of the dot due to poor focus or incorrect binary
threshold settings would have minimal effect on the accuracy of the centroid computation.
Due to the fairly low quantity of bubbles in each image, it was not necessary to use advanced
algorithms for determining bubble pairs between images. A nearest-neighbor approach was
used to match bubbles in image A to those in image B. Figure 6.4 shows the bubbles pairs
identified in image A and image B. For each region in image A, the distance between its
centroid and each region’s centroid in image B was calculated. The pair with the shortest
centroid-to-centroid distance was logged as a matched pair; their respective indexes in the
region-numbering scheme were logged in a separate table. Each matched pair was filtered
for violation of the area ratio threshold (area(A)/area(B) < threshold). This filter ensured that
only bubbles with similar area would be logged as pairs. Those regions that failed this test
82
were tagged orange and subsequently removed. If bubbles were tracked near the cavity in
subcooled conditions, the area ratio threshold was set high so that the collapse of a bubble
would not exclude it from analysis. If only the bubbles in the plume were of interest, then
this ratio was set low, as bubble collapse was not observed to occur in the plume.
(a)
(b)
Figure 6.4: Bubble pair area ratio criterion applied. If no violations of the area ratio criterion
occur, all regions pass and are paired (a) if an area ratio violation does occur, the
corresponding bubble pair is tagged orange (b) and excluded from the vector plot.
It is possible that two regions in image A are matched to a single region in image B. This often
occurred because the second region in image B was not resolved, likely because the bubble
moved out of the depth of field. In this case, the pair with the shortest centroid-to-centroid
distance was retained, as this pair most likely represented the slight shift in position the
bubble has made between image A and image B. Figure 6.5 shows a typical vector plot that
results from identification of bubbles in images A and B. The length of the vectors is
proportional to the square root of the bubble speed for better viewing of all vectors in the
speed range.
83
(a)
(b)
Figure 6.5: Velocity vector map obtained from bubble tracking velocimetry. Each bubble
could be annotated quantitatively with, speed, equivalent diameter, and volume information
as well as details on the images used (a), or qualitatively with vectors only (b). Vector
magnitudes are proportional to the square root of the bubble speed.
Because the shape of the individual regions was rarely circular, the diameter of each bubble
is computed an equivalent diameter for a circle with area equal to that of the region.
√
[
(6.1)
]
Equation 6.2 provides a good estimate of the diameter for the purpose of comparing the
data to existing models of bubble departure diameter. The bubble volume was computed
assuming the region represents a cross-section of an oblate spheroid (i.e. the minor axis is
the axis of symmetry for the volume):
(
) (
)[
]
(6.2)
The length of the major and minor axis used in Eq. 6.2 was computed automatically by
Matlab® and given in terms of pixels. This was deemed the most accurate method to
84
estimate bubble volume knowing only the shape of the cross section, and was considered
quite accurate for bubbles appearing circular or elliptical in shape. However, for greatly
deformed bubbles this is likely a meaningless quantity, as there is no axis of symmetry in
such a case.
Additional error is expected from the bubble diameter calculations, as there was no means of
determining the accuracy of the equivalent diameter estimates. Figure 6.6 shows a cropped
portion of an image with five regions as resolved by the gradient, Laplace, and Sobel filters
(red, green, and blue, respectively).
Figure 6.6: Estimated outline of bubbles as resolved by the filter. Shown is the gradient filter
(red), Laplace filter (green), and Sobel filter (blue).
The gradient and Sobel filters often agreed quite well, whereas the Laplace filter usually
resolved smaller regions. The parameters that controlled the size of the resolved region were
the binarization threshold and the filter operator size; lower binary thresholds allowed
fainter pixels to be included in the region and smaller filter operator dimensions increased
the sensitivity of the filter, meaning only sharp gradients would pass as edges. The most
effective binarization threshold and filter operator size were determined iteratively for each
filter, and kept constant for all experiments. However, the ambiguity as to which filter is
more accurate indicates that there is likely some error associated with estimating bubble size
using these filters. Since the final region size was a combined result of all three filters, in
85
most cases this simply meant the largest of the three. However, as there was no means of
comparing the resolved data to the true values, the uncertainty for this error could not be
computed. A flowchart describing the algorithm along with the purpose of the individual
functions used is provided in Appendix C.
6.2. Calibration and Uncertainty Analysis
In order to provide confidence in the reported results, all instrumentation was calibrated to
report accurate measurements if the capability existed. The instruments calibrated were the
high-speed camera and the four individual thermocouple probes. In addition to calibration,
due to the number of instruments and data processing procedures, a number of sources of
uncertainty were expected. Each measurement recorded was expected to have some error
associated with it, namely thermocouple measurements, heat flux measurements, motion
blur, timing of the camera frame rate, and distortion in the images acquired from the camera
caused by the lens as well as the temperature-induced density gradients over the test
section. Data processing errors were expected in image processing and use of the
thermocouple calibration curves. Some human sources of error were present as well,
specifically in determining the resolution of the image acquired by the camera. The following
uncertainties were not computed for the listed reasons:
Timing of Camera Frame Rate
No uncertainty specifications were available for the Phantom V310 camera’s frame rate.
However, given internal clock frequency in modern chips/computers, the uncertainties are
expected to be on the order of nanoseconds and are therefore orders of magnitude lower
than deemed significant.
Bubble position as a function of distortion due to duration of exposure
As the velocity measurements rely solely on the relative change in a bubble’s centroid
position from one frame to the next, unquantifiable errors in absolute centroid position due
to motion blur and edge detection errors were not expected to significantly influence this
value.
86
Bubble position as a function of image distortion due to gradients in fluid density
There was no available means to insert a standard optical grid into the 8 mm x 5 mm field of
view above the test section to compare with the in-situ resolution measurements.
Additionally, the mere presence of the optical grid would greatly affect the local temperature
gradients, rendering any calibration meaningless. Furthermore, the temperature gradientinduced distortion varies greatly with applied heat rate, fluid, and pool temperature; any
attempt to calibrate for this would require local temperature measurements during
calibration and experiments which would obscure the view of the test section.
Experimental Facility Acceleration
The accelerometer was not calibrated as it was not used for the work presented in this
thesis.
6.2.1. Calibration of High-Speed Camera
Calibrating the camera was necessary to compensate for any pixel offset on the CMOS chip
caused by the recording parameters set in the user interface. There are a number of
variables that can contribute to this pixel offset error, namely the temperature of the CMOS
chip, the frame exposure time, extended dynamic range (a function that sets a lower
exposure time for brighter pixels), frame rate, and resolution. Fortunately calibration could
be performed using the Phantom software’s built-in Current Session Reference calibration
function. This function allows the camera to adjust for pixel offset as it may be affected by
the aforementioned variables. To perform this calibration, the camera lens was covered and
the Current Session Reference button was clicked.
6.2.2. Calibration of Thermocouples
All four K-type thermocouples were calibrated using a Heating Refrigerated Water Bath
Circulator (Thermo Scientific Thermo Haake C30P) and an NIST standard (Omega PCL-1B with
PCL-HR-1 module and PR-11-2-100-1/8-9-E-TA4F RTD probe). As the intended range of
operation for this experimental facility was between room temperature and 100°C, the
87
thermocouples were calibrated between 5°C and 95°C in 10°C increments, relying on
extrapolation of the calibration curves to extend the calibration to 100°C.
The thermocouples were to be mounted in the circular port on the side of the chamber when
performing experiments in the upright or inverted orientation. To minimize dangling wires
and cables in the facility, the thermocouple leads which connect the probes to the data
acquisition unit (National Instruments USB-9211) were made as short as possible (Fig. 6.7a).
However, when performing experiments with the test section in the vertical orientation, this
port was occupied by the wafer holder, and the thermocouples were installed in the bottom
port. Since the existing leads were too short to reach the probes in this configuration, a set of
extension cables was made to bridge the gap, and were only used when the test section was
mounted in the vertical orientation (Fig. 6.7b). The addition of these cables meant that eight
calibration curves needed to be generated, four with the extension, and four without.
(a)
(b)
Figure 6.7: Use of thermocouple extension cables. The standard configuration was used for
when the test section was installed in the upright and inverted orientations (a), and the
extended configuration was used for when the test section was installed in the vertical
orientation (b).
The uncertainties for the four probes in both configurations (with or without extension cable)
were computed as the square root of the sum of squares. The uncertainty for each probe
88
was dominated by the uncertainty of the standard; average uncertainty was 0.3°C. Details of
the analysis and calibration curves are provided in Appendix D.
6.2.3. Uncertainty in Applied Heat Flux
Uncertainty in the applied heat flux was a function of the measured voltage and current, as
well as the area of the heated surface between the thermal isolation trenches. The
uncertainty in the heated area dimension measurements was estimated to be ± 0.05 mm,
and the uncertainties of the measured current and voltage were estimated based on the
maximum precision at which the DMM912 multimeters could provide a stable reading, ±
5x10-5 A and ± 5x10-4 V respectively.
