ANALYSIS OF SEVERE REACTIVITY EXCURSIONS IN FAST REACTORS by r ZACK T. PATE B.Sc., U.S. Naval Academy (1958) SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY March, 1970 Signature of Author Depafm-ent of Nuclear Engineering, March 12, 1970 Certified by Thesis Supervisor Accepted by Chairman, Departmental Committee on Graduate Students ArcTiiVei AMpSS. INST TCH APR A RV7O ClRp 1 ES 2 ANALYSIS OF SEVERE REACTIVITY EXCURSIONS IN FAST REACTORS by ZACK T. PATE Submitted to the Department of Nuclear Engineering on March 13, 1970, in partial fulfillment of the requirements for the degree of Doctor of Philosophy. ABSTRACT An investigation of the events which can occur during and as a result of a severe reactivity accident in a Liquid Metal Cooled Fast Breeder Reactor is undertaken. Nine phenomena which can affect reactivity are considered. Primary emphasis is placed on the sodium voiding and fuel motion effects which can occur after clad failure in some region of the core. An analysis of the energy exchange behavior of fuel and sodium in the region of clad failure is presented. Hydrodynamics calculations are employed to estimate the rate of sodium voiding which takes place subsequent to clad failure. It is shown that the rate of reactivity addition which results can be as much as an order of magnitude greater than the maximum rates expected in an intact core. Fuel motion effects are analyzed employing the ANISN multigroup code. In addition, a model which allows simple hand calculated estimates of reactivity perturbations in fast reactors is developed and employed in assessing fuel motion effects. It is concluded that fuel motion during the course of an excursion is of less importance than the sodium voiding effect unless (and until) pressures sufficient for wide scale core disassembly are generated. Thesis Supervisor: Title: Michael J. Driscoll Professor of Nuclear Engineering Thesis Supervisor: Title: Theos J. Thompson Professor of Nuclear Engineering (on leave of absence) 3 ACKNOWLEDGEMENTS The author is indebted to Professor M.J. Driscoll for his guidance throughout this work. Rarely did a discussion with Professor Driscoll end without my having gained new insight into succeeding steps of the thesis. Special thanks is extended to Dr. Theos J. Thompson for his assistance in giving initial direction to this work. His insight led the author into an interesting and fruitful area of research. Appreciation is extended to Professor D.D. Lanning for his valuable assistance and to Miss Rita Falco for her patience and competence in typing the final manuscript. All computer calculations were done at the MIT Computation Center, the cooperation of the staff is appreciated. 4 TABLE OF CONTENTS Page Abstract 2 Acknowledgements 3 Table of Contents 4 List of Figures 6 List of Tables 7 Chapter I: Introduction 1.1 Reactor Type Considered 1.2 Earlier Work 1.3 Objective of the Present Work 9 9 Chapter II: Background 2.1 Categories of Reactivity-Induced Accidents in LMFBR's 2.2 Description of Typical CATEGORY II Excursions 2.3 Reactivity Effects Considered Chapter III: 3.1 3.2 3.3 3.4 3.5 4.1 4.2 4.3 Chapter V: 5.1 5.2 5.3 5.4 5.5 13 20 28 Methods of Analysis Description of Computer Analysis Spatial and Spectral Effects Kinetics Model Used Effect of Severe Transients on the Neutron Energy Spectrum Summary Chapter IV: 11 34 37 0 47 49 Transient Heat Transfer Considerations of Phenomena in the Region of Clad Failure, " R. " Energy Exchange in the Region Summary 50 55 69 Reactivity Additions from Sodium Voiding Hydrodynamics of Sodium Voiding Estimates of Energy Addition Rates to Sodium Reactivity Addition Rates from Sodium Voiding Accidents Initiated by Sodium Voiding Summary 70 79 81 86 89 5 Page Chapter VI: 6.1 6.2 Effects Leading to Fuel Motion Reactivity Model for Fast Reactor Core Perturbations Applications of the Reactivity Model to Accident Analysis Core Compression and Expansion Effects on Reactivity Observations from Calculations and ANISN Results Reactivity Addition Rates from Fuel Motion Summary 6.3 6.4 6.5 6.6 6.7 Chapter VII: 7.4 7.5 7.6 7.7 Chapter VII: 8.3 8.4 92 95 105 119 129 137 152 Doppler Effects and Miscellaneous Reactivity Considerations Delays in Doppler Feedback Sources of Delays in Doppler Feedback Effect of Spectral Shifts (and the parameter "n") Doppler Dead Band Due to Heat of Fusion Positive Doppler Feedback from Fuel Cooling Homogeneity Effects Summary 7.1 7.2 7.3 8.1 8.2 Reactivity Effects Resulting from Core Rearrangements 156 166 170 171 174 175 176 Summary and Conclusions General Analytical Models Developed Previous and Future Work in Accident Analysis Comparison with Thermal (Water Cooled) Reactors References 178 182 184 186 192 Appendix A 1. Power History Calculations 198 Appendix B 1. 2. Heat Transfer Correlations Employed Supplemental Sodium Voiding Analysis 207 208 Appendix C 1. Spectral Effect Calculations 212 6 LIST OF FIGURES Figure Page 2-1 Power Level vs. Time 14 2-2 Reactivity vs. Time 15 2-3 Sketch of Core Following a CATEGORY II Excursion 24 3-1 Neutron Flux vs. 38 3-2 Expanded Plot of Figure 301 39 3-3 Total Group Flux vs. Energy 41 5-1 Schematic of Core Model for Sodium Voiding Calculations 71 5-2 Schematic of Temperature Behavior for a Fuel Fragment in Sodium 82 5-3 Pressure vs. Temperature for UO 2 88 6-1 Group III and Group VIII Axial Flux Profile with Central Sodium Voiding 102 6-2 Reactivity vs. Axial Sodium Voiding (20% Sodium Removal) 111 6-3 Reactivity vs. Axial Sodium Voiding (Total Sodium Removal) 112 6-4 Reactivity vs. Radial Sodium Voiding (20% Sodium Removal) 115 6-5 Differential Void Reactivity vs. Axial Position of Void 118 6-6 Reactivity vs. Unperturbed Flux (Cylindrical Geometry) 125 6-7 Reactivity vs. Unperturbed Flux (Spherical Geometry) 126 6-8 Schematic of Fuel Injection Model 139 7-1 Available Mechanical Work as a Function of Doppler Feedback 155 Axial Position 7 LIST OF TABLES Page Table 1-1 Reactor Model 10 3-1 Multigroup Constants for the Hansen Roach Cross Section Set 42 4-1 Vapor Pressures of UO 2 52 4-2 Fuel, Sodium, and Clad High Temperature Properties 60 5-1 Reactivity Addition Rates from Sodium Voiding 84 6-1 Comparison of ANISN and 'PS' 6-2 Reactivity Effects from Core Expansion or Contraction (5k) 124 6-3 Reactivity Effect of Core Rearrangements 130 6-4 Reactivity Addition Rates Resulting from Fuel Motion 149 7-1 Maximum Acceptable Delays in Doppler Feedback 165 Model Results 117 (-Cmax) 7-2 Variation of the Magnitude of Doppler Feedback with the Parameter 172 "n" 8-1 Summary of Reactivity Effects Considered 190 A-1 Approximate Results for Various Reactivity Insertion Rates 205 C-1 Cross-Sections and Spectral Parameters for 'PS' Model 216 8 Chapter I INTRODUCTION The advantages of a practical and safe breeder reactor have long been recognized. observed: As early as 1945, Enrico Fermi "The country which first develops a breeder reac- tor will have a great competitive advantage in atomic energy" (1). Although a great deal of thought and effort have been directed toward the breeder concept in recent years, the fact that much remains to be done is also evident. An exhaustive compilation of the tasks requiring completion is given in the United States Atomic Energy Commission's recently published Liquid Metal Fast Breeder Reactor (LMFBR) Program Plan (WASH-ll10) (1). This ten volume set of documents sets forth a plan of action for producing "a viable industrial capability which will provide LMFBR plants on a self-sustaining competitive basis, at minimum cost, and in a timely manner." The program plan establishes development of the LMFBR, with sodium coolant, as the AEC's priority effort toward large scale acceptance of the breeder. The present work focuses on the safety aspect of such reactors. Specifically, accidents are considered which are initiated by excessive reactivity insertions, or reactivity additions at an excessive rate, into a critical reactor. the course of the study, consideration was given to those problem areas outlined in Volume 10, SAFETY, of the AEC Program Plan (1). In 9 1 M1 1--_ 9 1.1 Reactor Type Considered A series of conceptual design and follow-on studies have been completed by four U.S. nuclear contractors (2-5). These studies, under sponsorship of the USAEC, considered various compositions and configurations for a 1000 megawatt electric (Mwe) fast breeder reactor. The present work has drawn on these studies in arriving at a "typical" LMFBR for use as a basis for analysis. 1.2 Earlier Work Numerous earlier studies in the literature have investi- gated severe reactivity accidents in sodium cooled fast reactors. Such studies are frequently referred to as "Meltdown Analyses" when carried to the point of core meltdown and disassembly. An important work in 1956 by H.A. Bethe and J.H. Tait (6), work. forms the basis of much of the later This well known analysis uses perturbation theory to calculate shutdown reactivity resulting from density changes as a reactor is disassembled by the high internal pressures generated in a severe power excursion. Later investigations have modified the Bethe-Tait approach and have incorporated other considerations into the analysis, such as inclusion of doppler feedback and treatment of zoned reactors (7-10). In addition, computer programs such as AX-1 (11) have been developed to perform coupled neutronic and hydrodynamic calculations. A common feature of these investigations is their 1..1m --=== moi mi is-1141== 111mit,..,.,amme ..11- ... 10 Table 1-1 REACTOR MODEL 1. Fuel Material PuO 2 /UO 2. Fertile/Fissile Ratio 6.5/1 3. Geometry Cylindrical 4. Diameter of Core 274 cm 5. Height of Core 100 cm 6. Blanket Dimensions 20 cm in thickness on all sides 7. Average Specific Power 150 kw/kg fuel 8. Composition (Vol. (a) (b) 2 %) Core Fuel Coolant (sodium) Structure (stainless steel) Blanket U02 Coolant Structure 9. 35 50 15 35 50 15 Clad/Thickness 316 SS/15 mils 10. Fuel Pin OD .25 in 11. Pitch/Diameter (Triangular Arrays) 1.33 12. Mean Centerline Na Temperature at Full Power 1000 0 F (800 0 K) 13. Maximum Fuel Centerline Temperature at Full Power 4700 0 F ( 2900 0 K) 14. Mean Fuel Temperature at Full Power and Location of Peak Flux 2700 0 F ( 1700 0 K) 15. Delayed Neutron Fraction: p .0033 16. Full Power 2500 MWt (1000 MWe) 17. Neutron Mean Generation Time: .A. 3.3 x 10~ sec 11 treatment of core fuel, clad, and structural materials as homogeneous throughout the excursion. The principal reac- tivity effects, and hence the severity of the accident, have often been analyzed without considering the detailed progression of events taking place during the course of the transient. The reactivity effects usually included are: an accident "ramp" reactivity insertion from an unspecified source (or from an unspecified process of sodium boiling); Doppler feedback*; and disassembly feedback. 1.3 Objective of the Present Work The objective of the present work is to consider the progression and interrelation of events taking place during severe power excursions. A particular effort is made to include every phenomena that might affect reactivity and thus, ultimately, the seriousness of the accident. A description of the accident and the various reactivity effects investigated is the subject of Chapter II. The detailed study which follows attempts to identify by analytical arguments the more important parameters governing each reactivity effect and to show clearly the overall importance of the effects considered. Chapters III through VII are de- voted to this effort. *In reactors of the type under investigation, Doppler broadening contributes most of the reactor's negative temperature coefficient and, as such, is the primary inherent mechanism which can provide the negative feedback necessary to prevent core disassembly in the event of a severe reactivity excursion. 12 The area of reactivity changes associated with fuel and other core material movements is the subject of considerable study in Chapter VI. The usual form of one group perturba- tion theory was found entirely inadequate for this purpose. Thus, the development of a simple analyticalreactivity model capable of handling local and global perturbations in an LMFBR, and therefore useful in analyzing accident conditions, is undertaken. The ANISN/DTR II multigroup code is employed in analyzing core rearrangements and for determining the accuracy of the simple reactivity model developed. As a final assessment of the significance of each phenomenon considered, a qualitative comparison between the comparable effects in thermal reactors (specifically H20 moderated PWR's and BWR's) and LMFBR's is given in the concluding chapter of the thesis. 13 Chapter II BACKGROUND In this chapter three categories of reactivity excursions in LMFBR's are defined; generally according to the severity of the accident. A typical excursion in the category of principal interest in the present work is described and reactivity effects which can arise during such an excursion are discussed. 2.1 Categories of Reactivity-Induced Accidents in LMFBR's The oscillatory behavior of prompt supercritical excur- sions in LMFBR's with negative Doppler feedback has been amply demonstrated. (7)(13)(15)(16)(17)(19) Figures 2-1 and 2-2 show the time history of power and excess reactivity for two "cycles" of a typical transient of this type. If no corrective action is taken, clearly any excursion induced by a continuous "ramp" of reactivity will result in clad failure and eventually in core destruction or disassembly. For pur- poses of discussion, it is somewhat arbitrarily assumed that external corrective action (such as scram or cutback) can "turn down" or terminate an excursion after about 200 milliseconds (55). For excursions considered in the present work, the effects cf phenomena of interest will be seen to take place well within the first 200 msec after initiation of the accident; thus the exact value of the scram delay time is not important. 2-I~ ri&u~ ~OLt4EJ~ To0 ~tFP L~N~.L-~ ---- 41 ro I I' 11 I I 3 £10 'ho 30 ~- 210 ------- 4* / 4 -- - CIA -- 10 I I 'S I--- / / 1 Is I 1 / -- -- I S 3 41 I %2- (Qnsac .) - 1) )o' * 2. A -00 3 JAI1 5" 1(6 11 18 19 :Zo FIG~URE Z-Z vs. TtN\E-. - _4______ o6 2 3 Q It l~z 13 JI A2/ I ! / In Chapter VII, a ramp rate of 10 $/sec is found to result in four cycles or power pulses in a 200 msec interval. Note that 100 msec are required to achieve prompt criticality; the four super-prompt critical power bursts then occur in the remaining 100 msec. If corrective action is taken after the fourth power oscillation, clad rupture is not expected to occur for such an excursion in the LMFBR's considered here (see Appendix A). For higher reactivity insertion rates, clad failure can occur within the first 200 msec and may occur after the first, second, or third power pulse. For purposes of discussion, three categories of such excursions are defined as follows: CATEGORY 1: An excursion which is not severe enough to cause clad rupture (or melt-through) in any part of the core within 200 msec after initiation of the accident reactivity ramp. Present models of transient heat transfer from intact fuel rods and resulting sodium void effects, Doppler effects, and the associated reactor kinetics models appear sufficiently accurate to predict the threshold of clad failure for such excursions with reasonable precision (8)(12)(13). An approximate analytical solution defining this threshold is given in Appendix A and discussed in Chapters VI and VII. 17 CATEGORY II: An excursion which leads to clad rupture in some region of the core but in which the power pulse causing clad rupture is "turned down" by Doppler feedback before pressures sufficient for core disassembly are generated. Clad failure may occur at any time within the first 200 msec: after one or more super-prompt critical power bursts. To simplify terminology, the power oscillation(s) occurring prior to clad failure will be referred to as the initial excursion. Transients or power pulses occurring after clad failure will be referred to as secondary excursions. Core disassembly may or may not occur subsequent to the initial excursion. For reactors of the type under consideration, Doppler induced reactivity changes are thought to be sufficient to turn down the first power transient before core disassembly, even for very severe excursions resulting from assumed reactivity insertion rates in excess of 100 $/sec (13)(15)(16)(17). (Note that the time to reach prompt criticality at a ramp rate of 100 $/sec is 10 msec. Again from Chapter VII, the time between successive power peaks is about 8 milliseconds.) Thus, CATEGORY 2 covers quite a wide range of reactivity accidents. 18 Clad rupture in this category will be followed by fuel fragmentation and/or dispersal into the adjacent coolant channels; a phenomena which is discussed in some detail in Chapters IV through VI. CATEGORY III: For extremely severe excursions result- ing from accident ramp rates in excess of approximately 200 $/sec, the first power pulse is not "turned down" by Doppler prior to core disassembly. In such a case the entire excursion, from the time of prompt criticality to shutdown by disassembly, is over in about five milliseconds or less (7)(18). The elapsed time from clad rupture until pressures sufficient to start core disassembly are generated is of the order of one millisecond (7). Calcula- tions in Chapters VI and VII indicate there is insufficient time for the phenomena which takes place between clad failure and core disassembly (other than disassembly itself) to alter the behavior of such a rapid excursion. Since the progression of events following clad failure 19 (and prior to core disassembly*) is of particular interest in the present work, CATEGORY III accidents will not be treated. This does not, of course, limit the severity of accidents studied but simply rules out of consideration an initial excursion of the severity here classified as CATEGORY III. CATEGORY II accidents, then, are the subject of the present study. It must be pointed out that this category covers the more credible severe excursions, and perhaps the complete range, since it can be realistically argued that the rates of reactivity insertion required to override Doppler and lead to core disassembly during the first power pulse of an excursion (CATEGORY III) are not credible in large oxide cores of the type under consideration (see Section 2.2 of the present chapter). In the important range of accidents here defined as CATEGORY II, however, the events following clad rupture may combine to produce a more severe secondary excursion, or to prevent complete Doppler *n , a distinct possibility of core compaction into a second critical configuration exists for certain geometries of large fast reactors. Such a compaction could be accomplished at a very rapid rate, thereby possibly increasing the severity of the initial excursion or causing a second excursion just as disassembly is terminating the first. Indeed, the rates of negative reactivity insertion during disassembly are estimated to be well in excess of 1000 $/sec (7). If the pressures which cause disassembly at such rates can conceivably cause re-assembly into a second critical configuration, the serious implications are evident. Reassemblies of divided sections of a core under the acceleration of gravity have been discussed frequently in the literature (2)(4)(7)(15). While such an accident appears no more credible than reassembly by differential internal pressures, the latter could pose a much greater hazard. Neither type of accident is the subject of the present investigation, however. 20 "turn down" of an initial 2.2 excursion. Description of Typical CATEGORY II Excursions In recent literature on LMFBR accidents, reactivity in- sertion rates in excess of about 66 $/sec are not given much credence in that this is the maximum value estimated for sodium voiding from intact cores (12)(13)(23)(24)(25); for core collapse under gravity (24); and is the maximum considered in investigating a number of initiating accidents such as control rod ejection, loss of flow, etc. (25). In refer- ence (23), 66 $/sec is used as the reactivity insertion rate in the Design Basis Accident (DBA) proposed by Atomics International (Aronstein) and in reference (24), 65 $/sec is the maximum rate estimated for "hypothetical accidents" considered by Combustion Engineering (Noyes). Furthermore, in reference (25), 60 $/sec is established as the "basis of the design" of the Belgian-Dutch-German Consortum LMFBR; the 300 Mwe SNR reactor. Thus, in the present work, 66 $/sec will be used as a basis of comparison for various reactivity effects considered. For the LMFBR considered here, a reac- tivity insertion rate of 66 $/sec will result in a CATEGORY II accident, as will be seen in succeeding sections of the present chapter. In the presence of Doppler feedback of the magnitude expected in large oxide fueled LMFBR's, as typified by the reactor of Table 1-1, initial excursions in the category of 21 interest are expected to be "turned down". That is, power will be reduced to its level at prompt criticality and reactivity reduced to less than (1+P) (18)(19). In fact, in the limiting case of zero generation time, Hafele has shown that reactivity is reduced below (l+P) after an initial transient by an amount equal to the maximum rise above (l+s) during the transient (19). As noted, if the initiating reac- tivity ramp continues, successive power peaks will occur until the accident is terminated by core disassembly. References (13), (15), (16) and (17) (representing studies in France, Japan, The United Kingdom, and the United States respectively) show detailed histories of excursions in which two or three power peaks occur before disassembly takes place. Various "representative" Doppler coefficients are used in these four studies with reactivity ramp rates of from 40 $/sec to 100 $/sec. The duration of the excursions, from prompt criticality until shutdown by disassembly, ranges from 25 to 70 milliseconds; with time intervals between power peaks of 10 to 20 milliseconds. Thus, these CATEGORY II ex- cursions are over in well under 200 msec and insufficient time is available for external corrective action by presently known methods. None of the four studies includes the fuel fragmentation/dispersal or fuel motion effects considered in the present work. During such accidents, clad rupture could occur after the first or any subsequent power peak. For 22 purposes of discussion, clad rupture will be assumed to occur during or immediately following the second power peak of an excursion typified by the time histories of power and reactivity shown in Figures 2-1 and 2-2. The region of clad failure following such a transient depends primarily on the accident reactivity ramp rate, the strength of the Doppler feedback, and the neutron flux shape. If the effects following clad rupture are not pre-emptive, and if the accident reactivity ramp continues, a third power pulse will occur as discussed above and as shown by the dashed curves of Figures 2-1 and 2-2. The excursion represented by Figures 2-1 and 2-2 assumes an accident reactivity ramp rate of 66 $/sec. Doppler feed- back reactivity behavior is assumed to be of the form: dK -ADO( ) , (2-1) T where ADOP = .003 to .008, n = 0.8 to 1.2. The parameter "n" is determined by the fissile/fertile ratio, the fuel temperature, and the neutron energy spectrum of the LMFBR in question. It is fairly close to unity in LMFBR of the type considered here. For the present estimate n was assumed equal to 1.0 and ADOP was taken as .004; hence: 23 5K where DO? T= - oo4n( , O)l(- (2-2) average fuel temperature at steady state full power. The flux shapes for the core considered were calculated by using the ANISN multigroup Figure 2-3. computer code and are shown in For the ramp rate, Doppler feedback, and flux shapes described, calculations described in Appendix A indicate that clad rupture and fuel fragmentation/dispersal will occur over a substantial region of the core ( " 30 volume percent). This result is depicted in Figure 2-3. The growth of the fuel rupture region, hereafter designated -19,, tinues until time t2 of Figure 2-1. con- The trend in large LMFBR's is toward designs which produce a relatively flat radial flux profile with steeper axial flux gradients (through the use of fuel zoning and pancake geometry) as indicated in Figure 2-3. Thus'ucan extend over a large radial area of the core while its axial propagation remains relatively small. The vapor pressure of UO 2 is a strong function of temperature at pressures near the threshold of clad failure, hence a transient which produces pressures just sufficient for clad rupture out to i z I = 30 cm in the axial direction could produce pressures several thousand psi higher near the core center (see Chapter IV for supporting calculations and further discussion). Thus, the fuel nearer the center is F i GuRV SkE--iL 24 Z- 3 o-E COFCORE FOLLOWiJ& A CA-Te&oaRY rL - Exrenoa4O - 4~ zOm. z I I AX IAL''-RE irLrr-c MQ' SM i Kts ~ 2o cm. I to / I I 137om . 10 CoAE is VEp c-TED A-r TIME Ag 1:u LvA SHAge ' ?A 10R -0T-o n%K or lrl&-2-L. *2 - Wi-ir 25 likely to be more finely fragmented or dispersed (20)(32). Presumably the clad will remain intact outside la; at least immediately following time t , 2 The fuel vapor pressure profile within the clad (prior to clad rupture) is not expected to be completely symmetric about the axial centerline of the core due to the fact that clad temperatures are higher in the upper half of an operating core (21). The deviation from symmetry and correspond- ing axial dislocation of regionR is expected to be small for severe excursions and of little significance in the arguments presented here. Thus, in Figure 2-3 and for purposes of discussion, the size and axial position of region IL have been determined from the axial flux profile and hence for an axially symmetric pressure distribution. For an excursion of the type described here and depicted in Figures 2-1 through 2-3, the following characteristics apply and will be of some importance in subsequent reasoning and developments: a) The time scale of events is of the order shown in Figures 2-1 and 2-2. In the absence of pre- emptive effects following clad rupture, as considered herein, or some external shutdown or cutback action, and if the accident reactivity ramp continues the time interval between successive power peaks will be about 2 tp or approximately 26 8 msec for the particular transient illustrated. Furthermore, the third power peak will occur about 20 msec after the time of prompt criticality. b) For this time scale of events, the heat conduction through the intact clad is a very small fraction of that generated in the fuel. It is, in fact, generally less than the energy deposited in the sodium by prompt neutron and gamma heating. Calculations verifying this assumption are given in Chapter IV. c) No appreciable melting of clad occurs in the regions where the clad remains intact for 20 milliseconds or so following the first transient. in &R is Clad rupture the result of excessive fuel vapor pres- sures; not clad melting. (The higher average temperature of the clad does, however, result in a significant reduction in stress required for rupture and this is taken into consideration.) d) The fertile constituent of the fuel (UO2 ) contributes the negative component of the Doppler feedback. Theory and experiment indicate that the reactivity contribution from the fissile isotope (PuO ) is 2 quite small by comparison (22). The Pu0 2 contri- bution is taken as zero in this work and the small 27 percentage of fissions occurring in the fertile component is ignored in discussions of Doppler feedback. Direct neutron and gamma heating of the fertile species is also ignored initially; this is discussed further in Chapter VII. e) The power pulses in such transients are of relatively short duration. In Figure 2-1, where clad failure is assumed to occur during the second power pulse, the pulse width is 2 msec. This is relative- ly short compared to the time between successive power pulses (8 msec in the present example). Thus the growth of region'd& is expected to be rapid with a distinct termination at time t , 2 f) The reactivity above prompt critical is unlikely to exceed 40, even for extreme initiating accidents, before the excursion is turned down by Doppler feedback or terminated by core disassembly. (This observation is discussed further in Chapter III). The total reactivity available from most of the phenomena considered in the present work exceeds 40i, however. Thus, the rate of reactivity addi- tion rather than the total reactivity available is found to be the more important consideration. g) The transients considered herein are assumed to start from one of two conditions of reactor 28 operation; full power or 10 percent power. In both cases, the average sodium temperature in the core is taken to be 800 0 K. For power levels below 10%, presumably plant operating procedures will require that plant temperatures will be lowered; resulting in a substantial increase in the energy required to produce sodium voiding and in a substantial improvement in the magnitude of Doppler feedback (see Eq. 2.3 (2-1)). Reactivity Effects Considered An examination of the possible progression of events in the excursion describedabove suggests the following reactivity effects should be considered. 