Professional Documents
Culture Documents
edu>
22.211 Nuclear Reactor Physics I
Prof. Kord Smith
Massachusetts Institute of Technology
February 8, 2012 - May 23, 2012
Contents
1 Introduction 11
1.1 Course Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2 Sample Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3 Summary: Reuss Ch 1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Summary: Reuss Ch 2 Intro to Nuclear Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Summary: Reuss Ch 3 Intro to Neutron Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
3
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
4
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
5
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
6
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
List of Figures
1 Flux vs. Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2 Flux vs. Lethargy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Spectrum, Added in Fission Neutron Emission Spectrum and Hydrogen Scattering XS . . . . . . 27
4 Thermal Scattering Probability vs. Energy for Hydrogen and Graphite . . . . . . . . . . . . . . . 29
5 Spectrum, Added in Thermal Scattering and 1/v Hydrogen Absorption XS . . . . . . . . . . . . 30
6 Target Motion Is Important for U238 Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7 Slowing Down in Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
8 Flux Between Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
9 Plots of the Psi, Chi Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
10 The 1/E Spectrum Above A Resonance Suggests Group XS Independent of Higher Energy Ab-
sorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
11 Impact of Temperature On Effective RI, Doppler Broadening . . . . . . . . . . . . . . . . . . . . 52
12 Impact of Energy Self-Shielding On Effective RI . . . . . . . . . . . . . . . . . . . . . . . . . . 54
13 Impact of Spectral Hardening On RI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
14 Desired Product of NJOY: Multi-Group Cross Sections . . . . . . . . . . . . . . . . . . . . . . . 58
15 Resonance Scatters Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
16 Comparison Of Wide and Narrow Resonance Approximation . . . . . . . . . . . . . . . . . . . . 66
17 Bells Factor for Various Geometris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
18 Dancoff Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
19 Algorithm for a Simple Pin Cell Code with Collision Probabilities . . . . . . . . . . . . . . . . . 81
20 Pin Cell Code Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
21 Pin Cell Code Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
22 Pin Cell Code Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
23 Pin Cell Code Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
24 Pin Cell Code Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
25 Purpose of Reactor Design Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
26 Different Ways to Define Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
27 Our Choice of Coordinate Systems and Area of Unit Surface . . . . . . . . . . . . . . . . . . . . 86
28 Setup of Infinite Slab Problem using Methods of Chords . . . . . . . . . . . . . . . . . . . . . . 87
29 Tube Presentation in Methods of Chords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
30 Setup of Direct 3D Volume Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
31 Sensitivities of Direct 3D Volume Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
32 Setup of Direct 3D Monte Carlo Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
33 Sensitivities of Direct 3D Monte Carlo Integration . . . . . . . . . . . . . . . . . . . . . . . . . . 92
34 Cylindrical Numerical Integral in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
35 Sensitivities in Cylindrical Numerical Integral in 3D . . . . . . . . . . . . . . . . . . . . . . . . 94
36 Cylindrical Monte Carlo Integral in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
37 Monte Carlo Pin Cell Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
38 Pin Cell Code Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
39 Pin Cell Code Boundary Conditions: Opposite Face Periodic and Rotational . . . . . . . . . . . . 98
40 Boundary Conditions and Interface Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
41 Flux Shape Depent On Diffusion Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
42 Plane Source in Finite Multiplying Medium with k < 1 . . . . . . . . . . . . . . . . . . . . . 135
43 1/M plot and Flux Shape In Approaching Criticality . . . . . . . . . . . . . . . . . . . . . . . . . 136
44 A Critical Reflected Slab Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
45 Energy Groups in Multi-Gruop Diffusion Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 146
46 Diagram Illustrating Neighboring Cells in Finite Difference Scheme . . . . . . . . . . . . . . . . 154
47 Destruction and Production Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
48 Power Iteration Convergence Rate (left) and Dominance Ratio (right) . . . . . . . . . . . . . . . 158
7
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
8
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
9
Lulu Li 22.211 Class Notes: February 8, 2012 - May 23, 2012
List of Tables
1 Topics Covered in This Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Main Reactions Undergone By Neutrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Slowing Down parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4 Comparison of Characteristics of Moderators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5 Volume Fraction/Isotopic Importance of Reactor Materials for Typical LWRs . . . . . . . . . . . 22
6 Interactions At Energy Ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7 Resonance Escape Probability Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
8 Resonance Escape Probability For A Hydrogen System . . . . . . . . . . . . . . . . . . . . . . . 49
9 Observation Concerning Self-Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
10 Comparison of Narrow and Wide Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
11 Physics and Their Corresponding Spectra in MC code . . . . . . . . . . . . . . . . . . . . . . . . 112
12 Subcritical System: Source and Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
13 Simple Geometry Laplacians . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
14 One Group Fundamental Mode Eigenvalues and Eigenvectors . . . . . . . . . . . . . . . . . . . 145
15 Nuclear Fuel Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
16 Finite Geometries With Zero Flux BC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
17 Yields and decay constants depend on fissioning species . . . . . . . . . . . . . . . . . . . . . . 235
18 Absolute Total Yields of Delayed Neutrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
19 8-group Fixed Decay Constans for All Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
20 Matrix Form of 2-Group PKE Diffusion Equations . . . . . . . . . . . . . . . . . . . . . . . . . 243
21 Common Reactivity Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
22 Typical Values for PKE Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
23 BWR Predictive Accuracy of Nodal Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
24 Summary of PWR Reactivity Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
25 BWRs Design Criteria and Acceptance Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
26 Units of Common Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
10
Lulu Li 22.211 Class Notes: February 8, 2012
1 Introduction
1.1 Course Coverage
Calculation of neutron spectra;
Resonance absorption and spatial self-shielding models;
Continuous energy transport, reduction to multi-group diffusion group;
Determination of few-group diffusion constants;
11
Lulu Li 22.211 Class Notes: February 8, 2012
12
Lulu Li 22.211 Class Notes: February 8, 2012
13
Lulu Li 22.211 Class Notes: February 8, 2012
2. Moderator choice: low neutron capture, light nuclei to slow down neutrons, sufficienty dense (1.5).
14
Lulu Li 22.211 Class Notes: February 8, 2012
14. Prompt neutron fission spectrum (2.10.1): Maxwell spectrum, Cranberg spectrum, etc. Energy released in
fission (2.10.3).
15
Lulu Li 22.211 Class Notes: February 13, 2012
e
Z
(~r) = [f (~r0 )(~r0 ) + s (~r0 )(~r0 )] d3 r0 (2)
4R2
Operators associated with the Boltzmann equation (3.2.3).
5. Neutron spectra (3.3.1):
Fast reactors: the spectrum is degraded with respect to the fission spectrum. Reason: slowing down
by inelastic scattering off heavy nuclei and elastic scattering off sodium.
Thermal reactors:
(a) a high energy hump. Reason: fission spectrum, degraded due to scattering;
(b) slight decrease in the epithermal region. Reason: resonant capture losses, esp. U238;
(c) a low energy hump. Reason: Maxwell distribution of the thermal agitation but a little harder
because the temperature equilibrium has not been perfectly achieved.
6. Four factor formula (3.3.2).
16
Lulu Li 22.211 Class Notes: February 13, 2012
Enf A2 + 1 + 2A cos 1
= = [1 + + (1 ) cos ] (3)
Eni (A + 1)2 2
1 + A cos
cos = (4)
A2 + 1 + 2A cos
(A 1)2
= = min ratio between final n energy and initial (5)
(A + 1)2
4. In CMCS, scattering is isotropic in solid angle (except very high energy), which implies that cos is
uniform, and Enf is uniform:
1 1
P () d = sin d = d| cos | (6)
2 2
dE
P (E) dE = (7)
(1 )Ei
2
5. In Lab system, scattering towards the front it favored. The mean of cos is = hcos i = 3A .
6. Lethargy, a unitless measurement of energy (the idea comes from laws of elastic collision governs an energy
ratio):
Eref
u = ln (8)
E
17
Lulu Li 22.211 Class Notes: February 13, 2012
Notice as time goes, neutrons slow down, u increases, making it like a measure of the age of the neutrons.
Then we know that the uniform distribution of energy becomes a decreasing exponential distribution for
lethargy gain:
1
w = uf ui = ln (1 + + (1 ) cos ) (9)
2
wmin = 0 (10)
wmax = = ln (11)
w
e
P (w) dw = dw (12)
1
hwi = = 1 (13)
1
is like the efficiency of slowing down by a nucleus. That is, neutrons advance by lethargy units on
average at each collision. Then to overcome the total lethargy interval U = ln E
E1 , the number of collisions
0
needed is:
U
n= (14)
7. Moderating power is the best measure of a materials ability to slow down neutrons. It has two forms:
A good moderating material should have: high slowing down (hence light nuclei), low capture (D, Be, C),
moderating power (take into account both high slowing down and high scattering xs). Water has the highest
moderating power, but it requires an enriched fuel (around 1%).
8. Laws of inelastic collision. Elastic scattering is important for moderators because they are light; inelas-
tic scattering is important for heavy materials like the fuel because they have almost no elastic collision.
Minimum energy of the neutron for an inelastic collision is:
A+1
Ethreshold = Q (16)
A
18
Lulu Li 22.211 Class Notes: February 13, 2012
9. Slowing down equations: a simplified Boltzmann equation performing neutron counts that depends on one
variable, either velocity or energy or lethargy. We pick lethargy. Then we assume an infinite, homogeneous
medium with a source uniform in space and time.
10. First form of the slowing down equation (7.1.9):
In the specific case of COM isotropic, monatomic elastic slowing down, the transfer probability is:
0
e(uu )
P (u0 u) = P (u u0 ) = (23)
1
11. Decay of neutron spectrum by successive scattering events (7.2.2): producing spectrum by adding up suc-
cessive scattering events;
If we plot flux vs. lethargy, all isotopes would fluctuate (Placzek transient) before reaching the asymptotic
flux value, except H which immediately reaches as .
13. Slowing down in hydrogen (7.2.4): uses Greens function; the Dirac distribution compensates for the source
in the equation; the Heaviside step function makes sure the neutrons scattered at least once scattered beyond,
not below, the original lethargy.
14. Slowing down with resonance absorptions (which is low energy) (7.2.5): we can approximate resonance
absorption of the following types
(a) Black resonance trap:
0 0
1 e(uu ) 1 e
Z Z
1p= du0 = (25)
0 u 1 (1 )
19
Lulu Li 22.211 Class Notes: February 13, 2012
(c) A set of narrow grey resonance traps: we express each resonance in exponential form, and find the
product of them:
Z Z
a (u) a (u)
pi = 1 du = exp du (27)
i t (u) i t (u)
" #
Y X Z a (u)
p= pi = exp du (28)
i i i t (u)
(d) General expression: Because the integrated function is zero outside of the resonance traps, we can
simply write,
Z
a (u)
p = exp du (29)
t (u)
This is the general expression for resonance escape probability.
15. Slowing down with low, slowly varying absorption (which is high energy) (7.2.6).
20
Lulu Li 22.211 Class Notes: February 13, 2012
(b) The source in the thermalisation equation Sslow-down (E) is not a true source; it is a density of
arrival at energies below the cutoff energy due to scattering events occurring in the last part of the
slowing down domain and transferring the neutron beyond the cutoff energy in the thermalisation
domain.
6. Thermal Spectrum(9.2): it is approximately reasonable well by Maxwell distribution, except,
0 2kT : real thermal spectrum is smaller due to absorption of neutrons;
2kT and above: Maxwell spectrum approaches zero quickly, whilst the real density falls slightly but
remains significant, due to the slowing down queue: neutrons coming from high energies slow down
and enter the thermal domain, compensating for the disappearances by absorption.
7. Comparing MOX and UOX spectrum: the two are close for the fast and epithermal domains, because they
essentially have the same quantity of moderator, U238, and cladding. Whereas in the thermal domain, the
number of neutrons in MOX is lower by a factor of 4 because of the high absorption by MOX fuel of thermal
neutrons.
8. Average cross section (9.2.3):
R
(E)(E) dE
= R (33)
(E) dE
Example: calculate the average xs for a Maxwell spectrum and a 1/v xs.
r
293
= (v0 ) = 2200 (34)
2 2 T
where 2 is the average of x = v/v0 on a Maxwell spectrum, and also the average of 1/x.