(6.3)
The total uncertainty in applied heat flux was estimated using the Kline and McClintock
method. Uncertainty values and details regarding the analysis are provided in Appendix D.
6.2.4. Uncertainty in Image Processing
There are a number of uncertainties in the imaging process that can affect the final velocity,
acceleration, and bubble diameter calculations. The quantifiable uncertainties in this system
are image distortion do to the optics, specifically the K2/SC microscope lens and CF-3
objective, uncertainty in the image resolution due to uncertainty in the scale of the reference
geometry in experiments, and the accuracy of the measurement of that reference geometry.
Reference geometry was needed as in-situ measurements of distance were required during
experiments to determine the resolution in pixels per mm of the acquired images.
Any vector ⃑ in the image in dimensions of mm was computed as follows:
⃑
⃑⃑⃑⃑⃑⃑
(6.4)
where ⃑⃑⃑⃑⃑⃑ is the measured distance between two points in pixels. The image resolution was
determined by visually observing the peaks of the farthest left and farthest right ratchets,
89
and counting the number of pixels between the two (Fig. 6.8). This number was then divided
by the known distance, computed as a multiple of the ratchet periods.
Figure 6.8: Method of calculating the image resolution. The number of pixels between the
farthest left and right ratchet peaks was counted and divided by the known distance (1 mm
per ratchet period).
*
+
(6.5)
Substituting Eq. 6.5 in Eq. 6.4 gives
⃑
⃑⃑⃑⃑⃑⃑(
)
[
]
(6.6)
Uncertainties exist in each of the variables in Eq. 6.6. The first uncertainty that is discussed is
the uncertainty in ⃑⃑⃑⃑⃑⃑; both the uncertainty in length and direction are considered. Before
considering uncertainty it is important to define the source of error in this measurement. The
error ⃑⃑ in ⃑⃑⃑⃑⃑⃑ can be represented in terms of the lens distortion error ⃑. It can be shown that
a uniform distortion error over the entire image has no effect on the magnitude or direction
of a resolved vector, as both points are shifted by the same amount of distortion in the same
direction (Fig. 6.9).
90
Figure 6.9: Vector distortion error in a uniform distortion field. The distortion field has no
effect on the resolved vector length or direction. The solid outlines represent resolved
positions and the dotted outlines represent the true positions.
Although the position of the vector has shifted slightly, this will not affect the results, since
length and direction are preserved. However, if a gradient in the distortion exists (i.e. the
magnitude of the distortion error varies throughout the image), the distortion error in one
location might be greater than the distortion error in another. In this case the length and
direction of the actual vector is different than that of the resolved vector (Fig. 6.10).
91
Figure 6.10: Vector distortion error in a variable distortion field. Length and the direction of
the resolved vector no longer correspond to the actual vector. The solid outlines represent
resolved positions and the dotted outlines represent the true positions.
Depending on the angle between the gradient in the distortion error and the resolved vector,
the error in the resolved vector becomes
⃑⃑⃑⃑⃑⃑⃑⃑
⃑
⃑
(
)
⃑
̂
(6.7)
Eq. 6.7 shows that the error is dependent on the angle. Since the uncertainty sought is a
global uncertainty, this angular dependence must be dropped. A conservative (worst case)
assumption that the resolved vector is in the direction of the gradient in the distortion error
(i.e.
( )
) simplifies Eq. 6.7 and produces a scalar value of the largest possible error in
the magnitude of ⃑⃑⃑⃑⃑⃑,
|⃑⃑⃑⃑⃑⃑| :
|⃑⃑⃑⃑⃑⃑|
where
⁄
(6.8)
is the magnitude of the gradient in the distortion error of the image.
Additionally, the direction of the actual vector can be computed:
92
(
)
(
(
)
)
(
(
(
)
)
(6.9)
(6.10)
)
The error in the angle of the new direction can be computed as the difference between
and
:
(
(
)
)
(6.11)
To convert this error into an uncertainty, the largest possible value must be found. In order
to calculate
, the distortion error gradient
⁄
must be found first. The distortion error
gradient was computed by first creating a distortion error map, and subsequently calculating
the gradient of that map. The distortion error map was created by using a 25 mm x 25 mm
fixed frequency grid distortion target with dots of 0.0625 mm diameter and 0.125 mm
spacing (Edmund Optics stock no. NT58-509) and comparing the recorded image to a perfect
square grid generated using Matlab®. The distortion target was placed between the lens and
the LED array, in the working distance range of the K2/SC lens with CF-3 objective (92 mm125 mm), which was the distance to the target from the lens in actual experiments. Several
frames were captured with various tilts of the grid to ensure there would be at least one
frame in which the grid appeared orthogonal. The most orthogonal image was then imported
in Matlab® and binarized such that each dot appeared as a distinct region. The individual
regions were then dilated to be slightly larger than the actual dot as it appeared in the image,
ensuring the region encompasses an area that would include some of the white background
around the dot in the original image. Because the contrast between the dot (black) and the
background (white) was high, it allowed the centroid location of each region to be computed
by intensity weighting all of the pixels in the image that corresponded to the pixels in any
particular region (black was weighted 255, and white was weighted zero). Intensity weighting
the pixels within the region ensured that any misrepresentation of the dot due to poor focus
93
or incorrect binary threshold settings would have minimal effect on the accuracy of the
centroid computation. After resolving all centroid locations, the average horizontal and
vertical spacing between dots was computed and used to generate a perfect square grid with
spacing that most closely matched the measured spacing. A complete grid of 84 x 52 was
resolved, and the perfect grid was then aligned with the resolved grid at the center
coordinate and the magnitude of the displacement between grid points of the exact and the
resolved grid were computed to provide the distortion error map (Fig. 6.11).
Figure 6.11: Magnitude of pixel displacement magnitude due to lens distortion. The
resolution of this map was restricted by the dot density on the optical grid in the field of view
(84 x 52). The center of the image was set as the origin (i.e. zero distortion).
94
Figure 6.12 shows the gradient of the distortion error map as it corresponds to
⁄
.A
fairly uniform distortion error gradient between 0.02 and 0.06 is observed for most of the
image, with higher values (as high as 0.17) at the corners.
Figure 6.12: Gradient of pixel displacement magnitude due to lens distortion. The majority of
the image shows a near uniform distortion gradient. The values of this map correspond to
the maximum velocity uncertainty in terms of fraction of bubble velocity for their respective
position in the image. Scale is in pixels/pixel (unitless). The resolution of this map was
restricted by the dot density on the optical grid.
Based on the distortion gradient map, a single RMS value was computed to represent the
error in the magnitude of any vector at all locations as fraction of the total magnitude (Eq.
6.8), keeping in mind that the higher uncertainties exist near the corners: (
0.0395; (
⁄
)
= 0.1721 Now that two values for
⁄
⁄
)
=
have been calculated, Eq.
6.11 is used to determine the maximum and root-mean-square angular uncertainty for any
vector (Fig. 6.13).
95
(a)
(b)
Figure 6.13: Results of the computation of the angular error in a vector (Eq. 6.11). Angular
error was computed for both the RMS error (a) and the maximum error (b).
Figure 6.13 shows that the maximum angular deviation between the resolved vector and the
true vector is no more than 1.1° for the majority of the image, and is a maximum of 4.5° in
the upper right-hand corner of the image where the distortion error gradient is highest. Such
a deviation can occur if the angle between the resolved vector and the distortion error
gradient is roughly 45°.
The uncertainty in
in Eq. 6.5 was a function of the uncertainty in the spacing of the
ratchets. The ratchet period was designed to be 1 mm; this was confirmed by placing an
optical grid (Edmund Optics stock no. NT58-509) over the test section, and recording the
image with a microscope (ken-a-vision T-1754). The side of the glass with the dots was faced
down to avoid any parallax errors. Figure 6.14 shows the image recorded by the microscope,
revealing an accurate 1 mm ratchet period. Viewing along all six ratchet trough periods, the
dot pattern did not phase shift relative to the ratchet peaks or troughs.
96
Figure 6.14: Optical grid placed over the test section to determine accuracy of ratchet period
spacing. Each dot is 0.0625 mm in diameter. Dot spacing is 0.125mm. The image shows an
accurate 1 mm ratchet period.
Additionally, the greatest precision estimated by viewing this image was half the dot
diameter (0.0313 mm). The uncertainty in the distance across the six troughs is therefore no
more than 0.0313 mm, and the total uncertainty in span can be represented as a fraction of
the measured span
(
The uncertainty in
)(
)*
+
(6.12)
in Eq. 6.4 was a function of human error. Therefore, a
number of trials were performed, each of which measured the distance across the six periods
to within one pixel of the mean. Subsequently, the pixel-counting method was determined to
be precise to within one pixel. Unfortunately there was no means of determining the bias
error in this approach. However, because this method was visual, any bias error was believed
to be much smaller than the precision error, and the precision error was taken as the total
uncertainty in the pixel measurement.