2.3.1 Increased Rate of Sodium Voiding Following Clad Rupture The fuel fragmentation or dispersal following clad rupture in region-R exposes higher temperature surfaces - and potentially a greater surface area - to sodium in region 1. Analyses which have not considered fuel failure have been used to calculate maximum reactivity addition rates of 15 to 65 $/sec as a result of sodium voiding (12)(23)(24)(25). If the heat transfer rate from hot fuel to sodium is appreciably increased in '6:L , a substantial increase in the rate of reactivity addition from this source is possible. 29 Furthermore, high fuel vapor pressures near the core center, in a region from which sodium has been voided, can further accelerate the voiding process. The strong depen- dence of fuel vapor pressure on temperature in the range of interest will be seen to make this a plausible effect. 2.3.2 Fuel Injection from Intact Clad Immediately following time t2 of Figure 2-1, the fuel outside " is constrained by its cladding except, for an "open end" or "split" somewhere in V., (see Figure 2-3). Since the fuel vapor pressure in region 6p. was sufficient to rupture the clad, the vapor pressure within the clad can be considerable some several inches outside 6R (or in a nonvented design, the fission gas pressure could also be considerable). Since the fuel even farther away from 'R is cooler and is eventually "tamped" by the axial blanket material in the continuous rod designs usually proposed, the potential exists for fuel movement or "injection" into the region V and hence into a higher worth region of the core. 2.3.3 Fuel Movement Under a General Inward Pressure Gradient If fuel dispersal and subsequent substantial fuel cooling takes place in region L, as discussed in 2.3.1 above, the potential exists for an inward pressure gradient; independent of restraint by the clad. During the time interval between an initial and secondary power excursion, fuel in 30 central regions ofltcould conceivably be subcooled by several hundred degrees relative to that outside or near the edge of "aL. If this occurs, fuel vapor pressures in'R will tend to "lag" the vapor pressures generated in adjacent regions, thus tending to accelerate fuel into higher worth regions of the core during some portion of a secondary transient. It is shown in Chapter IV that the fuel vapor pressure generated during such a secondary transient can rapidly exceed the sodium vapor pressure generated between the initial and secondary excursions. 2.3.4 Positive Doppler Effect The high temperatures in the fertile material, which produced the negative Doppler feedback required for turn down of the initial transient, can be reduced at a rapid rate in region - as the fertile material comes into more intimate contact with sodium coolant. Thus the cooling of fuel in region -R is a source of positive reactivity insertion and could significantly change the rate of reactivity insertion at the time of prompt criticality during the start of a secondary excursion. This positive Doppler feedback will continue until a secondary transient progresses to a point where the rate of heat addition from fission is sufficient to overcome the cooling effect described. 2.3.5 Doppler "Dead Band" Due to Heat of Fusion and Vaporization During the time UO2 is being melted, little or no Doppler 31 broadening is thought to occur in a macroscopic sense (26). The heat of fusion of UO 2 is 278 joules/gm0 K; or roughly equivalent to a temperature "dead band" of 700 0 K. During a secondary transient, which starts from a higher temperature datum, much more fuel could be affected by this "dead band" than is the case for an initial transient. If a significant portion of the fuel is in this dead band during particular portions of a secondary excursion, the effect could be substantial. The heat of vaporization can play a similar role but behavior of the system with regard to this phenomenon is not clear. This is discussed further in Chapter VII. 2.3.6 Doppler Reduction Due to Spectral and Temperature Effects As temperature increases, a reduction in the magnitude of Doppler feedback is predicted by Eq. (2-1) for all values of "n" of current interest. In addition, effects which harden the energy spectrum (such as sodium voiding) cause a reduction in the strength of Doppler feedback. The reduction has been represented by a decrease in ADOP (27) or by an increase in "n" of Eq. (2-1). Furthermore, an increase in "n" has been suggested as temperatures increase (28)(29). All of these effects tend to reduce Doppler effectiveness during a secondary transient. 2.3.7 Delay in Doppler Feedback If the fissile and fertile fuel materials are mixed as 32 powders with particle sizes on the order of the mean range of fission products in fuel ( ^j 10 microns) or larger, a finite period of time is required for transfer of the heat of fission from the fissile isotope to the fertile isotope. This results in a "delay" or "time lag" for Doppler feedback (9)(30). This delay will affect the initial transient, but could have a stronger effect on a secondary excursion since the fuel material will be mixed with sodium and structural materials. This dispersal or mixing reduces the direct fis- sion product heating of the fertile species. Additionally, if dispersion is fine enough, the introduction of sodium vapor between fissile and fertile particles and the resulting film temperature drop could increase the time required for conductive heat transfer. Both of these latter effects pro- mote a greater time delay between the fission event and Doppler broadening of UO . 2 2.3.8 Transient Induced Spectral Shift If the neutron distribution as a function of energy is altered as a result of a strong transient, a reactivity effect and an influence on the behavior of other phenomena, such as Doppler feedback and sodium voiding, can be induced. Specifically, the inclusion of "W/v" terms in the usual multigroup formulation is analogous to the insertion of a "l/v" absorber. This in turn can be expected to produce a spectral perturbation. I........ .. 33 2.3.9 Homogeneity Effects Although fast reactors are generally considered homogeneous in neutronic spatial calculations, a recent study (31) indicates the small degree of fuel "lumping" employed in current LMFBR designs is beneficial (i.e. results in increased reactivity). The gain in reactivity arises primarily from the increased "first flight" neutron flux within the fuel rods (31). Thus the tendency towards homogenization which results from severe accident conditions can induce a negative reactivity effect. Effects 2.3.1 through 2.3.9, then, could combine to substantially alter the behavior of a secondary excursion. The controlling parameters influencing these effects are discussed in succeeding chapters; with quantitative estimates of the maximum total reactivity change and reactivity insertion rates for some of the individual effects. The nine effects above are listed in Table 8-1 together with information developed in succeeding chapters. 34 Chapter III METHODS OF ANALYSIS The present chapter gives a description of the computer analyses employed in this research. The effect of severe perturbations or core rearrangements (simulating accident conditions) on the space and energy distribution of the neutron flux is analyzed, using multigroup computer calculations. The neutron kinetics equations employed in succeeding chapters are presented and the question, posed in Chapter II, as to whether a severe transient produces a significant spectral disturbance, in and of itself, is answered. Description of Computer Analysis 3.1 The ANISN/YDTF II Multigroup Code (56) was employed for various reactivity calculations and for investigation of spatial and spectral effects. The S-8 transport theory approxi- mation was used for the final values of runs; the diffusion theory approximation was employed in setting up the various problems of interest to save running time. Cross section sets employed included the Hansen Roach 16 group set (57) and the Russian ABBN 26 group set (58). Differences in reac- tivity effects predicted by the two sets were found to be quite small; comparative data is given in Chapter VI. The ANISN code is one dimensional, but buckling values (B ) can be specified in one or two additional dimensions to 35 simulate cylindrical or parallelpiped configurations. The code then calculates leakage in these additional directions by including an artificial absorption term, DB , in the 2x multigroup formulation for each Bx specified. In the present work, the following steps were employed to obtain a base case representation of the reactor described in Table 1-1. 1) Calculations were made in cylindrical geometry with a first estimate of axial buckling, B , to account for axial leakage. The core radius was adjusted to achieve a critical system. 2) The code was then modified to calculate the flux dependence in the axial direction using a value of B to characterize radial leakage. B chosen was that corresponding to the fundamen- The value of tal mode eigenvalue which gave the best fit to the radial flux shape calculated in step (1). This proved to be a very accurate method of estimating radial leakage because of the predominance of the fundamental mode shape over most of the core, as will be seen in the next section of the present chapter. 3) The fissile/fertile ratio was then modified slightly to obtain a critical system for a core height of exactly 100 cm. 36 4) Iterations of steps (1) through (3) were carried out until the changes in radial buckling and the fissile/fertile ratio were negligible. The re- sulting values are those given in Table 1-1. All subsequent calculations for the reactor of Table 1-1 were made in this pseudo-cylindrical geometry with the axial direction as the dimension of primary interest. In addition, a number of computer runs were made in spherical geometry for comparison purposes, using a reactor with the same composition as that of Table 1-1. The core was divided into 40 intervals of 2.5 cm each in the axial direction for cylindrical geometry and 20 radial intervals of equal volume in spherical geometry. The concen- tration or density of each core constituent could be varied in any interval(s) to represent various accident conditions. For example, sodium voiding in the central 10 cm of the cylindrical core could be represented by specifying zero sodium density in the four central 2.5 cm intervals. For convenience, a concentration factor, C, is defined as the ratio of the perturbed density of a given material to its density in the unperturbed critical core. C px perturbed px unperturbed , Thus, (3-1) where the subscript x denotes the core constituent; Na, Fe, or fuel. 37 3.2 Spatial and Spectral Effects Calculations with the ANISN code, made by using both 16 and 26 energy groups, indicate that the neutron flux shape over the central 80% of the core volume corresponds to the fundamental mode shape. Furthermore, the results obtained show that even severe local perturbations do not significantly alter the flux profile. Figure 3-1 shows a plot of neutron flux versus axial position in the core for four cases of interest. Figure 3-2 is an expanded plot of the same data showing percent deviation from the unperturbed flux profile. Curve 1 (in both figures) is the unperturbed profile. Curve 2 is a cosine curve, representing the axial fundamental mode, with an effective core height of 137 cm. In Figure 3-1, curves 1 and 2 coincide except for the region within 7 or 8 centimeters of the blanket. Curve 3 shows the results for complete sodium voiding (CNA=0) in the central 25 centimeters axially, and in the entire radial direction. The reactivity calculated for this degree of sodium voiding was + $3.70. Curve 4 shows the results for sodium voiding as above, but with all fuel (UO2 and PuO 2 ) from the 12.5 to 15 cm region (Cfuel=0 in interval 15~ ) moved inward and smeared uniformly over the central 25 cm of the core (Cfuel=1.2 in intervals Ll-4 - ). served. Notice that the total mass of fuel material is conIn Figure 3-2, observe that the maximum deviation from the unperturbed flux profile or from the fundamental 38 Lse.. ~ cr -J C U ZV F- C' ~E'4CEP- 7' U42t 2r X p I opl~4-r~-s scjLA- P/-or Oulr6lbr. z=:40c*, 4 PkU V2S%-. f-l CUAVC NAzVIDE u R4 i L 4 0te 31 ~)43-2 Ok2 bATA 2Poi.sr) Tr L, 2to Not)Iz IZ C.vi. S e A~s> 2. Scz(o ) Ar16ru~e-)?ADFD ?J-oTr salb rZ oa I E(7 F 9 3- 1 514DWiN& DE,11wrloo 4 0 V: C J Me -5 1 1-3.P MAD FRO M - HE NNbAMC-NIML IF&oAOrm___ , 3-2~~ MODE 5HIAPr= -Irr I Yr <p REctoN~ OF' C-o~ uM tA FIL A LoJ N 0RCuAVt' R~ CU o-IA'N VE / / / 01 7' 2 0 g a: .7 (~J A 0 A I 2 U (~. / / cp 1k ~~-/;~ Cf, 2o - N~S Cu~vcE 3 NA1j, '4JoIDE jriif, )Z}lL )2.5c-. - UNFOPb1L Sri~cZ~ 10 IC, 20 30 ~ ~z~j2.c~,&E~1oN Sb j 4o mode shape is about 1%. Also note that even in the 12.5 to 15 cm region where all fuel is removed, the flux dip is quite small. The reactivity calculated for this rearrangement was + $3.30; quite a substantial perturbation. This tendency of the flux shape to remain very close to the fundamental was observed to hold for all core perturbations or rearrangements considered. The flux shown in Figures 3-1 and 3-2 is the total flux (sum of the flux in all energy groups) at a given spatial position. The spatial shape of the flux in each individual energy group was observed to behave very much like the total flux, but there are, of course, local changes in the energy spectrum as will be discussed shortly. Reactivity perturbations in the hypothetical cases investigated ranged as high as + $6 (see Table 6-3 VI). of Chapter Recall from Chapter II that the net positive excess reactivity cannot realistically exceed about + $1.40. Thus, for perturbations of interest, the assumption of a fundamental mode flux shape in the central regions of the core should be quite valid. Figure 3-3 is a plot of the total flux in each energy group for the first nine groups of the Hansen Roach crosssection set (groups 10-16 are in the epithermal and thermal range and have very little influence on the behavior of the LMFBR under consideration; Table 3-1 shows the energy and lethargy range of each group). Curve 1 shows the unperturbed 3-3 Fua-ForAL d/o . 3 'I. CvE13: Iuoj 30- I 6&(Zou? FLJ WAOJ- ON SK c-rqw C- u Rv s-: , 44 - C,,=-I 4), Lv -S E c Lj = - 0 000 stC 2 N 4 to. E 0 40 4 0 Af 4 1~0 2. 7 0.0011j77 W. %-.,I tj I . a W. .4 t 0. 5- A V C kA(s E t R 0 u -,-' S S C- R Z- Y, Z:..* M C V. --f- J. I S' z 2.2.0 I 42 Table 3-1 MULTIGROUP CONSTANTS FOR THE HANSEN ROACH CROSS SECTION SET Group Neutron Energy Range Avg. Neutron Velocity, Au Fission Spectrum 108 cm/sec 28.5 0.204 0.762 19.9 0.344 0.9 - 1.4 Mev o.442 14.7 0.168 4 0.4 - 0.9 Mev o.811 11.0 0.180 5 0.1 - o.4 Mev 1.386 6.7 0.090 6 17 - 100 key 1.772 2.70 0.014 7 3 - 17 key 1.735 1.14 0 1.696 o.48o 0 1 3 - oMev 2 1.4 - 3 Mev 3 8 0.55 - 3 key 9 100 - 550 ev 1.705 O.206 0 10 30 - 100 ev 1.204 0.101 0 11 10 - 30 ev 1.099 0.0566 0 12 3 - 10 ev 1.204 0.0319 0 13 1 - 3 ev 1.099 0.0179 0 14 o.4 - 1 ev o.916 0.0109 0 15 0.1 - 0.4 ev 1.386 0.00606 0 0.00218 0 16 Thermal (0.025 ev) 43 energy spectrum calculated by ANISN. Curve 3 shows the hardened spectrum which results from voiding the entire core of sodium. Notice that the reduction in neutron flux in group 8 is roughly 50% and that the increase in group 3 is roughly 20%. Thus a comparison of Figures 3-1 and 3-3 shows a much more substantial qualitative change in the spectral shape than in the spatial shape for analogous perturbations. These results suggest that reactivity effects can be accurately calculated by assuming that the spatialflux profile corresponds to the fundamental mode, provided the spectral shift effect is properly accounted for. In particular, perturbation theory should be useful in dealing with the effects in question. This conclusion is given further support in Chapter VI and a simple method for accounting for spectral effects, within the framework of perturbation theory, is presented. In addition, since power density and energy density are directly proportional to neutron flux, the fundamental shape functions should be useful in analyzing effects associated with energy density; such as local temperatures and pressures. Use is made of this observation in Chapters IV and V. 3.3 Kinetics Model Used The simple spatial flux dependence shown in the previous section indicates that the point kinetics equations can be usefully employed in investigating time dependent effects. Only system behavior in the prompt critical range is considered since CATEGORY 2 excursions are expected to lead rapidly to prompt criticality. For such analyses, the simple point kinetics equations given below have been found to give accurate results provided proper initial conditions, calculated with consideration of delayed neutrons, are specified (7)(8). With the usual notations: 5 k dn (3-2) or, since in the reactor of interest, power level is everywhere proportional to neutron flux: dq__ k(t)) - dt - qo A(3-3) where q(t) = power density at time t bk(t) = reactivity above prompt critical A = neutron generation time. Time zero is taken as the time of prompt criticality and the powerlevel at this time is given by: (7)(8) 0o =ss where 2a q ss = delayed critical steady state power density P = delayed neutron fraction a = reactivity ramp rate in 5K/sec. 45 Because of the short neutron generation times, ramp reactivity additions are generally considered more realistic than step additions in fast reactors. The solution to Eq. (3-3) for a step insertion is given here, however, to illustrate a point: q(t) = qoexp( j) . (3-5) Typically, lifetimes in a thermal reactor are about a factor of 100 longer than in LMFBR's, whereas the delayed neutron fraction is roughly a factor of 2 larger. With these compara- tive values, if a step reactivity input of 40o above prompt critical is assumed for both systems, Eq. (3-5) gives: q(t) qoe 4 000t for the LMFBR and = qoe 8 0t q for the THERMAL system. In an LMFBR, as will be shown, time intervals of interest during an excursion are typically of the order of 10 milliseconds. If the systems described above are not altered for 10 msec following the step insertion, the power level will, in theory, reach q = qoe 10 3q for the LMFBR and q~)= qoe* 8 2.2q 0 for the THERMAL system. 46 Thus, the hypothetical power level achieved in the LMFBR is more than 1012 times greater than the level reached in the THERMAL system if an excess reactivity of ~%$1.40 is maintained in both systems for 10 msec. This indicates in a very quali- tative way that, while excess reactivities of several dollars may be of interest in THERMAL systems, such is not the case for LMFBR's. This conclusion is reached in a more rigorous manner in references (7) and (8) and substantiated by comments in (22). For the more realistic ramp input of reactivity as the accident initiating condition, Eq. (3-3) gives: E~t) oexp 2 - t d p + A15K 36 feedback The feedback reactivity is frequently written: SK~t + 5K =5K Doppler feedback Na voiding +5SK+.. disassembly (3-7) Such an expression may be misleading, however, as some of the effects may not be linearly additive. In Chapter VI, in fact, it is shown that the effect of a given amount of fuel motion or rearrangement depends strongly on the degree of sodium voiding present. The nonlinearity is shown to arise primarily from spectral effects. In the present analysis, feedbacks from the various effects listed in Table 8-1 are treated as part of the integral term of Eq. (3-6) whenever possible; they are not assumed to be linearly additive. 3.4 Effect of Severe Transients on the Neutron Energy Spectrum The question of whether a significant shift in the neutron energy spectrum is induced by a strong transient, independent of the effect producing the transient, is now examined. One straightforward method of doing this is to include an appropriate "temporal absorption" cross section, W /Vi, in each energy group of the usual multigroup formulation; where W is the inverse reactor period and V. is the appropriate velocity The "/v" for each group (see Table 3-1). terms can be intro- duced into an otherwise critical system which, for a positive (a , causes the code to calculate an "effective k" of less than 1.0. Alternately, the "M4/v" terms can be introduced to compensate for some other perturbation, such as sodium voiding. For a given degree of sodium voiding, if the proper W is chosen, clearly the code will calculate an effective k of 1.0. approaches were used in the present analysis. Both Values of J of from 1000 sec1 to 40,000 sec~ were employed and the cal- culated energy spectrum examined. The energy spectra genera- ted were compared with those calculated for an unperturbed critical system and with the spectrum which results from voiding the entire core of sodium. The spectral shift associated with sodium voiding produces relatively well known reactivity effects and therefore should form a good basis for comparison. 48 The spectral shifts induced by severe transients were found to be quite small for all values of W considered. 3-3 shows the results for 4a = 40,000 sec~ . Figure Curves 1 and 2 show the energy spectrum for an unperturbed system (k = 1.000) and one with Wa = 40,000 sec 1 (k = 0.998) respectively. Curve 3 shows the spectrum with sodium voided (k = 1.012) and curve 4 shows the results for sodium voiding with the simultaneous inclusion of an W' value of 40,000 sec~ ; which reduced the net k value to 1.000. Examination of Figure 3-3 shows that the shift introduced by an 3 of 40,000 sec~ 1 is negligible compared to that induced by sodium voiding. As should be expected, values of W of less than 40,000 sec1 produced proportionately smaller spectral shifts. It should be noted that an ) of 40,000 sec- corres- ponds to an asymptotic period of 25 microseconds and, for the values of neutron generation time and delayed fraction given in Table 1-1, this corresponds to a (point kinetics model) total excess reactivity of .013 or about + $3.90. (in fairly good agreement with that calculated by ANISN; i.e.: (K-l) = .012 or about + $3.65) Recall from Chapter II that the maximum realistic excess reactivity is about $1.40. Therefore an ta of 40,000 sec 1 corresponds to an appreciably shorter asymptotic period than can be realistically achieved in an LMFBR. Thus, clearly, the transient induced spectral shift can be considered negligible for purposes of accident analysis in the present work. -7 49 3.5 Summary The application of the multigroup code ANISN to the present work has been described. The code was employed in the present chapter to show the effects of core perturbations (which are exemplary of accident conditions) on the neutron space and energy distributions. A very minor influ- ence on the spatial distribution was observed whereas the spectral disturbance was found to be quite pronounced. The minor influence of severe core perturbations on the spatial flux shape was used to justify employing the point kinetics equations for analyzing time dependent behavior. An appropriate form of these equations for use in succeeding chapters was presented. Finally, it was shown that the spectral distortion or shift produced by a severe transient is minor; and negligible when compared to that produced by other effects of interest in the present analysis. 50 Chapter IV TRANSIENT HEAT TRANSFER The present chapter considers the extent of fuel fragmentation subsequent to clad rupture or failure. The mechanisms and rate of energy exchange between the hot fuel fragments and the relatively colder sodium are examined in detail. 4.1 Considerations of Phenomena in the Region of Clad Failure "-R" For rapid (CATEGORY I] excursions, no clad melting is expected to occur prior to the time of clad failure. During the time interval of interest for such an excursion, more heat is added to the clad and coolant by fast neutron and gamma energy deposition than by conduction. Thus, the time of clad failure is determined primarily by the rate of pressure buildup within the clad. Furthermore, since very little heat conduction takes place within the fuel during the power transient, the fuel can be considered to be heated adiabatically for purposes of analysis. These simplifications have been employed in earlier works (21) (43); their validity is also substantiated by results obtained in the present chapter. Clad failure during a CATEGORYII excursion may occur by longitudinal splitting or by general fragmentation of the clad. The degree of fuel and clad fragmentation and the method of clad failure are important in the present work. -I 51 The threshold of clad failure in a rapid transient corresponds to the integrated energy addition at which the hottest fuel in a given core location reaches a temperature high enough to produce vapor pressures sufficient to overstress the clad. At the core hot spot this requires an adiabatic energy addition of about 1080 joules/gram above the energy density of the fuel at steady state full power. As seen from Table 1-1, the maxi- mum hot spot steady state temperature is taken to be 2900 0K at the core center. Equation of state relationships for oxide fuel materials have been given by Meyer and Wolfe (33), Braess (9), and others (44), for various ranges of temperature and pressure. In the range of interest here, the pressure of the vapor in equilibrium with liquid UO 2 can be described by the following exponential equation; obtained by curve fitting the data given by Braess et. al. from work done at Karlsruhe. p = 8 x 107 exp L This gives: 6.7 x 10- with p in atmospheres T in degrees Kelvin. Valid from the melting point of U0 2 to 5400'K; roughly accurate to about 6200 0KK. (4-1) 52 Pressures predicted by this equation over the temperature range of interest are tabulated below for convenient reference. Table 4-1 VAPOR PRESSURES OF UO 2 T (K) p (psia) 3070 (melting point) 0.45 4000 63 4500 485 4800 1130 5000 1760 5500 5850 6000 16,200 The internal pressure required for clad rupture is estimated to be 1200 psia, (21) (62) corresponding to a peak fuel temperature of 485 0 0K. This rupture pressure is based on the clad thickness given in Table 1-1 and the mean clad temperature expected at time t2 of Figure 2-1. As can be seen from Table 4-1, however, the energy required to produce clad failure is relatively insensitive to the estimated internal pressure required for rupture. For example, a 47% increase in internal vapor pressure (to 1760 psia) requires only a 3% increase in temperature (to 5000 0 K) and therefore only about a 3% increase in energy density. 53 The energy density required to produce clad failure is calculated in Appendix A, using 4850 0 K as estimated above for the peak fuel temperature at the threshold of clad failure. The value obtained is 1855 joules/grams, in excellent agreement with an experimentally observed value of 1900 joules/gram (59). Both of these results take 273 0K as the "zero energy" reference temperature. The total energy density calculated corresponds to an energy addition of 1080 joules/gram to the fuel at the core center, as noted earlier. As shown in Chapter III, the axial fuel temperature distribution prior to clad rupture can be accurately described by a cosine function of the form: T = T cos . (4-2) In the absence of axial fuel motion, the fuel vapor pressure within the clad can be obtained by employing Eqs. (4-1) and (4-2). As noted in Chapter II, axial fuel motion is expected to be of little influence prior to clad failure for a transient of the severity considered here. Thus, inserting Eq. (4-2) in (4-1) gives the following approximate expression for the axial fuel vapor pressure distribution prior to clad failure: p = 8 x lOexp P(Z) 107 6.7 cis x 1j. VTZ (4-3) - -- ft"A"W"A - -0 idm""W" 54 A key observation in the present work follows from this relationship. If an excursion occurs which is severe enough to produce clad rupture within the region I z I ! 