9. Heterogeneous Configurations (9.2.4):
10. Approximate thermal neutron speeds(9.3.1): assume absorption xs are 1/v and scattering xs are constant,
we can approximate
r
T
v = 2200m/s (37)
293.15
11. Thermal utilisation factor (9.3.2): the fraction of thermal neutrons absorbed in the fuel,
disadvantage factor
z}|{
Vf a,f f 1 Vm m a,m
f= , 1= (38)
Vf a,f f + Vm a,m m + f Vf f a,f
|{z}
moderation ratio
21
Lulu Li 22.211 Class Notes: February 13, 2012
NAV molecule/mol
NcH = 0.55 1g/cm3 2/molecule = 3.68 1022 (40)
18g/mol
NAV molecule/mol
NcO = 0.55 1g/cm3 1/molecule = 1.84 1022 (41)
18g/mol
NAV molecule/mol
NfO = 0.35 10g/cm3 2/molecule = 1.56 1022 (42)
270g/mol
NAV molecule/mol
Nf238 = 0.95 0.35 10g/cm3 1/molecule = 1.48 1022 (43)
270g/mol
NAV molecule/mol
Nf235 = 0.05 0.35 10g/cm3 1/molecule = 0.079 1022 (44)
267g/mol
NAV molecule/mol
N Zr = 0.10 6.6g/cm3 1/molecule = 0.56 1022 (45)
90g/mol
That is, in a typical LWR the relative number densities come out to be around:
Notice hydrogens and oxygens number densities are about the same.
Abtracting Cross Sections See Figure 146, Figure 147 for common cross sections. Notice resonance absorp-
tion and scattering tend to line up in terms of energy.
Practice: Why can we use sodium as coolant if its capture cross section is so high?
Answer: Sodium has a high capture xs in thermal range, and a small capture xs in fast range. But we can still
build a thermal reactor with water as moderator and sodium as coolant, as long as sodiums volume is lower than
that of the H.
22
Lulu Li 22.211 Class Notes: February 13, 2012
2. From conservation equations in the COM and translating back to the energies
FIXME
3. Differential scattering cross section:
(E)
(E E 0 ) = (E)P (E E 0 ) = E E 0 E (48)
(1 )E
5. Introduce lethargy:
6. Logarithmic energy decrement:
ln()
=1+ (50)
1
ln(E2 /E1 )
n= (51)
Examples: for E2 = 2 106 eV, E1 = 0.1eV, the number of collisions is: nH = 17, nD = 23, nC =
106, nU = 2000.
7. Slowing down in non-absorbing, arbitrary mass, isotropic scattering material:
8. Flux vs. energy:
There are two ways to understand flux here.
23
Lulu Li 22.211 Class Notes: February 13, 2012
Particle based: we let all the neutrons to scatter till death; then for each energy interval, if we add up
the number of neutrons with energies in that interval; we get the flux corresponding to each interval.
Generation based (as in Fig. 1): we let each neutron scatter once in each generation, save all the
number of neutrons vs. energy data, and add up all the generations, this would effectively give us flux
vs. energy as well. Notice with insufficient number of generations (left plot), the flux tails down at
lower energies, meaning that some neutrons are not slowed down entirely yet. With sufficient number
of generations (right plot), we get a 1/E spectrum for COM elastic scattering only.
9. Flux vs. lethargy: as in Fig. 2, flux is a constant in lethargy space.
24
Lulu Li 22.211 Class Notes: February 13, 2012
25
Lulu Li 22.211 Class Notes: February 13, 2012
2
E
(E) dE = E exp dE T = 1.33 MeV for U235 (52)
(T )3/2 T
Simulation: constructing a cdf for Maxwell distribution (when constructing any cdf, it is a good practice to
sample immediately and double check with the pdf); for each neutron, generate a random number [0, 1], and
use = cdf (E) to reverse-lookup for E, and this is the fission emission energy of that neutron.
Characteristics of spectra after adding in (E) (caution: Fig. 3 is plotted with x-axis being energy, but it is
still in lethargy space):
1. The spectrum is independent on the number of incoming neutrons;
2. For one specie, the spectrum is independent of its cross section; for two species, only the relative cross
section matters.
3. The effect of (E): smooth out the shart edge at the highest energy cutoff (Fig. 3 middle plot).
4. A little bump at fast energy due to source neutrons. FIXME.
Z
1 S(E)
(E) = dE 0 S(E 0 ) + (53)
s (E)E E s (E)
5. Adding in real Hydrogen scattering cross section. So far we reach a constant spectrum in lethargy. To obtain
the fission source peak around 2 MeV, we have to divide the flux count by the hydrogen cross section:
XX XX 1
(E) Ngn gn = Ngn (54)
g n g
(Egn )
N
Recall the hydrogen scattering cross section decreases from 0 to 1 eV, stays constant between 1 eV to
10 keV, and decreases again afterwards. In the fast energies, the decreased hydrogen absorption cross
section increases the flux, causing the 1.7 MeV peak.
Notice hydrogens scattering cross section is not exactly flat in the eV to 0.1MeV region; it is slightly
decreasing, resulting in the spectrums middle region
26
Lulu Li 22.211 Class Notes: February 13, 2012
1e+06
100000
0.01 1 100 10000 1e+06 1e+08
Energy Associated with Lethargy Bins (eV)
Slowing Down in Hydrogen From Neutron Emission Spectrum
1e+07
50 generations, 1 million neutrons, 1000 bins
1e+06
Flux (not normalized)
100000
10000
1000
100
0.01 1 100 10000 1e+06 1e+08
Energy Associated with Lethargy Bins (eV)
Slowing Down in Hydrogen, Neutron Emission Spectrum, W/ Real H Elastic Scattering XS
1e+06
50 generations, 1 million neutrons, 1000 bins
100000
Flux (not normalized)
10000
1000
0.01 1 100 10000 1e+06 1e+08
Energy Associated with Lethargy Bins (eV)
Figure 3: Spectrum, Added in Fission Neutron Emission Spectrum and Hydrogen Scattering XS
27
Lulu Li 22.211 Class Notes: February 13, 2012
3. Resonance parameters:
For instance, Interpreted ENDF file for U238 resonance parameters is Reich-Moore data, thus we get
negative resonance energies etc.
U238 s-wave = 23 meV, and n changes anywhere from 1 to 70 meV. U238s resonances are
widely spaced (s-wave 21 eV spacing, p-wave 7 eV spacing) and are good for using SLBW because
resonances are widly spaced. Though typically odd isotope tends to widely spaced.
Resonance spacing is the spacing between two peaks; capture width is the width of a resonance.
Capture width tends to be constant (23 meV for U238), while neutron width tends to change (Hebert
thinks increasing as energy).
Unresolved resonances: it is not that resonances become tighter together (resonance spacing is almost
constant); it is that the xxxx gets too small to measure.
Doppler broadening is the same as thermal agitation and apparently they matter even more in high
energy range.
Resonance integral is a good measurement of the data: we do not care to compare data point by point,
but rather we care about the data in a range. For instance, ENDFB-VIIs resonance integral over the
8 energies are about 1% different from ENDFB-VI. This does not seem that much, but it leads to
100pcm in reactivity.
4. Example: U235 and Pu239 in fast range fission cross section does not change much while the capture cross
section decreases significantly, thus we are interested in operating these two isotopes in fast reactors.
5. How are cross sections measured? When cross sections become very large, neutrons cannot penetrate the
foil which is only a certain thickness, so the cross sections vs. energy we get are going to have a flat top for
the high cross section range. Thus it is common to do multiple measurements with different thickness. For
instance, U238s 6 eV is the largest and widest resonance, so we would want many points around there.
28
Lulu Li 22.211 Class Notes: February 13, 2012
where the upper signs are for E 0 > E and the lower signs for E 0 < E, and the , terms are:
A+1 A1
= = (56)
2 A 2 A
2. For a proton gas (A=1), we can simplify the energy transfer function to be (notice the function is continuous,
but not the derivative),
EE0 q
s0 e kT erf
E
E0 > E
s (E)fs (E E 0 ) = q kT (57)
E erf E0
E0 < E kT
A=1 Scattering Probability vs Final Energy A=12 Scattering Probability vs Final Energy
1 1
E = 1 kT E = 1 kT
E = 4 kT E = 4 kT
E = 25 kT E = 25 kT
E = 100 kT E = 100 kT
E = 200 kT E = 200 kT
0.8 E = 500 kT 0.8 E = 500 kT
Probability x Initial Energy x (1 - alpha)
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Scattered Energy Over Initial Energy Scattered Energy Over Initial Energy
Figure 4: Thermal Scattering Probability vs. Energy for Hydrogen and Graphite
Observations:
(a) The reason we make the y-axis P E(1 ) is because the asymptotic value would be normalized to
1.0 this way. Reason:
dE
P (E E 0 ) dE = P (E E 0 )E(1 ) = 1.0 (58)
(1 )E
29
Lulu Li 22.211 Class Notes: February 13, 2012
E
(b) , the ratio of the neutron energy to the medium temperature (energy), is what we care about. For
kT
instance, when we raise the neutron energy and the medium temperature both by a factor of 4, the
probability remains the same.
(c) In Fig. 4: E = 1kT to 25kT provides a good cover of range, and even at 1200K, 1kT = 0.1 eV, and
25kT = 2.5eV, so 4eV is a typical upper scattering cutoff. Although keep in mind that there is always
a very small probability to scatter to a much higher energy.
0
(d) In Fig. 4: the EE axis shows that there is a good probability that the neutron would gain energy from
the elastic scattering(the lower the initial neutron energy, the closer it is to the target nuclei thermal
agitation, the more probable is upper scattering). In fact this is the difference between this thermal
energy model and the previous high-energy model.
(e) We can add the thermal scattering into our MC model and observe a peak in the thermal range as in
the left plot of Fig. 5. To make the height more realistic, we add in the hydrogen absorption cross
section (a 1/v distribution, only affect the thermal range) as in the right plot of Fig. 5. The absorbing
strength is picked such that the absorption-to-scattering ratio is something reasonable, like 0.2.
Slowing Down and Thermalization in Hydrogen (no H absorption yet) Slowing Down and Thermalization in H w/ 1/v absorber
1e+07 1e+06
"spec.dat" using 1:2 "spec.dat" using 1:2
1e+06
100000
Flux (not normalized)
100000
10000
10000
1000 1000
0.01 0.1 1 10 100 1000 10000 100000 1e+06 1e+07 0.01 0.1 1 10 100 1000 10000 100000 1e+06 1e+07
Energy Associated with Lethargy Bins (eV) Energy Associated with Lethargy Bins (eV)
Adding Target in Motion (important for U238 resonances at hot fuel conditions). Thermal agitation does not
matter that much for non-resonance conditions; that is, when d/ dE is small, a small v hence a small E
does not change the cross section much. This is not true anymore for U238 resonances, because when cross
section changes very rapidly (especially true in the scattering cross section than the absorption cross section),
target motion/thermal agitation would affect up scattering quite significantly. As shown in Figure 6, it could be a
10% increase in LWR U238 Doppler.
We will cover resonance models later, for now we just need to know that the weird dips in scattering cross
section is a result of the interacting wave between the potential scattering and the compound nuclear scattering.
30
Lulu Li 22.211 Class Notes: February 13, 2012
31
Lulu Li 22.211 Class Notes: February 13, 2012
4. Slowing down happens through elastic scattering that is isotropic in the center of mass system (s-wave
elastic scattering), that is,
s (E 0 ) E
s (E 0 E) = for E < E 0 < (60)
(1 )E 0
where = (A 1/A + 1)2 . In assuming s-wave, the slowing down model only applies for neutrons
below 1MeV in light moderators and neutrons below 0.1 MeV in heavier materials, since the average
0 = 0.07A2/3 E [MeV]. For neutrons above 0.1 MeV, inelastic scattering from heavy nuclei becomes
important and cannot be ignored (but I dont think weve ever addressed inelastic scattering in this course).