[
]
(6.13)
Since all three errors are independent of each other, the Kline & McClintock method was
used to estimate the total uncertainty in ⃑.
97
√(
⃑
⃑⃑⃑⃑⃑⃑
⃑⃑⃑⃑⃑⃑ )
⃑
(
)
⃑
(
) [
(6.14)
]
Substituting Eqs. 6.8, 6.12, 6.13 in Eq. 6.14, and solving for the three partial derivatives gives
√(
)
(
)
) [
(
]
(6.15)
Rearranging Eq. 6.15 and factoring out | ⃑| (Eq. 6.4) gives the final uncertainty for any
dimension measurement in the image:
√(
)
(
)
(
(6.16)
)
Based on Eq. 6.16, the uncertainty in a point-to-point measurement is calculated for each
experimental condition discussed in this thesis as a function of uncertainty in span, pixel
count, and either the RMS or maximum lens distortion such that the RMS uncertainties can
be given for the majority of the image along with worst-case scenario maximum
uncertainties for measurements near the image corners.
Table 6.1: Uncertainty in any dimensional computation derived from images due to
uncertainty in ratchet spacing, pixel count, and lens distortion (maximum as well as RMS).
Uncertainty for each experiment was computed using the Kline and McClinktock method.
Test 1 RMS
Test 1 max
Test 2 RMS
Test 2 max
Test 3 RMS
Test 3 max
[
6
6
5
5
6
6
]
[
882
882
743
743
737
737
]
⁄
± 3.95%
± 17.21%
± 3.95%
± 17.21%
± 3.95%
± 17.21%
⁄
± 3.95%
± 17.21%
± 3.95%
± 17.21%
± 3.95%
± 17.21%
98
7. RESULTS AND DISCUSSION
This section covers the results of the experiments performed in the imaging chamber using
the test section. For all results discussed, the test section was in the upright orientation and
the infrared camera was not used. The results of three experiments are presented: first a
qualitative description of the bubble motion is presented for two subcooling cases, followed
by a quantitative discussion including bubble velocity, acceleration, diameter, and departure
frequency. Replication of the phenomenon under the high subcooling conditions is
demonstrated next. Lastly, a force balance analysis is presented to highlight some of the
mechanics involved.
7.1. Qualitative Visualization
Three experiments were carried out as part of the proof-of-concept study. The experimental
conditions for each of these tests are summarized in Table 7.1.
Table 7.1: Experimental conditions for the two tests performed in January (Test 1 and Test 2),
and the test performed in March (Test 3).
Test
1
2
3
ΔTsub (°C)
4.9 ± 0.3
20.0 ± 0.3
20.0 ± 0.3
Current
(A)
1.000
3.000
3.150
Voltage
(V)
2.310
7.458
5.444
Applied Heat
Flux (W/cm2)
1.96 ± 0.13
18.94 ± 1.26
14.52 ± 0.97
Interval
(µs)
312.50
312.50
188.68
Lateral motion
demonstrated
Slight
Significant
Significant
A qualitative visualization of the bubble dynamics is presented for each test as a series of
video frames captured by the high-speed camera. A specific clip was chosen from each fulllength video. These clips were chosen for the clearest view of the area of greatest activity, as
it provided the best conditions for the bubble-tracking algorithm to resolve the individual
bubbles. The view of a particular site was often obscured by bubbles in the foreground or
background or by strong density gradients caused by gradients in temperature which often
distorted the image to such an extent that resolving a bubble’s outline was not even possible
by eye. For this reason, the camera focus for each test was set to the row of cavities closest
to the camera, as thermal gradients over the test section made it difficult to focus on bubbles
farther back. The aperture was set to fully open, to allow for the shortest exposure times and
99
restrict the algorithm to resolving only this bubbles in the narrow plane of focus. The images
presented in this section are the raw images from the camera with the exception that the
image histogram has been stretched over the full grayscale range to enhance contrast for
easier viewing as described in Section 6.1.
7.1.1. Test 1 – Low Subcooling
The first test was conducted in near saturated conditions (
= 4.9°C) with an applied
heat flux of 1.96 W/cm2 (Fig. 7.1). For this reason, large bubbles are evident and the picture
is well-focused due to the weak temperature-induced density gradients at such low
subcooling.
100
Figure 7.1: Sequence of 20 frames at low subcooling (Test 1). q”applied = 1.96 ± 0.13 W/cm2;
ΔTsub= 4.9 ± 0.3°C; Δt = 312.5 µs; exposure time = 40 µs; Aperture 100% open; field of view =
8.61 mm x 5.38 mm; = 24°.
101
Stationary Bubbles
Direction of departing bubbles
t = 0.00 ms
t = 0.31 ms
t = 0.63 ms
t = 0.94 ms
t = 1.25 ms
t = 1.56 ms
Coalescence
t = 1.88 ms
t = 2.19 ms
Figure 7.1 (continued)
102
Stationary Bubbles
Direction of departing bubbles
t = 2.50 ms
t = 2.81 ms
t = 3.13 ms
t = 3.44 ms
t = 3.75 ms
Coalescence
t = 4.06 ms
t = 4.38 ms
t = 4.69 ms
Figure 7.1 (continued)
103
Stationary Bubbles
Direction of departing bubbles
t = 5.00 ms
t = 5.31 ms
Coalescence
t = 5.63 ms
t = 5.94 ms
Figure 7.1 (continued)
Figure 7.1 shows that large bubbles are being generated at the cavity locations, namely on
the second and third ratchet from the left. The bubbles grow normal to the 24° slope and
continue in that direction after detaching from the cavity opening. As the bubble originating
from the second cavity from the left departs, while still close to the test section it repeatedly
coalesces with multiple bubbles emerging from that same cavity before these bubbles reach
departure diameter. After a number of coalescence events, the bubble is eventually carried
away by buoyancy forces.
7.1.2. Test 2 – High Subcooling
The second test was performed at a high subcooling (
= 20.0°C) with an applied heat
2
flux of 18.94 W/cm (Fig. 7.2). Much more distortion is evident in these images as the high
subcooling creates temperature-induced density gradients, which makes it difficult if not
impossible to focus the lens on the desired area. Fortunately there were a few segments of
the video that provided a clear view of an active portion of the test section.
104
Figure 7.2: Sequence of 20 frames at high subcooling (Test 2). q”applied = 18.94 ± 1.26 W/cm2;
ΔTsub= 20.0 ± 0.3°C; Δt = 312.5 µs; exposure time = 40 µs; Aperture 100% open; field of view =
8.71 mm x 5.44 mm; = 24°.
105
Plume
Stationary Bubbles
Direction of departing bubble
t = 0.00 ms
t = 0.31 ms
Collapse
Collapse
Rapid acceleration
t = 0.63 ms
t = 0.94 ms
Collapse
Collapsed bubble
joins plume flow
Direction of bubble growth
t = 1.25 ms
t = 1.56 ms
t = 1.88 ms
t = 2.19 ms
Figure 7.2 (continued)
106
Stationary Bubbles
t = 2.50 ms
Plume
t = 2.81 ms
Collapse
Direction of departing bubble
growth
t = 3.13 ms
t = 3.44 ms
Rapid acceleration
t = 3.75 ms
t = 4.06 ms
Downward velocity component
t = 4.38 ms
Figure 7.2 (continued)
t = 4.69 ms
107
Stationary Bubbles
Plume
Direction of bubble growth
t = 5.00 ms
t = 5.31 ms
t = 5.63 ms
t = 5.94 ms
Figure 7.2 (continued)
Based on the images in Fig. 7.2 it can be seen that at least one cavity is generating vapor
bubbles on the shallow slope of the wafer, specifically the second from the left. Initially, the
bubble departs the cavity normal to the surface as seen in the low subcooling case. However,
during collapse, the bubble is seen moving to the right at a very high speed with very little
upward component. In fact, in some instances the bubble actually traveled to the right with a
slight downward component (specifically at 4.69 ms). This behavior was unanticipated, and
the cause was not understood at the time of the experiment. However, the bubble did not
continue at such a high speed indefinitely; after it traveled to just past the crest of the
adjacent ratchet, it slowed down quickly and was carried upward and to the right in a
continuous plume of small slow-moving bubbles. This plume still moved to the right at an
angle that was closer to horizontal than the bubble departure direction. It was believed that
the bubbles in the plume were merely carried along by the movement of the surrounding
liquid; this was hypothesized by observing the motion of thermally-induced density gradients
in the plume. There was some indication that cavities as active as the one pictured in Fig. 7.2
108
existed on other areas of the test section, but the surrounding bubbles and temperature
gradients made it difficult to resolve any bubbles that could be tracked by the algorithm.
Because not all cavities behaved the same way, a microscope image of each cavity was taken
to determine what, if any, differences there existed between each of them (Fig. 7.3).