30 cm; a pres- sure of 1200 psia and a centerline fuel temperature of 4850 0 K at Iz ;M 30 cm is implied. Equation (4-2) then predicts a peak fuel temperature of 6250 0 K at the axial core center and Eq. (4-3) predicts a potential peak pressure of approximately 26,000 psi. Of course, clad failure will occur at a much lower value and a pressure of 26,000 psia cannot actually be attained. The intent here, however, is to show the disruptive potential for fuel near the center of the core when clad failure occurs over a substantial portion of the reactor. Further appreciation of this effect can be gained by examination of Figure 5-3; showing a plot of UO2 vapor pressure versus temperature. Note the steepness of the curve at temperatures above 50000K. Early investigations suggest (20), and recent ex- perimental results indicate more conclusively (32), that the degree of fuel fragmentation or dispersal depends rather strongly on the excess energy above that required for clad failure. These experimental results and the above calculation indicate that a high degree of fuel dispersal is likely in the central regions of an LMFBR core when a severe excursion produces clad failure over a fairly large region. heretofore referred to as 6l, This region, may contain subregions varying from those with split clad and lightly fragmented fuel to 55 those where fuel, clad, and sodium are intimately mixed in a fine dispersion. In the following section behavior of the system subsequent to clad rupture will be seen to depend very strongly on this degree of dispersion. 4.2 Energy Exchange in the Region -& Analysis of the radial temperature profile in a fuel rod during a severe transient indicates that the percentage of fuel which is in the molten state at the threshold of clad failure varies from about 50% for a transient which starts with the core at full power to about 80% for one which starts from low ( ~ 10%) power levels (36) (59). Thus, the fuel which disperses into adjacent sodium coolant will have a mean temperature in excess of 3070 0 K at the instant of clad failure for all excursions considered. The temperature of the adja- cent sodium will be about 800 0 K at this time. It is shown in Chapter V that the maximum energy addition to the sodium in region'R during the time interval between an initial and secondary excursion is about 3600 joules per gram of sodium. For this upper limit in energy addition an examination of the properties of sodium in Appendix B of reference (35) shows that the heat transfer problem in region It following fuel fragmentation involves primarily transfer from solid fuel fragments to saturated liquid or two-phase sodium. A rough sketch of the phase diagram for sodium (using the data of reference (35)) is given below to clearly illustrate this important point. 56 PHASE DIAGRAM FOR SODIUM SKETCH V S. 'PAT N 3R cpaT= 365ot. -PArg 2 .Atoo (K) (pP S kI94 -- The energy required to heat the sodium to the saturated vapor (or critical) state along any of the three paths shown exceeds the 3600 j/gm expected to be available. found to hold for all possible paths. This condition was Thus, sodium in region -R will remain in the saturated liquid or two-phase state until a secondary excursion adds additional energy to the system. Heat transfer calculations in the remainder of the present chapter are predicated on this observation. From consideration 2.2.b of Chapter II, the sodium adjacent to intact fuel pins in regions outside R is not expected 57 to reach the boiling point within the several milliseconds under consideration, thus liquid sodium will exist in all regions of the core outside the vapor bubble generated by rapid heat transfer in 16L . shows that . Furthermore, work in Chapter V will not expand by more than about 10 cm in the axial direction before a secondary excursion occurs. To obtain an estimate of the heat transfer in region - under these conditions, it is assumed that the fuel frag- ments can be represented by spherical particles of some mean diameter. Additionally these particles are assumed to be at a uniform temperature at the instant of dispersion, at which time they are immersed in the surrounding sodium. The heat conduction equation for the fuel fragments with constant fuel material conductivity is qs(r,t) = pCp $ - KV9 (4-4) where 9 = T(0,t) T(r,t) and qs(r,t) = heat source within the fuel fragment. For the time interval of interest, namely between an initial and secondary excursion, the heat source (fission) is found to be negligible compared to the heat conducted to the surrounding sodium. For example, the power level at prompt 58 criticality in Figure 2-1 is found to be about 1500 joules/ gram sec of fuel; using Eq. (3-4). During the time interval between transients, the power level is approximately at this level, again as seen in Figure 2-1. By contrast, in Chapter V, it is shown that typically about 500 joules/gm fuel of energy will be transferred from the fuel to the sodium in a period of 5 milliseconds or less. The mean heat transfer rate during this interval, then, exceeds 100,000 joules/gram sec of fuel. Thus, qs of Eq. (4-4) can be set equal to zero with a negligible loss of accuracy in the present analysis. With this approximation the solution to Eq. (4-4), obtained by separation of variables, is: o 9(r,t) = sin7nr Am r exp K 2 pC ) n=1 _nt , (4-5) p where nwF 0 R = mean fuel particle radius, and K, p, and Cp are average properties of the fuel. For uniform fuel particle temperature as an initial condition, all harmonics are required for the complete solution. If it is assumed for the moment that the boundary layer or film heat transfer resistance at the oxide-sodium interface is negligible, the higher harmonics are seen to die rapidly; lw 59 giving rapid transfer of that quantity of energy associated with the higher harmonics. For aspherical fuel particle initially at uniform temperature, this is about 69% of the energy "available" to the surrounding sodium bath. energy available if theresidual words, In other the temperature dis- tribution assumes the fundamental mode profile with a central temperature of T of that available from the is only 3/v particle at uniform temperature T0 ; as can be verified from Eq. (4-5). The time behavior of the fundamental is given by = G0 e 9 (4-6) , where 9(t) =T (r=Ot) - T(r=R) '(4-7) and S 1 = pC K R P(-) 7r 2 2 (4-8) * For decay of the higher harmonics: . = -y- ; n = 1, 2, 3, ... . (4-9) For the oxide fuel, the following average properties from Table 4-2 are used to approximate behavior in the 2000K to 3000 0K temperature range: p = 10 gm/cm 3 K = 0.02 j/gm K sec Cp = 0.4Zj,/gm K 60 Table 4-2 FUEL, SODIUM, AND CLAD HIGH TEMPERATURE PROPERTIES A. FUEL: (33)(34) 1) Melting point (Tmelt) at atm pressure 3070 0 K 2) Heat of fusion (Ahfusion) 278 j/gm 3) Heat of vaporization (Ahvap) 1850 j/gm 4) Vaporization temperature (Tvap) at atm 6200 0 K pressure 5) 6) "Mean" specific heat of solid (cPs) "Mean" specific heat of liquid (cP) 7) "Mean" specific heat of vapor, by 13/2 (cv ) B. given 0K 0.33 j/gm 0K 0.42 j/gm 0.20 j/gmOK 8) Density of solid (ps) 0. 10 j/cm3 9) Conductivity (k) (at 3000 0 K) 0.02 j/cm 0 K sec SODIUM: (35) (properties at 1300 0 K) 1) Heat of vaporization (h ap) 1600 BTU/LB 2) Density of liquid (PL) 43 lb/ft 3 ( 0.7 gm/cm ) 3) Density of vapor (pV) 0.05 lb/ft 3 4) Viscosity of liquid (1L) 0.34 lbm/fthr 5) (4070 j/gm)* ) Specific Heat of liquid (c 0.32 BTU/LBMOF (1.47 j/gmOK) 6) Conductivity of liquid (KL) 26 BTU/hrftOF 7) Viscosity of vapor ( V) 0.04 BTU/hrft 8) Specific heat of vapor (c 9) Conductivity of vapor (KV) ) 0.30 BTU/LBMOF (1.4 j/gm0 K) 0.04 BTU/hrft0 F *Values are given in Metric and British units when both are required for a calculation given in the text. 61 For a mean dispersed fuel particle size of 500 microns (about one-tenth the intact fuel pellet diameter), which represents fairly fine fuel dispersal, and which may be a real- istic particle size in the central regions of : t'i = 15 milliseconds ' 2 = 3.75 msec r3 = 1.67 msec ... etc. For comparison, recall that the time interval between power peaks is on the order of 10 msec (Section 2.3 of Chapter II). For a mean particle size of 2000 microns (0.2 cm or about one-third of the intact fuel rod diameter) representing only minor fragmentation of the fuel pellets; 1 200 msec Z2 2 50 msec V3 2 22 msec ... etc. Clearly, if the assumption of low film or boundary layer resistance is reasonable, heat transfer of a large fraction of the energy available to the surrounding sodium will be extremely rapid; even for the case of relatively minor fragmentation. It is interesting to note that if R0 is chosen equal to the intact fuel rod radius, Eq. (4-8) gives: 62 1 = 1800 milliseconds, which is very nearly equal to the e-folding time for transient heat transfer from the intact cylindrical fuel rods; as should be expected. This substantiates statement 2.2.b of Chapter II (negligible heat transfer from intact fuel during the time of interest) but of more importance shows the effect of higher harmonics in producing an enormous increase in the initial rate of heat transfer as a result of only minor fuel fragmentation. The physical explanation of this "mathematical" phenomena lies simply in the fact that any degree of fuel fragmentation exposes extremely high temperature internal surfaces to cold sodium. The limitations imposed by boundary layer resistance and film boiling are now considered. Prior to Departure from Nucleate Boiling (DNB), the customary Nusselt number (Nu hD -- ) is used to represent the boundary layer heat transfer coefficient, whereas after DNB a film boiling heat transfer coefficient (hfb) is employed. It is not a foregone conclu- sion that DNB and film boiling will occur around all fuel fragments in the usual manner for a number of reasons: (1) Heat transfer in "R is an exceptionally rapid transient process; the usual rules for estimating the start of film boiling may not apply. (2) If fuel dispersal is fine, there may not be "room" for film growth around each fuel fragment to the film thicknesses usually associated with film 63 boiling until substantial voiding takes place. It is shown below, in fact, that about 10% of the sodium in region 1, must be removed or "voided" before film boiling can predominate due to this film thickness limitation. of highest heat transfer in 36 present in other regions of (3) The region will determine the pressure 3-. Thus, fine dispersal and high heat transfer rates in one relatively small region of may lead to sodium vapor pressures sufficient to significantly delay DNB in other regions of 4V. If q0 represents the energy loss per unit volume for a given fuel particle; the initial rate of heat transfer associated with the fundamental temperature mode for a pure conduction process is: fI where 9 = 3Kf 1K~) Go (4-10) , is defined in Eq. (4-7). The rate of energy trans- fer through a boundary layer is given by: q/A With q''' q'' = hBL = = 3 q'' R- (4-11) for a sphere and using the Nusselt number 0 for heat transfer prior to DNB, Eq. (4-11) can be written q11' = 3/2KNANu()BL . (4-12) 0 For a spherical fuel particle immersed in a pool of sodium, after all harmonics except the fundamental have 64 decayed, the heat transfer process is limited primarily by the conductivity of the oxide; as is the case for intact geometry in an LMFBR. In this case Q of Eq. (4-10) approximates the total temperature drop from the fuel particle center to the surrounding sodium bath. On the other hand, at the instant of fuel dispersion, before the harmonics decay, the entire temperature drop occurs across the boundary layer, or: 9 = 0 (t (t=0) oo) Thus, as a measure of the initial rate of heat transfer with respect to the rate associated with the fundamental temperature distribution (or "final" rate for an infinite pool of sodium) a ratio "m" Eqs. (4-10) M = may be defined as follows, making use of and (4-12): q0BL Nu KNA ,,C = - KF.E q 0COND (4-13) FUEL Note that the ratio m is independent of fuel particle size when written in terms of Nu. When m > >1 the process is not boundary layer limited and the assumption of rapid transfer of the energy associated with higher harmonics is correct. Typically, for solid oxide fuel and liquid sodium at the appropriate temperatures: KFUEL 0.02 g/cm0 K sec KNA 0 .50 j/cmoK sec. 65 From reference (41) p. 202, a lower limit on Nu for heat transfer from solid spheres to a stagnant liquid is 2; independent of the size of the spheres. For the process under discussion, one expects Nu to be greater than 2 until film boiling starts. Thus, at the instant of fuel dispersal, Eq. (4-13) gives: m > 25. Since m >>1, the values of thermal relaxation time, t , calculated above appear to provide a realistic picture of the energy exchange rates in regions of R where DNB has not occurred. The start of film boiling requires that a vapor blanket of sodium be formed around the fuel fragments and may, therefore, require that considerable sodium voiding take place before film boiling is an important part of the heat transfer process at the fuel-sodium interface for the majority of region R . It appears reasonable to assume, however, that the lowest rate of energy exchange between fuel and sodium in _& , for a given fuel to sodium temperature difference, will correspond to the occurrence of complete film boiling in the usual sense. The degree of voiding required to permit vapor film growth to a "limiting" thickness, 5, is considered first; then the limitations imposed on energy exchange rates after complete film boiling occurs is investigated. 66 Rewriting Eq. (4-13) in a form more appropriate for film boiling; 2KhfbD FUEL (4-14) where, hfb = film boiling heat transfer coefficient D = mean fuel particle diameter. For a vapor film around each particle of thickness 5, if 5 << D, the following relationship applies: hfb KNA VAPOR . (4-15) This simply expresses the film heat transfer coefficient in terms of the conductivity of the sodium vapor in the film and comes from the usual relations: q/A = K dT = K AT = hAT Using Eq. (4-15) in (4-14), the relation M = KNA VAPOR( ) F KFUEL is obtained. (4-16) Vapor conductivities for sodium are tabulated extensively in reference (35). For KNA VAPOR = 1.0 x 10- j/gm 0 K cm at a sodium temperature of 15000K (35), Eq. (4-16) R gives m = 0.05 -F . Treating m > 1 as an indication of the point at which the film temperature drop predominates and at - - I ! iiii ii!: "' - , - I ..- ,-- W - - 67 which the process is therefore limited by film boiling: 5 R 0.05 is required. (4-17) For the spherical fuel particles assumed, this result requires that about 7% of the region 'R consist of sodium vapor - or that 7% of the region be "voided" of sodium before the average film thickness around all fuel particles is such that the film boiling process is clearly limiting. Note that this vapor fraction is independent of the assumed size of the spherical fuel fragments. For very small fuel particles, the film thickness, however, becomes microscopic. For 200 micron fuel particles, for example, the film thickness is only 5 microns. It may be erroneous to equate conductivity of a sodium vapor film of this thickness with that of a macroscopic film. One might expect, however, that molecular exchange across a microscopic film would increase the energy exchange rate as compared with a thick film. If this is the case, Eq. (4-17) predicts a conservative result for an accident which results in fine fuel dispersion in the sense that a larger degree of sodium voiding will take place before film boiling can predominate. In any case, it i's difficult to conceive of a means by which a microscopic vapor film can result in a lower heat transfer rate than would otherwise be expected. In order to examine the limitations imposed on heat transfer rates by complete film boiling, two cases are considered. 68 First, natural convection of sodium around fixed fuel spheres, as represented by Eq. (5.64) of reference (42), is treated. Second, a reasonable sodium velocity across an equivalent cylinder of fuel material, as described by Eq. (9.30) of reference (41), is assumed. The relative velocity between the sodium and fuel particles is arbitrarily taken to be the same as that corresponding to normal core sodium flow (lOn/sec). This should give a reasonable estimate of hfb and thus of the heat transfer rate for a sodium/fuel mixture undergoing film boiling. In both cases, the properties of sodium are taken from reference (35). For both correlations, it is found that m, as defined by Eq. (4-14), is roughly proportional to particle size and that: m>l for D > 250 microns. The detailed calculations are given in Appendix B. These two latter approaches give a crude approximation for hfb at best and represent only an attempt to obtain a reasonable estimate of the heat transfer coefficient for a process for which a suitable correlation is not available (60). The results indicate that the energy exchange rate in 1 remains much higher than the rate outside Rk even after each fuel fragment is completely blanketed by sodium vapor. 69 4.3 Summary The strong dependence of fuel vapor pressure on position within the core during a CATEGORY 2 excursion has been shown. The resulting possibility of fine fuel fragmentation in central regions of the core, supported by some experimental evidence, has been indicated. Extremely rapid heat transfer from fragmented or dispersed fuel to the surrounding sodium has been demonstrated for regions where DNB has not occurred. This energy transfer rate has been shown to depend strongly on fuel fragment size, 12 namely a (R-) dependence. Furthermore, it was shown that a significant degree of sodium voiding must take place in the region of clad failure before a vapor film of the usual thickness around the fuel fragments can be produced. Even so, after complete film boiling in the usual sense occurs, the heat transfer rate between fuel fragments and sodium was seen to remain high in region VC as compared to the case for intact LMFBR geometry. The energy transfer characteristics presented in this chapter are employed in Chapters V and VI to predict the effect of subsequent events; particularly for estimating rates of sodium voiding induced by rapid additions of energy to the sodium in region - . __ I _WAWAM__ =_ -- I ,- 70 Chapter V REACTIVITY ADDITIONS FROM SODIUM VOIDING It is the purpose of this chapter to consider the magni- tude of possible sodium void reactivity insertion rates during a specific accident sequence. An excursion leading to clad rupture over a significant portion of the core is assumed to have occurred and to have been turned down by Doppler feedback (CATEGORY II excursion) as shown in Figures 2-1 and 2-3. If the excursion is initiated by some effect other than sodium voiding, the fuel in the region of failure mixes with liquid sodium and the heat transfer behavior considered in the previous chapter applies. If the accident is initiated by sodium voiding, fuel dispersal into a previously voided region may occur. The former condition is of primary inter- est in the present chapter; the latter is treated briefly, however. The description of energy exchange between hot fuel and relatively colder sodium developed in Chapter IV is used in conjunction with the hydrodynamic equations to calculate the rate of sodium vapor generation (or voiding) and, ultimately, the reactivity addition rates which result. 5.1 Hydrodynamics of Sodium Voiding The core described in Table 1-1 is assumed to have a large volume of sodium at the core exit, with a cover gas at a pressure of about 2 atmospheres, as depitced in Figure 5-1. This arrangement is typical of currently proposed LMFBR 71 FICuRE. 5'-l Sc HE\ATiC OF CooE r/t oti>Nr Coyca &AS -09 Sobirt d lf;ktAr cA~ iin ~~' F2La SoAFACL -4o. OUT Lcf NA NA caitis /7 L~t~P~ESS"'t4 UPW A&AFLoW p . . CORE _ L T DOWNWA RD plow ~RG.P ZE SEWetATr\V9 C ANNEL FOk F. SIN&,LF, $obiUt FI.oW, LIQUID TH- ICiANNEL! is THe pLow AREA ASSOCIA-rED Wi-R CotrE So~,,ur4 V=FEL 'Roo. EXPNOIPtMG AND vAPoa L . 5orN\ 0U.b I -F 72 designs (2A)(4A). The core model employed in considering the hydrodynamic effects associated with sodium voiding is shown schematically in Figure 5-1. in Table 1-1. Appropriate dimensions are given In the present analysis, the following assump- tions or simplifications are made. 1) Only upward flow of sodium is considered. Very little radial motion of sodium is expected compared to axial motion because of the close packed nature of LMFBR fuel rods and the employment of fuel rod wire wrapping and subassembly boxes in current designs. As shown in Figure 5-1, however, downward flow paths clearly exist. Further comment on this aspect is made in the next paragraph. 2) A constant pressure is assumed to exist at the core exit. This pressure consists of the cover gas pressure and the static liquid head above the core. The justification for this assumption rests on the following observations. a) The volume of the cover gas region is much greater than the size of the vapor bubbles considered; thus only slight compression of the cover gas is required to accommodate a large sodium void. b) The velocity of sound in liquid sodium is high, typically about 7000 ft/sec (35), so 73 that pressure waves generated at the core exit can affect (compress) the cover gas in time intervals on the order of 1 msec. c) The volume of sodium above the core is typically large compared to the size of the vapor bubble generated and the compressibility of liquid sodium at core exit temperatures is high. Thus, from the velocity of sound cited in (b) above, liquid sodium compression tends to relieve the pressure at the core exit in time intervals less than that required for pressure pulses to reach the cover gas region. Calculations in Appendix B, through (c), employing observations (a) further indicate that this assumption is reasonable. Downward flow is not included in the hydrodynamic analysis primarily because of the complexities duced. intro- The smaller volume of sodium at the core entrance and the more restricted access to the cover gas region can lead to high compressive pressures in the core inlet region (see Figure 5-1). The in- clusion of check valves at the core inlet or main coolant pump outlets in various designs further complicate the matter. For example, the question -~now 74 of the position of the valves at the start of the accident as well as their activation time arises. As seen in Figure 5-1, however, sodium forced from the bottom of the core by an expanding void can pass upward through the outer regions of the core and blanket. Thus, while the neglect of downward flow, in conjunction with the present assumption of a constant pressure at the top of the core, greatly simplifies the analysis; an underestimate in the severity of the sodium voiding effect calculated is expected. The principal conclusion of the present analysis will not be affected, however. 3) "Slug" or column flow in subcooled regions outside the vapor bubble generated by rapid heat transfer in "R is assumed. This implies that subcooled liq- uid sodium moves axially as a uniform slug of liqquid without entrained vapor bubbles. This assump- tion is in agreement with that of other authors (9)(12)(39) and some experimental work (40). 4) Choked or sonic flow in the core is not considered. The maximum flow velocity attained in the present analysis is 103 m/sec. Since sonic velocity in liquid sodium is about 2100 m/sec at the temperatures of interest, this assumption is clearly valid in the liquid regions. In the two phase - 09 000111=1 - I t 75 region the matter is not so clear, however, as insufficient information is available to predict the velocity at which choking will occur for the high pressures and low qualities of interest here (61). A development of the hydrodynamic equations employed is given in Appendix B. An outline of the development is given here for continuity and to show salient features of the analysis. The core friction drop is included by employing the usual relationships 4A APfriction ~ e p V2 c (5-1) and f = .o46(p-L) PDe V , (5-2) where I = core half height De = equivalent diameter for the flow area associated with one fuel rod. A representative core pressure drop at normal full flow is taken to be 80 psi. This value was determined,as explained in Appendix B, by averaging the pressure drops given by General Electric and Atomics International for proposed 1000 MWe LMFBR's; given in their respective follow-on-studies, 76 references (2A) and (5A). Then, since the properties of sodium in the subcooled region outside the vapor bubble (see Figure 5-1) change only slightly, the friction pressure drop can be written, employing Eqs. (5-1) and (5-2), as: 1.8 Apf = 'V'( - Z)(-) , (5-3) where Ap0 = reference half-core pressure drop at full flow; namely 40 psi VO = reference velocity for full sodium flow, taken as 10 m/sec (2A)(5A). In the present analysis, for the range of z(t) (see Figure 5-1) considered, the fraction ( 'Q-Iz(t)) varies only from 0.57 to .71, and an average value of 0.64 was therefore employed; that is: (- Z) = 0.64 in Eq. (5-3). The influ- ence of this approximation is seen to be negligible when the V 1.8 factor (V-) is observed to range from 1.0 to 51.5 (see Table 5-1). This approach properly includes end or exit effects since Ap 0 is the total (half) core pressure drop and since the functional dependence of the exit pressure drop is approximately the same as employed in Eq. V 1.). Thus, (5-3) (41) (Apexit~ V2 vice frictional effects are incorporated in the analy- sis by a method which is simple but which is in keeping with the accuracy of the theory on which the parent relations, Eqs. (5-1) and (5-2), are based. 77 Inertial forces are included through the usual conservation of momentum relationship: 2 Apinertial = p(X- Combining Eqs. (5-3) z(t)) dt and (5-4) (5-4) 4 and considering the pressures depicted in Figure 5-1, 2 1.8 p(t) = p(j - z(t)d)z + Apo(.64)(l- Xdz 1 (5-5) + p - dt If Q(t) is the energy added to the sodium in at any time t, Q(t) = c (Tf -To) + (h (t)-hf) , (5-6) where T = saturation temperature at time t, h(t) = enthalpy of two phase mixture at time t Te = initial h = enthalpy of the saturated liquid at Tf(t) cP p sodium temperature = mean specific heat of sodium It was shown in Chapter IV that sodium is expected to remain in the two phase state, thus the applicability of Eq. (5-6) is assured. The Clapeyron equation relating changes in enthalpy and specific volume for a reversible process in a two phase mixture is (38) 78 Ah = T AV (5-7) p dT The saturation pressure as a function of temperature for sodium is given by (35)(37) p = Be-A/T (5-8) , where B = .332 nts/m A = 11,950 T in 'K. A key to the present analysis is the employment of this empirical relation, Eq. (5-8), in the Clapeyron equation to eliminate the enthalpy term of Eq. (5-6). The appropriate manipulations give the following expression for Q(t) = (T -TO) + ET A A ) Q(t): vo( (5-9) Note that z is a measure of the extent of the axial void. zo Now, using Eq. (5-8), Bexp(TA ) (t) Eq. (5-5) 2 --p( t -z )d2 dt can be written: 1.8 + Ap (. )) 1.8 + p( . 0 (5-10) These latter two equations give Q(t) and f(z,T) and z = f(T) respectively. Thus, it is seen that if Q(t) is specified, the rate of void growth as a function of time can be determined. 79 The method employed to obtain a solution is described in Appendix B. The specification of Q(t) is the subject of the next section of the present chapter. Although the hydrodynamic analysis developed in this section is by no means exact, the inaccuracies generated are small compared to the uncertainties in Q(t) discussed in Chapter IV. The importance of Q(t) in the analysis is further indicated by the results of the next section. 