When we had the time dependent term t , flux changes with respect to time and is driven by source. Once
we go into steady state and take away the source, TE becomes an eigenvalue problem, and we need to add a k
on the bottom of the fission term. Notice that our TE only depends on energy now,
Z
(E)
Z
0 0 0
t (E)(E) = dE s (E E)(E ) + dE 0 (E 0 )f (E 0 )(E 0 ) (61)
0 k 0
If we re-write assuming the fission source is an external source of neutrons Sf (notice its energy independent):
Z
t (E)(E) = dE 0 s (E 0 E)(E 0 ) + (E)Sf (62)
0
1. Fast range (> 0.1MeV): Assume no upscattering, and any scattering event is likely to slow neutrons below
0.1 MeV. Then our TE becomes,
(E)Sf
t (E)(E) = (E)Sf (E) = (64)
t
(E) dominantes fast flux, because t is fairly flat in the fast range, and Sf = kf which is proportional
to (E) anyway (or say flux scales with source), so we are left with the distribution of prompt neutrons
dominant the flux in the fast range.
32
Lulu Li 22.211 Class Notes: February 13, 2012
2. Slowing down region/resonance region [1eV, 0.1MeV]. We define a slowing down density:
Again assume no upscattering. In the presence of scattering, all neutrons born from fission will reach E
unless absorbed,
Z Z
0 0
q(E) = (E )Sf dE a (E 0 )(E 0 ) dE 0 (66)
E E
R R
From the general expression 0 (E 0 ) dE 0 = 1, we can approximate 0.1
(E 0 ) dE 0 = 1, thus only
leaving the energy independent Sf term:
Z
q(E) = a (E 0 )(E 0 ) dE 0 + Sf (67)
E
Next we apply the second fundamental theorem of calculus, or the Newton-Leibniz axiom, which states
that,
Z b
F 0 (x) = f (x) F (b) F (a) = f (x) dx (68)
a
Be careful about the bounds as the second theorem is not simply taking the derivative of Eq. 67. In our
case, q = F, = f . We already have q(E), substitute E = in Eq. 67 we get,
q() = Sf (69)
Z
q() q(E) = a (E 0 )(E 0 ) dE 0 (70)
E
which is the second expression in the second fundamental theorem of calculus. Then by the theorem,
dq(E)
= a (E)(E) (71)
dE
Comments/notes:
If absorption increases, less neutrons would make it to E, then q(E) should be smaller.
q(E) changes proportional to a , because in our approximation the source term becomes indpendent
of energy.
If a = 0, q(E) = constant.
In reactor applications, most absorptions happen in the resonances of 238U. 238Us cross section changes by
order of magnitudes for each resonance, so we can assume that a = 0 between resonances (also because
33
Lulu Li 22.211 Class Notes: February 13, 2012
between resonances there is no quantum effect between compound nuclides). That is, q(E) = constant
between resonances. Consider our TE again with a = 0,
Z
s (E)(E) = s (E 0 E)(E 0 ) dE 0 (72)
0
0
s (E )
Plug in s (E 0 E) = s (E 0 )P (E 0 E) = (1)E 0 . Also assume: below birth E of fission neutrons,
q(E) q(u)
(E) = (u) = (77)
s (E)E s ()
34
Lulu Li 22.211 Class Notes: February 13, 2012
FIXME: assume hydrogen is purely scattering. Again in an infinite medium the spectrum is independent of
the absolute number density; it only depends on the relative number density.
Duderstadt p.399 plots of spectrum when we add in different effects. One question is, the plots end around
0.2 eV.
p.17: left plot: constant power. right plot: average thermal flux to be 1.
Adding in absorption: fast peak is slightly decreased, flat portion is slightly decreasing as we lower energy,
and thermal peak magnitude is very depressed. Observation: if we want high thermal flux, we want low absorption
xs (flux trap!) In the old days (e.g., Duderstadts plots), we normalize the thermal average to 1, then we get the
new flux is slightly lowered and the flat portion raises up slightly, which makes it hard to observe absorption
hardening.
Perturb leakage. More likely to leak out in fast energy.
1 a subtle point here is that the equations here are the asymptotic values. Recall the Placzek Transient is due to the discontinuity of the flux
at u = , and its derivative at u = 2 etc where is the maximum lethargy gain. The discontinuity is more noticable for heavy nuclides, and
disappear for hydrogen.
35
Lulu Li 22.211 Class Notes: February 13, 2012
where p(VA ) is the Maxwellian distribution, and the x (|Vn VA |) is the 0K cross section at the corrected velocity
Vn VA .
Essentially we correct the cross section using an average velocity distribution of the material at temperature
T. The effect on resonance range is,
In the thermal range, the convolution of a constant cross section would increase the thermal cross section as
temperature increases; more specifically, scattering cross section tends to be constant in low temperature, and
becomes 1/v as temperature increases; thus at higher temperatures it is harder to tell absorption cross section (1/v)
and scattering cross section apart.
Prof. Forget introduces SLBW here. The convolution integral will yield the expression in section 2.7.1 with
, definitions of 8.4.3.
term shows up in both absorption and scattering cross section terms and it gives the shape of the cross
section. term only shows up in the scattering cross section and it represents the interference effect. The sum of
and results in the dip before peaking that always shows up in the scattering cross section.
36
Lulu Li 22.211 Class Notes: February 13, 2012
The dependency on A suggests that the bound elastic scattering matters for light nuclei, and not so much for
heavy nuclei. In the case of hydrogen, the free gas model would still provide the right shape, but the probability
can be off by a factor of 10. In the case of graphite, the free gas model does not provide the right shape. The lower
energy you want to go, the more careful you need to be.
Elastic scattering for bound molecules can be characterized by,
r
d2 0 0 b E0
(E E , ) = e 2 S(, ) (81)
d dE 0 4kT E
where b is the bound scattering cross section for the material, kT is in eV, S(, ) is the symmetric form of the
thermal scattering law, since , depends on temperature, S depends on temperature too.
2 2
exp 4
+
S(, ) = (82)
4
where , depends on two terms:
The momentum transfer ,
E 0 + E 2 E 0 E cos ~2 2
= = (83)
AkT 2M kT
where A is the ratio of the mass of the scattering atom to the neutron mass.
The energy transfer ,
E0 E
= = (84)
kT kT
Typically thermal bound scattering in LWRs because the change in cross section is not large, absorption cross
sections are high, and water number density is low. Though bound scattering becomes important in scattering
experiment, cold neutrons etc.
Be careful that free carbon cross section is very different from the graphite cross section because the latter has
a crystal lattice structure which affects the phonon distribution tremendously.
FIXME today: p.31 U238 scattering kernel: makes sense: T is the temperature of the medium, and earlier
plots E = blah kT describes the neutron energy.
Free gas model: target at rest? no!!!!! abs & leakage makes sense: they have the same effect.
Flux trap:
Why doesnt CMS uses new ENDF libraries?
37
Lulu Li 22.211 Class Notes: February 15, 2012
2.11 Summary
1. Properties of isotropic elastic scattering/asymptotic downscattering (which we assume for larger than 4 eV).
See Table. 3 for common values.
(a) After each collision, the minimum energy is characterized by:
2
Emin A1
= = (85)
E0 A+1
(b) Mean log energy decrement: A , .
=1+ ln (86)
1
(c) Number of collisions required to slow a particle from Ei to Ef :
ln(Ei /Ef )
n= (87)
2. Not near sources or resonance (Duderstadt p.319):
1
(u) (88)
s (u)
S0
q(u) = S0 (u) = (94)
s (u)
38
Lulu Li 22.211 Class Notes: February 15, 2012
In Ch. 2, we end up with a spectrum of a thermal peak and a fast peak and a smooth transition in between. In
this section, we are going to focus on this middle resonance region. We will spend about 7 lectures in discussing
different aspects of modeling neutron spectra.
39
Lulu Li 22.211 Class Notes: February 15, 2012
3. Statistical resonance cross section model (2.7.2): n widths fluctuate greatly from one resonance to the
other, but stays about the same. This section also include expressions for the average distance between
resonances, and the Wigner probability distribution.
4. Absorption cross section is 1/v in the thermal domain.(2.7.3) Reason: Breit-Wigner states that absorption
cross section is (i = for radiative capture and f for fission etc):
2 n i
i = g (95)
(E E0 )2 + 2 /4
f , , etc are independent to energy E; n E (for s-wave);
2 1
E;
The denominator is approximately equal to the constant E02 assuming that E, are small compared
to E0 .
Thus f , c 1E v1 . Even if several resonances make a contribution, the reasoning remains valid. This
reasoning is not valid if E0 is close to zero.
40
Lulu Li 22.211 Class Notes: February 15, 2012
0
z }| {
(u) + S(u) = t (u)(u) (96)
q(u) q
(u) = = Constant (97)
s (u) s
Then the arrival density in the resonance is constant as well:
q
(u) = s (u)(u) = =C (98)
Plug back into the slowing down equation, we get the flux in the resonance:
q C
(u) = (99)
t (u) t (u)
a (u) du
Pabs = a (u)(u) du = (100)
t (u)
a du
p = 1 Pabs = 1 a (u)(u) du = exp (101)
t
(compare with the derivation we had in class)
4. Fine structure/self-shielding factor captures the detailed resonances, whereas macroscopit flux captures
everything else (8.1.3).
5. Fine structure in homogeneous mixture (8.2):
Nm
(a) We define dilution xs as: d = Nr m (8.2.1).
(b) Two models for transforming a heterogeneous situation into a homogeneous one: at very wide reso-
nance (low energy), and at very narrow resonance (high energy).
6. Distinguish between the different cases: truely homogeneous mixture, two-isotope heterogeneous configu-
ration, homogenized heterogeneous configuration, and arrays of rods.
41
Lulu Li 22.211 Class Notes: February 15, 2012
h i
N0 Ieff
Homogeneous p = exp ( s )m
VC N0 Ieff
General X(V )
p = exp
s i
i
7. When a particle coming out of fuel is not necessarily interacting with moderator, that is in the case of tight
rods, we use Dancoff factor C, which is the probability for a neutron leaving a fuel element of crossing the
moderator without a collision. (8.3.4).
8. Derivation of resonance escape probability in a heterogeneous situation as in Table 7 (8.3.5).
9. Doppler Effect (8.4):
(a) Origin: we can ignore thermal agitation in treating scattering, because scattering xs is fairly constant,
so a small change in velocity (hence energy) does not change xs much. Whereas in absorption, near
resonance peaks, a small change in velocity (hence energy) would cause a huge difference in absorp-
tion xs, hence taking into account thermal agitation of the fuel material would make a difference.
(b) Characteristics 1: as temperatures increase, the resonance widens, and the peak is lowered, but with a
constant resonance integral (area under the curve).
(c) Characteristics 2: although integral is constant, self-shielding says that the widening of the resonances
has a much greater effect than the lowering of the peaks. That is, Doppler effect as a whole leads to
an increase in resonant capture by U238.
(d) RIeff for capture by U238 varies approximately linearly with the square root of dilution xs; it also
varies linearly with the square root of absolute temperature.
42
Lulu Li 22.211 Class Notes: February 15, 2012
One place to get resonance data is from LANLs website. For instance, U238 Resonance Parameters. We can find
parameters like, scattering length AP, and the potential scattering cross section is just 4A2p = 11.293 for U238.
Other parameters: GN means N with unit eV, and GG means G in eV. Extract the energy range from 0 to 10
keV (ignore the negative resonance energies).
Model For Resolved Resonance Absorption
r
E0 2
(E, T ) = A(x, ) (102)
E
2
n (E, T ) = [A(x, ) + B(x, )] + potential (103)
To convert them into a form suitable for numerical calculation,
r
E0 n
(E, T ) = r(x, ) (104)
E
n n
n (E, T ) = [r(x, ) + q(x, )] + potential (105)
where
r
A 2(E E0 )
= x= (106)
4kT E0
h2 A + 1 2603911 A + 1
r= = q = rpotential (107)
2E0 A E0 A
= n + potential = 4R2 (108)
Notice there are many numerical representations of the , functions. For now, we can use,
" 2 #
x+i x+i
(x, ) = Re exp i erfc i (109)
2 2 2
" 2 #
x+i x+i
(x, ) = Im exp i erfc i (110)
2 2 2
43
Lulu Li 22.211 Class Notes: February 15, 2012
In principle, resonances at any energy contribute to all energies, even though their contributions can
be infinitesimally.