250 μm
Figure 7.3: Microscope pictures of each cavity in the test section. The diagram shows the
orientation of the cavities as well as the viewing direction for the high-seed camera (i.e. row
“a” is in the foreground, row “d” is in the background). Cavity 2a exhibited by far the greatest
lateral motion of any of the cavities. For reference, the thin-film heater traces lay directly
underneath the cavities on the opposite side of the wafer.
109
It can clearly be seen that obvious differences exist between some of the cavities. First of all,
it must be noted that the intended geometry was achieved for most of the cavities on the
test section. However, there exist a number of cavities in rows 1, 2, and 3 that did not
feature the designed geometry. For these cavities, the gold and tin solder wicked up into the
cavity and in some cases, particularly for cavities 1b, 2a, 2d, 3a, and 3d, filled the cavity
almost entirely. What is perhaps most surprising is that cavities 2a and 3a correspond to the
cavities that produced the well-resolved departure behavior in Fig. 7.1, and cavity 2a
produced the noticeable collapse, and acceleration behavior observed in Fig. 7.2, whereas no
such behavior was confirmed for any non-defective cavities. An additional note regarding
cavity 2a is that the top layer of the gold and tin solder appeared to be horizontal whereas it
appeared coplanar with the face of the ratchet for the other filled cavities. This was
concluded by observing the focus of the microscope image as it varied along the face of the
ratchet due to the narrow depth of field (Fig. 7.3). The texture of the gold-tin solder surface
appeared in focus with the lower edge of the cavity mouth, indicating the cavity geometry
may well resemble the geometry shown in Fig. 7.4. Nevertheless, this geometry cannot be
accurately confirmed. It is also unclear as to why bubbles emanating from the non-defective
cavities did not depart with a lateral velocity component.
Figure 7.4: Geometry for cavity 2a construed from a microscope image (Fig. 7.3). During the
bonding process the gold and tin solder used to bond the test section to the wafer was
wicked up into the cavity up to the lower edge of the cavity mouth.
7.2. Bubble Kinematics
Using the bubble tracking algorithm described in Section 6.1, the velocity and diameter were
computed for each of the three tests. Because the labels on the bubble regions greatly
110
obscure the vectors and bubble shapes, the velocity field is presented without labels (Fig. 7.5
and 7.6). For each test, a corresponding set of labeled velocity fields can be found in
Appendix E. As stated previously, the length of the velocity vectors shown in this section do
not linearly correspond to the bubble speed; the length of the vectors was set proportional
to the square root of the bubble speed to prevent any absurdly large vectors from cluttering
the image, and making shorter vectors more visible. Also as stated previously, bubbles could
not be resolved until after departure from the surface. The uncertainty for all diameter
measurements and velocity magnitudes presented in this section is nominally ± 3.95% with a
maximum of ± 17.21% near the upper-right hand corner of the images; vector direction
uncertainty is nominally ± 1.1° with a maximum of ± 4.5° as determined in Section 6.2.4.
7.2.1. Test 1 – Low Subcooling
As the majority of the bubbles seen during this experiment were quite large in comparison to
the high subcooling experiment, not much can be observed in regard to the flow of the
liquid, as buoyancy forces are quite large and carried the bubbles upward rapidly. The
number of vectors resolved is therefore quite small.
111
Figure 7.5: Bubble velocity field at low subcooling (Test 1). The velocity vectors are
proportional to the square root of the velocity of the bubble. q” applied = 1.96 ± 0.13 W/cm2;
ΔTsub= 4.9 ± 0.3°C; Δt = 312.5 µs; exposure time = 40 µs; Aperture 100% open; field of view =
8.61 mm x 5.38 mm; = 24°.
112
t = 0.00 ms
t = 0.31 ms
t = 0.63 ms
t = 0.94 ms
t = 1.25 ms
t = 1.56 ms
Figure 7.5 (continued)
113
t = 1.88 ms
t = 2.19 ms
t = 2.50 ms
t = 2.81 ms
t = 3.13 ms
t = 3.44 ms
Figure 7.5 (continued)
114
t = 3.75 ms
t = 4.06 ms
t = 4.38 ms
t = 4.69 ms
t = 5.00 ms
t = 5.31 ms
Figure 7.5 (continued)
115
t = 5.63 ms
t = 5.94 ms
Figure 7.5 (continued)
The departure diameter of the bubbles in Test 1 was typically resolved to be 0.70 mm, and
left the cavity with velocities in the range of 100-250 mm/s; the resolved velocity varied
greatly from frame to frame for bubbles just leaving the surface. The unsteady velocity was
attributed to the strong oscillations in the bubble surface caused by detachment or
coalescence with another bubble emanating from the cavity (Fig. 7.1 from t = 3.75 ms to t =
5.31 ms). These strong oscillations occurred in three dimensions and made centroid
estimation unreliable. The frequency of bubble departure was also difficult to determine as
some bubbles emanating from a cavity would coalesce with a bubble that had already
departed from the surface. After departing the surface the bubbles tended to collapse and
accelerate over the course of 9 ms. The bubble leaving the cavity consistently demonstrated
a velocity normal to the 24° ratchet face until it began collapsing. During and after collapse,
the bubble did not demonstrate consistent speed or direction.
7.2.2. Test 2 – High Subcooling
Because of the radically different behavior between the bubbles departing the cavities and
those in the plume over the test section, the bubbles in the plume are discussed separately
from the bubbles just departing the cavity. Additionally, the mean horizontal component of
the plume velocities was computed to provide insight into the flow generated over the test
section.
116
Figure 7.6: Bubble velocity field at high subcooling (Test 2). q”applied = 18.94 ± 1.26 W/cm2;
ΔTsub= 20.0 ± 0.3°C; Δt = 312.5 µs; exposure time = 40 µs; Aperture 100% open; field of view =
8.71 mm x 5.44 mm; = 24°.
117
t = 0.00 ms
t = 0.31 ms
t = 0.63 ms
t = 0.94 ms
t = 1.25 ms
t = 1.56 ms
Figure 7.6 (continued)
118
t = 1.88 ms
t = 2.19 ms
t = 2.50 ms
t = 2.81 ms
t = 3.13 ms
t = 3.44 ms
Figure 7.6 (continued)
119
t = 3.75 ms
t = 4.06 ms
t = 4.38 ms
t = 4.69 ms
t = 5.00 ms
t = 5.31 ms
Figure 7.6 (continued)
120
t = 5.63 ms
t = 5.94 ms
Figure 7.6 (continued)
The bubble departing the cavity is discussed first. The departure diameter of this bubble is
consistently resolved to be between 0.66 and 0.71 mm, and left the cavity with velocities in
the range of 170-200 mm/s. Though the bubble grew in a direction normal to the face of the
24° ratchet face, as soon as it detached from the surface, it would collapse and rapidly
accelerate to the right to speeds as high as 906 mm/s (Fig. 7.6 from t = 3.13 ms to t = 4.38
ms). Even higher speeds are believed to be achieved by these bubbles as the distance
traveled per time interval increased beyond the distance that produced the 906 mm/s value,
but the small size, high speed, and density gradients made the bubble difficult to resolve by
the algorithm. These high velocities prevented any bubbles near the cavity from coalescing,
and allowed the departure frequency
to be measured. Fortunately, the frame rate of the
camera (3200 frames per second) was an integer multiple of the departure frequency, as
every 10 frames showed each subsequent bubble in the exact same state of nucleation or
departure as the preceding bubble. The departure frequency was therefore determined to
be 320 Hz. Once the bubbles fully collapsed and traveled across the crest of the adjacent
ratchet, they slowed down rapidly to speeds around 200 mm/s (Fig. 7.6 t = 5.94 ms), after
which they would turn upwards and were carried away by the plume, gradually slowing to
speeds in the range of 30-60 mm/s. The bubbles resolved in the plume moved at a mean
horizontal velocity of 25 mm/s. The bubbles traveled at this speed and direction for their
duration in the field of view. A direct visual observation during experiments indicated that
the plume continued in this direction well outside the camera’s field of view. The maximum
121
tangential acceleration experienced by a bubble during its collapse was recorded to be 127g
(Fig. 7.6 from t = 4.38 ms to t = 5.00 ms), whereas a maximum deceleration of -167g occurred
after the bubble collapsed and crossed the crest of the adjacent ratchet (Fig. 7.6 from t =
3.44 ms to t = 4.06 ms). Acceleration was manually calculated as
⁄
for two consecutive
frame pairs
7.3. Repeatability
A third test was performed 58 days following tests 1 and 2 to demonstrate repeatability of
the experiment. This test was performed at the same subcooling as Test 2 (
= 20.0°C)
but with a lower applied heat flux of 14.52 W/cm2 (Fig. 7.2). Two things must be noted
regarding this test: although the supplied current was greater than the current supplied for
Test 2, the voltage was much lower. This decrease in voltage was due to a repair of a severed
trace on the test section’s thin-film heater; Colloidal Silver Liquid (Ted Pella Prod. No. 16031)
was applied over the trace to complete the circuit, but the reduced resistance in that area of
the heater resulted in a reduced the voltage relative to its previous condition for the same
current. Secondly, despite cleaning procedures the test section became more susceptible to
stagnant bubbles around its perimeter, making it more difficult to resolve events occurring
between the ratchets (Fig. 7.7).