5.2 Estimates of Energy Addition Rates to Sodium The following average properties of fuel material and sodium are used in the present analysis (35). 0 1.4 joules/gram K c Na Liquid = c Fuel = .4 j/gm K = .8 gm/cm3 PNa Liquid PFuel 3 = 10 gm/cm The heat of fusion for the fuel is assigned an equivalent temperature change given by: AT eq == fusion -. cp - 278 4j/no gK= 7000K. (5-11) Using this equivalent temperature and the initial fuel temperature distribution for full power operation (see Table 1-1), 61 the 80 mean equivalent fuel temperature at IzI 20 cm is 44000 K (Tactual = 3700 0 K) when the peak temperature is sufficient to produce clad failure within the region Jzj 4 20 cm. temperature in the central regions of R , The mean then, will be somewhat higher and an underestimate in the energy added to the sodium results if the mean equivalent temperature throughout the regionR is taken to be 44000K at the instant of clad rupture. Furthermore, if the initial excursion starts from a power level below full power and clad rupture over a region 1R similar in size occurs, the mean temperature of the fuel in 'A,will be much higher. This results from the lower initial fuel rod peak to mean (radial) temperature ratio at lower power levels. With a mean equivalent fuel temperature of 4400 0 K and an initial sodium temperature of 800 0 K, as given in Table 1-1, the energy transferred to sodium for three assumed cases, a, b and c, is calculated in Appendix B. For case (a) it is assumed that all energy associated with the higher harmonics of the fuel temperature distribution is transferred in a few milliseconds. As shown in Chapter IV, fuel fragmentation in- to particles of a mean diameter of 500 microns (about onetenth of the intact fuel pellet diameter) gave an e-folding time for the first higher harmonic of 3.75 milliseconds and of 1.67 msec or less for the next and higher harmonics. These times are relatively short compared to the time required to produce significant sodium voiding. The e-folding time of 81 the fundamental mode, on the other hand, was 15 msec; a period somewhat longer than the time required for significant sodium voiding under certain circumstances; as will be seen. Thus for mean particle sizes of 500 microns or less, it is reasonable to assume transfer of the energy associated with the higher harmonics in a few milliseconds. The temperature distribution effects associated with case (a) are depicted in Figure 5-1. The energy transferred by collapse of the higher harmonics is used as the value of Q(t) in case (a). As stated in Chapter IV, this is 69.6% of the total energy available from the fuel if the mixture is allowed to come to equilibrium at constant volume. For cases (b) and (c), respectively, values of Q(t) of 50% and 30% of the total energy available are employed to account for inaccuracies in the analysis and to show the trend in the behavior of the system for rapid transfer of smaller fractions of the available energy. 5.3 Reactivity Addition Rates from Sodium Voiding Using the values of Q(t) cited above, Eqs. (5-9) and (5-10) have been solved (see Appendix B) to obtain the rate of axial void growth as a function of time. given in Table 5-1. The results are The initial axial extent of region 1. was assumed to be the region i 4 20 cm but the resulting void growth rates were found to be relatively independent of the axial extent of &Rwithinthe Izi = 10 cm to jz[= 30 cm range. 82 f% C uR~ SC146-MA-ric ov: TECvPAavOzE A VuEi. PPA&r4r FO IN sooluwi FOCL PRA&FAENr 103 CA I C Ro t4lb I I oAk~4(kk#L UtWOPNM FXAGMNT ryUeAN Orem? £N-k&Xm A ssaUATED rrSii4 Akt4~c HAP-Rmics INFOEL 5:AAf1EM1r AT FULEL MET4 IUCL-rerrit. ApiTi IA-Y I T itMCum AFGR~ Dr'cA' OF HPAQMfONtet NtECAN I I I I 9 09iGI4NAL soblurf\ TEMVP. 83 The reactivity addition rates for the given rates of axial void growth were calculated for regions 'I which extend over 25% (note (2) of Table 5-1) and 50% (note (3) of Table 5-1) of the radial area of the core. The reactivity addition rate and net reactivity calculations were based on the results shown in Figure 6-2; the curve showing sodium voiding reactivity calculated by ANISN was employed. The reactivity addition rates obtained were found to be relatively independent of the axial extent of 6R in the Izi cm range. Beyond values of \ zi = 10 to izI = 20 = 20 cm for the initial size of '11,, the rates shown in Table 5-1 drop off sharply (see Figures 6-2 or 6-5). Observe that the reactivity addition rates estimated range from 62 $/see to 642 $/sec. For the case of transient heat transfer from intact fuel geometry, maximum sodium voiding reactivity rates of from 20 $/sec to 65 $/sec have been predicted (12)(23)(24)(25). In view of the heat transfer rates estimated and the comparisons with the case of intact geometry given in Chapter IV, the rates shown in Table 5-1 do not seem too surprising. Note further that for the larger t 3 (time to achieve void growth rate tabulated) for each case in Table 5-1, the friction pressure drop has become a substantial fraction of the peak pressure generated in -R . Since the friction pressure drop is proportional to V1.8 and the reactivity Table 5-1 REACTIVITY ADDITION RATES FROM SODIUM VOIDING - Case - Peak Pressure C b 1520ps a 4850 Ps i 360 s 1 msec 2 msec 2 msec 3 msec 4 msec 7 msec dz a at t3 51.5 m/sec 103 m/sec 33 m/sec 49 m/sec 15.2 m/sec 26.5 m/sec Az at t3 2.5 cm 10 cm 3.3 cm 7.5 cm 3.05 cm 9.3 380 psi 8oo psi 170 psi 355 psi 43 psi 115 psi 219 $/see 418 $/sec 136 $/see 197 $/see 62 $/sec 108 $/sec 328 $/sec 642 $/sec 212 $/sec 309 %/sec 96 $/sec 168 $/sec + 110' + 430' + 140' + 170' + 68U t3 (1) Apfriction at t3 dk (2) dk (3) dt (2) 5k t (3) 5 knet + 131' I+ 210' cm + 4o + 649( NOTES: 1) Time at which void growth rate and reactivity rate shown are reached. Figure 2-1. 2) 6k 3) R extends over 25% of radial core area extends over 50% of radial core area See also 85 addition rate is roughly proportional to V; the friction effect quickly becomes important for time intervals larger than those tabulated. Furthermore, as the bubble grows beyond the 7.5 cm to 10 cm Az value corresponding to the larger time intervals shown, differential reactivity of sodium voiding rapidly diminishes (see Figure 6-5). Thus, the larger reactivity rates shown for each case are approximately the maximum values expected within the limits imposed by the assumptions made. An examination of the influence of the more important assumptions follows. For all cases, the size of -A,grows axially from an initial dimension of 40 cm to between 42.5 and 50 cm; as seen in Table 5-1. Thus, sodium vapor fills between about 6% and 20% of the region -R over the time span considered. From the calculations of Chapter IV, then, the lower values of the reactivity addition rates estimated can be attained before the film thicknesses associated with complete film boiling (and therefore slower energy exchange) can be present. Perhaps of more impact is the observation that for pressures in region R below the critical pressure of sodium ( 5980 psi), the time required to accelerate sodium outward sufficiently to permit a vapor fraction of about 10% (by volume) is of the same order as the time required for transfer of a large fraction of the energy available if the mean dispersed fuel particle size is 500 microns or less. Thus, the sodium is confined by inertial and frictional effects long enough for extensive energy exchange between fuel and sodium. 86 As stated, only upward axial bubble growth is allowed in the calculations. As suggested in Figure 5-1, downward motion might be significant. low pressure (- (If the assumption of a constant 0) at the bottom of the core is employed in the present analysis; the rates of void growth and reactivity addition shown in Table 5-1 almost double in some cases an increase by over 150% in all cases.) In view of the small extent of axial bubble growth considered in Table 5-1 (Azmax = 10 cm), calculations in Appendix B indicate the assumption of a constant low pressure at the top of the core is realistic. Finally, for primary excursions which start from a low power level and result in clad failure over a region of the size considered here, the mean fuel temperature in the region of clad failure will be appreciably higher, as will be seen in the next section. This again will result in higher sodium voiding rates than estimated for the present case with full power as the power level before the initial excursion. Thus, the net effect of these latter three assumptions is likely to result in an underestimate of the sodium voiding reactivity addition rates which can follow CATEGORY II excursions. 5.4 Accidents Initiated by Sodium Voiding If the initial excursion is initiated by sodium voiding, it is possible that the region of clad rupture will be largely enveloped by sodium vapor. For the particular LMFBR of 87 interest, the specific volume available to the fuel in a region from which sodium is removed is 65.5 cm 3/gm mole (see Table 1-1). Vr This corresponds to a reduced volume of = c 655 = 0.73, (5-12) where V = critical volume for UO 2 = 90 cm 3/gm mole (44). Menzies tabulates the vapor pressure of UO 2 as a function of temperature for reduced volumes between 0.4 and 1.0 (44). Applicable data from this work is plotted in Figure 5-3, along with Eq. (4-1); the latter representing fuel vapor pressure in the intact clad (Vr := .33). The data given by Menzies for Vr in the range 0.7 to 1.0 essentially coincide for UO 2 temperatures below 7000 0K. high power level ( For initial excursions starting from a -- 100%), as assumed in the preceeding sec- tions of the present chapter, the mean fuel temperature in region R (Izt 1E 20 cm) is estimated to be 4000 0 K. As seen from Figure 5-3, this results in a fuel vapor pressure of about 100 psi; a value which is insufficient to substantially affect the outcome of the excursion. If, however, the tran- sient starts with the plant operating at a low power level (-~l0%), the mean fuel temperature in the region R will be about 5200 0 K. The difference, of course, results from the lower peak to average radial temperatures in a fuel rod at low power. For this case, examination of Figure 5-3 shows 88 Fr&oke S'-3 FOtR UOL A PPAD. C i-AD FAtLU(F. TE-MP. V APOp gR65 !SJR-E F (zo\ DATip S PLOT OF -Q(i4- 1) ctiove. 3 cu IM e z 112 'PAGE1SSUE P~ioR TrO e-Ab FAftuRE ANr SO AFr'ER CLIb (EALLLI(i - IV CoRE REMn96$ 10,000 FILLEO LiauiL wt-TH SobiUM) I I / / 5.000 /;0 MEN 'y uo.'' TO . zd ' E.sV ACoMac FOR 1. o O- 'T Y- '. Y<oETPLO-OR, -T~<?Doao 1000 5 5'If0i VI 1- (*K) -+ 1.0 89 that the fuel pressure will be about 1800 psi after clad failure in region - . As seen from Table 5-1, this peak pressure is comparable to that in case (b) for energy exchange with sodium. According to data given by Menzies (44), only a slight pressure drop will occur as the fuel expands to fill a volume corresponding to Vr = 1.0; or expansion by a factor of about 1.4. Thus for excursions which start from relatively low power levels, the problem examined in the present chapter is not eliminated even if sodium voiding is the cause of the initial accident. Furthermore, for excursions which start from low power levels or which result in initial clad rupture over a region larger than Iz ' 20 cm, the fuel vapor pressure itself can play an impor- tant role in initiating or increasing the rate of sodium voiding whether sodium is present in the region of clad failure or not. Note, however, that fuel vapor pressures were not considered in arriving at the reactivity addition rates cited in Table 5-1. 5.5 Summary To preface the following discussion it is reiterated that the rate of reactivity addition at approximately the time of prompt criticality is a major, if not the primary factor, in determining the consequences of a severe excursion; as discussed in Chapter II. Recall also that a reactivity 90 insertion rate of 66 $/sec is taken as a basis of comparison in the present analysis. An accident which results in clad rupture over a substantial portion of the core, namely a CATEGORY II excursion, is considered credible since such an accident can result from initial reactivity rates below the 66 $/sec basis; as shown in Chapter II. Given that such an accident can occur, the assumptions leading to the reactivity addition rates shown in Table 5-1 are realistic with one important qualification: the rate of heat transfer to sodium, which depends on the degree of fuel dispersal, is an area of great uncertainty. Insufficient information is available at present to properly estimate this effect. Fuel vapor pressure effects must be taken into account for a complete analysis of CATEGORY II excursions; particularly if the excursion is initiated by sodium voiding itself. In the present chapter the reactivity effects of fuel motion were not considered. As will be seen in Chapter VI, however, sodium voiding is the predominant effect until fuel motion becomes quite substantial. The small amount of outward motion in a close packed LMFBR lattice may be of little influence until pressures sufficient for destruction of the lattice (outside -R ) are generated and, in fact, a mechanism which produces inward fuel motion (into 6R. ), discussed in Chapter VI, may override any outward fuel motion occurring early in a CATEGORY II excursion. L The principal conclusion which must be drawn in the present chapter, then, is the following: If a large portion of the fuel in region _R fragments to particle sizes of mean diameter equal to one-tenth the intact pellet diameter or less, the reactivity addition rates resulting from the subsequent sodium voiding can exceed the 66 $/sec basis by as much as an order of magnitude. Some additional appreciation for the estimates obtained in this chapter can be gained by noting that the energy transferred to the sodium in cases a, b, and c of Table 5-1 is sufficient to produce sodium superheats of 1700 0 K, 1150 0K, and 610 0 K respectively. These values are based on a sodium boiling point temperature of 1000 0 K, typical of normal operating plant pressures. 92 Chapter VI REACTIVITY EFFECTS RESULTING FROM CORE REARRANGEMENTS The reactivity effects resulting from core rearrangements which can result from a severe accident are explored in the present chapter. Reactivity insertion rates which can arise during such rearrangements are estimated. A model is developed which successfully predicts the reactivity effects introduced by sodium voiding and fuel motion. The basis of the model is perturbation theory. In Chapter II it was shown that the spatial neutron flux shape remains very close to the unperturbed shape even for quite large perturbations. Thus, it is assumed that a function describing the unperturbed flux shape can be used to account for spatial effects. Use is made of the spectral characteri- zation work done by Shaeffer and Driscoll (45) to include the all important energy or spectral effects for a fast reactor. It is shown that removal of a quantity of core fuel material from any given location produces a stronger negative reactivity than the positive reactivity which would result from inserting the same quantity of material at the given position. The influence of this effect in ameliorating the hazard of possible fuel rearrangements during an accident is demonstrated. 6.1 Effects Leading to Fuel Motion The rupture of cladding over some region of the core has been discussed in some detail in Chapters II and IV. 93 Immediately following time "t2 " of Figure 2-1, the fuel in regions outside $, is constrained by intact clad except for the "open end" inside -R 2-3. , as shown schematically in Figure Movement in the axial direction away from blis preven- ted by colder intact fuel and blanket material. As noted in Chapter IV, 50 to 80% of the fuel at the edge of region 'Q is in the molten state (36). Since the fuel vapor pressure was sufficient to rupture the clad in 1R1, the vapor pressure within the clad will be considerable some several inches outside P% (in a non-vented design, the fission gas pressure could also be considerable). Thus, the potential exists for axial fuel movement or "injection" into bl, from regions adjacent to R . That such axial motion is plausible is indica- ted by recent experimental evidence (36)(46) and by calcula- tions in Section 6 of the present chapter. A second method by which fuel material can be moved inward or into potentially higher worth regions of the core is now outlined. In Chapter II it was shown that appreciable cooling of the fuel material in region IR could take place during the several milliseconds following an initial transient. Thus the mean fuel temperature in -R could be reduced below that in adjacent regions by several hundred degrees. Furthermore, fuel fragmentation reduces the ratio of peak fuel temperature to mean fuel temperature at any given location. From the developments in Chapters IV and V, it is - m 94 clear that the peak temperature in regions adjacent to could be over 500 K higher than the peak temperature in by the time a secondary prompt-critical excursion is initiated. Examination of Figure 5-3 shows that as temperatures in regions adjacent toA reach about 6500 0 K during a secondary transient, 500 0 K of subcooling in R leads to a pressure differential in excess of 5000 psi tending to induce inward fuel motion. Only one of the two phenomena described above is likely to have a significant effect on the overall excursion. If substantial cooling of the fuel in-R occurs, high sodium vapor pressures are generated in 'R , preventing the fuel injection mechanism induced by clad constraint. This is shown conclusively in Section 6.6 of this chapter. If this fuel cooling does not occur; then the conditions for an inward pressure gradient during a secondary excursion are not established and the second mechanism of general inward fuel motion cannot be appreciable. The second fuel motion phenomenon can take place subsequent to significant sodium voiding, however. As seen in Chapter V, the sodium vapor pressure is not expected to reach the critical pressure (5980 psi) during the time interval of interest whereas, from Figure 5-3, the fuel vapor pressure generated during a secondary excursion can quickly exceed the sodium critical point pressure (and therefore the maximum 95 pressure expected in region _R during the early phase of a secondary excursion). Thus, the initiation of a secondary excursion by sodium voiding can, in affect, lead to further reactivity additions by this fuel movement mechanism. Although these two phenomena leading to fuel movement into higher worth regions of the core may be of short duration and may result in fuel movement of only a few centimeters, the important question is whether such rearrangements can add on the order of 50 in reactivity at a rapid rate. The remainder of the chapter is devoted to this question. 6.2 Reactivity Model for Fast Reactor Core Perturbations Figures 3-1 and 3-2 show that the spatial shape of the neutron flux remains very close to the fundamental mode shape over most of the core, even for quite large perturbations. Conversely, Figure 3-3 shows that the shape of the energy spectrum shifts appreciably for comparable perturbations. While the former result suggests that spatial effects can be accounted for by the usual one-group perturbation theory formulism, the latter result indicates that the one or two group approach will be entirely inadequate for assessing such reactivity effects in fast reactors. Subsequent developments in the present chapter show that this indeed is the case. In recent work at MIT (45) directed toward the development of simple LMFBR core calculational methods, Shaeffer 96 and Driscoll have developed an accurate spectral characterization technique. The spectral characterization parameters employed in this technique have been found particularly useful in calculating the influence of spectral shifts on reactivity. A brief description of this work is given to set the stage for the present analysis. In reference (45) a one group method for calculation of neutron balances in the core of fast breeder reactors is developed and evaluated. The key feature of the method is the definition of two spectrum characterization parameters, S = (6-1) f + tr and R 1S 1tr r (6-2) where 7r = removal cross section (45). The former index enables correlation of all required microscopic cross sections except those for threshold fission in the form 0i k 0k riSk(63 97 where i = element in question k = type of cross section (r- = reference cross section for element i and cross section type k = correlation parameters for element i and cross section type k. For the threshold fission elements, the fission cross section is correlated in the form: I (6-4) R f A rapidly converging iterative procedure is presented (45) through which S and R can be determined for any practical core composition. Microscopic cross section data has been correlated employing Eqs. (6-3) and (6-4) for some 43 materials as of this writing; using the 26 group Russian ABBN multigroup set (58). (The same set employed with the ANISN multigroup computer code in much of the present work.) The one group model developed has been tested for some 45 different fast reactor compositions by comparing the results of one group calculations to 26-group fundamental mode calculations. The results have been found to agree with an average error of 1.77% in 98 material buckling, 0.218% in enrichment, + 0.588% in infinite multiplication factor, + 0.69% in reactivity, + 2.19% in core conversion ratio, and + 2.17% in the ratio of fertile to fissile fission. The present reactivity model is based on the selfadjoint perturbation theory formulation. Changes in macro- scopic cross sections are defined as follows: 5,= gAN+ 7AS (6-5) , where N = number density S = spectral characterization parameter from reference (45). The first term of Eq. (6-5) is simply the change in cross section resulting from a material concentration change, as usually employed in perturbation theory. The second term purports to account for the change in cross sections arising from a shift in the energy spectrum. From reference (45), all cross sections except the threshold fission cross sections can be written directly in terms of "S" as given by Eq. (6-3). Thus, the "spectral shift" term of Eq. (6-5) can be evaluated. For non-threshold cross sections, one obtains: 5Zspectral - I = 3 0 Sg)A = gZ( -). (6-6) 99 For the threshold cross sections, the following relation is used: = ( R )AS. (6-7) spectral Employing Eqs. (6-2) and (6-4) in this relationship and writing the result for U238, the only threshold fission element of interest here, gives: 28 6(7, + 28 V728 Cgf vf)gr )spectral + I- A - (6-8) where 1 1 = g gtr =z + 2 2 +gZtrtr tr tr (6-9) tr The change in diffusion length arising from a spectral shift is evaluated in a similar manner: D = 1 1 1 3,tr S 3(Z (6-10) 2 S + z tr + ... ) tr Thus : 5Dspectral = DAS = - 3D2( + g2 + ... )A or SD~ spectral = - t D( TD(~) where gtr is given by Eqs. (6-9). (6-11) -U 100 If it is assumed for the moment that the spectral shift referred to occurs in the perturbed region, that is; in the region where material concentrations are altered, and does not occur elsewhere, the desired result follows immediately from perturbation theory: + AJ [g + A dV + kAgtr - Vzf gtrD] - ( gi A 2dV (6-12) ?)T2 d where: A = F vz f dV core F = fraction of fissions occurring in the core 5 k= reactivity change calculated from the usual onegroup perturbation theory method. The two terms of Eq. (6-5) are separated in Eq. (6-12) simply for convenience. A major significance of the result achieved thus far is that the quantities in brackets in Eq. (6-12) need be evaluated only once for a given core. Only the quantity "AS" (and the usual 5#) must be determined for each perturbation of interest. Thus the quantities in brackets are evaluated and Eq. (6-12) becomes: 101 Sk =A a <P 2 + b( V~)2] dV + 5k, where V1 is the region in which material concentrations are altered. The method of evaluating "S" is given in reference (45) and "AS" is determined simply by calculating S' for the perturbed region (AS = St-S). An example of the calculation- al procedure, including a simplified method of determining S' for most cases of interest, is given in Appendix C. The energy spectrum, of course, does not abruptly change to the perturbed spectrum in V1 . Figure 6-1 in conjunction with Figure 3-3 shows the behavior of the spectral shift as a function of core position as calculated by ANISN. Figure 3-3 shows the energy spectrum for the unperturbed core and for the core with sodium removed. Figure 6-1 shows the shift in Group III and Group VIII fluxes as a function of axial distance from the core centerline for sodium removal from the central 30 centimeters of the core. Notice that the flux in Groups III and VIII is approximately that of a totally voided core in the central 20 cm, then changes gradually to the flux characteristic of the unperturbed core outside approximately Iz J = 30 cm. Similar behavior was found for other energy groups and other perturbations. Thus the accuracy expected from Eq. (6-12) for localized perturbations will depend on the extent to which the influence of the C- Flr~ L ,5c>_r Soolum ( T 1- - U rN ? f I Z r RCrQ & D A W LT-ACI Ait- MA av'r F ~x A L V 4RGLA-n F -I -V IS f-o~ N4 O R MC A L Z k o) ~ (;.Rovp I it. ALL N%.A IN . 3 iJ IA/ ///, /f/7~~ - ,- ,, ,ALL NA ip,4FLuy( - - RE-LA-VW E- 'I Di FLV)( FO&Z W46.E OCc,9E~ VOIDIK& ''I ~LL NAO~fr WA Ovr A" I 30 7- ( FJXAL -O-nON~Vv4 ) FLu~-.J' FLUX ,T 0 103 shaded areas in Figure 6-1 tend to cancel. That is, if the overestimate of the spectral shift in the perturbed region (area (1) of Figure 6-1) compensates, or nearly so, for the unaccounted for shift outside the perturbed region (area (3) of Figure 6-1), the accuracy of Eq. (6-12) in predicting reactivity changes arising from a spectral shift will be limited only by the accuracy of the spectral parameters "S" and "R" in correlating cross sections. "nneffect", Even without this Eq. (6-11) should give good results for perturbations involving large or "global" regions of the core. It is worth noting here that sodium voiding produces the strongest spectral shift of any of the plausible core rearrangements or perturbations examined, as will become clear from subsequent calculations in the present chapter. The cancellation effect just described is found to be quite beneficial for all perturbations considered and Eq. (6-12) has been found to give good results for localized perturbations. Furthermore, the "spectral" contribution to the total reactivity change predicted by Eq. (6-12) has been found to be substantial in most cases of interest. For the important case of sodium voiding, the spectral contribution predominates over much of the core. Before citing results predicted by Eq. (6-12), it is interesting to examine the components of the equation. For simplification in calculations employing Eq. (6-12) the core F 104 of interest is assumed to be composed of Pu239 U238, 02, Fe ANISN runs comparing calculations using these con- and Na. stituents with runs employing the actual core composition (316 stainless steel instead of Fe and including the small fractions of U235 and Pu240 expected to be present) show that this assumption introduces negligible errors for purposes of the present analysis. For the core described in Table 1-1, Eq. (6-12) then gives: 5(vZ f) 6k 5D a -4 Core Con stituent + 8.7xl0~ ]c2 + 1-.1161( v)LdV + 7.4xl0 4 J 92 + E-.018 0.0 + 0.25x10 4 + [+.0101( V<P)2j + A 0.0 ± + A 0.0 + 0.59xl0 = A I110.05x1 -8.4x1o~4 + Af +A + 6 j .k V2 0.26xl0-4B 2 + J2 + ( )Z U28 V Pu dV 02 -. 60] ( vp) 2JdV Na -. l06( vc) 2:dV Fe (6-13) Note that essentially all of the net contribution from 5(v7f) and 5(Za), and therefore the F 2 weighting, comes from U2. For the particular case of total sodium voiding Eq. (6-12) becomes: 105 i= A T(4.67 x 1o~4) ? 2 - 0.077 ($)1 5 2d spect 10 (6-14) + AfC(. 24 x 10~ )-4 2 - 0. 