Because , functions are very temperature dependent, so when we perform a monte carlo simulation it
is important to take into account that different pins are at different temperatures. However currently only
MC21 supports temperature gradients (by pre-generating many many tables using NJOY).
Model For Unresolved Resonance Absorption: We assume 25 eV resonance spacing, and assume
r
E
= 0.023eV n = 0.05 eV (111)
Elast
in which Elast means the last resonance energy used in the 14 resonance region. Notice that for the purpose of
this excercise we assume this region to be unresolved, whereas in reality they are resolved.
44
Lulu Li 22.211 Class Notes: February 22, 2012
RI depends on the normalization of the 1/E flux spectrum; it is also implicitly assumed that flux equals 1
when E = 1 through normalization;
RI is independent of energy bounds for isolated resonances;
RI is independent of temperature. This is because if the energy bound is larger enough, then as temperature
increases, the spectrum would broaden, but because the area under the curve remains the same, assume
the cross section is constant, then RI is essentially integrating the spectrum, which would not change upon
temperature change;
RI is useful for inter-comparing libraries or cross section models. It is a classic way to evaluate new
resonance data typically from 0.5 eV to 10 keV. We use RI to check our resonance data, in particularly the
three big resonances at 6.67, 20.9, 36.7, 66 eV. Numerical test of the SLBW RIs shows that the RI comes out
to be within 1% of ENDF/B-VII Reich-Moore data (Lec 6, slide 17, with 0.01 eV as spacing in histogram).
Group cross section is a similar but much more useful quantity. From its definition, we see g does not depend
on the flux normalization:
R E2
(E)(E) dE
g = E1R E2 (116)
E1
(E) dE
1
(E) (117)
E
R E2
E1
(E) E1 dE RIeff RIeff
g = R E2 = = (118)
1
dE ln(E2 ) ln(E1 ) ln(E2 /E1 )
E1 E
Notice3 :
Group cross section by definition depends on both cross section and flux spectrum.
3 Review here for exam
45
Lulu Li 22.211 Class Notes: February 22, 2012
Group cross section depends on the flux, but not on the normalization of flux (that is, only the shape matters,
not the magnitude);
Group cross section depend explicitly on energy bounds (widths) of the groups;
Effective RI can be computed from group cross sections and group energy bounds as in Eq. 119; As spec-
trum approaches 1/E, the effective RI computed from group cross sections will approach infinite RI.
46
Lulu Li 22.211 Class Notes: February 22, 2012
in lethargy. If we assume scattering into the resonance comes from this constant lethargy region, then the
resonance lethargy is constant as well.
Eq. 124 suggests that in infinite medium the flux shape near resonance depends only on the ratio of the number
density of the moderator to the resonance absorber and the moderator cross section. But once we move into a
finite medium or we take into account leakage, then the absolute number densities are needed.
To capture the ratio of number densities and the moderator cross section, we define dilution cross section as,
Nm m
d = (125)
Nr
47
Lulu Li 22.211 Class Notes: February 22, 2012
Figure 10: The 1/E Spectrum Above A Resonance Suggests Group XS Independent of Higher Energy Absorption
48
Lulu Li 22.211 Class Notes: February 22, 2012
Notice: (1) this expression is only an approximation; (2) p only depends on effective RI and dilution cross section.
For instance, in a hydrogen system with d = 200, we can tabulate the p values as in Table 8. The last entry tells
us that about 20% of neutrons are absorbed in U238 resonances.
Derivation of p expression:
1. Assumptions: only elastic down scatter above the resonance, a,m = 0, s =const, and only a single
resonance absorber.
49
Lulu Li 22.211 Class Notes: February 22, 2012
RIu1u2
eff
p12 = 1 (136)
d
or
RIeff Nr RIeff
p exp = exp (139)
d m
The motivation for this derivation is because back in when we had no idea what the cross sections are, as long as
we can measure the flux at a high energy and at a lower energy, we can calculate the resonance escape probability,
and the expression we derived to tell how it depends on dilution cross sections, and thus use the experimental data
for a different set of conditions.
50
Lulu Li 22.211 Class Notes: February 27, 2012
5 References: Reuss Section 8.4; Handbook of Nuclear Engineering Chapter 4 Section 3; Duderstadt p.337-338; Bell p.433 (points out the
area under the cross section curve is constant; hence the absorption rate is proportional to the magnitude of the flux in a sense.)
51
Lulu Li 22.211 Class Notes: February 27, 2012
52
Lulu Li 22.211 Class Notes: February 27, 2012
3.8 Self-Shielding
Self-shielding/flux depression is first observed through:
Observation Explaination
Spatial if natural uranium was made into a lump U238 in the center of a pin is shielded
surrounded by moderator, resonance ab- from neutrons in the moderator
sorption
Energy if U/H , absorption rate per atom (for neu- the magnitude of absorption is proportional
trons slowed down in a water & U mix) to the width of the resonance;
In a sentence, lumped configuration shields the fuel region from neutrons with E E0 , hence decreasing res-
onance absorption and increasing p. As fuel diameter increases, the RI and therefore the absorption decreases.
For a homogenized medium, the value of p for natural uranium is about 0.70. However, larger values exist
for the heterogeneous case. This is because of a phenomenon called spatial self-shielding. The idea is to keep
the fuel and moderator separate. Neutrons slow down in a stepwise manner. Some collisions will cause neutrons
to jump over the energies that correspond to the U238 resonances. Others will land in resonance energies. What
happens to these neutrons? Suppose a neutron has an energy close to that of a U-238 resonance.
If a neutron scatters off U238 it will lose only a small amount of energy. In this case, the neutron is left near
the resonance energy.
If it scatters off moderator it will lose a lot of its energy. In this case the neutron is removed from that
energy.
What then is the effect of homogeneous or heterogeneous fuel? If the fuel and moderator are separate and
the neutron is in the moderator, chances are it will collide with another moderator atom and be scattered out
of the resonance region. So, heterogeneous arrangements favor resonance escape. In contrast, if the fuel and
moderator are infinitely mixed (homogeneous) or if the fuel rod/moderator volumes repeat on small scales with
dimensions less than a diffusion length, the next collision may again be with fuel and the neutron will be absorbed.
Heterogeneous arrays are sometimes called lumped meaning that the fuel and moderator are separate, widely
spread entities.
Notice both f and p depends on the moderator-to-fuel ratio and there is a trade-off between the two6 .
Energy Self-shielding: in a sentence, the strong absorption of the resonance tends to shield the absorber
nuclei from neutrons with energy E0 , hence creating the flux depression at E0 .
Prof. Forget states energy self-shielding as: when pure scattering, we can assume the thermal flux to be
Maxwellian. But as we increase absorption, the spectrum would harden. Example: compare spectrum of UOX
and MOX. UOX is 4% U235, 96% U238. MOX is 8% Pu (2/3 fissile, 1/3 parasitic absorber), 92% U238. The
reason MOXs thermal spectrum is depressed is,
utilization. For f to be large, one seeks to maximize absorption in the fuel and hence minimize spatial self-shielding which could lead to
absorption in the moderator. Thus, there is a tradeoff between f and p. Both depend on the moderator-to-fuel volume ratio. Optimal designs
have both f and p at about 0.9.
53
Lulu Li 22.211 Class Notes: February 27, 2012
Prof. Smith approaches self-shielding using RI values vs. U/H ratio. As we increase the U/H ratio, the RI
decreases in the three big resonance regions, and we see big dips on the spectrum plot. There are so much U238
in the fuel, that there is no flux in the fuel anymore.
As we increase the number of uranium atoms by a factor of 10, the number of absorption per atom is
decreased by a factor of 3. That is, the total aborption still increases, but the absorption per atom decreases.
Figure 12 illustrates that RI are very dependent on the density of resonant materials.
Recap: energy self-shileing means the decrease in absorption per atom (thus the depression of flux). Though
the flux between resonances are trying to remain 1/E.
54
Lulu Li 22.211 Class Notes: February 27, 2012
As Shultis states(p.286-297),
As the temperature of the core material increases, the thermal neutrons maintain a Maxwellian energy dis-
tribution one but shifted increasingly towards higher energies. The thermal spectrum is said to harden. If all the
cross sections in the core had exactly a 1/v dependence, there would be no change in the thermal-neutron inter-
action rates. However, most heavy nuclides have resonances near the upper end of the thermal energy range and
their cross sections, consequently, are not exactly 1/v in the thermal energy region. For example 239Pu has a large
resonance at about 0.3 eV, and, as the thermal neutrons shift in energy towards this resonance, more neutrons are
absorbed by 239Pu and thus the fission rate increases, thus producing a positive reactivity feedback. By contrast,
hardening the thermal neutron spectrum causes a very slight decrease in absorption by 235U.
This hardening of the thermal neutron spectrum with increasing temperature is taken to an extreme in the
TRIGA class of reactors which use as fuel enriched 235U blended in zirconium hydride. As the fuel temperature
increases, vibrating hydrogen atoms trapped in the zirconium-hydride crystal lattice can transfer some of their
vibratiorial energy (about 0.13 eV) to thermal neutrons, thereby removing them from the thermal energy region
so they are less likely to be absorbed by the fuel. This very rapid negative reactivity feedback effect is the
reason these reactors can be operated in a pulse mode, in which a large positive reactivity is inserted into the core
by rapidly removing a control rod to make the reactor very super prompt critical. However, the ZrH negative
temperature feedback acts within a few ms to stop the runaway chain reaction, which has increased the reactor
power by many thousands of times the initial power, and brings the reactor power back to safe limits.
Impact On Design:
k = pf (141)
55
Lulu Li 22.211 Class Notes: February 27, 2012
is fast fission.
F
R
F +T f (E)F dE
= R (142)
T
F
f F dE
eta is
F
R
f F dE
= RT (144)
T
F
a F dE
f is
V F T F
R
a f dE
f= R R (145)
VF T
F
a F dE + VM T Ma M dE
Insert image here. Maximum k gives us the optimal design (in terms of moderation). In reality we always
operate at slightly under moderated. With Boron in it, we need to operate at even under moderation, and have to
limit Boron concentration to 2000 ppm. Because now if we lose water, we not only lose moderation, but also lose
absorption, and when Boron concentration is to high, we lose more absorption than moderation, which actually
create a positive coefficient.
**** FIXME ****
As we increase H/U, there is still a finite absorption, but the absorption per atom approaches zero.
56
Lulu Li 22.211 Class Notes: February 27, 2012
Flux shape in unresolved region: probability tables and analytic functions. Example:
X Pi (E)xi (E)
i
0 + ti (E)
x (E) = X (147)
Pi (E)
i
0 + ti (E)
3. UNRESR: unresolved range, we have probability table that tells you the distribution of resonance spacings
etc. Analytical flux depression.
4. HEATR: energy deficient. We do not really discuss this part.
5. THERMR: deals with S(, ) to treat chemical binding, etc that we skip.
6. GROUPR: important one. NJOY evaluates integral numerically, whereas We mimic this process using
Monte Carlo because MC is easier to set up though is not as accurate.
7. GAMINR: not that important.
We get multigroup cross sections out of NJOY. As illustrated in Fig. 14, the multigroup cross section may differ
by a factor of 10 depends on the resolution you run. Thus we cannot blindly use a set of multigroup cross sections
without evaluating it specifically to a problem. Unless you are using 100,000 energy groups, no matter how good
your spatial representation is, your results is no better than the multigroup cross sections you start out with.