122
Figure 7.7: Sequence of 20 frames at low subcooling to demonstrate repeatability of the
experiment (Test 3) 58 days after Tests 1 and 2. q”applied = 14.52 ± 0.97 W/cm2; ΔTsub= 20.0 ±
0.3°C; Δt = 188.68 µs; exposure time = 40 µs; Aperture 100% open; field of view = 8.14 x 4.85
mm; = 24°.
123
t = 0.00 ms
t = 0.19 ms
t = 0.38 ms
t = 0.57 ms
t = 0.75 ms
t = 0.94 ms
t = 1.13 ms
t = 1.32 ms
Figure 7.7 (continued)
124
t = 1.51 ms
t = 1.70 ms
t = 1.89 ms
t = 2.08 ms
t = 2.26 ms
t = 2.45 ms
t = 2.64 ms
t = 2.83 ms
Figure 7.7 (continued)
125
t = 3.02 ms
t = 3.21 ms
t = 3.40 ms
t = 3.58 ms
Figure 7.7 (continued)
Though the bubbles emerging from the cavities are too distorted to be resolved by the
algorithm, the observed behavior remains the same (i.e. bubbles accelerate to the right
during collapse, and decelerate once across the crest of the adjacent ratchet). The departure
frequency
was determined to be roughly 570 Hz, much higher than that observed in test
2. The Also, the plume can provide some indication of the fluid motion generated by the test
section such that the behavior of the plume in test 3 can be compared to that of the plume in
test 2. In general, the small collapsed bubbles in the pool over the test section still move in a
definite right-hand direction, at an angle farther to the right than merely normal to the 24°
ratchet face. Similar to test 2, this plume of small bubbles originates from roughly the same
area over the test section. Because the plume does exhibit behavior similar to the plume in
test 2 despite the reduced heat flux, repeatability of the experiment was demonstrated (Fig.
7.8).
126
Figure 7.8: Sequence of 20 frames at high subcooling to demonstrate repeatability of the
experiment (Test 3) 58 days after Tests 1 and 2. q”applied = 14.52 ± 0.97 W/cm2; ΔTsub= 20.0 ±
0.3°C; Δt = 188.68 µs; exposure time = 40 µs; Aperture 100% open; field of view = 8.14 x 4.85
mm; = 24°.
127
t = 0.00 ms
t = 0.19 ms
t = 0.38 ms
t = 0.57 ms
t = 0.75 ms
t = 0.94 ms
Figure 7.8 (continued)
128
t = 1.13 ms
t = 1.32 ms
t = 1.51 ms
t = 1.70 ms
t = 1.89 ms
t = 2.08 ms
Figure 7.8 (continued)
129
t = 2.26 ms
t = 2.45 ms
t = 2.64 ms
t = 2.83 ms
t = 3.02 ms
t = 3.21 ms
Figure 7.8 (continued)
130
t = 3.40 ms
t = 3.58 ms
Figure 7.8 (continued)
Though the bubbles emerging from the cavities are too distorted to be resolved by the
algorithm, the bubbles in the plume were successfully tracked and revealed continuous
bubble velocities in the range of 30-70 mm/s, confirming the repeatability of the experiment.
The higher departure frequency is possibly explained by the smaller departure diameter.
Although the departure diameter was not resolved by the algorithm, it was estimated for Fig.
7.7 (specifically at t = 0.19 ms, t = 1.89 ms, t = 3.58 ms) by manually determining the most
appropriate equivalent diameter of the bubble. The estimated departure diameter for the
bubble leaving the cavity in test 3 was between 0.41 and 0.45 mm.
7.4. Force Balance
Because the performance of cavity 2a has been well-quantified, the focus of this remaining
discussion will pertain to the possible flow mechanisms driving the bubbles that nucleate and
depart from this cavity. As discussed in Section 2.2, a number of forces exist on a bubble
during the growth and departure phase. The following analysis will attempt to estimate the
net momentum imparted on the fluid by the growth and departure of a single bubble.
Additionally, an attempt is made at characterizing the forces that cause the high acceleration
of the collapsing bubble in subcooled conditions. Figure 7.9 shows a qualitative
representation of the force imparted on the bubble’s liquid-vapor interface as a result of
growth.
131
Figure 7.9: Growth of the bubble normal to the surface imparts a net momentum on the
fluid.
Using Eq. 2.9 the total momentum imparted on the fluid during one growth and phase can be
computed discretely for each time step. However, because of the cavity, the contact
diameter
remains fixed and the bubble is assumed to be spherical for the majority of its
growth such that the drag coefficient can be assumed to be that of a sphere. This assumption
in turn means that this equation is only reasonable when
(
.
(7.1)
)
As the calculated bubble growth rate is an average between two frames, the diameter was
estimated as the average diameter between the same two frames. The Reynolds number
which was used to determine the drag coefficient
[41] was a function of the average
diameter and growth rate of the bubble:
̅
∑ ̅̅̅̅ (
) (
)
(7.2)
There are 28 cavities on the test section, each located on 1/28 of the total test section area
. The momentum ̅ from Eq. 7.2 is assumed to be imparted on a control volume whose
cross-sectional area corresponds to 1/28 of the projected test section area
(Fig. 7.10).
132
Figure 7.10: Control volume for estimation of liquid plume velocity as it depends on the
momentum imparted on the fluid during one bubble growth cycle
A momentum balance can be performed on the control volume as follows
̅
̅
(
( ) ̅
)( ̅ )
(7.3)
Rearranging Eq. 7.3 and substituting Eq. 7.2 leads to a conservative estimate of ̅ :
̅
√
( )
∑ ̅̅̅̅ (
) (
)
(7.4)
In order to compute the liquid velocity as a result of the momentum imparted during one
growth cycle, the diameters of the bubbles are recorded for each frame up to departure.
Because the bubbles are still attached to the surface during this phase, the algorithm was
unable to resolve it, and the bubble diameters were determined manually by counting the
number of pixels and converting to mm. Table 7.2 shows the estimated diameters from
bubbles emanating from cavity 2a in Fig. 7.1 (from t = 1.88 ms to t = 4.38 ms) and Fig. 7.2
(from t = 1.25 ms to t = 2.50 ms).
133
Table 7.2: Bubble diameter during growth before departure. Data is shown for low
subcooling based on Fig. 7.1 (from t = 1.88 ms to t = 4.38 ms) and high subcooling based on
Fig. 7.2 (from t = 1.25 ms to t = 2.50 ms).
Low Subcooling
Time [ms] d [mm]
1.88
0.19
2.19
0.36
2.50
0.43
2.81
0.47
3.13
0.52
3.44
0.53
3.75
0.56
4.06
0.59
4.38
0.59
High Subcooling
Time [ms] d [mm]
1.25
0.40
1.56
0.69
1.88
0.80
2.19
0.81
2.50
0.82
In low subcooling, using Eq. 7.4, the liquid velocity normal to the 24° face of the ratchet was
estimated to be 17 mm/s. Because the strong jets developed by these growing bubbles are
normal to the 24° face, there is a component of the jet in the horizontal direction, specifically
7 mm/s. In high subcooling, the liquid velocity normal to the 24° face was estimated to be 45
mm/s with a horizontal component of 18 mm/s. These values can be compared to the plume
flow if the velocity of the bubbles matches the velocity of the surrounding liquid as was
hypothesized earlier in this chapter. To confirm this, the Stokes number was calculated using
the bubble time constant [42] and a flow time constant based on the diameter of the bubble:
(7.5)
For resolved bubble velocities of 30-70 mm/s, and a conservative diameter of 0.1 mm, the
Stokes number is 3.5x10-4-8.2x10-4. Because
<<1 for the entire range, bubbles are assumed
to follow the flow very well, meaning the resolved bubble velocities are very close to the
plume velocity. The relatively close match between velocities of the bubbles in the plume in
Fig. 7.6 (30-60 mm/s with a mean horizontal component of 25-35 mm/s) and those
calculated using Eq. 7.4 (45 mm/s with a horizontal component of 18 mm/s), suggests this
may actually be the driving mechanism for the plume flow. However, for the low subcooling
case, the lack of any trackable bubbles in the location where the plume would be expected
thwarts any attempt at comparing the model to data.
134
The second flow mechanism considered is that which drives the bubbles to the right at such
a high acceleration. Figure 7.11 shows a possible reason for the observed behavior. Because
of the high subcooling conditions, thermal gradients exist over the test section, and have
been observed to look similar to those pictured in Fig. 7.11 (indicated by the different shades
and small arrows).
Figure 7.11: Preliminary reasoning for observed behavior of bubbles in subcooled conditions.
