ARGONNE CODE CENTER
MASTER COPY
REFERENCE.
ABSTRACT_22?_ ___
NUCLEAR SCIENCE AND ENGINEERING: 23, 58-73 (1965)
Criticality Calculations for Uniform Water-Moderated
L. E. Strawbridge and R. F. Barry
Westinghouse Electric Corporation, Atomic Power Division, Box 355, Pittsburgh 30, Posylvanio
Received August 7, 1964
Revised February 11, 1965
‘A procedure for calculating the neutron multiplication factor and few -group
constants for water-moderated reactors has beon developed. The intent of this
development was to produce # calculational procedure which could be used with
7 ‘confidence in the engineering design of water-moderated reactors. Analytio pro~
‘edures that require large amounts of computer time were excluded from consid
eration.
‘The proposed procedure includes a heterogeneous treatment of the fast-fission
effect, resonance absorptions and the thermal utilization, The results of the pro~
edure agree well with more rigorous calculations requiring orders of magnitude
more computer time, ‘The procedures have been applied to 116 U-motal and UO2
‘experimental lattices covering a wide range of parameters. ‘The multiplication
factor is
INTRODUCTION
‘The engineering design of a nuclear reacior
core requires hundreds and even thousands of
criticality calculations. The neutron multiplication
factor (#) Is undoubtedly the most frequently
calculated parameter in the physics design of a
reactor core. Regardless of whether one Is
studying fuel depletion effects, reactivity coetfi-
cients, methods of control, or analyzing critical
experiments, the neutron multiplication factor is
of utmost importance.
‘There are probably at least as many techniques
for calculating few-group constants and multipli-
cation factors as there are physics design groups
in the industry. The methods that have been
proposed range from techniques that are amenable
to hand calculations to those that require many
hours of computational time on a high-speed
computer. Both of these extremeg may serve a
useful purpose In the Industry, For example, it
might be argued tat the hand calculations give the
user a better uncerstanding of the physical phe-
nomena involved. The extremely complex and
expensive techniyues may be justified, if the
This work was performed under Contract AT(20-1)~
3269 for the Atomic Energy Commission.
‘oulated for all cases with a standard deviation about the mean of 0.86%.
results’are accurate enough to negate the need for
a critical experiment; thus the expensive calcula~
tion might be used as a critical experiment’ is
used — as 2 basis for developing analytic tech-
niques,
For engineering design calculations, neither of
the extreme methods is satisfactory. The hand
calculations must ordinarily be simplified to the
point where many of the physical phenomena can-
not be adequately represented. Empiricism is
strongly relled upon and the success of the method
often depends upon the cancellation of errors.
With a method of this type, It is extremely
Important that the method be used only within the
range of variables for which it was empirically
developed. On the other hand, the most refined
caleulational methods are not Sultable for engi-
neering design work because of the large compu-
tational expense involved, Thus, a strong necd
exists for a method that is between these ex-
tremes. The work described here is an attempt to
fulfill this need.
‘The method that has been developed is believed
to be orders of magnitude better than hand calcu-
lations in terms of accurate representation of the
physical phenomena involved. The computational
time is orders of magnitude less than the most
detailed calculations (computational time on theCRITICALITY CALCULATIONS
order of 20 or 30 seu on the IBM-7094), Compari-
sons with a large range of critical experiments
and with more rigorous calculations justify the
use of the method for engineering design.
THEORETICAL METHODS
Over-All Description of Method
‘Tne proposed procedure relies upon the fa-
miliar Mutt! and Sofocate? codes in the calculation
of nonthermal and thermal effects, respectively.
The Muft-Sofocate model has previously been used
at WAPD for criticality calculations", Mult is a
54-group Fourier-transform slowing-down code,
whieh ineludes the energy range from 10 MeV to
0.625 eV. ‘The Greuling-Goertzel and B, approx!
mations are used. The Sofocate code calculates
the thermal-group constants averaged over a
Wigner-Wilkins spectrum with a cutolf energy of
0.625 ev.
Both Muft and Sofocate perform homogeneous
calculations; i.e, there is no explicit representa-
tion of the fuel, clad and moderator. Consequently,
any effects that are dependent on the heterogeneity
of the lattice must be accounted for by modifica-
tions of the homogeneous results, In the past, for
example, the homogeneous number densities input
to Sofocate were flux-weighted to account for the
different thermal flux in the fuel and moderator.