57
Lulu Li 22.211 Class Notes: February 27, 2012
58
Lulu Li 22.211 Class Notes: February 27, 2012
NJOY Flux Spectra Model and Resonance Model We assume a homogeneous system comprised of a
moderator (m) and a resonance material (r) with a fission source. Then we can write the balance equation of
consumption equals fission and scattering source:
Z E/R Z E/M
[NR tR (E) + NM tM (E)](E) = (E) + NR sR (E 0 0
E)(E ) dE + 0
NM sM (E 0 E)(E 0 ) dE 0
E E
(148)
s (E 0 )
Further assume scattering is isotropic in CMCS, we write s (E 0 E) = (1)E 0 :
E/R E/M
sR (E 0 ) sM (E 0 )
Z Z
[NR tR (E) + NM tM (E)](E) = (E) + NR (E 0 ) dE 0 + NM (E 0 ) dE 0
E (1 R )E 0 E (1 M )E 0
(149)
Assume the resonance region starts well below the fission neutron emission energy region, (E) 0, and the
moderator scattering xs is independent of energy, that is,sM (E 0 ) sM , (E 0 ) E10 ,
E/R E/M
sR (E 0 ) sM
Z Z
1 1
[NR tR (E) + NM tM (E)](E) = NR 0
(E 0 ) dE 0 + NM dE 0
E (1 R )E 1 M E E0 E0
E/R
sR (E 0 ) sM
Z
0 0
= NR (E ) dE + N M (150)
E (1 R )E 0 E
NR M
Recall the definition of background cross section b = ,
NR
E/R
sR (E 0 )
Z
b
[tR (E) + b ](E) = 0
(E 0 ) dE 0 + (152)
E (1 R )E E
59
Lulu Li 22.211 Class Notes: February 27, 2012
60
Lulu Li 22.211 Class Notes: February 29, 2012
R
As energy increases, E , thus we use NR model for the unresolved resonance energy range.
To start, recall the assumptions we made for slowing down equation in Section ??:
Steady state.
s (E 0 ) E
s (E 0 E) = for E < E 0 < (153)
(1 )E 0
The moderator has a constant s and a negligible a (at least over the resonance range):
M M M
t (E) s (E) = po (154)
61
Lulu Li 22.211 Class Notes: February 29, 2012
2. We facterize (u) = (u)(u), where (u) is the fine structure/self-shielding factor that models the
resonances, (u) is the flux outside the resonance (Hebert calls it macroscopic flux, and it represents the
asymptotic behavior of flux between resonances).
3. For the resonant material (R) and the moderating material (M), we define the slowing down operator using
the zeroth Legendre moment of the differential scattering cross section,
Z Z
RR = du0 Rs0 (u0
u)(0
) R M
= du0 M 0 0
s0 (u u)( ) (156)
0 0
But for RM , we apply the 1/E flux between resonances and get rid of the (u) term,
RM (u) = M
t (u) (159)
5. Then we can write (omitting u dependency for simplicity) and cancel out on both sides,
(M R R M
t + t ) = R + t (160)
(M
t + R
t )
R
=R + M
t (161)
(d + tR ) = rR + d (162)
62
Lulu Li 22.211 Class Notes: February 29, 2012
R R (E) d 1
R (E)+ E dE
Z Z
d E2 a
RIeff = R (u)W R (u) du = R (u) R
du = ln R d 1
d
a (u) + d E1 R (E)+ E dE
a d
(165)
Comments:
WRs sR (u) term on the bottom is the resonant materials scattering cross section including both
the potential scattering (hard sphere collision) and the compound nuclide scattering. That is, NR only
accounts for the potential scattering cross section, whereas WR uses the entire scattering cross section.
WR RIeff only depends on cross sections: d is assumed to be based on material composition of the
reactor, and r (E) are from libraries. There is no need for flux spectrum.
WR ignores U238 scattering, as there is no U238 scattering cross section in Eq. 164. Interpretation:
the resonance width is large compare with the approximately 1% energy lose upon a scattering, thus
we can ignore the U238 scattering. To improve this approximation, sometimes people add the potential
scattering of 11.39 barns.
WR approximations match direct MC results near infinite dilution.
2. Narrow resonance approximations assume resonance is very narrow, and a neutron is going to be out of the
resonance upon one scattering, thus cross section remains the same, thus (E) 1/E, (u) =constant.
Thus we can write,
Z u
1 0
rR (u) = R
e(uu ) sR (u0 )(u0 ) du0 (166)
u R 1
where (u0 ) = 1, sR (u0 ) po
R
potential scattering, also
Z u
1 0
R
e(uu ) du0 = 1 (167)
u 1
Z
RIeff = R (u)N R (u) du (170)
R
pot + d
Z
= R (u)
R R +
du (171)
a (u) + pot d
R
R R (E) pot +d 1
E2 R (E)+ R + E dE
a pot d
= ln R +
pot
(172)
E1 R d 1
dE
R (E)+ R +d E
a pot
Comments:
NR approximation is good for higher energy. For instance, at 100 keV, scattering lose 1 keV, which is
larger than the 25 eV spacing of 238U. For fast reactor applications for instance narrow resonance is
used a lot.
However, the results on the slides come out to be that narrow resonance approximation is not really any
better than the wide resonance approximation at higher energy, so there must be some other physics
going on, see next section.
63
Lulu Li 22.211 Class Notes: February 29, 2012
1. NR: assume the resonance width is small compared to the average energy loss in a collision with the
R
resonant material (since the average energy loss (12 )E0 , one can expect that NR is valid for high
energy):
R 1 R
p E = E0 (174)
2
Then the resonant scattering integral can be approximated by replacing (E) by its asymptotic form:
E/R R E/R R
R (E 0 )(E 0 ) dE 0 1
Z Z
p p
dE 0 a = (175)
E (1 R )E 0 1 R E E0 E0 E
M R
s + p M
s + p
R
t (E)(E) = N R (E) = (176)
E t (E)E
Or with Prof. Smiths notation, using microscopic cross section and background cross section,
R
pot + b 1
(E) = R
(177)
t (E) + b E
2. WR: assume the resonance width is wide compared to the energy loss:
R 1 R M
E = E0 p E (180)
2
WR can be referred to as NR infinite mass absorber (NRIM): resonant material is infinitely massive such
that neutrons suffer no energy loss in a collision with the resonant materials. That is, as AR , R =
(AR 1/AR + 1)2 1, and
E/R 0 0 0 E/R
R dE 0
Z Z
s (E )(E ) dE
lim R 0
R
s (E)(E) lim = R
s (E)(E) (181)
R 1 E (1 )E R 1 E 1 R
64
Lulu Li 22.211 Class Notes: February 29, 2012
M
s Ms
t (E)(E) = R
s (E)(E) + W R = (182)
E (t (E) Ms (E))E
Prof. Smith introduces WR model as Narrow Resonance Infinite Mass model, that is, we approximate the
resonance material as having an infinite mass, thus R 1, thus we can evaluate the scattering integral in
Eq. 173:
E/R Z E/R
sR (E 0 )
Z
0 0 1 1
lim R
(E ) dE = s (E)(E) dE 0 (183)
R 1 E (1 R )E 0 E 1 R E
1 1 E 1 1 1
= sR (E)(E) E = sR (E)(E) E E sR (E)(E) (184)
E 1 R E 1 R
Then our balance equation becomes,
b 1 b
[aR (E) + b ](E) = (E) = (185)
E E aR (E) + b
WR: intuitively, you cannot get in, and cannot get out, so the two sides cancel out. Once you get in,
you pretty much just get absorbed.
3. IR/Goldstein-Cohen Model: since isotopes are going to be somewhere between the two extremes, we con-
sider a linear combination of N R , W R ,
g po + d
GC (u) = (186)
to (u) (1 g )so (u) + d
where g depends on isotope and energy band. When g = 1 we get the N R , g = 0 we get the W R .
Alternatively, we can discretize our energy group very finely, then we would not need any resonance models.
4. CASMO Intermediate Resonance Model: assume fixed temperature (900 K), assume fixed background
cross sections (either 30b or 60b),
Ig (T = 900K, b = 60b) Ig (T = 900K, b = 30b, X = 30b)
1 g (M ) = (187)
Ig (T = 900K, b = 60b) Ig (T = 900K, b = 30b)
where g is the IR parameter, X represents the isotope of atomic mass M , and X is the contribution of
isotope X to the background cross section. The numerator of the above equation corresponds to the effect
of 30 barn of hydrogen to the resonance integral. The denominator correponds to the effect of 30 barn
of isotope X to the resonance integral. Therefore the above g can be interpreted as the effectiveness of
isotope X in terms of neutron slowing down capability relative to hydrogen. Hence g is called the
hydrogen-equivalence parameter. Actual IR model parameters g in CASMO:
Are based on 30 and 60 barn calculations with Monte Carlo.
Are tabulated for each mass of scatter ( dependence).
Are tabulated for each actinide.
Are tabulated for each resonance group (41 groups in CASMO).
g does not depend on temperature nor background cross section.
65
Lulu Li 22.211 Class Notes: February 29, 2012
2. As in Figure 16, WR agrees pretty well with MC results without U238 potential scattering; NR agrees
alright with the MC results with U238 potential scattering. Notice adding in scattering would change the
low dilution factor results.
3. Compare with MC results using real U238 and U235 xs data, the real results fall between the wide and the
narrow approximations for the high dilution factor; but for the low dilution factor, that is, when there are
lots of resonance material (uranium), the physics gets complicated:
Non 1/E slowing down sources;
Non constant moderator xs;
Resonance scattering alters sources;
We may have multiple resonance in each RI group; that is, each of our RI does not include a single
resonance anymore;
Resonance interference between 235U and 238U; eg: 238Us resonance spacing is 25 eV, whereas 235Us
resonance spacing is 5 eV, that is, for each 238U resonance, there are multiple 235U resonances on top
of it.
4. Helstrand Resonance Integral Measurements (for isolated rod): LWR has a radius of 0.4 to 0.5cm, but
a lattice of pins can be approximated with fatter isolated pins. So CASMO-5 results are about 2% different
from Helstrand RI measurements, which is equivalent to about 1% difference in Doppler coefficients.
66
Lulu Li 22.211 Class Notes: February 29, 2012
There are two possibilites in using NR approximation to calculate the condensed cross section depends on the
number of resonant isotopes:
1. One resonant isotope. Assume we have H2 O and 235U.
NH H + NO O
d = (190)
NU
We evaluate using narrow resonance approximation:
po + d
N R (u) = (191)
d + to (u)
Then the flux-weighted resonant material cross section is,
R
U (u)(u) du
U,g = R (192)
(u) du
235
2. Multiple resonant isotopes. Assume we have H2 O, U, and 238U.
Solution 1 (Livolant-Jeanpierre): The idea is, we have to approximate d now that there are two resonant
isotopes. The reason we emphasize on d is because in production tool we resolve the spectrum/group cross
section for a number of dilution cross sections, and the data provided to the downstream applications would
be a function of the diluion cross sections.
(a) Assume 235U is the only resonant isotope, then we solve an energy independent d by
N238 238 + NH H + NO O
d = (193)
N235
Then we solve for (u):
po,235 + d
N R (u) = (194)
d + to,235 (u)
Then calculate the collapsed 235 :
R
235 (u)(u) du
235 = R (195)
(u) du
7 the only energy/lethargy dependent term on the RHS should be to (u).
67
Lulu Li 22.211 Class Notes: February 29, 2012
N235 235 + NH H + NO O
d = (196)
N238
Then we solve for (u):
po,238 + d
N R (u) = (197)
d + to,238 (u)
Then we feed 238 back into step 1 and repeat the process until 235 , 238 converge, which happen fairly
fast with a simple resonance.
Solution 2 (Dederstadt):
In Eq. 195, (u) shows up in both the top and the bottom of the equation, so if we scale (u) by any
non-energy dependent term as they would just cancel out, the solution would not change. Recall (u) as in
Eq. 194, the top po,r + d is energy-independent, and Nr is energy-independent as well, that is,
po,r + d 1 1 1
N R (u) = X
d + to,r (u) d + to,r (u) Nr d + Nr to,r (u) Nm m + Nr to,r (u)
moderators
(199)
Then we dont have to compute d anymore. For each resonant isotope, the three steps becomes two steps:
calculating using last steps condensed cross section, and update this steps condensed cross section with
-collapsed cross section.