During collapse, thermal gradients (indicated by shading and small arrows) cause Marangoni
convection that pushes the bubbles in the direction of increasing temperature.
As described in Section 2.2, Marangoni flow around a bubble travels from regions of low
surface tension (high temperature) to regions of high surface tension (low temperature). In
the case where a bubble has detached from the surface, this jet can propel the bubble in the
direction of increasing liquid temperature. As the bubble detaches normal from the surface
and travels to the right, it encounters a relatively sharp temperature gradient. This gradient
facilitates Marangoni flow to propel the bubble in this direction. A force balance is set up to
determine if this is a possible cause for the observed behavior. The forces considered are
that due to Marangoni flow (Eq. 2.5), and the drag force associated with the bubble’s speed
(Eq. 2.8):
(7.6)
(
)(
)
(
)
(7.7)
135
⁄
Rearranging Eq. 7.7 and solving for
yields
(
)(
(7.8)
)
Equation 7.8 requires the velocity, acceleration, and diameter of the near-cavity bubble,
which was determined from Fig. 7.6 (t = 0.00 ms to t = 3.13 ms) and summarized in Table 7.3.
Table 7.3: Bubble accelerations calculated to determine if Marangoni flow plays a role in the
strong acceleration of bubbles immediately after departure. Data obtained from Fig. 7.6 (t =
3.13 ms to t = 5.00 ms).
Time
[ms]
3.13
3.44
3.75
4.06
4.38
4.69
5.00
Diameter
[mm]
0.65
0.53
0.50
0.39
0.27
0.12
0.10
Speed
[mm/s]
Tangential
Acceleration [g]
171
3
127
34
75
-167
181
571
675
902
394
Equation 7.8 was used to determine the necessary temperature gradient to produce the
+127g and -167g accelerations shown in Table 7.3, all properties were assumed to be of
water at saturation temperature (
[
⁄
⁄
[ ⁄
],
[
⁄
],
]). During the +127g acceleration, the bubble was moving at an average
speed of 376 mm/s (Re = 639 [41]; CD = 0.53). To effect such acceleration at this speed, a
temperature gradient of 903°C/mm would need to exist in the direction of the bubble
velocity. During the -167g acceleration, the bubble was moving at an average speed of 650
mm/s (Re = 265 [41]; CD = 0.72). To effect such acceleration at this speed, a temperature
gradient of 42°C/mm would need to exist in the direction of the bubble velocity, as drag
alone would cause an even greater acceleration. These hypothetical temperature gradients
(the 903°C/mm value in particular) are unrealistic for the conditions imposed, and as such
the accelerations observed during collapse of the departed bubble are not due to Marangoni
flow alone, leaving questions for future investigation.
136
8. CONCLUSIONS
The hypothesis for this thesis work was that heat applied to an asymmetrically structured
surface could generate fluid motion parallel to the surface in the nucleate boiling regime. The
results presented in Chapter 7 clearly show that thermally actuated lateral flow was
demonstrated in the nucleate boiling regime, and that it was repeatable. At a high subcooling
of 20°C and a heat flux of 14-18 W/cm2, bubble tracking resolved large plume velocities in
the range of 30-70 mm/s, with a mean horizontal component between 25 and 35 mm/s.
These values correlated fairly well with the predicted velocities obtained from a conservative
semi-empirical momentum balance model based on the drag induced by growing bubbles on
the surrounding fluid. In lower subcooling conditions, substantial lateral flow was not
observed as there were no small bubbles present to track the flow. However, extrapolating
the from the bubble diameters during growth, plume velocities were predicted to be in the
range of 17 mm/s. It must be noted that these plume flows, as best can be deduced from the
data, were produced by only one or two active cavities on the test section, indicating that
much higher plume velocities may have been achieved if all cavities were active. Also, the
fact that the only active cavities were those with severe manufacturing defects indicates that
for the conditions studied, the designed cavity geometry was not ideal, and perhaps required
a greater heat flux to be active than that which could be provided by the thin-film heater.
The chamber and experimental facility procured for this work have proven extremely
reliable, effective, and user-friendly. The electrical connections and control panel worked
without fail, the custom LED array proved invaluable in creating a bright and uniformly
backlit image for the high-speed camera, allowing for exposure times as low as 40 µs, and
with all modules securely mounted to the aluminum rack, the repeatability of experimental
conditions was easily met. The silicon wafer holder securely mounted the fragile test section
in the chamber in any of three orientations, while maintaining an excellent seal and
providing an unobstructed side view for the high-speed camera. The ability for the facility to
trigger the camera as well as temperature data acquisition using the experimental facility’s
accelerometer greatly improves the probability that data can be acquired in a microgravity
environment, even if these conditions exist for only a few seconds at a time, as no user
137
action is required to trigger data acquisition when this mode is enabled; this indicates that
the experimental facility should perform well when flying parabolic arcs, capturing data only
when microgravity conditions exist. Additionally, the chamber demonstrated its ability to
accommodate Particle Image Velocimetry, Laser-Induced Fluorescence, and infrared
thermography measurement techniques, which could be employed in follow-up work or in
other experiments that use this facility.
The bubble tracking algorithm demonstrated its ability to track bubbles reliably for a variety
of experimental conditions, providing valuable bubble velocity, acceleration and diameter
data that could be compared with semi-empirical models. Vector plots were presented in a
clear and intuitive manner, providing an instantaneous glimpse of the bubble behavior near
the test section, and the flow of the plume over the test section. In fact, the algorithm
successfully resolved velocities as high as 1 m/s and accelerations as high as 167g.
Although experiments were performed with FC-72, no data was reported as FC-72 was highly
susceptible to creating a greatly distorted image over the heated test section in high
subcooling conditions, particularly near the surface, which made bubble tracking extremely
difficult.
138
9. RECOMMENDATIONS FOR FUTURE INVESTIGATION
As the plume velocity in the subcooled case was attributed primarily to the drag induced on
the fluid during bubble growth on the sloped ratchet face, future designs should focus on
generating favorable nucleation sites on this face without creating complex reentrant
cavities, as the process to create these resulted in high cost and long procurement times for
the test section. The algorithm should also be improved to resolve the diameters of bubbles
still attached to the surface, as this will greatly increase the precision of the induced drag
calculations, which in turn will predict the performance more accurately.
The test section should be more durable and have the capability for greater heat fluxes than
those that can be provided by a thin-film heater. As gravitational dependence has not yet
been studied, a number of experiments should be performed in the vertical orientation as
well as in microgravity. As part of preparation for these experiments, the experimental
facility should be outfitted to fully comply with the appropriate microgravity flight
experimental facility regulations. Such additions will likely include a double containment
barrier around the chamber, an accumulator to accommodate excess vapor generated
without causing an increase in pressure, and a single emergency power disconnect switch for
all systems. As a precaution, a pressure transducer should be included to monitor chamber
pressure in the event of a failure of the accumulator or other systems. The power supply that
will power the test section should also be fixed to the experimental facility. Further upgrades
to the facility should include:

Stirrer to thoroughly mix the pool to insure no temperature gradients exist prior to
data acquisition.

High capacity filter with pore size less than 10 µm as the current filter requires
flushing prior to each experiment.

Calibration of the accelerometer should be performed. At the time of the writing of
this thesis no precise calibration had been performed because it was not required for
the work presented. The accelerometer was powered directly from the experimental
facility power supply, and supply voltage changed slightly whenever any other
139
system connected to the power supply was switched on or off (primarily the pump or
LED array). Since the accelerometer implemented was an analog device, the output
voltage was directly affected by the input voltage.

If active cooling of the pool becomes necessary, a thermoelectric cooling device
should be used to maintain compliance with the relevant experimental facility
requirements for flight experiments.
140
REFERENCES
[1] Incropera F. P., DeWitt D. P., Bergman T. L., and Lavine A. S., 2006, Introduction to Heat
Transfer, John Wiley & Sons, Hoboken, NJ, pp. 860
[2] Wolf D. H., Incropera F. P., and Viskanta R., 1993, “Jet Impingement Boiling,” Elsevier,
pp. 1-132.
[3] Grover G. M., Cotter T. P., and Erickson G. F., 1964, “Structures of Very High Thermal
Conductance,” J. Appl. Phys., 35(6), p. 1990.
[4] Vasiliev L. L., 2005, “Heat pipes in modern heat exchangers,” Applied Thermal
Engineering, 25(1), pp. 1-19.
[5] Birur G. C., and O’Donnell T. P., 2001, “Advanced thermal control technologies for space
science missions at Jet Propulsion Laboratory,” AIP Conference Proceedings,
Albuquerque, NM, pp. 263-270.
[6] Watts B. D., 2001, The Military Use of Space: A Diagnostic Assessment, Center for
Strategic and Budgetary Assessments, Washington, DC.
[7] Incropera F. P., DeWitt D. P., Bergman T. L., and Lavine A. S., 2006, Introduction to Heat
Transfer, John Wiley & Sons, Hoboken, NJ, pp. 587-588.