48( 2~ dV usutatlo Inspection of this equation shows the dominance of the strong spectral effect in the central regions of the core where the flux gradient is small. Equation (6-14) in conjunction with (6-13) shows the important role of the fertile isotope in producing strong positive sodium void coefficients in LMFBR's. It should also be noted that the decrease in absorption by U 238 produces essentially as strong an influence on void reactivity as the increase in fission which results from the spectral hardening accompanying sodium voiding. The contribu- tion of each core constituent to the sodium void coefficient can be readily analyzed with the present theory. It may thus be possible to employ the theory to help minimize the positive sodium void reactivity within a given set of design criteria. The reactivity model derived in this section will be referred to hereafter as the " PS Model " for brevity. 6.3 Applications of the Reactivity Model to Accident Analysis The PS model developed above and expressed in component form in Eq. (6-13) is written here in its condensed form for the core under investigation: [(18.85 x 10~)9 2 k =A + I .31( o)p )1 V (6-15) 1o6 Note that this result is applicable for all perturbations and in any geometry. The simplicity of the final result for a given core composition is one of the nicer features of the model. Equation (6-15) is now solved for a number of plausi- ble core rearrangements during an accident condition and the results compared with ANISN runs. While primary emphasis is on the pseudocylindrical geometry described in Chapter II, the equation is also solved and compared with ANISN in spherical geometry. The following cases of core rearrangement are considered: Condition or Rearrangement Case I (percentages shown are relative to the unperturbed core) 20% Sodium Voiding: Sodium density is reduced by 20% in the region of interest. II Total Sodium Voiding: Sodium density is reduced to zero in the region of interest. III 20% Fuel Addition: Fuel (U28 , Pu49, 02) density is in- creased by 20% in the region of interest. IV 20% Fuel Removal: Fuel density is reduced by 20% in the region of interest. V Total Na Voiding and 20% Fuel Addition: Sodium density is reduced to zero and fuel density simultaneously increased by 20% in the region of interest. 107 VI Total Na Voiding and 20% Fuel Removal. VII 20% Na Voiding and 20% Fuel Addition. VIII 20% Na Voiding and 20% Fuel Removal. Equation (6-15) for the eight cases of interest becomes: 20% Na Voiding ( AS = .0484): Case I: 5k = A (.91 x 10- 4 ) y 2 - .015( spectral v<r )27 dV effect (6-16) + A Case II: p L(. 0 5 x 104 ) 2 - AS (-s Y Total Na Voiding (4.67 x 104) <T 2 k = .080( usual perturbation vqp7)]dV .248): spectral .077( vp )23dV effect (6-17) + A usual perturbation r(.24 x 10~4) ( 2 - .48( Vc?)2jdV As seen from these results and as noted earlier, the spectral effect strongly dominates in the central regions of the core for the case of sodium voiding. Case III: 6 = A = .055): 20% Fuel Addition($ 16 j(l.04 x 10~) + AI,(3.00 x 10 2 ) 2 .06(V ]dV <P ) dV -150( .10( 7rpJdV - spectral effect { (6-18) usual perturbation 108 20% Fuel Removal (A Case IV: = A = - .073): \-<p)22dV (- 1.37 x 10~4) cp 2 + .023( spectral (6-19) + A 3.00 x 10~ ) Q9 - .190( V')jdV erturbation Two important points are clear from Eqs. (6-18) and (6-19). First, the spectral shift plays an important role in the reacSecond, the negative tivity effects produced by fuel motion. reactivity induced by fuel removal is stronger than the positive effect induced by the same amount of fuel addition. In this case the primary reason for the difference is the stronger spectral shift induced by fuel removal (=AS - .073) as com- = + .055). pared to that induced by fuel addition (A This important phenomenon is discussed further later in the present chapter. Case V: Total Na Voiding with 20% Fuel Addition( .A= .265): S AJ c) 2 - .082( (4.99 x 1 04 2 dV spectral effect (6-20) + A Case VI: 4 L(3.24 x 10 ) 2 - .240( \7 )2JdV Total Na Voiding with 20% Fuel Removal ( Sk = A (4.13 x 10 4 ) < 2 - .68( .'-\ perturbation = .219): spectral ) 2 dV 'eJLeffect (6-21) + AJL(- 2.76 x 10~ ) cp 2 - .830( v'e)JdV usual perturbation I- -9 - ION-- -1 1-- 109 Note for case VI that the overall reactivity effect near the core center is positive, that is; sodium removal has a stronger positive influence via the induced spectral shift than the concurrent negative effect of reducing fuel density by 20%. This is indicative of the strong influence sodium voiding has in such a reactor; a behavior which becomes more apparent as the present development continues. Additionally, for case VI, note the strong negative coefficient of the leakage (( 'q')2) term as compared with cases II and V. Section 6.4 of the present chapter considers the significance of this behavior in detail. Case VII: 20% Na Voiding with 20% Fuel Addition ( gv)2)dV 7 5= A kc [(1.98 x 10-4) q 2 - .0326( . 4Oj = .105): fspectral effect (6-22) + A [(3.05 x 10-4) q 2 .o8o(wy)2]dV - usual (perturbation Case VIII: A k= k 20% Na Voiding with 20% Fuel Removal ( ( - .374 x 10-) 2 + .oo6( p L 2dV J = - .0198: spectral effect -- (6-23) + A ( - 2.95 x 10~)7 2 - .270( 9q)ldV usual perturbation L Comparisons between the reactivity predicted by the ANISN computer code and the reactivity model developed above are now made. For the pseudocylindrical geometry described in 110 Chapter II, the axial flux shape is taken to be cO q=0cos 7z As shown by Figures 3-1 and 3-2 this gives an exceptionally good fit to the axial flux distribution predicted by ANISN over the core region of interest. cal geometry a flux shape of 'f= For comparison in spheri< 0 El - (1)2 is assumed. This was found to give a reasonably good fit to the ANISN data for runs in spherical geometry. The reactivity predicted by Eq. (6-16) (case I: 20% Sodium Voiding) is compared with that predicted by ANISN in Figure 6-?. The void extends over the entire radial area of the core and is assumed to expand symmetrically about the axial core centerline. The 26 group Russian (ABBN) cross sec- tion set in the S-8 transport theory approximation was employed for these ANISN runs. case is seen to be quite good. Success of the model in this In Figure 6-3 a similar comparison is made for total sodium voiding (case II). In AS this case the spectral shift is severe (.1 = .248). Since S derivation of the spectral reactivity model employed partial derivatives with respect to S, the results are not expected to be accurate for large values of . An improvement in the results for each case investigated was achieved when an average value is used for the quantity S. For example, for the case of total sodium voiding, the following values were calculated: FIGuI~e GZ -ZE:A C F IV -f DIPP 00 *009 1004DT .003 0 :Z0 7- CS%7-r= of: 6oWUM '40\6 . 2W-% Sot>%UtA 007; 3. 4'0 112 vs. Cr~p N =.0 ?4'E&IGO 01F .03O1 -000-Ai A-- l's't%~L '4IT'A MVOjL '010?Sl (AIrr$ 5 )4AN6E4 JRoMC.e-H \1okt> -rzowrt'4 xs <pmrae I X (5)ZF- OV: :~ ~SOt )OrA -30 \JOb; Al- :Sc2-CV,)AA 9EMfOVO; W% ~r 113 S = .3530 for the critical unperturbed reactor St = .4531 for total sodium voiding = S + = .403 AS .1001 . AS .1001 .248 The dashed curve in Figure 6-3 results from using this latter Points are also shown in Figure 6-3 quantity in Eq. (6-15). for calculation based on the unperturbed S values. The im- provement in accuracy in this case is typical of that for The use of other cases calculated. -$ for large spectral S found to give more consistent perturbations was, in fact, results than the employment of AS - or AS . If the change in cross sections is evaluated by means of a Taylor expansion instead of differentials, the use of AS- can be more rigorously justified. For example, by taking differentials, it was found that: 5E = gZ (6-6) AS Employing the Taylor expansion in evaluating 67 gives: 5z= ZO(st)g - z'(S)g =Zo (S+AS)g - S , 114 where 2 +... lAS + g(g-2)S- (S+AS)6 = Sg + gS By taking the first three terms of the expansion g~[S r is obtained. (g-1) g(AS +S 2 ]I Notice that the first term of this latter equation gives the result obtained in Eq. (6-6) by taking derivatives; as expected. By including one additional term, better accuracy for larger values of i is expected. As shown in Appendix C, the "average" value of g for all constituents used in the current LMFBR is considerably less than unity. Then if (- 1/2) is taken as the coefficient of AS 2 (T) in the above expression for 5z = gz 0SD AS = 1 AS 2 - 8 U.; the result becomes: S-. + l/2ASSJ = az SC( and the advantage of employing As for large perturbations in the present analysis is seen not to be entirely fortuitous. - -S = 0.0484, and for For the case of 20% sodium voiding, where other cases where A is small, the averaging technique is clearly unnecessary. Figure 6-4 shows the results for global sodium voiding in spherical geometry. The deviation of the ANISN and 115 IE (2.o~SoD~uH\ RernoviE0 ,009 '003, 000 2.0 ((ZADV L 30 t 'SlZrz. OF 401b- 20,2,0 SDt %tjrA 70 ~-i1i~1~ 116 calculated curves beyond the 50 cm radial position is due in part to the fact that the parabolic flux shape assumed did not give a perfect fit to the unperturbed ANISN flux data. In particular, the fit was not as good as that achieved by using the simple cosine shape for the axial direction in cylindrical geometry. Since the curves in Figure 6-4 are integral curves and the error in flux shape is weighted by the core volume; the error generated in spherical geometry might be expected to be appreciable in the outer regions of the core. Table 6-1 shows the reactivity predicted by ANISN and the spectral reactivity model for several cases of core rearrangement. Success of the model in handling the cases tabulated was good except for RUN 407. In this particular run the calculated result is the algebraic difference in the large positive reactivity calculated for 120% fuel density in the 0-15 cm interval = + .0278) and the large negative reactivity calculated for 80% fuel density in the 15-30 cm interval (5k = -.0240). RUN 407 was included to show this characteristic, but expected, weakness of the model. Each of the comparisons given thus far has involved perturbations which are uniform over fairly large regions of the core. As cited earlier, the model is not expected to be highly accurate for localized perturbations. in handling such perturbations Its usefulness is considered in Figure 6-5. Table 6-1 COMPARISON OF ANISN AND 'PSt MODEL RESULTS Condition or Rearrangement For Run Run 401 CNa = 0 in 402 CNa = 0 in 403 CNa (= 405 407 0 in interval [0-15 cm] int. {O-30 cm] interval 10-15 cm] (2) Cfuel = 1.2 in int. [0-15 cml Cfuel = 0.8 in int. [15-30 cm] CNa = 0 in interval [0-30 cml Cfuel = 1.2 in int. [0-15 cm] Cfuel = 0.8 in int. [15-30 cm] Cfuel = 1.2 in interval [0-15 cm] Cfuel = 0.8 in 4o8 . CNa = 0 in int. ANISN Results with with 16 groups 26 groups I I T PS' Model Results .0156 .01520 .0172 .0245 .0239 .0275 .0241 .0236 .0220 .0286 .0282 .0292 .oo625 .0038(3) [15-30 cmj interval [0-15 cm Cfuel = 1.2 in int. Cfuel = 0.8 in int. [7.5-15 cm] +.01515 [0-7.5 cm3 density after perturbation unperturbed density 2) Intervals are symmetric about axial centerline. Thus [0-15 cm3 represents a void in the central 30 cm of the core. 3) The calculated result here is the difference in two large numbers. See text for additional discussion. NOTE: 1) C F-i 118 OF' Wk-t~4 -=2-S CwA~. Pia or . 003 INTER~EST-~ I'ObO-L rA-rA '002, ENISe Co(I. CCuvR- co5 K. loot 0 -. 001 410S 0 A)(AL ?oP0Sj-r0N 04C c6-eA-IO' c- ~ .~ mm 119 Local axial sodium voids of 2.5 centimeters in width were employed in ANISN and in Eq. (6-17). Figure 6-5 shows that the success of the model in predicting the magnitude and shape of the resulting reactivity changes is relatively good. Observe that this particular series of calculations involves the limits of the model in two respects; the value of As is large (0.248) and the size of the region perturbed is quite small (2.5 cm). As can be seen from the values listed above for cases I-VIII, sodium voiding produces about as strong a spectral shift as any of the rearrangements considered. As a result of the reasonable success of the PS model in handling spectral shifts of such magnitude; its perhaps fortuitous accuracy for localized perturbations; and, more importantly, the excellent results achieved for less severe global perturbations, it has been found highly useful in analyzing reactivity effects associated with accident conditions. 6.4 Core Compression and Expansion Effects on Reactivity The ultimate shutdown mechanism for an extremely severe accident is the negative reactivity inserted as a result of core expansion or disassembly. In analyzing such effects it has frequently been assumed that the core remains homogeneous (in the sense that the volume fraction of each constituent remains constant in all locations) throughout the disassembly process. (7)(8)(9)(13)(18)(47) In such cases reactivity mu 120 effects can be realistically expressed as a function of the homogenized core density for a fast reactor. This is the method used in the Bethe-Tait approach employed in references (7), (8), (9), (13), (18), and (47) and in numerous other places in the literature. In the present section, a more general relationship than that given by Bethe-Tait is derived for predicting reactivity changes as a function of core density changes. The present result shows the aforementioned behavior of density reductions in producing a stronger negative reactivity insertion than the positive reactivity added by a comparable density increase. A comparison of the present result and the Bethe-Tait method is given in Appendix C and it is shown that the Bethe-Tait model cannot predict the anomaly just cited. A key to the success of the present analysis is the observation that the spectral parameter "S" defined by Eq. (6-1) is independent of homogenized core density. Thus a core compression or expansion does not cause a spectral shift and the second term of Eq. (6-5) is zero. This implies that the usual one group self-adjoint perturbation theory result should be adequate for analyzing compression and expansion effects. A negligible spectral effect has previously been assumed but not demonstrated in various applications of the Bethe-Tait method. -U 121 The following relationships are employed in the present application of perturbation theory: C= Pf final or perturbed density initia1 or unperturbed density P (3-1) (C-1)Za (6-24) 5z = (C-1)Zf (6-25) 6,a D= (6-26) -C-)D With these definitions, the usual perturbation theory result can be written as follows: 5k = A (C ( ~)cp2 + D( 7qp)2dV. (6-27) In order to see the effect of compressions and expansions more clearly, it was found convenient to assume a cosine flux shape. Note that this does not restrict the applicabil- ity of the analysis to a particular geometry; the result will still be valid to the extent that a cosine function adequately describes the actual shape. cp = qpocosBx This simplification gives: for a slab (or the axial direction in a cylinder), 1 S= B = YQcos Br He or e for a sphere, . 122 Thus: ( \7q)2 ( Vq)2 = 1B = B(P2_ 2) (6-28) for a slab and 20 2) for a sphere (6-29) . Using these results and the one-group relationship: ( v-,f - in Eq. (6-25) 6= A 2a) = DB 2 (6-30) , we obtain: Cl 1 + ( )2(C-1) dV (6-31) in slab geometry and A + 1= 1 () (C - )jdV (6-32) in spherical geometry, where A FDB2 = l (6-33) 2f dV and F = fraction of fissions occurring in the core ~ .94 for the case of Table 1; as calculated by ANISN. Equations (6-31) and (6-32) show clearly the effect of density changes on reactivity. If density is reduced to 80% of the original value in some region of the core, the quantity = - On the other hand, if density is somehow increased to 120% of its original value, + 1= In 123 addition to the results shown by Eqs. (6-31) and (6-32), relationships have been worked out for spherical geometry using the more correct parabolic flux shape, q = <o 1l (r) and for the pseudocylindrical geometry employed in ANISN calculations as discussed in Chapter II. The resulting equations are: Sk = AoF4) 2 ) + C( (l- (6-34) )dV in spherical geometry and Ao C= 1 + ( )(C in pseudocylindrical geometry. - .246)1 dV (6-35) This latter expression con- siders only axial compressions and expansions for the cylindrical core under consideration. Table 6-2 shows the results predicted by Eqs. (6-32) through (6-35) for localized density increases or decreases at various positions in the core. Note that in each case except at the core center in spherical geometry, removal of a given quantity of material produces a stronger negative effect than the positive effect produced by addition of an identical amount of material. Figures 6-6 and 6-7 compare these results with those predicted by ANISN. Examination of these figures shows that the ANISN results substantiate the conclusion drawn from Table 6-2. From Figure 6-6, if 20% of homogenized core material is removed from a small volume element at position (a), it must be moved inward closer to the core center than position (b) Table 6-2 REACTIVITY EFFECTS FROM CORE EXPANSION OR CONTRACTION 5 ( k) Axial Expansion or Contraction in "Cylindrical" Spherical Geometry 2 (q )2 990 Cosine Flux C=l.2 Parabolic Flux C=l.2 C=.8 C=.8 1 + .20A 0 - .20A .7 + .151AO - .156A 0 .5 + .120A 0 - .130A 0 ± .25 + .o83A - .100A 0 + .o8l2Ao 1.0 0 + .20A 0 .121A 0 - .20A0 - .131A Geometry C=.8 C=l.2 0 - .097A0 + .33Ao - + .25A 0 - .325A 0 + .208A 0 - .288A 0 .40AO ro 125 ' I&L rm. G- ( uwraarupo3hO FLoyx ( SQuA r.Eb) FOe L.OC.ALIZ E D(2,5-e^. INTERVAL) OF ?RTAGTO4 )-Jt1orsoNizEaD ~12 ~.004~ CC (-- TAA -- .002 ((b) .DOO 001/4 ,& (2. l'4' .0031W" N....or -. 0He o v..... ..... SA TCLA.A T~ . 126 f f &0RE r;, - REiACT-w iTrY uNprVERU rGE) FLux FO(P LoLCALtr.6E OF )4oMO6ENIZEt t~ascr -C..= rUFSP7'ods Cotr 1, orn O .004j A- 00Z A - \ N ATE C ALC~ L.ATEb DATA ,001 0 1,0 v.'g 0., &.- .,q 03 P,.2 0.1 -oo\ -,002 --- 00 -.10o5 -. R- I ( a m 127 before a positive reactivity effect is produced. If it is inserted between (a) and (b), a negative reactivity effect results from moving core material inward. It is, of course, unrealistic to speak of adding 20% of homogenized core material at a given location. The tendency shown in the present analysis is exhibited, however, in physically realizable rearrangements. Consider, for example, cases II, V and VI above (Eqs. (6-17), (6-20) and (6-21)). For case II, total sodium voiding, the coefficients of the leakage term is - .557. For case V, total sodium voiding with 20% fuel addition, this term is - .322, while for case VI, total sodium voiding with 20% fuel removal it is - .898. Thus if sodium is voided from a region of the core and then 20% of the fuel in part of this region is injected into another region (removal of sodium now makes room for the fuel), the difference in leakage coefficients clearly shows the density effect cited. Namely for fuel addition we ob- tain a change in the coefficient from - .577 to - .322 or + .235 whereas for fuel removal we obtain a change from - .557 to -.898 or - .341. This is identical to the behavior for the homogenized compression and expansion effects shown by Eqs. (6-31) through (6-35). AP Notice that the coefficient C-1 can be written as C Pfinal The behavior of this quantity in ameliorating the effect of density shifts on reactivity is analogous to the behavior of 128 the quantity S on reactivity. in ameliorating the effect of spectral shifts These two apparently independent phenomena play a highly significant role in reducing the reactivity additions from all core rearrangements investigated. An example of the combined influence of these parameters can be deduced from data in Table 6-1. Consider the ANISN results with the 26 group cross-section set (column 2). In RUN 407 the reactivity introduced by fuel movement alone is seen to be + .00625. In RUN 402 the reactivity introduced by sodium voiding alone is seen to be + .0239. In RUN 405, the same extent ofsodium voiding and fuel movement taken simultaneously gives a reactivity change of + .0282. Thus, if the fuel movement described in RUN 407 occurs after sodium voiding of the region of interest, the reactivity added by fuel motion is: .0282 - .0239 = + .0043 and not the + .00625 calculated in RUN 407. The same con- clusion is readily reached by comparing the hand calculated results for the same runs in Table 6-1. A more impressive example can be shown by considering RUNS 401 and 408 of Table 6-1. In RUN 401 the reactivity introduced by voiding sodium in the central 30 cm of the core is seen to be + .01520. In RUN 408, fuel is moved in- ward while the same degree of sodium voiding is present. - U- 129 The result is a slight reduction in reactivity to + .01515. Inward fuel motion results in a slight shutdown effect. Further examples showing the combined influence of the density ( AP) and spectral (-S ) effects are given in succeeding sections of the present chapter. 6.5 Observations from Calculations and ANISN Results The reactivity changes resulting from a number of core perturbations or rearrangements are shown in Table 6-3. The data presented was obtained from ANISN computer program runs. As noted earlier, the runs listed in Table 6-2 were calculated with the PS Model as well. The following observations from the data of Table 6-3 are of interest: The compression of core material in one (1) region and expansion in another shown by RUNS 18, 20, 22, 24 and 25 results in an overall negative reactivity effect. served. Note that in each case material is con- Note also that RUNS 24 and 25 are the reverse of RUNS 18 and 20 respectively. As would be expected, when the general material movement is away from the core center, the shutdown effect is stronger. (2) In the runs cited above, such uncompensated compression of the core is, in general, physically unrealistic. In RUNS 212, 213 and 214, however, room is available for fuel inward motion as a result of sodium 130 Table 6-3 REACTIVITY EFFECT OF CORE REARRANGEMENTS Condition or Rearrangement for Run Run 18 20 22 24 25 C = 1.5 in interval [0-17.5 cm] C = 0.5 in int. 17.5-35 cm] C = 1.5 in int. [7.5-10 cml C = 0.5 in int. C = 1.1 in int. C = 0.5 in int.f12.5-15 cm] C = 0.5 in int. C = 1.5 in int.117.5-35 cm1 C = 0.5 in int. [7.5-10 cmJ [10-12.5 Net Reactivity Effect - .00199 - .00350 cmj - .00240 O-12.5 cm] 10-17.5 cm] (run 18 reversed) (run 20 - .03080 - .0044 reversed) C = 1.5 in 212 = 0 11O-12.5 cm] in int. Cfuel =2.0 in int. [20-22.5 in int. L22.5-25 cm] in int. E10-12.5 cm] Cfuel = 2.0 in int. f10-12.5 cm] in int. CNa Cfuel =0 213 int. CNa = 0 Cfuel = 0 20-22.5 cm] - .ooo8 cmj [12.5-15 cm] - .0012 -1: 131 Table 6-3 (Continued) Run 214 CNa = 0 in C fuel = 2.0 in 218 int. [35-37.5 cmI int. [35-37.5 cm] in int. in int. C fuel = o.8 in int. C fuel = 1.2 in int. C fuel = 0.8 in int. C fuel = 0 217 Net Reactivity Effect Condition or Rearrangement For Run C fuel = 1.2 - .ooo6 p37.5-40 cm) [20-22.5 cm]i - .0017 [22.5-25 cm] E10-12.5 cmI - .0020 J12.5-15 cm] 219 CNa = 0 in int. [0-12.5 cm] + .0131 223 CNa = 0 in int. [0-12.5 cm] ± C fuel = 1.2 in int. T0-12.5 cm] 224 225 C fuel = 0 in int. [12.5-15 cmJ C Na = 0 in int. '0-12.5 cm] C fuel = 1.5 in int. {0-12.5 cmj C fuel = 0.5 in int. [12.5-25 cmj = 0 int. (0-12.5 C fe1 = 1.5 in int. EO-12.5 cml C fuel = .5 in int. 112.5-15 cmJ C fuel = 0 in int. [15-20 cmj CNa in cmI .00170 + .0271 + .01863 132 Table 6-3 (Continued) Condition or Rearrangement Run 226 227 228 For Run in int. in int. [25-27.5 cm] in C fuel =1.1 int. 0-45 cm] C fuel =-0.1 in int. £45-50 cm] 1.1 Cfuel fuel Cfuel =0 CNa = 0 in int. CNa = 2.0 in int. Cfuel+ss = 2 .0 in int. Cfuel+ss = 0 in int. [ 0-25 cmj [0-12.5 cml Net Reactivity Effect + .00256 + .02007 + .o165 112.5-25 cm] [ 0-12.5 cm] [12.5-25 cm] 401 CNa =0 in int. [0-15 cm~ + .0152 402 C = 0 in int. [0-30 cm] + .0239 403 CNa 0 in int. EO-15 cmJ + .0236 C fue 1 = 1.2 in int. [0 -15 C fuel = 0.8 in int. [15-30 cml = 0 in int. 0-30 cm] in int. 0-15 cm] Cfuel = 0.8 in int. 15-30 cmJ 405 CNa Cfuel = 1.2 cmI ± .0282 133 Table 6-3 (Continued) Condition or Rearrangement for Run Run 407 408 int. [0-15 cm} Cfuel = 0.8 in int. 115-30 cm} in int. 1.2 in int. Cfuel = 0.8 in int. [7.5-15 cm] Cfuel = 1.2 in CNa =0 Cfuel = [0-15 cm]i Net Reactivity Effect + .00625 + .01515 0-7.5 cm] 409 CNa = 0.8 in int. £ 0-15 cm} + .00346 412 CNa = 0.8 in int. (0-30 cm] + .00562 414 CNa = 0.8 in int. 0-15 cm] Cfuel = 1.2 in int. [0-15 cm] Cfuel = 0.8 in int. 15-30 cm] CNa = 0.8 in int. [ 0-30 cm] Cfuel = 1.2 in int. L0-15 cm] Cfuel = 0.8 in int. [15-30 cm] = 0.8 in int. [O-15 cm~] = 1.2 in int. E0-15 cmI Cfuel = o.4 in int. [15-20 cmj 415 420 CNa + .oo6o + .0083 + .00565 134 Table 6-3 (Continued) Condition or Rearrangement for Run Run 421 422 0.5 in int. £0-15 cmj C fuel = 1.5 in int. E5.0-15 cmj C fuel = 0.0 in int. 115-20 cm] = 0.8 in int. [0-25 cm] C fuel = 1.2 in int. [17.5-25 cmj o.4 in int. CNa CNa = C fuel = Net Reactivity Effect + .00744 + .00576 [25-27.5 cml NOTES: 1) C = density after perturbation unperturbed density 2) C with no subscript implies overall core density (homogenized). 3) In all cases all material except sodium is conserved. That is, only sodium is allowed to cross the core boundaries. 4) All runs in this table are in pseudocylindrical geometry. All rearrangements are symmetric with respect to the core axial centerline. 5) The Hansen Roach 16 group cross section set was used in all In this series the Russian 26 runs except the 400 series. group set was used. The S-8 transport theory approximation was used in all runs. - U- 135 voiding. The same tendency in producing an overall shutdown effect is seen in these runs. (3) For realistic rearrangements affecting larger regions of the core, consider 401 and 408; runs which were compared earlier in Section 6.4 of the present chapter. As noted, the overall effect of inward fuel motion in RUN 408 is a slight reduction in the reactivity present from sodium voiding alone in RUN 401. RUNS 219 and 223 show a second example of this behavior. In RUN 223, overall reactivity is re- duced substantially when inward fuel motion follows sodium voiding. (4) Comparison of RUNS 224 and 225 show the strong effect of a local fuel "void". The same degree of sodium voiding and inward fuel injection is present in both cases. In RUN 225, however, the fuel for inward motion is obtained by completely removing fuel in the 15-20 cm region. This results in a substan- tial reduction in the reactivity inserted as compared with RUN 224. (5) The reactivity induced by fuel motion cannot be calculated separately and added to that produced by sodium voiding. invalid. In other words, superposition is This is a general observation from the pres- ent study and can be seen by comparing RUNS 402, 405, 136 and 407. The sum of the reactivities from RUNS 402 and 407, with sodium voiding and fuel motion taken separately, is + .03015, not the result obtained in RUN 403 for simultaneous voiding and fuel motion, + .02820. The non-linearity of these effects is further evidenced by the spectral shifts predicted in cases I-VIII of Section 6.4 above. For example, 20% sodium voiding produces a AS value of + .0484; 20% sodium voiding with simultaneous 20% fuel removal gives a value of g- of - .0198; whereas 20% fuel re- moval alone produces a value of - .073. Then the sum obtained by adding the separate spectral shifts is + .o484 - .0730 = - .0246; not the more correct value of - .0198. (6) In general, when 20% or less of the fuel in a given region is moved to some other core location, the reactivity effect appears to be small in comparison to the effect of sodium removal (voiding) from a region of comparable size. Even when a large fraction of the fuel in a localized region is shifted inward through an appreciable distance, the reactivity effect is of the same order as that expected from extensive sodium voiding. RUNS 212-214, 225 and 228 are indicative of these observations. 137 (7) Comparison of RUNS 226 and 227 is particu- larly interesting. The rearrangement considered for these runs is intended to simulate a collapse of fuel within the clad to 100% theoretical density; approximately 110% of the normal density. In RUN 226 the central half of the core is collapsed toward the center. In RUN 227 the entire core is collapsed toward the center. Note that the reactivity added in RUN 227 is about eight times as great as that added in RUN 226, although the extent of fuel movement is only about twice as great. RUN 226, in essence, is a rearrangement of the central region of the core and is somewhat similar to the rearrangements of interest in the present work. RUN 227, on the other hand, is more representative of a core compaction into a second critical configuration in the sense that material on the edge of the core is involved in the compaction. This result is somewhat incidental to the present study and its implications were not pursued in detail. It does suggest, however, that a small amount of compartmentalization within an LMFBR fuel element might substantially limit reactivity insertions due to fuel motion. 