So effectively narrow resonance suggest,
1
(E) = (201)
Et (E)
68
Lulu Li 22.211 Class Notes: February 29, 2012
3. Evaluate pin-wise Dancoff factor from pin diameter, lattice pitch, etc.
4. Compute coefficients of rational expressions, e.g., Carviks 2 term:
4A+6
5A + 6 A2 + 36A + 36 5A + 6 + A2 + 36A + 36 A+1 1 1C
1 = , 2 = = A=
2(A + 1) 2(A + 1) 2 1 C
(203)
69
Lulu Li 22.211 Class Notes: February 29, 2012
5. Compute the escape cross section for each resonance absorber from pin diameter (be careful e s unit is
b/resonant atom):
e 1 Sf 1
e = = = = (204)
Nr lNr 4Vf Nr 2rNr
6. Compute potential xs using Intermediate Range s and isotopic number densities and non-resonant multi-
group cross section. FIXME.
7. Interpolate in RI tables and add up terms to get final RIeff and multi-group xs for each resonant nuclide:
u2 F u2 F
pot + 1 e pot + 2 e
Z Z
RIeff = r (u) F F +
d + (1 ) r (u) F F +
d (205)
u1 r (u) + pot 1 e u1 r (u) + pot 2 e
For real application, it is rarely possible to just use someone elses cross section table; we almost certainly need
to generate xs suitable for our application.
70
Lulu Li 22.211 Class Notes: March 5, 2012
e
2 (207)
4R2
The probability of going from A to any point in region 2 and colliding is:
e
Z
2 dV2 (208)
4R2
The probability of a uniform spatial neutron source in region 1 going to any point in region 2 and colliding
is:
e
Z Z
dV1 2 dV2 (209)
4R2
First flight probability Pij describes the probability that a neutron born in Vi will make its first collision
in Vj (the average probability per unit volume of a uniform source neutron going from region 1 to 2 and
colliding):
e
Z Z
2
P12 = dV1 dV2 (210)
V1 4R2
71
Lulu Li 22.211 Class Notes: March 5, 2012
Likewise,
e
Z Z
1
P21 = dV2 dV1 (211)
V2 4R2
Since the two integrals are symmetric (that is, PAB and PBA are essentially the same), we can write,
V1 V2
P12 = P21 (212)
2 1
Hence we reach the reciprocity condition:
This is entirely general in geometry so far (maybe require convex geometry?) as we are just saying that
the attenuation coefficient is the same forward and backward.
3. Two Region Balance Equations. First collision reaction rate balance, where t,f (u) = r,f + pot,f (the
resonance absorption xs plus the potential scattering xs):
fuel region lost neutrons fuel region gain neutrons from slowing down sources
z }| { z }| {
Nf t,f (u)(u)Vf = Qf Vf Pf f + Qm Vm Pmf (214)
Because slowing down sources in both fuel and moderator are assumed to arise from the 1/E flux above the
resonance energy, we can write
Plug the slowing down sources back into the reaction rate balance equation, and recall (u) = (u)(u)
where (u) is the fine structure energy shape of the flux for homogeneous mixture, we extand the definition
to define f (u) as the heterogeneous energy shape of flux in fuel,
Pmf Nm m Vm = Pf m Nf f Vf (217)
Divide both sides by the number of neutrons, we get a balance equation in unit of cross section:
That is, if we know Pf f , we can solve for the spatially averaged energy shape of the flux in the fuel.
4. Alternative derivation (approached using resonance scattering operators Rf , Rm ) presented by Prof. Forget:
We start by writing two region balance equation:
72
Lulu Li 22.211 Class Notes: March 5, 2012
Vf Rf Pf f + Vm tm Pmf = Vf tf (224)
Rf tf 1 Pf f tf
+ = (225)
Nf Nf Pf Nf P f f
That is,
rf + d = (tf + d ) (226)
tf (1Pf f )
which is the same form as the homogeneous model, except with different definition of d = Pf f .
73
Lulu Li 22.211 Class Notes: March 5, 2012
pot,f + e pot,f + e
f (u) = = (231)
t,f (u) + e r,f (u) + pot,f + e
This is important as it shows that the only energy dependency is from r,f (u).
Heterogeneous/Homogeneous Equivalence suggests that the effects of heterogeneity are equivalent to some
dilution cross section in a homogeneous mixture:
pot,f + d Nm m
Homogeneous mixture energy shape of flux (u) = d = (232)
r,f (u) + pot,f + d Nr
pot,f + e Sf
Heterogeneous energy shape of flux in the fuel f (u) = e = (233)
r,f (u) + pot,f + e 4Vf Nr
Notice in the homogeneous case, moderator (hydrogen for instance) is dominant. Whereas in the heterogeneous
case, we are dependent on hydrogen at all. The reason is because if we have a pin surrounded by an infinite
sea of moderator, then as soon as neutrons get out, they are not likely to go back, so it does not matter what the
moderator cross section is.
Examples:
238
1. Calculate corresponding cross sections. Given a PWR with a 0.41 cm radius pellet, U number density
of 4.0 1022 cm3 , enriched to 3%, and fuel density is 10 g/cc.
If the system is homogeneous uranium and hydrogen, then the dilution cross section looks like:
Nm m 1
d = = 20 barn = 100 barn/resonance atom (234)
Nr 0.2
If the system is 2 region, and we homogenize the fuel region, then we can calculate the escape cross
section (be careful about the unit!)
Sf 2rh 1 1 1barn
e = = = = = 55 barn/resonance atom
4Vf Nr 2
4r hNr 2rNr (2)(0.41cm)(2.2 10 cm ) 1024 cm2
22 3
(235)
74
Lulu Li 22.211 Class Notes: March 5, 2012
and for the potential scattering cross section we just add up that from every isotope,
bhom = 0.05pot
235 O
+ 2pot = 0.05(11.3) + 2(4.0) = 9 barn (236)
bhet = bhom + e = 64 barn (237)
sF = R
pot + bhom = pot
238
+ 235
0.05pot + O
2pot = 11.4 + 0.05(11.3) + 2(4.0) = 20 barn (238)
2. (12 Qual #4) Calculate escape probability. Given a cylinder of radius r = 0.5cm, infinite height, and
= 10cm1 . Assume uniform source inside, calculate the probability of neutrons making first collision
outside the cylinder.
4V 4r 2 H F 10
Answer: we calculate chord length l = S = 2rH = 2r, so PF F = 1 = = 91%, then
F + l
10 + 1
the probability of it making it out should be 9%.
75
Lulu Li 22.211 Class Notes: March 5, 2012
[FIXME]: add materials from 2013 fall. We would typically compute a table of PF F as a function of opacity,
and sample from it on the fly.
76
Lulu Li 22.211 Class Notes: March 5, 2012
Collision probability only gives us the average behavior in space and in energy. For instance, in CASMO
when we run a 2D case we only have 8 energy groups.
[FIXME] Lecture 9, p.6. An error in PF F leads to a bigger error in flux. Group flux is not very sensitive
on the resonances (because the width of the resonances are small), so we need accurate group cross sections to
capture the resonance behavior.
77
Lulu Li 22.211 Class Notes: March 5, 2012
In Fig. 18, the x-axis is the radius of the rods (in units of the mean free path of the moderators), and the curve also
depends on the lattice size/radius of the rods (know that in a uniform square lattice, the lattice size is the pitch).
Notice if the moderator is opaque, the Dancoff factor becomes zero; if transparent, the Danoff factor trends
(1C)b
C changes the coefficient in front of the escape cross section from e 1C+Cb e . For a typical PWR pin, that
is reduced by 0.7 to 0.8.
D f N
d = + (244)
lN0 NN
+
(1C)b +
where b = (1C)+Cb + where b is the b factor for a single pin cell.
Example: Find Dancoff Factor. Given water at 1 g/cc with a number density of 6.6 1024 H atoms/cc,
hydrogen xs of 20 barns, a PWR pin with 0.41 cm radius, a lattice pitch of 1.26cm.
We first find the mean free path:
1 1 1
mf p = = = = 0.76cm (245)
N 0.066 10 20 1024
24
0.41cm
Then we find the parameter one, pellet radius over mfp = 0.76cm = 0.54, the other parameter is the lattice
1.26cm
size over pellet radius 0.41cm = 3.07. Using these two parameters, we can read off the plot that C = 0.3.
For a Bell factor of about 1.1, the escape xs is reduced by,
(1 C)b
= 0.774 (246)
(1 C) + Cb
78
Lulu Li 22.211 Class Notes: March 5, 2012
Remember 60 barns as the approximate average cross section for PWRs. See slide 22 of Lecture 8.
From 22.212 HW7, Dancoff factor typically assumes the fuel to be a black absorber (with source placed on
the surface), because ultimately we care about the scattering from moderator to fuel. Production tools typically
generate tables of Dancoff factors as a function of energy, fuel cross section (sub-group methods: as a function of
the band of cross sections, that is, the width of resonance cross section energy) etc.
79
Lulu Li 22.211 Class Notes: March 5, 2012
t,f (1 )t,f
Pf f + (247)
1 + e + t,f 2 e + t,f
where
4A+6
5A + 6 A2 + 36A + 36 5A + 6 + A2 + 36A + 36 A+1 1 1C
1 = , 2 = = A=
2(A + 1) 2(A + 1) 2 1 C
(248)
Carlviks two-term approximation is what is actually used in production tools nowaday. To use it, we essentially
look up the resonance integral (or group xs) twice, once with e multiplied by 1 and the second time with e
multiplied by 2 , each of them are simple functions of the Dancoff factor:
Z u2 Z u2
pot,f + 1 e pot,f + 2 e
RIeff = r (u) du + (1 ) r (u) du (249)
u1 r,f (u) + pot,f + 1 e u1 r,f (u) + pot,f + 2 e
To double check,
C 1, 1 0, 1, e.g., there is no escape from fuel.
C 0, 1 2, 2 3, 2, e.g., isolated rod.
For intermediate values of C, each coefficient is a monotonic function.
The beauty of this method is that the formula stays the same, we just have pre-tabulated table of the first and
second terms depend on background xs. So given a background xs, we just need to table look-up the two terms
and add them up.
Recap: moderator volume, cross section, density do matter when we come to arrys of rods through Dancoff factor.
80
Lulu Li 22.211 Class Notes: March 5, 2012
Figure 19: Algorithm for a Simple Pin Cell Code with Collision Probabilities
1. With no oxygen or zirconium: flux dips in moderator are mild because there is only elastic scattering. See
Fig. 20.
2. Adding in Zr: Zr has massive elastic scattering resonances, hence creating fluctuations in both spectrums
most noticable in energies right above the resonance energies. The reactivity change is small (positive 560
pcm). See Fig. 21.
3. Adding in O: the fast flux looks distorded in both fuel and moderator, this is because the oxygen xs has
weird dips in high energy xs. The reactivity change is not large (negative 680 pcm). See Fig. 22.
4. Removing hydrogen absorption: the spectrum shape does not change much, but the reactivity change is
huge (8808 pcm), because there are so much hydrogen, and that hydrogen xs is 1/v down to 100 keV. This
is why LWR fuel has to be enriched. See Fig. 23.
5. Remove U238 fission: reduce reactivity by 2% (negative -1910 pcm). Spectrum shape does not change.
See Fig. 24.
6. PWR enrichment worth 4.5% k per %.
7. Doppler reactivity example: PWR Doppler worth = -2.7 pcm/K. It is important to generate the Dopplar
coefficient when preparing cross section data.
8. Boron reactivity: Borons capture cross section is almost entirely (n, ) cross section which produces He.
Boron worth: 16 pcm/ppm.
Looking at the flux in the moderator divided by the flux in the fuel, we see that:
Above resonance energies: fuel flux is depressed because absorption is high compared with fission;
In resonance energies: larger than 1 because of fuel spectrums dips for the resonances; especially at the
interval containing 6.7 eV;
Below resonance energies: the ratio is around 1, till we hit the really low energy (less than 0.1 eV), that
there is a 1/E tail that the fuel flux gets depressed.
81
Lulu Li 22.211 Class Notes: March 5, 2012
82
Lulu Li 22.211 Class Notes: March 5, 2012
83
Lulu Li 22.211 Class Notes: March 5, 2012
Clad;
Coolant;
Other in-assembly moderators;
Structural can;
Moderator structure/tank.
In real life, there is no unique model exists in 1D; real analysis is done in 2D or 3D now.