[8] Carey V. P., 2007, Liquid-vapor phase-change phenomena: an introduction to the
thermophysics of vaporization and condensation processes in heat transfer equipment,
Taylor and Francis.
[9] Wu W., Bostanci H., Chow L. C., Hong Y., Su M., and Kizito J. P., 2010, “Nucleate boiling
heat transfer enhancement for water and FC-72 on titanium oxide and silicon oxide
surfaces,” International Journal of Heat and Mass Transfer, 53(9-10), pp. 1773-1777.
[10] Zhang N., and Chao D. F., 1999, “Models for enhanced boiling heat transfer by unusual
Marangoni effects under microgravity conditions,” International Communications in
Heat and Mass Transfer, 26(8), pp. 1081-1090.
[11] Wang H., Peng X., Garimella S. V., and Christopher D. M., 2007, “Microbubble return
phenomena during subcooled boiling on small wires,” International Journal of Heat and
Mass Transfer, 50(1-2), pp. 163-172.
[12] Zeng L., Klausner J., and Mei R., 1993, “A Unified Model for the Prediction of Bubble
Detachment Diameters in Boiling Systems. 1. Pool Boiling,” International Journal of Heat
and Mass Transfer, 36(9), pp. 2261-2270.
[13] Lee H., Merte H., and Chiaramonte F., 1997, “Pool boiling curve in microgravity,” Journal
of Thermophysics and Heat Transfer, 11(2), pp. 216-222.
[14] Lee H., 2002, “Mechanisms of steady-state nucleate pool boiling in microgravity,”
Microgravity Transport Processes in Fluid, Thermal, Biological, and Materials Sciences,
974, pp. 447-462.
141
[15] Ohta H., 1997, “Experiments on microgravity boiling heat transfer by using transparent
heaters,” Nuclear Engineering and Design, 175(1-2), pp. 167-180.
[16] Kim J., and Benton J., 2002, “Highly subcooled pool boiling heat transfer at various
gravity levels,” International Journal of Heat and Fluid Flow, 23(4), pp. 497-508.
[17] Straub J., 1999, “Physics of boiling and bubble dynamics with and without gravity:
Conclusions of space experiments,” Inorganic Materials, 35(6), pp. 615-621.
[18] Merte H., 2004, “Momentum effects in steady nucleate pool boiling during
microgravity,” Transport Phenomena in Microgravity, 1027, pp. 196-216.
[19] Das A. K., Das P. K., and Saha P., 2009, “Performance of different structured surfaces in
nucleate pool boiling,” Applied Thermal Engineering, 29(17-18), pp. 3643-3653.
[20] Das A. K., Das P. K., and Saha P., 2010, “Some investigations on the enhancement of
boiling heat transfer from planer surface embedded with continuous open tunnels,”
Experimental Thermal and Fluid Science, 34(8), pp. 1422-1431.
[21] Chyu M. C., and Fei J., 1991, “Enhanced Nucleate Boiling in an Angular Geometry Found
in Structured Surfaces,” International Journal of Heat and Mass Transfer, 34(2), pp. 437448.
[22] Liter S. G., and Kaviany M., 2001, “Pool-boiling CHF enhancement by modulated porouslayer coating: theory and experiment,” International Journal of Heat and Mass Transfer,
44(22), pp. 4287-4311.
[23] Stroock A. D., Ismagilov R. F., Stone H. A., and Whitesides G. M., 2003, “Fluidic Ratchet
Based on Marangoni−Bénard Convection,” Langmuir, 19(10), pp. 4358-4362.
[24] Jo M. C., 2008, “Thermally actuated pumping of a single-phase fluid using surface
asymmetry,” Thesis, Oregon State University.
[25] Linke H., Alemán B. J., Melling L. D., Taormina M. J., Francis M. J., Dow-Hygelund C. C.,
Narayanan V., Taylor R. P., and Stout A., 2006, “Self-Propelled Leidenfrost Droplets,”
Phys. Rev. Lett., 96(15), p. 154502.
[26] Ok J., Lopez-Ona E., Nikitopoulos D., Wong H., and Park S., 2011, “Propulsion of droplets
on micro- and sub-micron ratchet surfaces in the Leidenfrost temperature regime,”
MICROFLUIDICS AND NANOFLUIDICS, 10(5), pp. 1045-1054.
[27] Incropera F. P., DeWitt D. P., Bergman T. L., and Lavine A. S., 2006, Introduction to Heat
Transfer, John Wiley & Sons, Hoboken, NJ, pp. 587-588.
[28] Watts B. D., 2001, The Military Use of Space: A Diagnostic Assessment, Center for
Strategic and Budgetary Assessments, Washington, DC.
[29] Brusstar M., and Merte H., 1994, “Effects of Buoyancy on the Critical Heat-Flux in
Forced-Convection,” Journal of Thermophysics and Heat Transfer, 8(2), pp. 322-328.
[30] Zhang H., Mudawar I., and Hasan M. M., 2004, “A Method for Assessing the Importance
of Body Force on Flow Boiling CHF,” J. Heat Transfer, 126(2), pp. 161-168.
[31] 2006, “Zero Gravity Research Facility,” NASA Glenn Research Center, Cleveland, OH.
[32] 2008, “2.2 Second Drop Tower,” NASA Glenn Research Center, Cleveland, OH.
142
[33] 2010, “Black Brant XII Launch Vehicle (40.XXX),” NASA Goddard Space Flight Center,
Wallops Island, VA
[34] 2007, “Experiment Design Requirements and Guidelines NASA 932 C-9B (AOD 33897),”
NASA NASA Lyndon B. Johnson Space Center, Houston, TX
[35] 2010, “Review of NASA’s Microgravity Flight Services,” NASA Office of Inspector
General, Washington, DC.
[36] Kim J., and Benton J., 2002, “Highly subcooled pool boiling heat transfer at various
gravity levels,” International Journal of Heat and Fluid Flow, 23(4), pp. 497-508.
[37] 2007, “Interface Control Document NASA 932 C-9B (AOD 33912),” NASA Lyndon B.
Johnson Space Center, Houston, TX
[38] Straub J., 1999, “Physics of boiling and bubble dynamics with and without gravity:
Conclusions of space experiments,” INORGANIC MATERIALS, 35(6), pp. 615-621.
[39] 2007, “Experiment Design Requirements and Guidelines NASA 932 C-9B.”
[40] “Refractive index for various materials.”
[41] “drag on a sphere - Wolfram|Alpha.”
[42] Adrian R. J., and Westerweel J., 2010, Particle Image Velocimetry, Cambridge University
Press.
143
APPENDICES
144
Appendix A. Derivation of Marangoni Flow Reaction Force
The following is a derivation of the reaction force on a bubble due to Marangoni flow based
on the shear stress on the liquid vapor interface.
(
(
[
[
(
)(
)(
)(
)
(A.1)
( )
(A.2)
)
(
)(
)(
( )
( )
)
)∫
(A.3)
( )
]
( )] [
)
( ) (
)(
⁄
( )
][
( )]
(A.4)
(A.5)
(A.6)
⁄
(
(
)(
)(
)
)
(A.7)
(A.8)
145
Appendix B. Experimental Procedures Specific to FC-72
B.1. Cleaning the Chamber
After rinsing the chamber with deionized water, both the fill and drain fluid loop were blown
out using filtered compressed air and the chamber was left to dry for at least 12 hours to
insure that water did not contaminate the experiment.
B.2. Filling and Degassing
 The syphoning procedure was not performed as the low viscosity allowed the pump to
draw it from the reservoir without trouble.
 A 7-micron filter was used as the low viscosity allowed a smaller pore size.
 The PID controller set temperature for degassing was set above 57°C.
146
Appendix C. Bubble Tracking Algorithm
In this appendix a flow chart showing the scripts used in the algorithm is presented, along
with the actual Matlab® M-files (not all scripts used are included in the flow chart)
Velocity_Track.m
 Define image file to use
 Define recording parameters
 Initialize Image Processing Variables
 Call Bubble_Filter.m to resolve bubbles as
distinct binary regions
Velocity_Track.m (after Bubble_Filter.m)
 Call Delete_Incompletes.m to delete regions
not representative of a bubble
Velocity_Track.m (after Delete_Incompletes.m)
 Perform intensity-weighted centroid
calculations on remaining regions
 Call PTV.m
Delete_Multiples.m
 Determine any overlapping region pairs
 Identify pair with longer
centroid-centroid distance
Velocity_Track.m
 Plot velocity vectors
 Annotate plot with information identified
at top of script
Bubble_Filter.m
 Create Gradient, Laplace, Sobel filter masks
 Apply masks to image
 Binarize image
 Remove very small regions
 Fill hollow regions (2X)
 Remove small regions
 Clear image border regions
 Outline regions on original image
Delete_Incompletes.m
 Delete regions w/ low solidity
 Delete regions w/ high eccentricity
 Tag remaining regions light yellow
PTV.m
 Compute all possible bubble centroid-centroid
distances for each region in first image to all
regions in the second image.