6.6 Reactivity Addition Rates from Fuel Motion In Section 6.1 of the present chapter mechanisms which can lead to fuel motion or rearrangement were discussed. In 138 Sections 6.2 through 6.4 methods of predicting overall reactivity changes resulting from various core rearrangements were investigated. In Section 6.5 a tabulation (Table 6-3) of core rearrangements and the resulting reactivity changes is presented and discussed. In the present section a method of estimating reactivity addition rates which can result as fuel motion (leading to the tabulated rearrangements) takes place is developed. The fuel motion mechanism treated is axial fuel injection from intact clad as discussed in Section 6.1. The following sequence of events is postulated. Refer to Figure 6-8: (a) A CATEGORY II excursion occurs producing clad failure as shown in Figures 2-3 and 6-8. The region 6R- extends to position z1 of Figure 6-8. (b) Immediately after clad failure (time t=O in the present analysis) the pressure in region R. drops to a low value, say - 20 psia. This value is typical of the pressure at the core center in normal operation. Such a pressure can be expected to be reached quickly as the molten fuel in the center of the fuel rods in region-R is expanded and cooled. A relatively small degree of cooling is required to sharply decrease the pressure in region -R as can be seen from Figure 5-3. This low pressure will then persist until sufficient energy is added to the sodium in region IR to produce 139 MAooc- fr:-E6LDr FOE i - 01P INCR06AEv, IVE.N~ir 4 <D 0 I 0,: T- Moe (1')CPN T-EmP. r 00'*V, E~ (:t ~ol e 14r l- 'J C. -T -P-PICAL YPILALLS A'TOLTrzN 1:0 EL &WND'1ua P~tSSI~E-~I f~l EtAN L. 00 I F ~ z RZO F .F oP1 e-.A P(14 Goo - ASsurme PtzL-ssuc spj I mMFTELyr 2007- 0 -fl DI1MGN6IO")$ [L'&Pr.Nt ON" Rom~ CONS%_tAjqb (AC-VAL 14o high sodium vapor pressures; as discussed in Chapter V. Quantitative comparisons of the fuel motion process and the sodium voiding process are given later in the present section. (c) The fuel vapor pressure in region (B) of Figure 6-8 remains high initially. In the case shown the pressure at z2 in the center of region (B) is 600 psia at t=0. (d) As a result of (b) and (c) above, a pres- sure gradient is established between points in region (B) and iR ; resulting in fuel motion into-R . Fuel motion away fromR 3is prevented by the cooler solid fuel and eventually by the much cooler blanket material as shown in Figure 6-8. (e) These events produce a decrease in the fuel density in region (B) and an increase in (A); resulting in a reactivity change. The hydrodynamics and kinetics analysis of the events described in (a) through (d) above are clearly complex. Some considerable insight into the significance of the fuel injection mechanism can be gained, however, by invoking a few reasonable assumptions about the time dependent behavior of the process and by making use of the fact that the overall reactivity change for a given static rearrangement can be accurately determined. 141 The hydrodynamic analysis employed is somewhat similar to that developed in Chapter V for investigation of the sodium voiding process. The necessary equations are listed here and discussed momentarily. The relation: 2 7 (6-36) d(z,t) ,7p~.,t)== pPdt 7 expresses conservation of momentum for the fuel material in region (a) and (b) of Figure 6-8 in the absence of frictional effects. From Chapter V: Q(t) = ~c'P(T(t)-TO) + (h(t)-hf) , (5-9) which is the energy conservation equation for the two-phase fuel in region (B). Equation (4-1), repeated here for con- venience, gives the vapor pressure of fuel material as a function of temperature: 7 exp p = 8 x 10 6.7 x 10 4 T( K) (4-1) Figure 5-3 is a plot of this was emEquation (4-2), T(Z) = Tocos w2 ,17 where p is in atmospheres. relationship. ployed in Eq. (4-1) to obtain the axial pressure profile in the central core regions. In Figure 6-8, the dashed curve in the "p vs. z" plot shows this dependence schematically. The Clapeyron relation Ah Tdp TV= T (57) U U ~- 142 is used to eliminate the enthalpy term in Eq. (5-9) as in chapter V and as discussed below. The following steps, including pertinent assumptions, apply the above equations to the problem described and depicted in Figure 6-8. (1) In Eq. (5-9), Q(t) is the fission heat source and is found to be negligible in the present application. In Chapter VII it is shown that the average power level between an initial and secondary excursion is of For c = .42 j/gm 0 K for p the fuel (Table 4-2); the rate of temperature rise of the order of 1500 j/gm sec. the fuel is seen to be about 3.60 K/msec. In the pres- ent analysis the maximum time interval considered is 8 msec; corresponding to the time between power peaks in the typical excursion of Figures 2-1 and 2-2. the approximation Q(t) - Thus, 0 is quite reasonable in that the maximum temperature error which can result is about 290 K. From Eq. (5-9) this approximation gives: po(T~OT(t) = Ah (2) . (6-37) The expansion process in region (B) of Fig- ure 6-8 is assumed to be reversible; generally a reasonable approximation in evaluating the expansion work done by a two phase substance. With this assumption Eq. (4-1) can be used in (5-7) to evaluate the enthalpy 143 The result obtained is: term in Eq. (6-37). S(T-T p ) (tW = B() rJo l - 1) (6-38) , where B = 8 x lo A = 6.7 x 104oK atm= 8 x 10 dynes/cm2 3 = volume of region of interest (cM) (3) Two cases for the condition of fuel in region (B) are employed. (B) is taken as In the first case 10% of the fuel in V-0 of Eq. (6-38) and the mean tempera- ture of the region is estimated to be 4600 0 K. From Eq. (4-1) this gives an initial "mean" pressure of 600 psia. In the second case 20% of the fuel in region (B) is taken as Tr 0. The mean temperature of the region is then found to be 4400 0 K; leading to a mean initial pressure of 290 psia. If smaller values of V-0 are assumed (and therefore, very slightly higher mean temperatures), it is found that the pressure in region (B) drops rapidly as expansion takes place; resulting in a lower effective pressure for producing fuel motion over the time interval of interest. If volumes of larger than 20% are assumed; the initial mean temperature is lower and the result again is a lower "time average" driving force for fuel motion. Thus, the two - U - 144 cases employed result in "worst case" behavior for a number of combinations considered. Recall also that the clad rupture pressure is about 1200 psia so that, in view of Figure 6-8, a pressure of 600 psia some several centimeters outside 6-R is about as high an initial pressure as can be reasonably expected. (4) No frictional effects are included. In five of the seven cases of rearrangement considered only motion of molten fuel in region (B) is assumed to occur. Of these five, fractions of from 20% to 60% of the fuel in region (B) is assumed to be injected inward. In the two remaining cases essentially all of the fuel in region (B) is assumed to be injected. (For purposes of calculation a minimum of 10% of the fuel in (B), that is; 9 , is retained in the region. Thus the wide range of cases considered allows, in a sense, for a wide range of influence of "friction" in determining how much fuel moves inward. method of including this effect, An analytical for which no experi- mental guidance is available, could not be realistically envisioned. The short distance travelled by the fuel material inside the clad rods (see Figure 6-8) and the very low velocities attained by the fuel while within the clad (a maximum of about 10 m/sec) indicate, however, that the frictional effect in the usual sense 145 (see Eq. (5-1) et. seq. in Chapter V) is a minor consideration compared to the question of how much fuel in region (B) actually moves inward as the pressure in (B) is relieved. (5) In Eq. (6-36) the pressure gradient is assumed to be linear as shown by the solid curve in Figure 6-8. The gradient is established by assuming a pressure of zero at some location in (A) and a pres- sure in (B) determined by inserting the appropriate (time dependent) temperature in Eq. (4-1). The dis- tance over which the pressure is dissipated, Az, corresponds to the "width" of region (A). For example, in RUN 420 of Table 6-3; Az = 15 cm and the pressure gradient is written as: Vp Az0(6-39) The following events occur in regions (A) and (B): The vapor pressure in (B) forces molten fuel in that region to expand toward the open end of the clad with the result that fuel from (B) is accelerated by the pressure gradient of Eq. (6-39) until the extent of axial motion specified for the RUN of interest is accomplished. Employing Eq. (6-39) in Eq. (6-36) gives: 2 dz ~()=p(Az) p.,(6-40) i46 and with Eq. (4-1) one obtains: B 1 exp(- A1 -T) = p(Az) d z t dt ( -1 (6-41) Equations (6-38) and (6-41) thus give T(t) f( and T(t) = f(z(t)) respectively. The relation between \ and z(t) is linear and is determined by conserva- tion of material for each particular RUN considered. of the fuel in (B) : when corresponding to 10% \T In Run 420, for example, with = 6, region (B) remains; as specified. 15 cm is required. 40% of the fuel in At this time, z(t) In other words, the fuel which is forced from (B) initially must be accelerated 15 cm into (A). The solution to Eqs. (6-38) and (6-40) is obtained in a manner similar to that described in Appendix B.2 The hand calculational procedure is quite simple in the present case, however, since the pressure in region (B) does not change by more than about 10% in the time interval of interest. Thus, from Eq. (6-41) one obtains: 2 constant =C dt dz =tC1 and t2 z(t) =Cl 2 (6-42) 147 as a first approximation. Relatively minor corrections are required as the fuel motion progresses. (6) The reactivity change for a given RUN is ob- tained from Table 6-3 and Figures 6-2 and 6-3. Again using RUN 420 as an example; the total reactivity addition for the rearrangement, consisting of sodium voiding plus inward fuel motion, is 5k = + .00565. In the present analysis fuel motion is assumed to occur after the specified degree of sodium voiding has taken place. Thus to obtain the reactivity addition from the fuel motion specified, the reactivity change wrought by sodium voiding must be subtracted. In this case, from Figure 6-2, the reactivity addition from 20% sodium voiding in the region i z \ 4 15 cm is 5 k= + .0034. The net change is therefore: knet ktotal rearrangement = 0.00565 - 0.00340 Na voiding = 0.00125.* Since the reactivity change given by this method is for motion over the entire radial area of the core; statistical weighting was invoked to obtain the reactivity effect for fuel motion over more reasonable *Note that this method does not violate the caution in Section 7.4 against indiscriminately adding and subtracting reactivity effects. Here the motion of fuel is superimposed on an existing degree of sodium voiding and, therefore, so is the resulting reactivity effect. U. 148 radial areas, namely 25% and 50% as employed in Chapter V. Selected rearrangements representative of the fuel injection mechanism, were chosen from Table 6-3 and steps (1) through (6) above were applied to calculate the data presented in Table 6-4. In this table, the appropriate RUN 3 from Table 6-3 is given in column 1; the radial core area affected is listed in column 2; and the total reactivity change which results when the rearrangement is complete is given in column 3. In column 4 the time required to complete the specified rearrangement is given. Note that in many cases this exceeds the 8 msec assumed to be available. Columns 5-7 give the reactivity insertion rates at time intervals of 2, 5 and 8 msec after clad failure in region 7R.. given are rounded off to the nearest $/sec. The values Pertinent obser- vations from the data in Table 6-4 include: (a) The RUNS with the fuel in (B) at T %0 corresponding to 10% of = 4600 0 K produce the highest reactivity addition rates in all cases. (b) RUNS 224, 225, 403 and 420 all involve inward fuel motion through a distance between 10 and 15 cm (see Table 6-3). RUNS 4o8, 421 and 422 involve inward motion between 7 and 10 cm. The latter RUNS produce substantially smaller reactivity insertion rates. RUN 421, in fact, results in a negative reactivity insertion. Table 6-4 REACTIVITY ADDITION RATES RESULTING FROM FUEL MOTION Fuel Density in (A) Increased by 20% Mean Fuel Temp. in Central 10% of (B) = 4 6000K Run Core Radial Area Affected 403 25% 50% 408 420 422 403 408 420 422 Max. Avail. 6k $1.04 $1.65 Time tl; Rearrangement is Complete (msec) 10.4 msec 10.4 Reactivity Insertion Rate After t = ($/sec) 2 msec 38 $/sec 61 5 msec 93 $/sec 149 8 msec 146 $/sec 232 -(1) 5.2 4 10 50% 30' 4.80' 5.2 7 17 25% 27s' 10.5 10 50% 430' 10.5 15 24 28 25% 5.3 50% 9.60' 15.5' 14 22 25% Fuel Density in (A) Increased by 2o% Mean Fuel Temp. in Central 20% of (B) = 44 000K 41 $/sec 18 $/sec $1.04 15.2 msec 50% $1.65 15.2 28 66 25% 30' 2 50% 4.80' 7.6 7.6 3 5 7 25% 27' 15.1 5 11 50% 430' 15.1 7 18 17 28 25% 9.60' 7.8 7.8 7 10 17 23 -(1) -(1) 25% 50% 15.5' 5.3 34 54 36 55 -(l) 68 $/sec 104 Table 6-4 (Continued) Fuel Density in (A) Increased by 50%' Mean Fuel Temp. in Central 10% of (B) = 4 6000K Run Core Radial Area A ffected 224 25% 50% 225 (2) 25% 50% 421 (2) 25% 50% 224 25% 50% 225 (2) 25% 50% 421 (2 25% tl; Max. Avail. 5k Time Rearrange- ment is Complete (msec) $1.77 $2.80 9.2 msec 741' $1.17 8.8 8.8 (-)7.2i -11.51 9.2 7.0 7.0 Reactivity Insertion Rate After t = ($/sec) 2 msec 8 msec 5 msec 81 $/s ec 130 39 61 (-)6 (-)9 180 $/sec 290 89 147 (-)14 280 $/sec 420 $/sec 142 223 -(1) - (1) (-)23 Fuel Density in (A) Increased by 50%' Mean Fuel Temp. in Central 20% of (B) = 4 4000K 42 $/sec[ 13.6 msec $1.77 102 $/sec 154 $/see $2.80 66 13.6 232 157 74' $1.17 12.7 18 42 12.7 29 71 (-)7.21 10.1 -3 68 102 -7 -9 10.1 -4 50% -10 -11.-5,1 -16 NOTES: 1) The new configuration is reached in less than 8 msec for these cases 2) In these two RUNS the complete rearrangement require zero fuel in part of region (B). In the calculation the fuel expansion was not allowed to proceed beyond 10% fuel in (B). No rates are cited for the complete rearrangement when t = t; hence the rates cited are valid. H4J - U 151 This observation is in agreement with the behavior discussed in Section 6.4 of the present chapter; namely, that slight inward motion toward presumably higher worth regions of the core can induce a shutdown effect and that for rearrangements where a positive reactivity is induced the effect is strongly ameliorated. (c) Only RUNS 224, 225 and 403 result in reactiv- ity addition rates greater than the 66 $/sec basis discussed in Chapter II. As noted and as seen from Table 6-3, the rearrangements postulated for these RUNS involve large axial regions of the core. In particular, the axial region affected,considering that identical events are assumed to occur on each side of the core axial centerline, is 50 cm, 40 cm and 60 cm for RUNS 224, 225 and 403 respectively. The axial motion postulated is assumed to occur over the entire radial region of interest (25% or 50%) simultaneously. For the perhaps more realistic axial fuel motions assumed in the remaining RUNS, the maximum reactivity addition rate is seen to be about 55 $/sec occurring after 8 msec in RUN 420 or 54 $/sec occurring after 5 msec in RUN 422. If the postulated region of fuel motion is further limited to 25% of the radial area of the core, these maximums are seen to be 36 $/sec and 152 34 $/sec respectively. (d) The time intervals required to produce high rates of reactivity additions are on the same order as those required to produce high rates of sodium voiding, as determined in Chapter V (see Table 5-1). Thus, the events following clad rupture cannot produce both the fuel injection mechanism considered here and the sodium voiding process studied in Chapter V. Specifically, if energy exchange between the fuel and sodium in region is sufficiently rapid to produce high sodium vapor pressures and rapid sodium voiding; the present process cannot occur. Note that the sodium vapor pressures in Table 5-1 are generally higher than the 600 psia maximum fuel vapor pressure employed in the present analysis. Conversely, if the fuel/sodium energy exchange rate in e is a relatively slow process, the present mechanism can be more significant than sodium voiding. 6.7 Summary Mechanisms which can lead to fuel motion into higher worth regions of the core have been discussed and the net reactivity change induced by a number of rearrangements which could result from such motion has been determined. The 'PS' Model for predicting the reactivity change arising from such core rearrangements or perturbations was developed and compared with ANISN multigroup computer calculations. The model employs one group perturbation theory in - U ~.- 153 conjunction with recent fast reactor spectral characterization work (45). It was shown to be quite accurate for small perturbations involving global regions of the core and reasonably accurate for large localized perturbations as severe as total sodium voiding from a region with simultaneous alterations in local fuel densities by as much as 20%. The model should prove useful in sodium voiding analyses. Use of the model is quite simple once the parameters in brackets in Eq. (6-12) are evaluated for a particular core. The importance of the parameters ( ) and (AS), the so- called density and spectral effects, in ameliorating the reactivity affect of various internal core rearrangements has been amply demonstrated. Finally, the reactivity addition rates which can take place as fuel motion occurs have been roughly estimated for the fuel-injection process described in Section 6.6. In general, the rates obtained are less than those estimated in Chapter V from sodium voiding which occur after a CATEGORY II excursion. If the analysis in Section 6.6 is limited to the more plausible cases or RUNS considered, the fuel injection process results in much lower reactivity addition rates than those estimated for the sodium voiding process. 154 Chapter VII DOPPLER EFFECTS AND MISCELLANEOUS REACTIVITY CONSIDERATIONS As discussed in Chapter II, Doppler broadening is the source of the major inherent negative feedback mechanism in As such, a great deal of attention has been given an LMFBR. to this important phenomenon. (13), (15), (18), (16), (19), References (7), (9), (30), (10), (47), (48) and (49) report the results of a number of the more recent investigations in this area. Equation (2-1) from Chapter II, which describes the temperature dependence of the Doppler reactivity insertion, is repeated here for convenience: dk ADOP( = ) (2-1) T (7-1) As stated earlier, a value of "n" near unity is generally employed in LMFBR analyses. Using this form of the equation, the importance of a Doppler coefficient, "ADOP", of about .003 or greater has been well established. (9)(13)(15)(22) (47) Doppler feedback of this magnitude or stronger has been shown to have a pronounced effect in reducing the energy released as a result of a given accident. Figure 7-1, taken from reference (22) shows typical examples of this behavior. The calculations on which Figure 7-1 is based assume that Doppler feedback is instantaneous, namely '55 AVAA~LF*QLrE ((I)cARWCAL F%)NC1\Or4 or- i00fPLER. RV~tKSG:(RnoN. AS \.oAV, rEt~ac-w A IFQO^ NC)RtnL ~,p~gfriN4-coN~r~r~s.(-o1 ~22,) to 40 bse qj 73 3e 0 0O00z? 156 that Doppler broadening of the absorption resonances of the appropriate isotopes, and the resulting increase in neutron capture, occurs just as rapidly as the integrated neutron flux level (or integrated power level) increases. Sections 7.1 and 7.2 of the present chapter deal with the validity of this assumption. Additional reactivity effects are treated, primarily qualitatively, in Sections 7.3 through 7.6. respectively: The influence of the parameter These include, "n" of Eq. (7-1); the significance of the heat of fusion of UO 2 in producing a Doppler "dead band".; the possibility of positive Doppler feedback during certain portions of an excurfinally, the reactivity effect induced by core sion; and, homogenization which can occur during a severe accident. 7.1 Delays in Doppler Feedback In Chapter III, Eq. (3-6) was derived from the point kinetics model to show the time history of reactor power during a ramp reactivity input with appropriate feedbacks. When only Doppler feedback is considered, this equation becomes: q = oexp [yt2 + 6kDOPdt' L_1 , (7-2) - 1 2U I - . - -- -- - - - -- - 157 where 7= a= reactivity ramp rate, A-= neutron generation time. Inserting Eq. (2-1) into Eq. (7-1) with n=l, one obtains: q(t) =oe [ APf A 2 ln(T ))dt' (7-3) An equation for q(t) as a function of time only can be readily obtained by assuming a constant specific heat over the temperature range of interest, so that: T(t) _ ) + T , (7-4) and by noting that: = Q(t) q(t')dt'. (7-5) The direct substitution of Eqs. (7-4) and (7-5) into (7-3) gives a double integral expression which can be written in differential form as: d dt 2Q.(t) + A ln(_ + 1) . (7-6) cPT0 This equation cannot be solved analytically in closed form. - Uw 158 Even the assumption of values of "n" near but not equal to unity, thus eliminating the logarithmic term from Eqs. (7-3) and (7-6), will not simplify the problem. tial Doppler feedback, $, If the differen- is written as a constant average value, however, Eq. (7-6) becomes much more manageable. This approximation takes the following form for the temperature range of interest: T dk (m) Ak DOP DOP A DOpln(7 =- T - T0 f where Tf = final or maximum temperature considered T = initial temperature. Employing a final temperature of 39000K and an initial temperature of 17000K in Eq. (7-7) gives an average Doppler feedback of: (d) =- D = - 1.5 x 10-6 5k/ 0K. (7-8) The final temperature employed is the mean fuel temperature at a location (normally near the core center) where the peak fuel temperature is sufficient to reach the threshold of clad failure. The initial temperature employed is the mean fuel temperature at the core center during steady state full power operation. With () = - , Eq. (7-3) can be written: 159 ft= qoexp[7t2 - b (7-9) Q(t')dt' where b= D max (q(t)) power peak valid until solution, an approximate and is reached, can be obtained. The result, derived in Appen- dix A, is: 1 q(t) 1 2 =q exp -7yt bq -7- 22 2 1 2 + (7xt2 )+-.-T- 2 12 ( 2t 3 + +0 (7-10) (valid only from time zero to the time of peak power, tp.) The reader should be advised that while the series in brackets can be shown to converge (by the ratio test, for example), it converges very slowly. For ramp rates in the 10 $/sec to 200 $/sec range, a minimum of seven terms is needed for reasonable accuracy (i.e. + 10%). The result expressed by Eq. (7-10) can be employed in a straightforward way in determining the effect of a delay in Doppler feedback, however, and an extremely useful result obtained. The time of peak power is determined by requiring = 0 and solving for tp. as shown in Appendix A: From Eq. (7-10) this gives, 160 1Yt 2 (2?tr ln( ) ) ) + ln(t 1 bq0 (7_11) , 7-1 2 where tp = time of peak power. For all cases of interest, it is found that >> (2) bq0 (1y 2) P and therefore an excellent approximation for Eq. (7-11) is: ) ( 7t = ln( ) + 1nn( 2) (7-12) . Suppose, now, that the time interval between the production of energy in the reactor and the introduction of reactivity from Doppler feedback can be represented by an "average" or "characteristic" time delay, C' Then for Eq. (7-9) . one can write: ~ t q t) = 7t2 - b The requirement that (-) 7y(t-'C)2 = ln( 2 2 p Q(t--' -oexp 2 bqo )dt (7-13) . 0 -0 ) + ln in this relation leads to: 7 2 L (t --C')2(l P _ J (7-14) where tp is the time of peak power when a delay is present. 161 ) is ignored for the moment in this , t -t equation, the result is identical to that of Eq. (7-11) If the quantity ( but with: tp = (t -V ) (7-15) is found to be on the order of 10-2 for Typically ( -5) t I-'c delay times Bf interest, and it will be shown that dropping this term from Eq.(7-14) has a negligible influence on the result obtained. Equation (7-15) can thus be used in Eq. (7-9) to determine the ratio of peak power with a delay to that without a delay. qpeak e with delay This substitution gives: (7-16) . qpeak without delay Presumably, a characteristic feedback delay which allows a significantly higher peak power is unacceptable; thus from Eq. (7-16) 7t-C << 1 is required. This condition can be rewritten, using Eq. (7-12), as: 1 52 FSqo <(7-17) + lnln( )y 0 on the quantity This relationship, dependence of 2( ) is seen to be extremely weak. Recall from (bq - Mu- 162 Chapter III that the power level at prompt criticality, q, is related to the steady state power level, qss, by: q = (7-18) 2&c a. qss (3-4) Substitution of this expression into (7-17) gives: 1 .707 L2 lc 3 1 3 1 -22 1 2 2 + lnln(l.77qss p ln( .77qs_ [ ~ l s PD ss _& D (7-19) Equation (7-19) then shows the extremely weak influence of initial power level and Doppler coefficient in determining whether a given delay is acceptable. (Doppler, of course, plays a major role in determining the peak power itself.) The reactor lifetime and reactivity insertion rate are seen to be the dominant parameters with respect to the affect of a given delay time. For the reactor of Table 1-1 with reactivity ramp rates of from 20 $/sec to 200 $/sec, the quantity in brackets in Eq. (7-19) was found to vary from 2.06 to 2.68; for an 'average" value of 2.37. (7-19) becomes: With this average value, Eq. U U w- 163 1 0. 3 (-) (7-20) In Eq. (7-16), if the more specific requirement is imposed that power overshoot with a delay not exceed 120% of the value without a delay, then 'ytV-' 0.182 is required. - p Equations (7-19) and (7-20) then become, respectively: .129(L) 3 ln(.77qss P _ L and s 1 3 1 ) + lnln(l.77qss P -2 2 ( +75 s c~(7-21) 1 '' < .0545 (-) (7-22) . This latter result is almost identical to that given in recently published work by Kohler (49); in which the limiting relation cited, 'Z:4 .05(A) 1/2, was based on the require- ment that energy release with a delay not exceed 110% of the energy release without a delay. Examination of Eqs. (4) and (5) of reference (49), however, show that a delay which leads to a 20% power overshoot (as employed here) is exactly equivalent to the 10% energy increase employed by Kohler. The parametric relationship, ()1/2, given in reference (49) was obtained by using approximate equivalency relationships for ramp and step induced excursions; and the coeffici- ent (.05) was determined by "comparing numerical solutions 164 for step and ramp induced super-prompt-critical excursion In the present analysis, Eq. (7-21) in fast reactors." was obtained by an analytical technique, and thus shows the influence of all parameters involved. For reactivity ramp rates in the range of interest, limiting values of the allowable delay time, C , from Eq. (7-22) (for a 20% power overshoot or, roughly, for a 10% increase in energy release) are given in Table 7-1. The time of peak power, tp, and the maximum reactivity inserted before Doppler "turn down" are also tabulated. Recall from Chapter II that the time of peak power as employed here is the time interval between prompt criticality and turn down of the transient ( - time tp of Figure 2-1). Note that 't varies from 13 to 120 microseconds and is less than 2% of t in all cases. I= 11 - Then the validity of the inequality is confirmed and Eq. (7-14) can be written t -c tp iR the form: 1Y ( t - ' ')2 2?bq = ln ( 2 Y ) + l n 0 1 7( t - ')2 _ Z' t p (7-23) Comparison of this result with Eq. (7-11) shows that the real test of the accuracy of the above analysis is the validityof the inequality: 165 Table 7-1 MAXIMUM ACCEPTABLE DELAYS IN DOPPLER FEEDBACK ('c'mx max) (2) (1) edmax cmax(msec) 5kmax t t p(msec) ~7t ( a $/sec 20 4.40 6.55 .120 .0183 66 16s 4.81 4.oo .070 .0175 100 23 5.76 3.40 .054 .0159 200 42d 7.00 2.04 .037 .0179 $1.10 9.90 1.40 .013 .0093 l000(3) NOTES: 1) The time of peak power for an excursion with the maximum acceptable delay is simply t' = t + 'm mx p p 2) The maximum acceptable delay is defined as that which allows a maximum increase of 10% in the energy release calculated with a zero delay (or, equivalently, a 20% increase in peak power, see text.) 3) The quantity in brackets in Eq. (7-21) is recalculated for a = 1000 $/sec and found to be 3.06. The 2.37 average value was employed in the 20 $/sec to 200 $/sec range. 