84
Lulu Li 22.211 Class Notes: March 7, 2012
F
pot F
pot VF
fF () = PF F + PF M fM () = PM M + PF M (250)
Ft M V M
Sensitivity: a 0.03 average error in PF F would lead to 3% over-prediction in U-238 capture; consider that
U-238 capture makes up 20% LWR capture, we are looking at an keff error of: -20000 pcm times 0.03
which is -600pcm.
3. We can also compute Bells factor/Bell function as a function of fuel pin opacity:
OF () 1
PF F = b(O) = OF () 1 (251)
b + OF () PFtrue
F
Before we start, notice there are multiple conventions about defining coordinate systems as illustrated in Fig. 26.
We choose to use Fig. 27.
85
Lulu Li 22.211 Class Notes: March 7, 2012
Figure 27: Our Choice of Coordinate Systems and Area of Unit Surface
86
Lulu Li 22.211 Class Notes: March 7, 2012
1. The probability that a neutron will escape from the region without a collision is:
Probability of escape = es( ~r , ) (252)
in volume dV is,
2. The probability that a neutron is generated in d about
d dV
Probability of generation = (253)
4 V
3. The escape probability for neutrons born in volume V is the integral of the product of the two probabilities
derived above over all directions and volume:
Z Z
1
Pesc = es( ~r , ) d dV (254)
4V
87
Lulu Li 22.211 Class Notes: March 7, 2012
, with a typical
4. By graphical illlustration, we can express volume as parallel tubes to a fixed direction
n
tube of length R and a cross sectional area dA:
dV = ( n ) dA dR
(255)
Alternatively we can also consider the volume: regardless of the orientation, we can generalize that the
n
projection of a curve surface dA onto the direction axis of the tube is dA
2R n
V = dA = 2R dA (256)
n
Notice the orientation of the tubes determined how many of them there are to represent the geometry as
illustrated in the left three plots of Fig. 29.
5. If we plug Eq. 255 int Eq. 254 and integrate over the path s, we get
ZZ
1 (1 eRs ) d dA
Pesc =
n (257)
4V
>0
n
88
Lulu Li 22.211 Class Notes: March 7, 2012
expression by defining:
Next we simplify the n
2R p p
= x = 1 2 cos y = 1 2 sin z = (263)
n
Then we consider n
to be perpendicular to the planar surface, that is, n
= (1, 0, 0), then
p
1 2 sin p
n= = 1 2 sin (264)
Plug back into Eq. 262 and
1
Z 1 Z /2 p 2
Pesc = d d 1 2 sin (1 e2R/( 1 sin ) ) (265)
R) 0 0
(a) Analytical solution: see Bell & Glasstone 2.112-2.117 for details,
1 1
Pesc = E3 (2R) (266)
2R 2
(b) Numerically integrate, we replace each integral with a summation with a weight:
X X p 2
Pesc = 1 2m sin (1 e2R/( 1m sin n ) ) (267)
R
m n
2 2 2R
where = nm , = n . If we let m,n = p , then
1 2m sin n
X X p
Pesc = 1 2m sin (1 em,n ) PF F = 1 Pesc (268)
R
m n
7. Pesc implies escape without collision. If we further assume that the media is purely absorbing, that is, once
a neutron leaves the fuel it would never returns, then we can write Pesc = PF M .
8. Note: in polar angle, instead of uniformally pick , we uniformally pick = cos . We just sample u/#
polar angles. In the azimuthal angle, we sample 2/# azimuthal angles.
Results: the more polar angles and azimuthal angles we have, the better agreement we have with the Bell factor
curve. 20 polar angles and 20 azimuthal angles seem to be sufficient for this simple infinite slab problem.
89
Lulu Li 22.211 Class Notes: March 7, 2012
1x 1x
Loop through all [0, /2] and set = sin for sin > 0, or = sin for sin 0.
X F
2
Accumulate PF F = e R / 1 .
Results: with 100 spatial mesh, 10 polar angles, 20 azimuthal angles we get good agreement. Notice the sensitiv-
ity on the number of spatial mesh: it takes 60 spatial meshes to get a close agreement. That is, the advantage of
method of chords is that by doing math we remove dependency on large number of spatial meshes.
90
Lulu Li 22.211 Class Notes: March 7, 2012
Results: 100k neutrons per material generate wavy curve; having 1,000,000 neutrons generate smooth curve.
91
Lulu Li 22.211 Class Notes: March 7, 2012
92
Lulu Li 22.211 Class Notes: March 7, 2012
For each pin, loop over all isotopes, interpolate the tables for two terms, and then multiply the corresponding
or 1 that only depends on Dancoff Factors. The pre-computed table stores RI as a function of background
cross section, isotope type, energy.
93
Lulu Li 22.211 Class Notes: March 7, 2012
94
Lulu Li 22.211 Class Notes: March 7, 2012
95
Lulu Li 22.211 Class Notes: March 7, 2012
2. Energy fine structure and a piece of the geometry treatment: Dancoff factor. We can tabulate Dancoff
factors and PF F as a function of coolant cross sections.
3. For this model, we do not have any IR (intermediate range) s and no isotropic number densities.
4. Tracklength sampling: let f (x) be the probability of starting at 0 and collide in dx about x,
Z x
f (x) = et x dx F (x) = f (x0 ) dx0 = 1 et x (269)
0
ln
x= (270)
t
96
Lulu Li 22.211 Class Notes: March 7, 2012
7. History-based variance estimators cannot be incremented and squared until this history ends:
!
2 1 X 2
s = x x2 (271)
N 1 i i
(a) Multiple cell typically uses natural BC as illustrated in green in Fig. 38.
(b) Reflective: illustrated in blue in Fig. 38.
(c) Opposite face periodic BC: illustrated in yellow in Fig. 38. This BC is good for repeating geometries,
especially checker-board repeating geometries the ABCD pattern in the left plot of Fig. 39. For more
specific geometries, full core periodic BC and quarter core mirrow BC are equivalent.
(d) Right-hand rotational BC: all 1st cycle PWR are octance symmetric, then we shuffle fuels, and place
the depleted fuel assemblies on the axis.
Our core is not diagonally symmetric, and 1/4 core rotational symmetry introduces lots of flex-
ibility (e.g., allows us to balance reactivity along major axes instead of placing two fresh fuels
next to each other).
97
Lulu Li 22.211 Class Notes: March 7, 2012
Notice 1/4 mirros is not possible any more; instead we simulate 7 cells (as surrounded by the dark
red lines in Fig. 39), where these 7 cells can be arbitrary.
The advantage of keeping full nodes (instead of modeling the geometry marked by CLon the
graph, that is, keeping 1/2 node on the axis and 1/4 on the corner) is to preserve the truncation
error and thus get the same results regardless of whether we are solving quadrature, quarter or
full core.
Figure 39: Pin Cell Code Boundary Conditions: Opposite Face Periodic and Rotational
9. k is the number of absorption should equal to the total number of neutrons; in our case, we are losing a
small portion of neutrons to the cutoff energy; we can cheat by pushing those neutrons back up.
10. We almost always tally reaction rate or flux. Then at the end of the tallying, k and effective cross sections
can be calculated from:
X X
f xg g
k = X xg = X (274)
a g
11. Flux disadvantage factor: (flux of the fuel times fuel volume) / (flux of the moderator times moderator
volume).
12. For debugging purpose, test the following limits: compute single rod PF F for F large and small; computed
single rod and rod array PF F for thick pitch limit; compute rod array Dancoff factor for thick pitch limit,
M = 0, .
Results: our Monte Carlo tools predict LWR parameters reasonable accurately,
1. k has a bias of about -700pcm relative to CASMO-5;
2. Reactivity (differential) effects are very accurate:
Enrichment worth is within a couple of percent;
Doppler worth is within a couple of percent;
Boron worth is within about 5 percent;
3. We should be able to use this tool to generate accurate differential reactivity effects and cross section data
for simple thermal reactor lattices;
98
Lulu Li 22.211 Class Notes: March 7, 2012
99
Lulu Li 22.211 Class Notes: March 7, 2012
ax2 + by 2 + f xy + px + qy + d = 0 (275)
100
Lulu Li 22.211 Class Notes: March 14, 2012
3. In an elastic scattering
2
Emin A1
= = <1 (278)
E0 A+1
ln(Ei /Ef )
5. Number of collisions to slow neutrons from Ei Ef : n = .
s
6. Moderating power = a , a measurement of the slowing down ability.
M
7. Flux disadvantage factor (HW4) = , and thermal group has > 1, fast group has < 1.
F
8. Spectrum summary:
(a) E < 1 eV (thermal range): a hardened Maxwellian distribution (hardened because the tempera-
ture equilibrium has not been perfectly achieved) plus a 1/E correction at higher energies, (E) =
M (E, Tn ) + (E/kT
E
n)
.
1
(b) 1eV < E < 50 keV (slowing-down domain): (E) (E)t (E)E .
(E)
(c) E > 0.5 MeV: (E) = t (E) .
9. Spectrum peaks:
(a) A hump in high energy. Reason: fission spectrum, degraded due to scattering;
(b) A slight decrease in the epithermal region. Reason: resonant capture losses, esp. U238;
(c) A hump in low energy. Reason: Maxwell distribution of the thermal agitation but a little harder
because the temperature equilibrium has not been perfectly achieved.
101
Lulu Li 22.211 Class Notes: March 14, 2012
3. CMCS: cos is uniformly distributed, that is, elastic scattering is isotropic in CMCS.
1
P (Ei Ef ) = Ef [Ei , Ei ] (281)
(1 )Ei
102
Lulu Li 22.211 Class Notes: March 14, 2012
(E)Sf
t = Sf (E) = (284)
t
dq(E)
= a (E)(E) (286)
dE
For regions that we can assume a = 0 (e.g., between resonances),
q(E)
q(E) = constant, (E) = (287)
s (E)E
4. Thermal range: assume thermal equilibrium, pure scattering xs, infinite medium, we get a Maxwellian
spectrum. See next section for more details.
103
Lulu Li 22.211 Class Notes: March 14, 2012
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Scattered Energy Over Initial Energy Scattered Energy Over Initial Energy
12 Exam 1 # 8: we are given a pdf symmetric about E 0 /E = 1. Whats the probability that a 0.0253 eV
neutron would be scattered to a higher energy in a collision with this thermal gas at 293K?
Answer: 0.5, because of the symmetry.
12 Exam 1 # 9: the cdf of this kernel predicts a probability of 0.0333 for a 0.025 eV neutron to be scattered
to energies greater than 0.05 eV in a material at 300K. If one uses the same curve to model the gas at 1200K,
what would be the probability that 0.10 eV neutrons would be scattered above 0.20 eV in the same gas at
1200K?
Answer: 0.0333. The ratio is all it matters.
12 Exam 1 # 10: draw and label the scattering kernel that you would predict for kT = 20.
Answer: the plot is probability vs. Eout /Ein , and it should be a box.
2. Add target motion/Doppler temperature effects: 12 Exam 1 #4,5: If the neutron cross section is independent
of energy at a temperature of 0K, what energy shape do you expect for the neutron xs at 1200K? Why?
Answer: 1/v shape, thermal motion/vibration.
2
A+1
3. Add chemical bond. bound = free . Know the S(, ) term.
A
104
Lulu Li 22.211 Class Notes: March 14, 2012
(b) Finite dilution: U/H , RIeff because of energy self-shielding: as U/H increases, more and more
self-shielding, the flux is more depressed, result in smaller RIeff . A second consequence is that the
higher energy range becomes more important. See Section 3.9.
(c) Temperature effects: T , RIeff becasue the broadened resonance increases the energy range over
which abosprtion occurs. There is less flux depression, whereas the area under the cross section is
about the same, result in an increase in the RIeff .
Spring 2012 Exam 1 #16: What is the value of the 300K infinite dilute RI for absorbium integrated
from 0 to 20 keV? (we calculated in the previous problem the RI for 0K.)
Answer: the infinite dilute RI is independent of temperature because in Doppler broadening the
area under the resonance is constant.
105
Lulu Li 22.211 Class Notes: March 14, 2012
Infinite dilute implies that here is no neutrons being absorbed by the resonances. Recall that in HW2 when
our U/H is small, the flux is not perturbed by U238 resonances at all, suggesting that the resonance escape
probability is one.