In the resonance region of Mult, self-shielding
factors for resonance absorbers were supplied to
account for the flux depression at resonance
energies. ‘The treatment of the heterogeneous
effects basically distinguishes one calculational
technique from another. The proposed method
treats heterogencities in the calculation of the
thermal utilization, the resonance absorption and
the fast fission. The procedures are described in
the following sections.
The cross-section library, of course,
portant in determining the success of
tional procedure. It is impractical to present the
{W, BORL, E. M. GELBARD and G. H. RYAN, “MUFT-4
Fast Neutron Spectrum Code for the IBM-104,"" WAPD=
‘TM-T2, Westinghouse Bettis Atomle Power Laboratory
Guly 1859,
‘WH. ANSTER and R. SUAREZ, “The Calculation of
‘Thermal Constants Averaged over a Wigner-Wilkins Flux
‘Spectrum/Deseription of the Sofocate Code,” WAPD-TM-
28, Mestnghause Betis etomle Pover Laboratory (Jamary
1957).
AW. H. ARNOLD, Jr, “Critical Masses and Lattice
Parameters of H,0-UO, Critical Evperiments; A Compa
fon of Theory and Exper! nert,"" YAEC-152, Westinghouse
Atoniic Power Departiser. (November 1959).
‘4y. 3. BICH and W. F_KOVACIK, “Reactivity and Neu-
tron Flux Distribation Studies in’ Multi-Region Loaded
Reactor Cores," WCAP-1433, Westinghouse Atomic Power
Department (June 1961).
59
library here since the various cross sections for
each nuclide are tabulated for 54 Mult groups and
172 Sofoeate groups (up to the cutoff energy). The
cross-section libraries were compiled at Westing-
house Atomic Power Division based on the re-
ported data. It is noteworthy that all of the cross
sections are credible and are within the range of
the reported cross-section measurements; i.e.
none of the cross sections were arbitrarily ‘ad-
justed’ to improve the agreement with critical
experiments.
‘The multi-group Muit-Sofocate calculations are
ordinarily édited to few-group results for use in
the standard diffusion codes. When four-group
results are used, the energy breakdown is: group
1, 10 MeV —- 0,821 MeV; group 2, 821 keV — 5.53
keV; group 3, 5530 eV — 0.625 eV; group 4, 0.625
eV 0eV. When two-group results are used, the
breakpoint is at 0.625 eV.
Calculations at Thermal Energies
Consider a fuel-rod unit cell, which can be
approximated as three concentric regions as
shown in Fig. 1. The three regions correspond to
the fuel, moderator and clad, and will be referred
MODERATOR
SUBSCRIPT REFERS TO
0 FUEL
1 MODERATOR
2 CLAD
Fig. 1, Fuol-rod unit cell,60 :
to by subscripts 0, 1, and 2 respectively in the
following discussion. If a void exists between the
fuel and clad, it will be assumed that the void and
clad can be homogenized.
‘The most significant thermal parameter in
eriticality calculations is nf,the number of fission
neutrons released per thermal absorption. Tt can
be calculated from group constants averaged over
the thermal spectrum:
Eye vB ja Yo bo
= Sie * Dag Vado + Bai Vi Ge? Daa aga
me)
Vo
where
Let, Ve and dy are the macroscopic absorption
exon ss fad fac a throat,
respectively, in region &.
Tr Sipe hy cre apcton and te
oe ranag actors (8a) are enere
tepene eommon pant ben to eles
the energy dependence of the disadvantage factors
2 ay cau in tte mse ely
the Maxwellian distribution. The success of this
method is a result of a cancellation of errors; the
assumption of a constant spectrum in the fuel and
moderator overestimates 1f, while the use of
Maxwellian cross sections in calculating disad-
vantage factors underestimates nf.
‘The proposed calculational procedure attempts
to circumvent thése errors by eliminating the as-
ce aitdvatage core nip
energy. As proposed by Greebler etal,*, disad-
vantage factors are calculated at every Sofocate
energy level and are inherently included in the
Spe leance, Te deseo fe a
culation can be divided into two parts: 1) the cal-
ti Oe Soaurtipel Gaatnage tor
and 2) the calculation of the spectrum and group
‘The Calculation of Monoenergetic
Disadvantage Factors
‘The calculation o” monoenergetic disadvantage
factors follows the :ntegral transport method as
proposed by Amouya. and Benolst*, Modifications
have been made to avcount for the clad, which was
"Pp, GREEBLER, W. TARKER and J. HARRIMAN, Nucl,
Set. Eng, 6, 128 (1859).
“‘ACAMOUYAL, P, Bi NOIST and J, HOROWITZ, J. Nucl.