166 1 ~2~ 2 2 4 . (7-24) p From Table 7-1 the left hand side of this equation is seen to vary from 4.2 x 10~ $/sec. at 20 $/sec to 9.8 x 10~4 at 1000 Clearly Eq. (7-24) is satisfied in the range of interest. 7.2 Sources of Delays in Doppler Feedback In Chapter II it was noted that while the fissile iso- tope is the primary source of fission energy in an LMFBR, it contributes a relatively small portion of the Doppler feedback. To obtain feedback of the magnitude expressed by Eq. (7-1) fission energy must be transferred to the fertile isotope. Additionally, neutrons from a given generation must slow to the energy range associated with the resonance region in order for increased resonance absorption to occur. Thus, mechanisms can be envisioned which can contribute to a delay in Doppler feedback. Those considered in the pres- ent study include: (1) The delay associated with the slowing down time of fission products. Energy for Doppler broaden- ing is not available until fission products of a given generation slow sufficiently to give up most of their kinetic energy. 167 (2) The delay associated with the time required for fission neutrons to slow from their velocity at birth to velocities at which capture by Doppler broadened resonances can occur. (3) The delay associated with heat transfer from the fissile to the fertile isotope. If the fuel con- stituents are mixed in powder form and the individual powder particles are larger in diameter than the mean range of fission products in the fuel ( ~ 10 microns), heat must be transferred to the fertile species primarily by conduction. Concern for the first two mechanisms can be dispelled quickly, with the aid of Table 7-1. The range of fission products in oxide fuel materials is about 10 microns (9). From Figure 6.1 of reference (50) and for the 10 micron fission product range, it is evident that fission products lose essentially all kinetic energy to the surrounding fuel material in 10-12 seconds or less. Clearly, from the values of 't cited in Table 7-1, this process has no influence in delaying Doppler feedback. The time required for an average fission neutron to slow down to the average velocity associated with the U-238 Doppler resonance region has been calculated to be about 2 x 10-6 seconds (49). While this is substantially longer 168 than the neutron generation time of 3.3 x 10~ seconds, it is again much shorter than the values of t in Table 7-1, even for ramp rates as high as 1000 $/sec. This this second mechanism can be ignored in considerations of Doppler feedback. With respect to the third mechanism, the thermal relaxation time for an average size particle of the fertile fuel material can be considered as a measure of the Doppler delay associated with conductive heat transfer. This assumes that the mean size of the fuel particles is substantially larger than the range of fission products in fuel and, as mentioned earlier, it ignores prompt neutron and gamma heating and the small fraction of fissions occurring in the fertile species. With these assumptions invoked, the characteristic heat transfer time needed has already been calculated. In Chapter IV (Eq. (4-8)) it was found that the relaxation time for a spherical U02 particle with no heat sources is given by: R 2 '= kL(~-) (7-25) (4-8) where R is the mean particle radius. For fuel particles in the size range of interest, this relationship gives the values of T" listed below: 169 Particle Diameter (Microns) 'C (Milliseconds) 10 4t .005 msec 20 .020 30 .045 4o .080 50 .125 100 .500 14o 1.00 Comparison of these results with the values of ' max given in Table 7-1 indicates that fuel powder particle sizes must be less than about 40 microns in diameter if the maximum credible reactivity insertion rate is taken to be 66 $/sec. If rates up to 1000 $/sec are credible, then the fuel particle sizes must be on the order of the range of fission products in fuel. These latter results corroborate a recent study of the heat transfer delay mechanism at General Electric, Sunnyvale, (48) in which machine calculations were employed to follow the time history of transients with and without feedback delays. In the General Electric study, the fissile/fertile fraction employed was 0.43 (vice the 0.15 used in the present work) and direct fission product heating of the UO 2 was assumed to be 14.5% of the total fission energy generated. Both differences tend to 170 reduce the effect of a given delay. Figure 1 of the General Electric study indicates that particle sizes of about 50 microns or less are acceptable for ramp rates up to 100 $/sec. In addition, the thermal relaxation times calculated from Eq. (7-25) and listed above corroborate time constants given in work by Braess et. al. (9). For example, reference (9) gives time constants of 0.13 msec and 0.53 msec for 60 and 110 micron particles respectively. These values are seen to compare quite favorably with those listed above. To illustrate the importance of these considerations, suppose a system is constructed with a mean fuel particle size of 140 microns (-t' = 1 msec and an excursion is initi- ated by an accident ramp rate of 100 $/sec.) For the LMFBR described in Table 1-1, Eq. (7-16) gives a peak power 35 times greater than the peak power with no delay in feedback. 7.3 Effect of Spectral Shifts (and the Parameter "n") on Doppler Feedback A great deal of attention in the literature has been given to the calculation of the Doppler coefficient, ADOP' of Eq. (7-1). Much less attention seems to have been focused on accurately determining "n" of Eq. (7-1), particularly as regards the behavior of n at high temperatures. Spectral hardening is known to result in an effective increase in n and thus in weaker Doppler feedback (18)(51). 171 This effect has been estimated to result in a change in n of from 0.8 to 1.05 for the spectral shift accompanying total sodium voiding (51). Some accident analyses have accounted for the change in Doppler feedback with sodium voiding by a change in ADOP or a change in To (13). For severe excursions, however, a change in n is of overriding importance. Table 7-2 shows the total Doppler feedback available for excursions resulting in maximum temperature of 6000 0 K and l0,000 0 K, with values of n between 0.8 and 1.2. The tabulated values of feedback reactivity were cal- culated using Eq. (7-1) with ADOP = - .004 and T 0 = 3000 0 K. Note that for a maximum temperature of 6000 0 K the total feedback available varies from - 14.5d for n = 1.2 to - $4.25 for n = 0.8. An increase in ADOP of 100% is negligible by comparison. For example, for the case of n = 1.2, increas- ing ADOP by a factor of two changes the total reactivity available to - 29' vice - 14.5' Thus the arbitrary extrapola- tion of Doppler behavior determined at low temperatures into the range of temperatures of interest in severe accident analyses must be given more attention. 7.4 Doppler Dead Band Due to Heat of Fusion The heat of fusion of UO 2 was given in Chapter IV as 278 j/gm and the melting temperature as 3070 0 K. From the time the fuel at a given location reaches 3070 0 K until melting at that location is complete, no Doppler broadening is expected to occur (26). As shown in Appendix A, this dead 172 Table 7-2 VARIATION OF THE MAGNITUDE OF DOPPLER FEEDBACK WITH THE PARAMETER "n" T max 5k n P dT T0 Tmax 60000 K .8 A - .014 - $4.25) - 851) 1.0 6000 0 K - 1.2 6ooo0 K - .00048 - 14.5 ) - .026 - $7.90) - .0048 - $1.45) - .00082 - 25e) .0028 00K .8 10,0000 K 1.0 10,0000 K 1.2 A DOP = -.oo4 T =0 .0033 T0 = 3000 0K band is reached at the core hot spot after an energy addition of about 56 j/gm to the fuel. Equation (4-4) of reference (8) can be used to calculate the time required to add the 278 j/gm necessary to heat the fuel at that location through the Doppler dead band. For an accident ramp rate of 100 $/sec, a time increment of 0.65 milliseconds is obtained. 173 (This is the time interval between the addition of 56 j/gm and 278 + 56 = 334 j/gm.) If this time increment is thought of as a delay in Doppler feedback for fuel at the location in question, comparison with the results of Table 7-1 shows 0.65 msec to be an unacceptable delay. Fortunately, not all the fuel material in the core is in this dead band at the same time. Additionally, the time required to add 278 j/gm depends strongly on the initial temperature of the fuel in question for a given reactivity insertion rate. For example, fuel material which is at 2600 0 K (vice the 2900 K employed above for fuel at the hot spot) is heated through the dead band in about .34 msec for the 100 $/sec reactivity insertion rate employed in the preceeding calculation. (While the reason for this behavior is not imme- diately apparent from a physical standpoint, it is readily apparent from Eq. (4-4) of reference (8)). The 278 j/gm "dead band" is roughly equivalent to a temperature rise of 700 0 K. Because of the relatively flat flux profile in a large LMFBR, a significant fraction of the fuel can be in this dead band during somephase of a severe accident. Additionally, as seen in Chapter IV, fuel dispersal leads to more uniform "whole core" fuel temperatures. Thus, if an initial excursion produces fuel disper- sal over a large region of the core, a substantially higher fraction of the fuel material could be in the dead band during a secondary excursion. 174 Finally, excursions which start from low power levels, and consequently more uniform radial temperatures in the fuel rods, will suffer more from the dead band affect. A combination of the latter two effects could conceivably place the bulk of the fuel material in a large LMFBR in the dead band at the start of a secondary excursion. 7.5 Positive Doppler Feedback from Fuel Cooling Recall from Chapters II and IV that the fuel temperature in regions outside lR remains essentially constant during the interval between an initial and a secondary transient. The fuel in 'R, however, will be cooling rapidly, as seen in Chapter IV. This cooling introduces a positive reactivity due to the Doppler effect which may have an important bearing on the rate of reactivity addition at the start of a secondary transient and thus on the cumulative severity of the overall accident. In Chapter II it was seen that fuel dispersal into mean fragment sizes of 500 microns ( -- 1/10 intact fuel pellet diameter) could result in a 700 0K reduction in the mean fuel temperature in region 'R in about 3.75 msec (a decrease from a mean temperature of 4400 0 K to about 3700 0 K). If the importance weighted effect of region -9 is 0.5 ( 69 extant over about 30% of the core), Eq. (7-1) shows that this cooling effect will result in reactivity additions of + ll for n = 1.0 and + 7611 for n = .8. If these 175 reactivity changes are assumed to be introduced linearly over the 3.75 msec cooling period, the resulting reactivity addition rates are 33 $/sec and 202 $/sec respectively. These rates, of course, are substantially increased for finer fuel dispersion and strongly decreased for a smaller degree of fuel fragmentation. 7.6 Homogeneity Effects Recent calculations at MIT (31) indicate that a gain in reactivity of about 70$' results from the core arrangement employed in the typical LMFBR considered here, as compared with a completely homogeneous system. The gain arises pri- marily from the increased "first flight" neutron flux within the fuel rods, although other effects contribute (31). Thus, some reactivity effect arises as a result of the homogenization which takes place during fuel dispersal. If the region 1R. extends over 30% of the core, a maximum negative reactivity insertion of about 351 could accompany complete mixing with sodium in region V. - -' 176 7.7 Summary A limiting expression for the effect of delays in Doppler feedback has been derived analytically and found to be in good agreement with a recently published result based on a more empirical analysis (49). The expression obtained shows the relationship of all parameters involved in determining an acceptable delay time. Of the possible delay mechanisms considered, only conductive heat transfer times have been found significant. The effect of this delay mechanism can be eliminated or minimized by chemical co-precipitation of the fuel materials or by mechanical processing which insures that fuel particle sizes are sufficiently small (see Table 7-1 and page and that the powders are uniformly mixed (52). Survey calculations have been applied to three additional effects, primarily in an effort to show whether such effects should be included in analyses of severe reactivity excursions. The need for more accurate knowledge of the magnitude of Doppler feedback, particularly at high temperatures (possibly by more accurate specification of "n" of Eq. (7-11)) is indicated by the results in Table 7-2. The influence of the Doppler dead band and of positive Doppler feedback has been indicated. These effects should be included in accident analyses involving a secondary transient in LMFBR's. 177 Homogenization effects in LMFBR's which can produce reactivity effects opposite to those encountered in thermal reactors, should be considered in core meltdown studies. The influence of a number of the effects considered in the present chapter depend strongly on the degree of fuel dispersal or fragmentation during the accident. The impor- tance of accurate, and presently unavailable, knowledge in this area is again indicated. 178 Chapter VIII SUMMARY AND CONCLUSIONS 8.1 General The progression of events during a severe excursion in an LMFBR has been considered in some detail. The importance of distinguishing between initial and "secondary" excursions has been established. (A "secondary" excursion was defined in Chapter II as one which occurs after, and possibly as a result of, substantial cladding failure in the core.) Nine phenomena affecting reactivity, primarily during secondary excursions, have been identified as pertinent and investigated. A summary of these investigations showing the influ- ence of the phenomena considered is given in Table 8-1. The following observations from the results shown in Table 8-1 are noteworthy. The letters below correspond to those in columns (a)-(f) of the table: (a) Most of the effects considered can become significant in a time interval of less than 10 milliseconds. As seen in Chapter II and Figure 2-1, the time interval between an initial and secondary transient, in the absence of pre-emptive reactivity effects, is typically about 10 milliseconds. Thus, sufficient time is available for the reactivity effects considered to exert their influence. 179 (b) Of the effects which can produce rapid reac- tivity insertions, all but one (item 2) depend on the degree of fuel fragmentation or dispersal in the region of clad failure. Item 2 (fuel injection from intact clad) depends, as noted, on the type of cladding failure and on frictional effects within intact clad. Both of the latter topics are areas in which present knowledge is inadequate. (c) The maximum total reactivity change avail- able from effects 1 through 4 is sufficient to produce a severe secondary accident if the total available change is added rapidly enough. As previously indica- ted, in the absence of external action an LMFBR remains near the prompt critical reactivity level after an initial prompt critical transient. With the system in this condition, if more than about 40og in reactivity is available, the rate of addition rather than the total change available is of greater importance. (d) At least three and possibly as many as five of the effects considered can produce reactivity insertion rates greater than the 66 $/sec "basis" discussed in Chapter II. (e) Of the six effects estimated to have a significant influence on a severe accident, all except the Doppler dead band effect (item 5) are of importance primarily during a secondary transient. 180 (f) Two effects (items 7 and 8) are concluded to be insignificant. Numerous calculations in the thesis, in fact, hinge on the conclusion that the energy spectrum is independent of the severity of the transient. More specifically, phenomena such as sodium voiding and fuel motion were found to have a much stronger influence on the energy spectrum than the transient itself. All other Doppler delay mechanisms were found to be of less importance than the delay associated with conductive heat transfer; and the latter was found to be insignificant for properly processed fuel (52). The importance of quality control in insuring that this effect is insignificant in mechanically mixed fuels is clearly indicated in Chapter VII, however. The need for accurate knowledge of the degree of fuel fragmentation under severe accident conditions is evident from Table 8-1 and paragraph (b) above. This is perhaps the most compelling conclusion of the present work. Recall from Chapter IV that the energy exchange rate between mixed fuel and sodium was shown to be proportional to the inverse square of fuel fragment diameter prior to the incidence of film boiling and roughly proportional to the inverse of fragment diameter after film boiling is dominant. For 181 refined analyses, accurate values of the Nusselt Number for the heat transfer process between fuel and sodium under accident conditions is needed. In view of the complex nature of this heat transfer problem, particularly with respect to the fact that environmental conditions (including fuel fragment size) can change rapidly during the course of an accident, it may not be possible to obtain an acceptable solution analytically. Experimentation with actual fuel clusters and sodium coolant in a severe overpower condition may well be mandatory. The primary argument in favor of the so-called "pancake geometry" is a reduction in the severity of the sodium voiding problem by promoting increased axial leakage. This necessarily increases axial buckling, however, which increases the severity of the fuel injection effect (item 2). Thus, at some point, pancaking may actually increase the susceptibility of a design to severe accident effects. While experimental work to date indicates that significant fuel motion within intact clad can take place during power excursions, (36)(46) much additional knowledge of the rates and extent of fuel motion is required before the effect can be incorporated in accident analyses. and clad are continuous So long as fuel rods (i.e. non-compartmented), as employ- ed in current designs, and clad rupture pressures are relatively high, this effect must be dealt with, however. In -U 182 particular, proposals for high strength clad designs to allow high fuel burnups should be examined in light of this phenomena since the reactivity addition rates estimated for this effect depend almost linearly on the internal clad failure pressure. It must be added, however, that the sodium voiding effect remains as the area of major concern with respect to LMFBR safety. The work in Chapters V and VI points inexorably to this conclusion. The reactivity addition rates estimated to result from fuel motion (Table 6-4) were smaller in general than those estimated to result from sodium voiding (Table 5-1). For cases involving more realistic assumptions with regard to fuel motion, the potential effect is much less severe than is found for sodium voiding. spectral The strong influence of the density ( A) P AS (-s-) and behavior in reducing the reactivity effects of fuel motion within the core is an essential factor in this observation. 8.2 Analytical Models Developed Two new analytical models were developed in the course of the present work which should prove useful in other applications and in future work in LMFBR accident analysis. In Chapter VI the tPS' model for predicting reactivity effects in fast reactors is derived. The model is based on 183 perturbation theory and recent spectral characterization work (45). It allows hand calculations of reactivity effects with a high degree of accuracy for moderate perturbations which extend over substantial portions of the core. Fairly accurate results are obtained even for strong localized perturbations. The model is useful in analyzing the various factors which contribute to a given reactivity change. For example, the core constituents or isotopes which provide the greatest contribution to a spectral shift can be identified. Furthermore, for a given perturbation, the contribution of the spectral shift to the overall reactivity charge can be determined. The PS model provides a relatively accurate method for calculating sodium void reactivities. More noteworthy, this is the first simple analytical method for calculating the sodium void effect which is derived directly from basic principles, and is not merely a curve-fit to multi-group calculations. In Chapter VII, an expression was developed for the maximum Doppler feedback delay acceptable under accident conditions. The expression, Eq. (7-19), is a limiting relation for Doppler delay in terms of well known core parameters and the accident reactivity ramp rate assumed. It is not limited to an LMFBR or, for that matter, to fast 184 reactors. It is, however, limited to situations where Dop- pler feedback over the temperature range of interest can be approximated by a constant value for ( 1) 8.3 . DOP Previous and Future Work in Accident Analysis Specific areas for future work have been enumerated in the preceding two sections. The present section comments on past work in fast reactor accident analysis and concludes with a general recommendation for future work. Eleven recent studies, which were reviewed extensively in developing the present work*, have considered severe transients in LMFBRs. Numerous earlier studies have consid- ered severe transients in fast reactors of various types, some including the LMFBR. In general the phenomena presen- ted here and reviewed in Table 8-1 have not been included in these works (item 7, delay in Doppler feedback, is an exception: In the present work, this effect is found to be insignificant.). A number of the recent works cited have taken "ADOP of Eq. (7-1) as a variable of primary interest. Several have compared the results of using various equations of state for the fuel material. The effect of zoning (two or more regions of different fuel enrichments) has been considered. Some of the studies have compared the results of employing different geometries. *References (7), (9), In cylindrical geometry, (10),(12), (13), (15-18), (24) and (47). 185 the effect of various L/D ratios has been investigated. In addition, the effect of varying enrichments and varying sodium/fuel volume fractions have been considered. the exception of the parameter these considerations "ADOp", With the influence of is relatively small. Typically, for a given quantity of fuel and a given reactivity ramp rate, the change in energy release resulting from varying the other parameters listed above has been less than about 40%. By contrast, several of the effects considered in the present work could alter the energy release by a substantial factor. A combination of these effects could completely alter the character of the accident. Unfortunately, the difficulty of formulating a precise description of these effects appears to increase with their possible importance to the accident sequence. Thus, in the past many phenomena of lesser importance may have been included more because they are more amenable to analysis than because of any inherent priority of importance. In general, no attempt has been made in previous works to deal with the secondary excursion as defined herein. Ultimately, the phenomena identified and found significant in the present work must be considered both singly and in combination. A related question which naturally arises with regard to the effects cited in the present study is whether any are of comparable significance in thermal reactors. The 186 next and final section is addressed to this topic. 8.4 Comparison with Thermal (Water Cooled) Reactors In the course of the present study each phenomena in Table 8-1 was also considered with respect to its influence in a water cooled (thermal) reactor. Most of the phenomena considered are expected to result in an increase in the severity of an LMFBR accident. By contrast, all nine effects are clearly insignificant or actually produce an opposite (shutdown) effect in water cooled reactors. In the case of item 1 (increased coolant voiding due to clad failure) "voiding" due to the formation of steam or radiolytic gas produces a substantial shutdown effect in water cooled reactors, primarily due to the corresponding increase in leakage and increased resonance capture (53). This difference in thermal and fast systems is being reduced however,by the use of soluble poisons in PWR's. The effect of fuel motion in large PWR's and BWR's (comparable in output to the 1000 MWe LMFBR considered) is substantially reduced by the much lower fuel "worth" in such reactors. The axial buckling is typically so small that fuel movements of a few centimeters produce a negligible reactivity change in comparison with such movements in an LMFBR (item 2, fuel injection from intact clad and item 3, fuel movement under a general inward pressure gradient). In reference (53) the rate of energy exchange between hot fuel and water under accident conditions was found to be 187 inherently limited by the physical properties of water. Thus, the rate is substantially reduced in comparison with the present sodium cooled system. This reduces the effect of item 4 (positive Doppler feedback due to fuel cooling) and tends to further reduce the significance of item 3 (fuel movement under a general inward pressure gradient). When fuel is dispersed or substantially fragmented, a large increase in the fuel surface to volume ratio takes place. In a thermal reactor, this strongly reduces the self-shielding effect and produces a corresponding increase in resonance capture. Thus, this "homogenization" provides an important shutdown mechanism for thermal reactors (53) but is a minor effect in an LMFBR (item 9). In the LWR reactor, the negative reactivity introduced by this reduction in self-shielding is, in fact, expected to override the positive Doppler effect which results from fuel cooling in the region of fuel dispersal (54), further ameliorating the effect of item 4 in a thermal reactor. Item 5 (Doppler Dead Band Effect) is expected to play a similar role in both reactor types. Its significance in a thermal reactor is considerably reduced, however, because of the less important role of Doppler in such reactors. As noted earlier, Doppler feedback is the only inherent negative feedback mechanism of consequence in an LMFBR. 