 Identify pairs with shortest distances between
the two images as bubble pairs
 Call Delete_Multiples.m to delete pairs that
reference the same image
PTV.m (after Delete_Multiples.m)
 Identify bubble pairs with greatly mismatched
area ratios.
 Remove all regions identified from binary
image to retain only valid bubble pairs.
 Create table with coordinates of bubble pairs
Figure C.1: Flowchart for bubble tracking algorithm functions.
147
Appendix D. Data Analysis
D.1. Thermocouple Calibration and Uncertainty Analysis
The thermocouple probes were held in the center of the heater bath, still mounted in the
circular chamber port cover. Keeping the thermocouples mounted ensured the probes would
not shift during calibration. The NIST standard RTD probe (uncertainty 0.3°C) was inserted
into the bath between the probes. Once the pool temperature stabilized, the NIST standard
temperature was recorded manually and thermocouple readings were recorded at a rate of
2.4 Hz. A total of 100 data points were recorded for each probe at each temperature. At 95%
confidence this equates to a student T-factor of 1.984.
The uncertainty for each thermocouple was computed as function of the uncertainty due to
the random error of sampling, curve fit error, and NIST bias error. The random sampling
error for each data point was computed as
√
where
̅ )
∑(
denotes the individual sample, and
(D.1)
denotes the individual data point. The
maximum random sampling error of the ten data points was used as the global sampling
error for the entire calibration range:
([
]
)
(D.2)
The curve fit error was computed as the standard deviation of the fit error:
√
∑( ( ̅ )
)
(D.3)
148
where
is the slope of the linear fit and
the standard was included as
is the y-axis intercept. Finally, the uncertainty of
. The total uncertainty was computed as a root
sum of squares:
√(
)
(
)
(
)
(D.4)
The uncertainties for each probe including the total uncertainty are summarized in Table D.1;
calibration curves are provided in Fig. D.1.
Table D.1: Thermocouple uncertainty measurements and calibration curve constants. Values
are presented for thermocouples both with and without extension cables.
Probe Type Extension Slope
TC0
K
No
1.0034
TC1
K
No
1.0026
TC2
K
No
1.0024
TC3
K
No
1.0032
TC0
K
Yes
1.0038
TC1
K
Yes
1.0030
TC2
K
Yes
1.0028
TC3
K
Yes
1.0037
Intercept
-0.0969
-0.1557
-0.1527
-0.2171
-0.1082
-0.1701
-0.1735
-0.2505
0.0742
0.0307
0.0445
0.0431
0.0632
0.0305
0.0423
0.0412
0.0134
0.0131
0.0153
0.0143
0.0152
0.0147
0.0163
0.0156
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
149
Figure D.1: Calibration curves for each thermocouple. * denotes extension cable was used.
150
Figure D.1 (Continued)
151
D.2. Applied Heat Flux
Heat flux applied to the silicon wafer by the thin film heater:
(D.5)
The total uncertainty in applied heat flux was estimated using the Kline and McClintock
method:
√(
)
(
)
(
)
(
(D.6)
)
Substituting Eq. D.5 in Eq. D.6 the uncertainty in applied heat flux becomes
√(
)
(
)
(
)
(
)
(D.7)
Based on Eq. D.7 the uncertainty in the applied heat flux is calculated for each experimental
condition discussed in this thesis (Table D.2).
Table D.2: Uncertainty of applied heat flux due to uncertainties in heated surface area and
applied power. Uncertainty for each experiment was computed using the Kline and
McClinktock method.
Heated area
Heated area
Current (A)
Voltage (V)
width (cm)
length (cm)
-5
-4
(W/cm2)
Test U = ± 5x10
U = ± 5x10
U = ± 0.005
U = ± 0.005
1
3.000
7.458
0.93
1.27
18.94 ± 1.26
2
1.000
2.310
0.93
1.27
1.96 ± 0.13
3
3.150
5.444
0.93
1.27
14.52 ± 0.97
152
Appendix E. Annotated Vector Plots for Tests 1, 2, 3
The following labeled vector plots were used to obtain the quantitative velocity and diameter
information for the data presented in Chapter 7. Labels show from top to bottom: bubble
speed in mm/s, bubble equivalent diameter in mm, and estimated bubble volume in mm3.
153
Figure E.1: Labeled vector plots for Test 1. q”applied = 1.96 ± 0.13 W/cm2; ΔTsub= 4.9 ± 0.3°C; Δt
= 312.5 µs; exposure time = 40 µs; Aperture 100% open; field of view = 8.61 mm x 5.38 mm;
= 24°.
154
t = 0.00 ms
t = 0.31 ms
t = 0.63 ms
t = 0.94 ms
t = 1.25 ms
t = 1.56 ms
Figure E.1 (continued)
155
t = 1.88 ms
t = 2.19 ms
t = 2.50 ms
t = 2.81 ms
t = 3.13 ms
t = 3.44 ms
Figure E.1 (continued)
156
t = 3.75 ms
t = 4.06 ms
t = 4.38 ms
t = 4.69 ms
t = 5.00 ms
t = 5.31 ms
Figure E.1 (continued)
157
t = 5.63 ms
Figure E.1 (continued)
t = 5.94 ms
158
Figure E.2: Labeled vector plots for Test 2. q”applied = 18.94 ± 1.26 W/cm2; ΔTsub= 20.0 ± 0.3°C;
Δt = 312.5 µs; exposure time = 40 µs; Aperture 100% open; field of view = 8.71 mm x 5.44
mm; = 24°.
159
t = 0.00 ms
t = 0.31 ms
t = 0.63 ms
t = 0.94 ms
t = 1.25 ms
t = 1.56 ms
Figure E.2 (continued)
160
t = 1.88 ms
t = 2.19 ms
t = 2.50 ms
t = 2.81 ms
t = 3.13 ms
t = 3.44 ms
Figure E.2 (continued)
161
t = 3.75 ms
t = 4.06 ms
t = 4.38 ms
t = 4.69 ms
t = 5.00 ms
t = 5.31 ms
Figure E.2 (continued)
162
t = 5.63 ms
Figure E.2 (continued)
t = 5.94 ms
163
Figure E.3: Labeled vector plots for Test 3. q”applied = 14.52 ± 0.97 W/cm2; ΔTsub= 20.0 ± 0.3°C;
Δt = 188.68 µs; exposure time = 40 µs; Aperture 100% open; field of view = 8.14 x 4.85 mm;
= 24°.
164
t = 0.00 ms
t = 0.19 ms
t = 0.38 ms
t = 0.57 ms
t = 0.75 ms
t = 0.94 ms
Figure E.3 (continued)
165
t = 1.13 ms
t = 1.32 ms
t = 1.51 ms
t = 1.70 ms
t = 1.89 ms
t = 2.08 ms
Figure E.3 (continued)
166
t = 2.26 ms
t = 2.45 ms
t = 2.64 ms
t = 2.83 ms
t = 3.02 ms
t = 3.21 ms
Figure E.3 (continued)
167
t = 3.40 ms
Figure E.3 (continued)
t = 3.58 ms
168
Appendix F. Part Drawings
All parts were designed in SolidWorks, and were machined at Oregon State University with
the exception of the stainless steel chamber.
169
Figure F.1: Assembly drawing of chamber.
170
Figure F.2: Drawing of chamber wall with laser port cutout and bolt flanges.
171
Figure F.3: Drawing of chamber wall with wafer holder port cutout and bolt flanges.
172
Figure F.4: Drawing of bottom chamber wall with wafer holder cutout.
173
Figure F.5: Drawing of top chamber wall with wafer holder cutout.
174
Figure F.6: Drawing of side chamber wall with window cutout and thru-holes for condensing
equipment and cartridge heaters.
175
Figure F.7: Drawing of side chamber wall with window cutout.
176
Figure F.8: Drawing of window clamp.
177
Figure F.9: Drawing of circular port cover.
178
Figure F.10: Drawing of laser port used for experiments in the vertical orientation.
179
Figure F.11: Assembly drawing of laser port used for experiments in the upright and inverted
orientation.
180
Figure F.12: Drawing of base flange for laser port.
181
Figure F.13: Drawing of side wall of laser port.
182
Figure F.14: Drawing of bottom wall for laser port.
183
Figure F.15: Drawing of top wall for laser port.
184
Figure F.16: Drawing of window mounting plate for laser port.
185
Figure F.17: Bracket for mounting the rotational platform for the prism to the laser port.
186
Figure F.18: Drawing of wafer holder.
187
Figure F.19: Drawing of wafer clamp.
188
Figure F.20: Drawing of arms used to support the high-speed camera.
189
Figure F.21: Drawing of high-speed camera mounting plate.
190
Figure F.22: Drawing of high-speed camera lens support bracket.