Energy, 8, 79 (1957).
STRAWBRIDGE AND BARRY
assumed to be a void in the original derivation.
Diffusion theory is used in the moderator and the
method of successive generations ts applied in the
fuel region; a transport boundary condition is
applied at the inside of the moderator.
"Amouyal and Benoist report that for a range of
moderator-to-fuel ratios, their method reproduces
41/4 to within 0.5% of the DS: monoenergetic
transport calculation. Furthermore, the spherical
harmonie approximation (,) approaches their
calculation as the order of the approximation in-
creases from Pi to Ps.
‘The Amouyal-Benoist equations are not derived
here. However, the final resclts are presonted
along with a discussion of the modifications made
to the original equations.
‘The clad disadvantage factor is given by
fe bs 5 Ame
em Bm”
where 4, is the flux at the fuel surface and A¢: 1s
the flux rise across the clad region.
‘The flux rise across the clad is calculated for
a thin annulus by a diffusion theory approximation,
(2)
Bb. qty Be
Ment Ee, 6
where « is the inverse thermal diffusion length
andf is the thickness of the region.
‘The factor $ in Eq. (8) is a result of assuming
the flix rise across the clad Is linear so that the
average flux in the clad equals the fuel surface
flux plus } the total flux rise across the clad. The
ratio of surface flux to average fuel flux is calcu-
lated by,
Hee aati eee
Pere par
fice Bates x) ] w
cross sections are transport-corrected to account
for anisotropy so that
Exo® = Bsol( 1-1) (5)
Trot = Eso(1-2) + Zao (6)
‘The colfision probability in a cylinder (function of
RoEro*) 1s Pe and is tabulated by Case et al."; a
and f are functions of ReZie* and are plotted in
Fig, 2.
Ero
"K. M. CASE, F. de HOFFMAN and G, PLACZEK, “‘In-
troduction to the Thaory of Neutron Diffuston,” Los Alamos
Sclentific Laboratory (1953).CRITICALITY ‘CALCULATIONS at
orp
w 20 «wo ~«5O.
30
Fo%te
Fig. 2, Functions alMeSts) end P(ReEte) used to
calculate thermal disadvantage Tactors,
‘The moderator disadvantage factor is det
mined from the following expression:
Oa, Ae yd + Ya Zea $2)
Ge go 72 bo (t Vo Fao 3)
2x 1, 2
{ae TER tT (™ ~ »)] ieee
where X is a function of xR, and xy and is given
in terms of modified Bessel functions
"0p, Rat
»[ es eva noma
Tr Gai) Riles) ~ Ga) cia ane
where x is the inverse diffusion length. The term
Ai {8 the extrapolation length (in terms of moder
ator mean free paths) for a black cylinder and is
plotted as a funetion of Rau: in Fig. 3. The term
Fok v4 nw» ane» AB, SP
m
fox 30H
a 3
a
me
-NUREAN FREE PATHS)
, th
Reker
Fig, 8, Extrapolation length A for a black cylinder.
Dus is the transport eross section of the moder-
ator.
Th Faq. (7), Zao in the original derivation has
been replaced by Eae(1+ ¥2 $e? #). ‘Thus the
fuel absorption cross section has, in effect, been
Increased to account for both fuel and clad ab-
sorption. ‘The blackness of the nonmoderator
region, a6 seen from the moderator, is thereby
preserved in calculating the flux rise across the
Troderator region.
Once the flix ratios $/% and é/& are
determined, the fluxes are normalized to 2 volume
average of tnity. Hence
Voge + Vidi + Vaga = 2, )
where Vj is the volume fraction in region &.
The above calculations are performed for
‘cross sections corresponding to each energy level
In Sofocate,
‘The Calculation of Spectrum -Averaged
Cross Sections
The Sofocate spectrum is determined by the
macroscople absorption cross sections at each
energy level. (There are 172 energy levels in
Sofocate up to the normal cutoff energy of 0.62562
eV). The macroscopic cross section at energy
level ¢ is calculated by summing the flux-weighted
cross sections and number densities over all
elements (j) and all regions (4) in the untt cell
Bie DE Nao! gat. (0)
‘The Wigner-Wilkins spectrum is calculated by
Sofocate and the average Sofocate flux in the cell
y? is determined at each energy level. The
macroscople cross sections at each energy level,
which include the effect of the disadvantage fac-
tors (64'), are weighted by the spectrum flux yf to
obtain the absorption eross section for the core:
Eel y! ae
- ae ay
The term yj" is calculated in an identical
fashion, with p-fission cross sections replacing
absorption cross sections in Eqs, (10) and (11).