188 The formation of steam or gas voids in a thermal reactor results in spectral hardening, as in the case for sodium voiding in an LMFBR. In a thermal system, the spectral shift is "toward" the resonance region, however, whereas the shift is "away" from the resonance region in an LMFBR. Thus, while sodium voiding results in a decrease in the magnitude of Doppler feedback in an LMFBR (item 6), the formation of steam bubbles in a water cooled system actually improves the Doppler shutdown effectiveness (53) by enhancing absorption in the resonance region. The problem of delays in Doppler feedback (item 7) is minimized in thermal reactors by the longer prompt neutron generation times of such systems. In Eq. (7-26), Vm max is 1/2 propotionl to , thus in the typical thermal system, to a_ proportional acceptable delays in Doppler feedback are about an order of magnitude longer. Additionally, in uranium fueled thermal reactors any heat transfer delay from the fissile to the fertile species is eliminated by the atomic scale mixing of the U2 3 5 and U2 3 8 isotopes. The effect of item 8 (transient induced spectral shift) is found to produce slight spectral hardening in both thermal (54) and fast (Chapter III) reactors. Although the overall influence is small in both cases, the effect is the same as that produced by "voiding" in both systems as discussed above. The thermal system benefits from the induced 189 spectral shifts, again in contrast to the effect in an LMFBR. Finally, in view of the results shown in column (c) of Table 8-1, the fact that the total reactivity change necessary to produce severe results in a thermal system is much larger than that required in a fast system, as discussed in Chapter III, must be reiterated. Rapid addition of several dollars of reactivity to a thermal system which is near prompt criticality is required to produce an excursion of similar severity to that which results from the rapid addition of about 40 critical level. to a fast system near the prompt Recall also from Chapter III that acci- dents which exceed about 40' in excess reactivity above prompt critical are extremely severe in an LMFBR. Note from column (a) of Table 8-1 that most of the maximum reactivity estimates given fall between these limits for thermal and fast systems. Thus, from this standpoint, the phenomena considered are of more crucial importancein an LMFBR than in thermal systems. Table 8-1 SUMMARY OF REACTIVITY EFFECTS CONSIDERED Effect Brief Description Pertinent Chapter a) When Significant b) Principal Parameter on Which Dependent(2) 1 Increased sodium voiding rate due to clad failure V 1-10 msec after initial transient Degree of fuel fragmentation 2 Fuel injection inZ from intact clad rods VI 2-8 msec after initial transient Type of clad failure; frictional effects w/in clad 3 Fuel movement under general inward pressure gradient VI 2-6 msec after start of secondary transient Time interval between initial and secondary transients 4 Positive Doppler feedback due to fuel cooling VII 1-10msec after initial transient "n" of Eq. (7-1) 5 Doppler dead band VII During initial Fraction of core fuel secondary transiin the 700 0 K dead ent. Likely more band at a given time important during secondary transient 6 Doppler reduction at high temps. and as result of' sodium voiding VII During secondary transient "n" of Eq. 7 Delay in Doppler feedback VII any transient fuel grain or powder particle size 8 Transient induced spectral shift III insignificant Homogenization effect VII 9 (7-1) 0 During secondary transient Size of region of fuel dispersal, Table 8-1 (Continued) c) Max. Estimated 6k Change d) Max. Estimated dk/dt e) Principally Secondary Transient Effect 1 1$ >-66 $/sec Yes Major 2 50 -2$ 66 $/sec Yes Possibly substantial 3 500-2$ Yes Not obtained Yes Possibly substantial Effect 4 501 but strongly dependent on "n not ascertained '- 66 $/sec 5 N.A. N.A. 6 N.A. N.A. 7 N.A. N.A. 8 negligible N.A. 9 (-) 30' Unknown but possibly quite high f) Estimated Overall Significance Substantial Likely More Important During Secondary Transient Yes To Slight Extent Substantial Insignificant if fuel co-precipitated minor if fuel properly processed mechanically No Insignificant Yes Possibly substantial NOTES: 1) N.A. implies Not Applicable 2) Effects (1) through (4) depend strongly on the size of the region of clad failure., R, as well as the parameters listed. H .- _400NOW4 . .............. 192 REFERENCES 1. WASH-illO, U.S. Atomic Energy Commission Liquid Metal Fast Breeder Reactor Program Plan, August 196d. 2. General Electric Co Liquid Metal Fast Breeder Reactor Design Study, GEAP-4 4 18, January 19b4 2A*. 3. 3A*. 4. General Electric Co., Summary Description of 1000 MWe LMFBR, G.E., Sunnyvale, November 1968. Westinghouse Electric Corp., Liquid Metal Fast Breeder Reactor Design Study, USAEC Report WCAP-3251-1, January 1964. Westinghouse Electric Corp., Summary Description of Conceptual Plant Design, November 9, 19b6. Combustion Engineering, Inc., Liquid Metal Fast Breeder Reactor Design Study, USAEC Report CEND-200, January 1964. 4A*. Combustion Engineering, Inc., 1000 MWe Liquid Metal Fast Breeder Reactor Follow-On Study, November 19b6. 5. Atomics International, Liquid Metal Fast Breeder Reactor Design Study, January 1964. 5A*. Atomics International, 1000 MWe LMFBR Follow-On Study, November 1968. 6. H.A. Bethe and J.H. Tait, "An Estimate of the Order of Magnitude of the Explosion When the Core of a Fast Reactor Collapses", British Report UKAEA-RHM (56)/113 (1956). *References (2A) through (4A) were presented at the Int. Conference on Sodium Technology and Large Fast Reactor Design, Argonne National Laboratory, ANL-7520, November 1968. 193 7. Nicholson, R.B., Methods for Determining the Energy Release in Hypothetical Fast Reactor Meltdown Accidents, NSE 18, 207-219 (1964). 8. McCarthy, W.J. and Okrent, D., The Technology of Nuclear Reactor Safety, Vol. I, Ch. 10, 1964, Ed: Thompson and Beckerley. 9. Braess, D. et. al., Improvements in Second Excursion Calculations, Proc. of Inter. Conf. on Safety of Fast Reactors, Aie-en-Province, Article 111-2, September 1967. 10. Sha, W.T. and Nicholson, R.B., Maximum Accident of Zoned Fast Reactors, ANL-7410, pp. 28b-2b9, January 19b9. 11. Okrent, D. et. al., AX-1, A Computing Program for Coupled Neutronic-Hydronamics Calculations, USAEC Report ANL-5977, (1959). 12. McFarlane, D.R. et. al., Theoretical Studies of the Response of Fast Reactors During Sodium Boiling Accidents, ANL-7310, pp. 310-317, January 1966. 13. Renard, A. and Stievernart, M., Evaluation of the Energy Release in Case of a Severe Accident for a Fast Reactor with Important Feedbacks, Proc. of Int. Conf. on the Safety of Fast Reactors, Aix-en-Province, pp. III-1-1 to 11, September 1967. 14. Carter, J.C., The Phenomenology of Fast Reactor Accidents, ANL-7410, pp. 290-292, January 19b9. 15. Oyama, A., et. al., Analysis of Fast Reactor Core Meltdown Accidents, Univ. of Tokyo, Proc. of Int. Conf. on Safety of Fast Reactors, Aix-en-Province, pp. 111-4-1 to 15, September 1967. 16. Hicks, E.P. and Menzies, D.C., Theoretical Studies on the Fast Reactor Maximum Accident, Dounreay Exp. Reactor Establishment; Thurso, Caithness, Scotland, ANL-7120, pp. 654-670, October 1965. 17. Meyer, R.A. et. al., A Parameter Study of Large Fast Reactor Meltdown Accidents, General Electric Co., ibid, pp. 671-bb5. 18. Storrer, F. et. al., Quasi-Static Model for the Analysis of Reactivity Accidents in Fast Neutron Reactors, Proc. of Int. Conf. on Safety of Fast Reactors, Aix-en-Province, pp. 111-5-1 to 10 (1967). 194 19. Hafele, VonWolf, Prompt Uberkritische Leistungsekwisionen in Schnellen Reaktores, Nucleonik, 5, Band 5, Heft, pp. 201-20d (1963). 20. Swift, D.L. and Baker, L., Experimental Studies of the High Temperature Interaction of Fuel and Cladding Materials with Liquid Sodium, ANL-7120, p. b39, October 1965. 21. Sanathanan, C.K. and Carter, J.C., Phenomena Leading to Fuel Casing Rupture During Transients, ANL-7410, pp. 293-304, January 1969. 22. Okrent, D., Design and Safety in Large Fast Power Reactors, review article prepared for publication in Atomic Energy Review. To be published. 23. Aronstein, R.E. et. al., Summary Description of Reference Plant, Proc. of Conference on Sodium Technology and Large Fast Reactor Design, ANL-7520, November 1968. 24. Noyes, R.C. et. al., Parametric Studies and Core Performance, ibid. 25. Harde, R., Design Considerations and Experimental Program for the Common Development of a 300 MWe Sodium Cooled Fast Breeder Prototype SNR by a Belgium-Dutch, German Consortum, ibid. 26. Schenter, G.E. and Gibbs, A.G., Binding Effects on Reactivity Change for Transitions from Solid to Liquid Phase in a Fast Reactor System, BNWL-717, April 19b6. 27. COO-279, An Evaluation of Four Design Studies of a 1000 MWe Ceramic Fueled Fast Breeder Reactor, Chicago Operations Office, USAEC, December 1964. 28. Hwang, R.N. and Ott, K., Comparison and Analysis of Theoretical Doppler Coefficient Results for Fast Reactors, ANL-7269 (19b6). 29. Nicholson, R.B. and Fischer, E.A., The Doppler Effect in Fast Reactors, Advances in Nuclear Science and Technology, Academic Press, New York (1968). 30. Petersen, R.F., Goldsmith, S., Fast Reactor Safety Considerations Related to Fuel Macro-Structure, GNWLSA-346, October 1965. 195 31. Westlake, W.J., Heterogeneous Effects in LMFBR Fuel Elements, M.S. Thesis, MIT, to be published. 32. Barghusen, J.J. et. al., Behavior of Zircaloy Clad UO0 Fuel During Nuclear Transients in TREAT, TANS 12-2, November 19b9. 33. Meyer, R.A. and Wolfe, B., High Temperature Equation of State of UO2 , TANS, Vol. 7, No. 1, p. 111, June 19b4. 34. Etherington, H., Ed. Nuclear Engineering Handbook, p. 11-7 (1958). 35. Golden, G.H. and Tokar, J.V., Thermophysical Properties of Sodium, ANL-7323, August 1967. 36. Dickerman, C.E. et. al., First TREAT Loop Experiment on Oxide Fuel Meltdown, TANS, 12-2, November 1969. 37. Judd, A.M., Sodium Boiling and Fast Reactor Safety Analysis, UKAEA, AEEW-R561 (1967). 38. Hatsopoulos and Keenan, Principles of General Thermodynamics, Wiley and Sons, New York (1964). 39. Descamps, C. et. al., Analysis of Thermal and Hydraulic Behavior of Sodium During a Prompt Power Insertion in a Fuel Assembly, Belgonucleaire, Bruxelles, Proc. of Inter. Conf. on Safety of Fast Reactors, Aix-enProvince, pp. 1-3-11 to 18, September 1967. 40. LeGonidec, B. et. al., Experimental Studies on Sodium Boiling, ibid, Article 1-3. 41. Rohsenow, W.M. and Choi, H., Heat, Mass, and Momentum Transfer, Prentice Hall Inc. (1961). 42. Vance, R.W., Ed. Cryogenic Technology, Wiley and Sons, New York (1963). 43. Horst, K.M., Fast Oxide Breeder-Stress Considerations in Fuel Rod Design, GEAP-3347. 44. Menzies, D.C., The Equation of State of Uranium Dioxide at High Temperatures and Pressures, TRG Report 1119(D), February 1966. U U 196 45. Sheaffer, M.K. and Driscoll, M.J., A One Group Method for Fast Reactor Calculations, MITNE-lO and Sheaffer, M.K., PhD thesis (MIT) to be published. 46. Hikido, T., Field, J.H., Molten Fuel Movement in Transient Overpower Tests of Irradiated Oxide Fuel, TANS 12-2, November 1969. 47. Lord, R.M., Effect of Core Configuration on the Explosive Yields from Large Sodium Cooled Fast Reactors, UKAEA, Int. Conf. on Safety of Fast Reactors, Aix-enProvince, Art. 111-3, September 1967. 48. Bailey, H.S. et. al., Effect of Fast Reactor Fuel Homogeneity on Transient Behavior, TANS 12-2 (G.E. Sunnyvale), November 1969. 49. Kohler, W.H., The Effect of Short Delay Times in SuperPrompt-Critical Excursions in Fast Reactors, TANS 12-2, November 1969. 50. Evans, R.D., The Atomic Nucleus, McGraw-Hill (1955). 51. Till, C.E. et. al., Doppler Coefficient Temperature Dependence and the Effect of Sodium Voiding, TANS 10(1), p. 335, June 1967. 52. HW-81603 Quarterly Report, October-December 1964, Hanford Atomic Products Operation, Richland, Washington. 53. Elbaum, G.J., Rapid Excursions in Water Reactors Involving Fuel Element Rupture, PhD Thesis, MIT, June 1967. 54. Buckman, F., Severe Reactivity Transients in Boiling Water Reactors, ScD Thesis, MIT, February 1970. 56. ANISN/DTF II Conversion to the IBM System/360 performed by Atomics International, AI-66, Memo 171, undated. 57. ANL-5800, Reactor Physics Constants, July 1963. 58. et. al. (ABBN) Group Constants for Bondarenko, I.I. Nuclear Reactor Calculations, pub. Consultants Bureau, New York (1964). 59. Agrawal, A.K., On the Analysis of Fuel Meltdown Studies with TREAT, TANS 12-2, November 19b9. 197 60. Comment based on private communication with Professor W.M. Rohsenow of MIT, co-author of reference (41). The two heat transfer correlations (from references (41) and (42)) employed in Section IV.2 were suggested by Professor Rohsenow. 61. Fauske, H.K., Two-Phase Compressibility in Liquid Metal Systems, ANL, Int. Conf. on the Safety of Fast Reactors, Aix-en-Province, Art. IVa-l, September 1967. 62. Horst, K.M., Fast Oxide Breeder - Stress Considerations in Fuel Rod Design, GEAP-3347, March 19b0. 198 APPENDIX A 1. Power History Calculations An approximate solution to Eq. (7-9), valid until the time peak power is reached, is derived. Equation (7-9), repeated here for convenience is: 1 q(t) =oexp 7t 2 - b (A-1) Q(tl)dt'J where b= cp a = accident ramp rate in 5k/sec = neutron generation time D= Doppler feedback defined in Eq. (7-8). As seen in Table 7-1, the inequality yt 2>> times of interest. 1 holds for all For this condition, Eq. (4-2) et. seq. of reference (8) gives the result for power level growth with a ramp reactivity input and no feedback. It is found that a good approximation for power level behavior with feedback up until the time of peak power can be obtained from Eq. (A-1) by employing an expression for power level without feedback in the integral term of Eq. (A-1). This is analogous to approximations made in references (7) and (8). term of Eq. (A-1) thus becomes: The integral 199 t 0Q(t')dt' b O q(t' = b )dt t b t'= g qee p(} '2 )dt. The first integral of this result has been evaluated in series form (8). A good approximation for times when 17t2 obtained by taking the first term of the series. t b valid for This gives: t (t I ) dt ' ~'--bqo b fQ 1 is 7-t2>> 1. A2 1p}7 '2)d O The desired expression for the condi- tion of peak power can now be obtained. By requiring d = 0 in Eq. (A-1) one obtains the condition: d _ 2 r o A Fp -7t 0 2) (t , dt' = 0 0, or 7 t2 -exp(yt ) (A-3) p Rearranging and employing t for the time of peak power, as done in Chapter VII, Eq. (A-3) becomes: (l7t) = + ln(7t ) , ln Dq0 identical with Eq. (7-11), as desired. (A-4) 200 Equation (A-2) can be integrated analytically by substituting the appropriate series for the exponential term (i.e.: el/27t2 = 1 + 1 (/2Yt 2) 2 and integrating term by term. + ... ), dividing by 7t as shown, This approach, while not ex- plicit, gives a usable result as follows: 2 2 1 q(t) = qoexp t ( o 1 2 27 274 + 2.2. 3 + 3.31 ) + (A-5) valid for ( 2 )> *t1 and only until the time of peak power. If the power burst is assumed to be symmetric about the time of peak power, as in references (18) and (19), Eq. (A-5) can be employed to trace the power history and thus determine the energy release for one "cycle" of an excursion. Equation (A-5) has been graphically integrated for several cases of interest and found to be in good agreement with the QuasiStatic Model presented in reference (19). Figure 2-1, in fact, is a plot of Eq. (A-5) for a reactivity insertion rate of 66 $/sec and using the averaged parameters for fuel properties given in Table A-1. In the Quasi-Static Model (19), a method for calculating the time of the start of a power rise (the quantity At1 of Figures 2-1 and 2-2) is given. A method for calculating the power pulse width is not given, however, and this is cited as a deficiency in the QuasiStatic model. From Figure 2-1 or 2-2, clearly if the time of peak power, tP, is known and At1 is known, the pulse width 201 is determined. A method for calculating t was C given in p Chapter VII of the present work. On rearrangement of Eq. 14 of reference (19) one obtains: 1 At = (A-6) P) 2ln(c On rearranging Eq. (7-12) of the present work there results: 1 2 tp=S~) 2c in _ F) 2 ac nn(A-7) Dq 0 D-q Thus the parametric dependence of tp and At are identical but tp is larger due to the factor 2 in the "ln" term and the additional "ln ln" term. For the example of Figure 2-1, t is 4 msec and At 1 is 3.05 msec. As can be seen from the figure, At 1 is the time at which the power level reaches the asymptotic power level, qAS, a quantity representing a mean of the oscillating level, and which is also the value to which the power level trends as the oscillations are damped (19). In each of the several cases investigated, good correlation between At1 (calculated from the Quasi-Static vJA3 o6+Aa,vec1. Model) and tp (calculated from Eq. (A-7),(see, for example, Figure 2-1). This provides further substantiation of the Doppler delay analysis, Eq. (7-20) et. seq., of Chapter VII. The Quasi-Static Model, with calculations of tp by the methods of the present analysis, was employed to estimate the behavior of a number of transients of interest for the LMFBR of Table 1-1 and, consequently, to form the basis for 202 some of the qualitative reasoning, particularly in Chapters II and III, of the present work. The results of calculations employing this method are presented in Table A-1. The quan- tities qAS and 5kmax were computed directly from the appropriate equations in reference (19), namely: ac qAS= and ac P+ 5k =qf-al?2 ln max Dq0 The quantity t 21n was computed from Eq. (A-7) above. The total fuel temperature rise associated with one power burst, AT.., was calculated from Eq. (13) of reference (19) and from the observation (Figure 2 of reference (19)) that the WT given by Eq. (13) is linearly related to the temperature rise in the interval required for one power cycle; namely 2tP as shown in Figure 2-1. Finally, the time to clad failure is estimated by employing the asymptotic power level to calculate the time required to reach the threshold of clad failure. As shown in Chapter III, the peak fuel temperature at clad failure is about 4850 0 K. For fuel initially at room temperature (293 0 K, internal energy taken as zero), the energy addition required to reach the clad failure threshold can be estimated as follows: 203 AT = .30(3070-293) (to melting n c p3(00C point) AQ(heat of = Ah j/gm = 831 = 278 j/gm - fusion) AQ(melting point = cPLAT = .42(4850-3070) = 747 j/gm = 1855 i/gm to clad failure) AQtotal (A-8) As noted in Chapter II, this calculated value is in good agreement with an experimentally observed value of 1900 j/gm. For 100% reactor power, the initial fuel temperature in the center of fuel rods near the core center is 29000K (see Table 1-1). Thus, the energy addition, starting from this power level, is 1080 j/gm, as is readily verified from Eq. (A-8). Similarly, from a steady state power level of 10%, the required addition is estimated to be 1590 j/gm. These values, then, were used in conjunction with the tabulated values of qAS to estimate the time to clad failure, tF, in Table A-i. The reader must be cautioned that while the results of Table A-1 are useful for purposes of discussion and in discerning a number of important behavioral characteristics of the LMFBR excursions of interest, the uniform cyclic behavior on which the results are based cannot be expected to persist for more than a few cycles. It is pointed out in reference (19) that 204 even in the absence of pre-emptive effects, each successive power oscillation will be damped. The present work, of course, also discusses a number of potential pre-emptive effects. f Table A-i APPROXIMATE RESULTS FOR VARIOUS REACTIVITY INSERTION RATES c' ($/sec) 10 $/sec 66 $/sec 20 $/sec 100 $/sec 200 $/sec 1000 $/sec full 10 100 24 16.2 13.1 AT (OK) 410 0 720' qAS (j/gm msec) 6.5 6.5 13 tF (msec) 240 166 122 kmax (g) At1 (msec) 7s' s s(power) (2tP) msec 10 100 10' 10 100 10 10 100 100 10 100 10 8 8.1 6.8 6.2 4.1 3.1 2.8 13500 9500 19000 13400 28oo0 2500' 75000 60000 13 40 4o 66 66 132 132 660 660 83 39.7 27 24 16.3 12 8.2 2.4 1.64 23U' 165 23,1 48g' 421' $1.35 $1.10 3.05 3.25 2.90 (2t ) = time interval between successive power pulses as shown in Figure 2-1. AT = q AS= tF total fuel temperature rise resulting from one power pulse. "asymptotic" power level. = time from prompt criticality until clad failure. -1. This value is the same for each uniform power pulse for a given initial = k 5k m p39er and reactivity insertion rate. max R) 0 206 Table A-1 (Continued) Average properties used in calculating the results tabulated were: D = 1.5 x 10-6 5k/sec = 0.3 joules /gmK C = 150 j/gm sec of fuel at 100% power qss = 15 j/gm sec of fuel at 10% power '= 3.3 x 10~ sec = .0033 Notes: (1) D is based on the average value given by Eq. (7-7) between the mean fuel temperature at steady state and that at clad failure. Tf = T = 12000 K 47000oK These values are: ~at 10% power Tf = 3900 K f= T0 = 17000K at 100% power Purely by coincidence Eq. (7-7) gives the same D for both calculations. (2) The missing data in the table corresponds to conditions at which the Quasi-Static Model is not applicable. 207 APPENDIX B 1. Heat Transfer Correlations Employed The heat transfer correlations employed for the film boiling conditions described in Chapter IV were taken from references (41) and (42). correlations.* Both equations are experimental These are: 1 h fb = 2 .7 V Kf pq (h fg+ o. 4c 0 D 0 ATf f f 0 AT) AT ) where (B-1) (Eq. 9.30 of reference 41) V0o t 2 TJgD , and hf =0.14 fb p(p0-p.)g cPL 2 -Po L42 f h )(1 + 0.5 p AT0 c AT h ) fg K , (B-2) (Eq. 5.64 of reference 42) where the subscripts .9 and f refer to the liquid and vapor states respectively and the sodium properties used in evaluating hfb in each case were taken from reference (35) at a "mean" temperature of 1760OR: *The use of these two equations was recommended by Professor W.M. Rohsenow (MIT). - im 208 P1 = 43 lb/ft 3 [f Pf = 0.05 lb/ft 3 = 0.05 lbm/ft-hr Kf = 0.04 BTU/ D = 2R0 = fuel particle diameter cp = 0.28 BTU/lbm0 F 4 = 0.34 lbm/ft-hr AT = 10000 F cp = 0.32 BTU/lbm0 F K = 26 BTU/hr-ft0 F g = 32 ft/sec = 1600 BTU/lb h In Eq. (B-1) a value of V corresponding to the normal flow of sodium through the core is assumed ( -'10 m/sec). The results obtained for hfb were employed in Chapter IV in the hfbD . relationship m = 2K Equation (B-1) gives m = 2 for fuel D = 250 microns and Eq. (B-2) gives m ~ 12 for D = 25 microns. This substantiates the conclusion in Chapter IV that m > 1 for D > 250 microns. 2. Supplemental Sodium Voiding Analysis Figure 5-1 is a sketch of the core and channel model employed here and in Chapter V for a hydrodynamic analysis of the sodium voiding process. From Figure 5-1, P(t) is taken as the pressure in the vapor bubble as a function of time and is assumed to be uniform throughout the vapor bubble. In the present analysis, the axial extent of the vapor bubble never exceeds z = 30 cm. As stated, p 0 is the pressure at the core exit (top) and is assumed to be constant. The reasonable nature of this assumption can be MMWA 209 justified as follows. Using the General Electric LMFBR design as an example (2A), the primary system sodium volume is 45,300 ft 3 , about 23,000 ft 3 of which is in the reactor vessel (see Table 1 and Figure 4 of reference 2A). The bulk of the reactor vessel volume in typical current designs is the region containing a "pool" of sodium above the core and the cover gas at the top of the vessel (see Figure 5-1). The volume of this pool above the core is assumed to be 15,000 ft 3 ( -- 420 m 3 ) for purposes of the present analysis. From Table 5-1 the maximum size of the sodium void generated in the present analysis is about 0.3 m. The velocity of sound in liquid sodium at the core exit temperature is 2250 m/sec (35). Thus, if 3 msec (a time interval typical of the values calculated in Table 5-1) is allowed for generation of the 0.3 m3 vapor bubble, a compression wave originating at the core exit can affect sodium at a radial distance of 6.7 m, or essentially all sodium in the "pool" above the core. The constant temperature compressibility of liquid sodium at the core exit temperature is PT AV (35): 2.7 x 10-5 -/atm (B-3) For the present analysis, we have AV hence Eq. (B-3) gives AP = 26 atm. 0.3 m3 and V = 420 m3 Notice that this pressure is relatively low compared to the pressures generated inside the core vapor bubble. Thus, even without access to the 210 cover gas region, compression of the large volume of liquid sodium at the core exit tends to relieve the pressure buildup in this region. The present analysis is intended only to show that the assumption of a low constant pressure at the core exit (top) is not unreasonable. To compensate for errors introduced by this assumption, only upward flow of sodium from the core was allowed. Then, from Figure 5-1, with p0 as the core exit pressure, the applicable hydrodynamics equation, in its simplest form, is: (t) in vapor bubble APfriction + APinertial + po (B-4) . The appropriate relations for Apfriction and Apinertial are given in Chapter V and lead immediately to the result: Bexp(T) (t)) dz 1.8 d2 d22 + Ap (.64)(V-) + p0 dt2 . (B-5) - 1) (B-6) 000 The relation: Q(t) = FP(Tt)-TO) + B( A)exp(- T was obtained in Chapter V. )'() It is found that an approximate solution to these equations can readily be obtained by hand calculations according to the following steps. Q(t) employed are given in Chapter V): (Values of 211 At t=O, z(t) = Zo, (1) hence since Q(0 ) is specified (in Chapter V) T is determined from Eq. (B-6). (2) This value of T d z ( with (-) (3) (4) is used in Eq. (B-5) to obtain = 0 at t=0). 2 d d2 Using this value of d dand z are deterdtz T dt d2 2 Z(t) aedtr mined at time t1 by taking d as a constant dt over the next time step. Steps (l)-(3) are then repeated until the times shown in Table 5-1 are reached. The procedure is simplified considerably by the fact 2 that dz) does not change appreciably from its initial dt value for the time scale of events considered in the present analysis. of Tt, (t) As a check on the method of solution, the values z, , and z - \~)-'d dt2 obtained at the times cited in Table 5-1 were substituted in Eqs. (B-5) and (B-6) and corrections made to the iterative procedure, if necessary, to achieve + 10% accuracy in d dt 212 APPENDIX C 1. Spectral Effect Calculations The method for calculating the spectral parameters "S" and "R" is given in reference (45). A sample calculation for the LMFBR of Table 1-1 is given below. Equations (6-1) and (6-2) are repeated here for convenience: 1) tr S =V (C-2) tr S R = 1-S7 (6-2) Step 1: (45) Estimate S and calculate R. S is shown in reference to be in the 0.3 to 0.5 range for oxide fueled LMFBR's. An initial estimate of Sl = 0.4 is chosen. As will be seen, the initial choice is not too important. On simple substitution of the values from Table C-1: ZR = 1 Ni 5R. = 0.03652 1 With S1 = 0.4, Vtr = : * tr i~ tr is (C-3) given by: by: N, (o. 4) = 0.012042 (C-4) from Eq. (C-2) Thus, R = 0.2126. 1-1-S1 ( 067)=0 )012042 (C-5) 213 Step 2: With the estimated S1 and R 1 from Eq. (C-5), calcu- late Yf : TZ f= (vN a 0S1 D) for + (vN R (C-6) ) for U28 f = 0.0067115 is obtained. Substituting from Table (C-1), Thus, from Eq. (C-1) S 0+.067115 . 2 = 0.0067115 + 0.0120T2 Step 3: estimate. Repeat steps (1) = and (2) This will give S3 = 0.3578 . (C-7) using S2 as the initial 0.354 and R3 = 0.1855. As noted in reference (45), only one iteration is required for reasonable initial estimates of S. At most, two iterations are necessary. The final values of S and R thus obtained are then used to calculate cross sections for the core of interest. For the present LMFBR the values obtained are: Macroscopic Cross Sections Element vzf a tr 0.005709 0.00241 0.00854 0.000937 0.00261 0.05400 0 3.7 x 10-5 0.04825 Na 2.4 x l0-5 0.04893 Fe 10.8 x 10-5 0.04814 214 For calculations of S', for use in the quantity in the PS Model, the following simplified technique can be used. For total sodium voiding, for example, the major parameter involved in the change is Ztr . The re- Ztr' , can be obtained by subtracting quired new value, as given by Eq. (C-4). z Na from The value ob- tained is: z tr Ftr ~ N7trNa = 0.008014 Thus: S= = v2;:+ 0.453 (Y and AS S .099 0.283 o7.3547 AS .099 = 0.248. If the iterative procedure given by Steps (1) through (3) above is repeated, the value of S' calculated is less than 4% larger. This is readily seen to be small compared to the uncertainty generated by the difference in the quantities AS and AS Recall from Chapter VI that total sodium voiding is an extremely strong spectral perturbation compared to the other cases of interest. For the less severe cases, 215 the error introduced in S' by not repeating the iterative procedure is even smaller, including those cases where fuel rearrangement is involved. Thus, once S is calculated for a given core composiAS AS tion, AS and the quantity -5- (or -- for estimating the influ- S ence of large perturbations) can be quickly obtained for perturbations of interest. -4 Table C-1 CROSS SECTIONS AND SPECTRAL PARAMETERS FOR 'PS' MODEL (taken from reference 45) Cross Sections and "g" Values Number 24 Density(xo~ ) (x102 4 )(cm2) Absorption Fission Transport Go g Low mean energy decrement Removal cross section R Element N(-/cm3 ) PU49 0.0010 1.713 -. 1472 1.813 -0.3067 6.661 -0.2725 0.0083 3.040 U28 0.0065 0.1972 0.7911 0.2147 -o.6249 6.331 -0.280 o.oo84 2.236 0 0.0150 0.002468 o.6728 3.307 -0.0269 0.120 0.1859 Na 0.0127 0.00064 -1.073 3.273 -0.1585 0.0845 0.3875 Fe 0.0127 0.00490 -0.5456 2.829 -0.2844 0.0353 0.6584 g g Notes 1) For Pu 9, v-= 2.953. For U , y = 2.806. N)