‘The other thermal parameter needed to de-
scribe the system is the diffusion coefficient for
the core (D), At energy level i we calculate a
diffusion coefficient Including the effect of the
disadvantage factors,
Eacee
1
Dey 12)
35 E Na ow! on te
Sofocate then weights Dé over the spectrum:
core, L DIY! dE
pe as)
Calculations at Resonance Energies
In the resonance energy range, the only impor-
tant heterogeneity is the lumping effect of reso-
nance absorbers and the consequent reduction in
absorption, A self-shielding factor is applied to
the resonance cross sections to account for the
reduced absorption. In slightly enriched U-metal
and UO; lattices, the only isotope for which self-
shielding is important is U"*, The concentration
of U™ is relatively low and calculations indicate
the resonance self-shielding factor for U"* can be
changed by a factor of 2 with a negligible change
in the multiplication factor. This is due to the
fact that both fission and absorption rates change
In U®, whereas only the absorption rate changes
in U"""with the self-svielding factor. Consequent-
ly all resonance self- shielding is neglected except
that in U™*.
In Mult, the self- shielding factor aiso includes
the Doppler effect tince the Mult cross-section
brary contains no te uperature dependence.
The calculation of the resonance absorption is
made in three step::, First the U* resonance
escape probability (1%) is calculated for the
lattice; then a self-sl.ielding factor is obtained for
STRAWBRIDGE AND BARRY
1; finally, the self-shtelding factor is used to
determine the Muft spectrum and fast few-group
constants. The three steps are described in
detail.
Calculation of p*
‘The Ui" resonance escape probability p* is
calculated from the Wigner expression,
= a Rit
Pes 2 |e ,
swhere N™ and ££, are homogenized values,
‘The resonance integral for U™* (RJ**) is cal-
culated from a correlation developed as part of
this work and referred to as the ‘metal-oxide cor-
relation.’ It has been found to agree with Hell-
strand’s U-metal® and UO, (Ref. 9) correlations
for isolated rods. The correlation is also con-
sistent with Helistrand’s temperature correlations
for U-metal and UO, (Ref. 10). Thus, a single
correlation, applicable to both U-metal and UO:
rods at any temperature, replaces the four Hell~
strand correlations.
In close-packed lattices a shadowing correction
(usually called the Dancoff correction) must be
applied. ‘This correction is simply the effective
blackness of the moderator and is designated Derr
When 1.08 barns is included for 1/v absorp-
tion, the metal-oxide correlation is
RI = 2:16x + 2,86 + (0.0279 - 0.0537) Tex,
aay
(15)
where
Tet 1s in degrees Kelvin and
4
Zagh Dett 16)
fs Petey st] (18)
the scattering cross section of the
fuel. The microscopic scattering cross
sections to be used are 10.7 barns for
uranium and 3.8 barns for oxygen
Saf ts
Ne is the U"* number density in the fuel re-
gion
fo 1s the meat. chord length in fuel (to=
4Vo/Se = 2Re for fuel rod)
Pp = 1-P where P; is a function of Ese! Ro
and is tabulated”
the effective shielding factor for the
lattice.
is
'E, HELLSTRAND and G, LUNDGREN, Nucl. Set. Bry.
42, 435 (1952).
#5, HELLSTRAND, J. Appl. Phys. 28, 1499 (1957).
5, HELLSTRAND, P. BLOMBERG and 8, HORNER,
Nucl. Sei. Eng. 8, 497 (1960),CRITICALITY CALCULATIONS
‘The blackness of the moderator (1s D) iscalcu-
lated by the method of Sauer" with the following
equations:
ean) = 1 SPT on
or a square lattice
olf (B* - Boo. a)
For a hexagonal lattice
je C#] 84 Boe, an
where
Zi Is the total cross section of the moderator
fi is the mean chord length of the moderator
vy
( = to
Vi is the volume fraction of region #
‘Moderator’ is defined as the homogenization of
ail nonfuel regions,
Levine™ has shown that the Dancoff correction
calculated for a black rod should be modified to
account for partial transparency. Aisu'® has inte-
grated the monoenergetic escape probabilities
over a Breit-Wigner resonance to obtain the cor-
rection for transparency. In both eases, the cor-
rection factor is the same:
Dar 01 1B) * tad
‘The metal-oxide correlation is plotted In Fig. 4
for three effective resonance temperatures. Also
plotted are many points from Hellstrand’s corre-
lations including his temperature corrections. It
1s apparent that the metal-oxide correlation 1s in
excellent agreement with the Hellstrand correla-
tions,
The metal-oxide correlation uses an equiva-
lence parameter rather similar to those proposed
earlier as, for example, by Dresner"’, However,
the factor Ps has becn included in the volume term
of Eq. (16) to account for the flux dip within the
fuel rod. ‘The volume absorption ts assumed to be
due to neutrons, which enter the fuel from the
moderator at energies just above the resonance
energies and then are scattered into a resonance
BAL SAUER, Nucl. Se’, Eng., 16, 329 (1963).
BM. M. LEVINE, Nuc! Sct. Euy., 16, 271 (1383).
SH. AISU and G. H. MINTON, Nucl. Set, Bng., 19, 468
(3368),
ML. DRESNER, Rescuance Absorption in Nuclear Reac~
tors, Pergamon Press (1960).
63
“+ HRUsTRANO'S conRaLATION FOR UD, 2005
“+ RUSTRANOS CORRELATONFOR UNETAL RODS
aL — weaonte conmeuron
Fig. 4, Comparisonof metal-oxide resonance integral
correlation with Hellstrand’s correlations for isolated
reds.
line in the fuel. The term Po accounts for the
depression in this flux within the fuel rod; from the
reciprocity relation, Ps is the ratio of average
fuel flux to surface ‘flux for neutrons entering the
fuel rod. Since the scattering cross section of the
fuel includes the scattering of the absorber atoms
In the correlation, the narrow resonance approxi-
mation is implied. This form of the resonance
Integral 1s not rigorous; however, its usefulness
is demonstrated in Fig. 4.
‘The fret two terms in the metal-oxide corre-
lation gives the resonance Integral at zero de-
grees absolute, and the last two terms provide the
temperature correction from O°K (-273°C) to Test
(the effective resonance temperature). Although
Dopplet correlations usually use room tempera-
ture as a reference, this choice is arbitrary and
is usually selected to coincide with the lowest
temperature at which the experiments are per-
formed, ‘The use of 0°K (-273°C) as the reference
temperature does not affect the results in any
way; however, it does tend to aid in understanding4
the magnitude of the resonance integral correc-
tion which results (rom Doppler broadening.
‘The metal-oxide correlation appears to be
more generat than the name implies. The corre-
lation has been used to calculate the resonance
integral for a wide range of homogeneous experi-
ments reported in the early literature’ The
homogeneous experiments include pure UO, and
UsOs, mixtures of carbon with UO; or U;Os, and
mixtures of UO, with H,O, D2O or sugar in varying
‘concentrations.
‘At room temperature, if the 1/y absorption is
excluded, the metal-oxide correlation (Eq. 15)
reduces to
RI™ = 2,638 + 0.561. (2)
For the homogeneous ease
(22)
‘The results from this correlation are plotted in
Fig. 5 along with the experimental results for the
homogeneous cases (modified to account for a
uy 25a
®
yl rw, 2009
amy a3
SUCAR .
a
5
5
1
ql eh et
ay
Fig, 8. Metal-o ide eorielation for homogeneous
mistures,
8C.G. ALLAN ef al ,“*Resonince Absorption of Uranium
tee on athe, telana Unversity (940.
Wp, CREUTZ, H. JU °NIK,T. SNYDER and E. P. WIGNER,
J. Appl, Phys, 26, 271 (1958),
STRAWBRIDGE AND BARRY
change in the accepted’ infinitely dilute resonance
integral since the experiments were performed).
‘The agreement is excellent except when the
scattering cross section per absorber atom is
greater than several hundred barns. For very
dilute homogeneous eases, the metal-oxide corre-
lation overestimates the resonance integral, The
success in calculating the resonance integrals for
the homogeneous cases increases confidence in
using the correlation for other fuel types (e.g
uranium carbides or uranium alloys)
Calculation of the U"* Self-Shielding Factor
‘A very important nonthermal lattice parameter
is w, defined as the ratio of nonthermal neutrons
captured in U"* to those thermalized, This para~
meter ts obviously important in deterniining the
maltiplieation of the core as well as the conver-
Sion ratio. In a reactor which has no leakage and
no captures except those in U"", w can be calcu-
lated very simply and, for this Special case, will
be called w*.
(23)
The term w* is calculated from the results of
step 1, A ‘modified Muft’ ts then run in which the
brary is modified so that the caplure cross
section in all elements except U"** is set equal to
zero and the leakage is made zero by setting
B?=0, The modified Mutt performs iterations on
the self-shielding factor for U'* until w* is
satisfied (w* is readily calculated from the Mutt
output). The U"* self-shielding factor in Muft that
satisfies w* is then used in step 3.
Calculation of Nonthermal Group Constants
It is assumed that the self-shielding factor in
U"" is unaffected by the absorption in other
isotopes and by the fast-neutron leakage. This is
a reasonable assumption since the resonances of
various isotopes interact only, slightly: Monte
Carlo studies” indicate the U'* self-shielding
factor changes by 1% or less for each weight per-
cent change in U™ enrichment. Note that this
‘assumption does not mean that the reaction rate in
U®* is unaffected by absorption in other Isotopes
and leakage; the reaction rate will reflect changes
in flux level and spectrum which result from other
absorption and leakage. However, these effects
should cause little, if any, change in the shielding
in U"" due to lumping, or the Doppler — which are
My, A. RISTY, “Unit Cell Homogenization for Reactor
Depletion Studies," WCAP-6060, Westinghouse Atomic
Power Division (1564).CRITICALITY CALCULATIONS 6
' the two phenomena the self-shielding factor must
account for.
‘Using the U" self-shielding factor calculated
from step 2 and self-shielding factors of unity for
all other isotopes, @ normal Muft is run, ‘This
Muft has the regular cross sections for all ele-
ments, and the leakage is accounted for by using
the critical buckling. The fast few-group constants
for the lattice are obtained from the Mutt edit.
(This procedure is modified slightly as explained
In the next section.)
Although the above method for treailng reso-
nances may appear rather involved, the procedure
can readily be automated; since Muft requires
only a few seconds of computing time, the several
Muft’s required in the above procedure present no
computational difficulties,
Calculations at Fast-Fission Energies
- The fast-fission reaction rate depends on the
high-energy flux within the fuel region. Since fis-
sion neutrons are all born in the fuel region, one
would expect the high-energy flux to be greater in
the fuel region than in the moderator region. If
this is true, the homogeneous Muft calculation
would underpredict the fast-fission rate in U™*. It
is not obvious that this effect is important since,
at a given energy, the source of neutrons is
comprised of the direct fission source and the
slowing-down source from all higher energies.
This slowing-down source, along with the exchange
‘of neutrons between moderator and fuel as gov~
erned by the escape probabilities, may tend to
equalize the flux throughout a unit cell. At some
energies, these considerations may even lead to a
higher flux in the moderator.
Since the situation [5 complicated, it is well to
derive a heterogeneous correction and investigate
its effect. Essentially all fast fission in U™*
occurs in the first ten energy groups, spanning the
energy range from 10 MeV to 0.821 MeV, Conse-
quently, a heterogeneous correction is calculated
for each of the first ten groups. The heterogene-
‘ous correction is in terms of a flux ‘advantage
factor’, which is the ratio of fuel fux to moderator
flux. The procedure is a variation of the method
proposed by Hellens"*,
We consider a fuel. rod mit cell consisting of
two regions, fuel (wil subscript 0) and nonfuel
(with subscript 1). By analogy with the treatment
of the thermal utilizatic a factor, we define the flux
ratio in terms of the ‘fst utilization factor’ of the
fuel, fo. In the therm :d group, only absorptions
are counted in determi ing f; however, in the fast
RL, HELLENS and F.C. HONECK, IAEA Tech. Rep.
Ser. No, 12, 27 (1962).
groups, all reactions that result In neutron re-
moval from the microgroup must be included,
Hence, all absorptions and inelastic scatterings
and a fraction of the elastic scatterings must be
counted. The fraction of the elastic scatterings
that results in removals is a function of the
logarithmic energy decrement (£) and the micro-
group lethargy width (41). The total cross section
(including all elastic scatters) 1s designated 2,
and the cross section for removal from the group
is designated E,. We define
1-cez, @)
For any microgroup, we can write
z Ero Veo
Po" Fe Voges Dnt Vig 5)
or
te Eats fo
$1 Ere Ifo
Ea (l-C)Va
. fe
Zee (1=Co) Vo
Ifo *
‘The flux corrections are normalized In each group
so that
foVor oi Vin de (26)
The flux correction factors f and 6 are
multiplied by the Muft group flux to obtain the
actual flux in the fuel and nonfuel regions, respec
tively, All reaction rates are thereby adjusted to
account for the heterogeneity.
‘The calculation of the fast utilization factor, fo,
Ig based on the method of successive generations,
using escape probabilities from the moderator and
fuel regions. The escape probability from the fuel
is based on tables by Case, ef al.’; in the moder-
ator, the method of Sauer" described previously
1s used, The derivation of fo appears in the
appendix.
4 RESULTS
Comparison with More Rigorous Calculations
Comparisons for several experimental lattices
have been made between the proposed calculation-
al procedure described in this paper and a more
rigorous procedure evaluated at WAPD". The
more rigorous procedure used Monte Carlo meth
ads to calculate the resonance absorption”. This
absorption is forced into Muft, which then calcu-
lates the slowing-down spectrum and the fast-
group constants. The thermal constants are cal-
'y. J, EICH and R, B, RADCLIFFE, Westinghouse
Atomic Power Division, Personal Communication (1963).
“"R, D. RICHTMYER, “Monte Carlo Study of Resonance
Escape in Hexagonal Reactor Laltices,"” NYO-G179, New
‘York University (1955).68 :
culated by a ‘scattering ring’ Thermos", which is
a unit-cell thermalization transport code. The
total computer time required for a eriticality
calculation using the more rigorous procedures {8
approximately 20 min on the IBM-7094, or ap-
proximately 50 times as long as the proposed
method,
Some significant parameters are compared in
Table I. The heterogeneous treatment of the
resonance absorption can be evaluated through
‘comparisons of the parameter w, defined as the
ratio of U** nonthermal capture rate to the
thermalization (slowing-down) rate. ‘The statisti-
eal uncertainty in the Monte Carlo calculations
results in a probable error of approximately 1.5%
in w; the proposed calculation generally falls
within this uncertainty band. It should be pointed
out that any difference in the calculated values of
w results in a smaller percentage difference in &.
For example, changing w by 1% causes a change
in & of approximately 1%. Therefore, a negligible
difference in & should result from using the
Proposed method rather than the Monte Carlo
method to treat heterogeneities in resonance ab-
sorption.
‘The most significant parameter in the evalua-
tion of thermal energy heterogeneities is mf, the
number of fission neutrons per thermal absorp-
tion. The following table indicates excellent agree~
ment between the proposed method and the trans-
port unit-cell calculation using the Thermos code.
‘The largest deviation in nf is less than 3%.
"Iq, C, HONECK, “THERMOS, A Thermalization Trans-
STRAWBIIDGE AND BARRY
‘The fast-fission heterogeneous effect was eval-
uated by comparing the flux advantage factors
calculated by the proposed method with those
calculated by the Monte Carlo code, Moca™. The
flux advantage factor is defined as the ratio of flux
in fuel region to flux in nonfuel regions. A
comparison of the advantage factor (averaged over
energies 0.8 to 10.0 MeV) for two U-metal lattices
is made in Table Il.
Although some library differences exist be-
tween the calculations, and the Monte Carlo cal-
culations have a statistical uncertainty estimated
to be several percent in terms of flux advantage
factor, it is evident that the proposed method does
not predict as large a heterogeneous effect as the
Monte Carlo calculations. However, a major por-
tion of the heterogeneous effect is accounted for
by the proposed method, and the calculation is
obviously much better than disregarding the het-
erogeneity entirely (which would imply an advan-
tage factor of 1.0). ‘The difference in fast fissions
resulting from differences in the advantage factor
calculations should not cause an error in the
multiplication factor greater than 0.1% in the
oxide lattices; in U-metal lattices, the errors
could be somewhat larger and might approach 1%
in extreme cases (large rod size and high W/U).
Because of the relative insensitivity of the
multiplication factor to the fast-flux advantage
factor, the proposed method is judged to be
adequate for engineering design purposes and the
"4. RIEF, “An IBM-104 Monte Carlo Code to Calculate
Fast Fission Effects in Homogeneous and Heterogeneous
port Code for Reactor Lattice Calculations,” BNL-8626, Systoms,"" BNL-647, Brookhaven National Laboratory
Brookhaven National Laboratory (1961). 360).
TABLET
Comparison of Proposed Calculations with
Moate Carlo-Thermos Calealations
ot v
Serre Monte Proposed Proposed
Carlo Method Thermos Method
=
(w%)
21 Steet Clad 0.358 1.481
22 Steel Clad 0.383, 1423
37 Steel Clad 0.364 160
37 Steel Clad 0.194 14s
32 Steel Clad 0.194 1286
40 1 Clad 0.334 1682
40 Steel Clad 0.334 1284
1s Alclad 9.280
25 al Clad 0.239,
‘tw ratlo of nonthe mal capture rate In U" to the neutron thermalization rate.
Su,0:U-Borated,CRITICALITY CALCULATIONS 8r
TABLE
Comparison of Monte Carlo and
Proposed Fast-Fission Advantage Factor
Lattice Deseription . Advantage Factor
um [_rod, Uemetal | Monte | Proposed
0:0 | (wt) [~in [em] Carlo | “Method
x0 [ro [os | us | 109 | 10s
additional computer time {approximately 10 min
on IBM-7094) required for Monte Carlo calcula=
tions is not justified. However, the 5- to 10-sec
increase in computer time to perform the pro-
posed heterogeneous calculations Is justified, par-
ticularly if U-metal lattices are considered,
Comparison with Experiments
Data for a large number of critical and expo-
neftial lattices using U-metal or UO, fuel have
been collected from the literature, Only uniform
lattices have been considered; i.e. cases containing
heterogeneities due to fuel element cans, water
slots, control rods or followers were not included
in these studies. ‘The cases were limited (with a
few exceptions) to lattices for which experimental
bucklings have been reported. Using the experi-
mental buckling to represent leakage, the multi-
plication factor can be calculated ‘in a ‘point
model,’ without performing any diffusion calcula
tions.
‘The experimental lattices used in this study
represent a severe test of any calculational pro-
cedure. Of the 116 lattices studied, 55 used UO,
fuet and 61 used U-metal fuel. The U"* enrich-
ments varied from 1.0 to 4.0%; Hz0:U varied from
‘approximately 1.0 to 12; boron concentrations in
the moderator as high as 3500 parts/10° were
considered and as much as 85% of the moderator
was D,O in some cases. Fuel rod diameters from
0.175 in, (0,445 em) to 0.925 in, (2.35 cm) were
‘used; both stainless steel and aluminum clad were
studied as well as both square and hexagonal
lattice arrays.
‘The data for all the UO, cases, along with the
multiplication factors calculated by the proposed
procedures, are given in Table III. Similar data
for’the U-metal lattices are presented in Table IV.
Considering all UO, and U-metal cases to-
gether, the calculaticns resuit in
‘de= 0,£331 + 0,0086 (116 cases),
where the quoted er-or corresponds to one stand-
ard deviation of a sample about the mean,
ao | vo | o6 | 18 | ust] 120.:}-
If the UO, and U-metal cases are considered
separately, we have
For UO; cases
Je= 0.9899 0,008 (55 cases)
For U-metal cases
A= 0.9960 + 0.0072 (61 cases).
Perhaps it is more instructive to consider the
results grouped. according to the major labor:
tories whieh provided the experimental data.
‘Table V provides this information.
An interesting observation from this table is
that the mean & is approximately the same (0.994
to 0.997) for all groups of experiments except the
Babcock and Wilcox (B&W) group, which is 1%
lower. Part of this discrepancy may be due to
failures in the proposed calculational procedure
for the B&W cases. However, a part may also be
due to a systematic difference in the experimental
‘buckling determination. Reference 28 reports the
results of B&W calculations for a number of their
critical experiments, It was observed that diffu-
sion calculations resulted in a mean & that was
more than 0.6% higher than the mean & from
‘point model’ calculations using the experimental
bucklings, Differences in several cases were In
excess of 1.5%. Consequently, either the buckling
measurements or the diffusion calculations appear
to be questionable. If our calculated 4's for the
B&W cases were increased by 0.6% (applying the
bias determined by B&W), the mean & for those
cases would be raised to approximately 0.992,
placing them much more in line with all the
others. This would also have a major effect in
reducing the standard deviation for all cases,
since the B&W experiments account for nearly
one-half of the total deviation.
‘We have also compared radial diffusion calcu~
lations with point calculations for several B&W
cases, In general, the agreement is better thar
that reported by B&W. However, case 37 stands
out as having a very large discrepancy between
the point calculation and the diffusion calculation.
For that case, the diffusion calculation results in
a that is more than 2% lower than the point
ealculation. Thus the reported buckling for case
37 appears to be inconsistent with the other
buckling measurements. This would explain why
ease 37 has the highest calculated & in Table Ill
and why it is inconsistent with the other B&W
cages, which are generally underpredicted by the
proposed calculational method,
‘The reported experimental buckling uncertain-
ties often result in an uncertainty in & of less than
0.1% and rarely in excess of 0.25%. However, the
uncertainties in the UO, cases. 51 through 55 are
larger than normal, ranging from 0.6 to 1.5%. No