3. Dilution factors/dilution cross section/background cross section, where r is resonance material, m is mod-
Nm m
erator, d = .
Nr
4. Resonance Integral
(a) Calculation, 2012 Spring Exam 1, #13,#15, notice is given independent of E for simplicity,
Z u2 Z E2
1 E2
RIeff = (u) du = (E) dE = ln (291)
u1 E1 E E1
where is the resonance xs, not including potential xs. If there are multiple resonances, sum them up
with flux-weighting.
(b) RI represents the average absorption xs characterizing the resonance, averaged over the flux within
the resonance. It conserves the reaction rate/collision density.
(c) Relating to g , d :
Z u2
E2 d
RIeff = g ln RIeff u1u2 = aR (u) R
du (292)
E1 u1 a (u) + d
(d) Narrow resonance RIeff approximation: assume any scattering with the resonant material scatters
neutrons below the resonance energy. Spring 2012 Exam 1 #18, assuming 1/E spectrum, homoge-
neous mixture of a scattering material and a resonance absorpting material.
Z E2
pr + d 1 pr + d E2
RIN R = r (E) dE = r ln (293)
E1 r (E) + pr + d E r + pr + d E1
(e) Wide resonance RIeff approximation: assume scattering with the resonance material leaves the neu-
tron within the resonance energy and they will be absorbed. Spring 2012 Exam 1 #19.
Z E2
d 1 d E2
RIW R = r (E) dE = r ln (294)
E1 r (E) + d E r + d E1
106
Lulu Li 22.211 Class Notes: March 14, 2012
t,f (u)
Pf f = (296)
e + t,f (u)
5. Bells refinement of the Wigners rational models, where b depends on geometry and material (opacity) as
plotted in Fig. 17.
t,f (u)
Pf f = (297)
be + t,f (u)
1
where escape cross section is calculated for a cylinder to be e = 2r . Upon getting Pf f , we find Pf m =
1 Pf f , then use reciprocity to find Pmf . See Spring 2012 Exam 1: #20.
6. Dancoff factors: ratio of probability of next collision happens in a different rod over probability of next
collision in any rod. See Fig. 18 for plot.
If the moderator is opaque, the Dancoff factor becomes zero; if transparent, the Danoff factor trends
to 1.0.
r , C .
P/D C .
Spring 2012 Final #22 Using Bells approximation, estimate the ratio of escape cross section of an isolated
0.5 cm radius rod to the fuel escape cross section in a uniform hexagonal lattice with a Dancoff factor of
0.6?
107
Lulu Li 22.211 Class Notes: March 14, 2012
(1 C)b
Answer: Essentially we are asked to approximate . We are already given the Dancoff factor
(1 C) + Cb
C = 0.6. Recall the plot of Bells factor for various geometries, we approximate b as 1.1 (as in the examples
in lectures), then
e (1 C) + Cb 0.4 + 0.6b
(1C)b
= = = 2.4 (299)
(1 C)b 0.4b
(1C)+Cb e
7. Solving for spatially averaged energy shape of flux in the fuel8 from the first collision rate balance equation.
We construct a first collision rate balance equation (next collision for neutrons in fuel from slowing down
source and uncollided flux):
where Q is the slowing down source, and it is assumed to arise from the 1/E constant lethargy flux above
the resonance energy,
Substuiting and dividing by , we get an expression of the fine structure energy shape of flux ff = f /,
Comparing with the narrow resonance expression for the homogeneous mixtures energy shape of flux, we
notice the only difference the two e are replaced by d .
108
Lulu Li 22.211 Class Notes: March 14, 2012
(c) Simple statistics: need enough to provide resolution, but not too many that each bin does not get
enough particles.
2. Pin-cell code:
M
(a) Flux disadvantage factors: F
.
(b) Fine structure effects.
(c) Scattering kernel: for larger than 4eV, asymptotic down scattering. For less than 4eV, thermal scatter-
ing, approximaetd by elastic scattering for Monatomic/Maxwellian free gas.
(d) Data:
Source from fission spectrum: chi, Maxwell spectrum.
sH : read in.
aH : 1/v absorber.
sU : resonance scattering cross section using , .
Spring 2012 Exam 1: if the neutron cross section is independent of energy (flat) at a temperature of 0K,
what energy shape would you expect for the neutron cross section at 1200K? why?
(E) 1/v.
Phenomenon: neutron energy is on the same order as the thermal vibration of the interaction material.
109
Lulu Li 22.211 Class Notes: March 14, 2012
4. 1/E fluxes: true except when resonant absorber is present (recall we keep using 1/E for spectrum above the
resonance).
It is characteristic of the scalar flux in the slowing down energy range, as long as there is no up-
scattering and no fission source.
It arises basically from the kinematics of the scattering interaction (asymptotic elastic scattering, more
specifically). However it gets distorted by the energy behavior of the scattering cross sections and by
neutron-absorption processes (Henry, p. 90).
Stacy (p.103) derives from slowing-down equation that, with Hydorgen as the only moderator, and
2
very heavy nuclei with j = A1 A+1 :
X Z E/j j (E 0 )(E 0 )
dE 0
Z
0
t (E)(E) = H
s (E ) + s
dE 0 (306)
E E0 E E 0 (1 )
j
j6=H
dE 0
Z X 1
0
= H
s (E ) + j (E)(E) (307)
E E 0 j s
j6=H
(E 0 )
Z
H H dE 0
a (E) + s (E) = s (308)
E E0
" Z #
E1
(a (E1 ) + H
s )E1 (E1 ) a (E 0 ) dE 0
(E) = exp (309)
(a (E) + Hs )E E (a (E 0 ) + H
s )E
0
1
suggesting that the neutron energy distribution varies with energy as (E) (a (E)+H
.
s )E
5. Maxwellian shapes.
Fission emission spectrum: peak occurs at 1.7 MeV, average occurs at 2 MeV.
Thermal flux spectrum: for an infintie source-free medium with small absorption cross section (as
long as a (E) is small compare to s (E)), the thermal spectrum is Maxwellian. The higher energy
portion of the thermal region is approximately 1/E with slowing down source and absorption present
(Henry, p.98). Stacy (p. 109) provides a more detailed reasoning for why neutron energy distribution
in the thermal range can be approximately by Maxwellian distribution: the neutron balance equation
in the thermal energy range is,
Z Eth
t (E)(E) = s (E 0 E)(E 0 ) dE 0 + S(E) (310)
0
For the thermal range, we assume no absorption a (E) and no slowing-down source S(E), and we
extend the upper limit on the integral to infinity under the assumption that the scattering to energies
greater than Eth is zero; then the neutron flux balance is,
Z
s (E)(E) = s (E 0 E)(E 0 ) dE 0 (311)
0
It can be shown that a Maxwellian distribution satisfies the above equation. However, absorption,
leakage and a slowing-down source would distort the actual spectrum from a Maxwellian. Since
most absorption xs vary as 1/v, absorption preferentially removes lower-energy neutrons, effectively
shifting the spectrum to higher energies than a Maxwellian at the moderator temperature T.
6. Group cross sections: the contribution of the resonance to a flux-weighted multigroup cross section for
reaction type :
R Eg1
E
(E)(E) dE
g
= gR Eg1 (312)
Eg
(E) dE
110
Lulu Li 22.211 Class Notes: March 14, 2012
1
Assume E,
X
RIi
RIeff ig
g = = E
(313)
ln(E2 /E1 ) ln Eg1
g
111
Lulu Li 22.211 Class Notes: March 14, 2012
1e+06
100000
0.01 1 100 10000
Energy Associated with Lethargy Bins (eV)
1e+06 1e+08
asymptotic elastic scattering in hydrogen with source neutrons from 2 MeV
Slowing Down in Hydrogen From Neutron Emission Spectrum
1e+07
50 generations, 1 million neutrons, 1000 bins
1e+06
Flux (not normalized)
100000
10000
1000
100
0.01 1 100 10000
Energy Associated with Lethargy Bins (eV)
1e+06 1e+08
replace source neutrons with neutron emission spectrum
Slowing Down in Hydrogen, Neutron Emission Spectrum, W/ Real H Elastic Scattering XS
1e+06
50 generations, 1 million neutrons, 1000 bins
100000
Flux (not normalized)
10000
1000
0.01 1 100 10000
Energy Associated with Lethargy Bins (eV)
1e+06 1e+08
add in real H elastic scattering xs (to divide the flux counter by)
Slowing Down and Thermalization in Hydrogen (no H absorption yet)
1e+07
"spec.dat" using 1:2
1e+06
Flux (not normalized)
100000
10000
1000
0.01 0.1 1 10 100 1000 10000
Energy Associated with Lethargy Bins (eV)
100000 1e+06 1e+07
add in thermalization in hydrogen (< 4 eV, thermal scattering)
Slowing Down and Thermalization in H w/ 1/v absorber
1e+06
"spec.dat" using 1:2
100000
Flux (not normalized)
10000
1000
0.01 0.1 1 10 100 1000 10000
Energy Associated with Lethargy Bins (eV)
100000 1e+06 1e+07
add in 1/v H absorber (effective only in the thermal range)
Flux vs Lethargy, U/H = 1
1e+07
5 million particles, 10,000 bins
Flux (not normalized)
1e+06
100000
0.01 0.1 1 10 100 1000 10000
Energy Associated with Lethargy Bins (eV)
100000 1e+06 1e+07
add in U238 resonance absorption cross section
112
Lulu Li 22.211 Class Notes: March 14, 2012
The only difference between 3D pin approach and 3D noval approach is to treat a pin as a node vs. to
treat an assembly as a node. For BWRs, thats about 2-3% difference in power; for PWRs, thats only 1%
difference in power.
113
Lulu Li 22.211 Class Notes: March 14, 2012
f
a = 0 (315)
k
Rearranging, we get
f f
k = = (316)
a a
Observations:
k is simply the ratio of total neutron production to destruction rate;
k does not depend on flux magnitude, normalization, etc;
k depends only on integrated cross sections;
Cross sections for down stream codes which treat pin-cells as homogenized will preserve exactly our
MC fuel reactivity for infinite repeating lattices.
Notice k assumes no leakage; keff would add an leakage term,
f
keff = (317)
DB 2 + a
2. k from Two-Group Cross Section. Again we start with the neutron balance equations (assume all neutrons
born in fast group):
f1 f2
1 a1 1 s12 1 + 2 + s21 2 = 0 (318)
k k
114
Lulu Li 22.211 Class Notes: March 14, 2012
2 s12
= (324)
1 a2
Plug into the first row, we get,
f1
s12 + f 2 s12 = 0
a1 (325)
k k a2
s12
f1 + f2 a2
k = (326)
s12
a1 +
Notice that,
Know how to derive the above expression;
Two-Group k depends only on the intergrated cross sections;
Although it looks like a chicken-to-egg problem that we include 21 in s12 , my understanding
is that we either know flux and needs to find cross section, or the other way around, hence we do
not need to worry about this term.
s12 is effectively s12 .
If no upscattering,
3. k from Balance Equation. For two region problem, we can also solve k from the neutron balance equa-
tion. We can reduce down scattering cross section from balance equation (instead of from group-to-froup
tallies),
f1
k a1
s12 =
(327)
f2
1 k a2
The k from the three methods above should agree, because reaction rates conserves.
115
Lulu Li 22.211 Class Notes: March 14, 2012
Notice the scattering term is discontinuous, that is, a neutron at E 0 , ~ 0 can jump to E,
~ ; the distri-
bution Zis a delta function not a smooth continuous distribution. Notice in the fission term on the RHS
(E)
dE 0 f (E 0 )(~r 0 , E 0 ), is the scalar flux as we can see the angular dependency is gone al-
4
ready, and the 4 there is to balance in an equation of .
4
5
z Z }| { Z z }| {
(E) 0 0 0 0 ~)
+ dE f (E )(~r , E ) + dS(~r , E, (330)
4 4
The scattering kernel is nasty, because a) s depends on energy, hence we have to integrate over all energy;
b) s depends on angle, hence we have to integrate over all angle; c) s depends on isotope, hence we have
to integrate over all isotopes.
3. Simplify the angular integrals, assuming that the source is isotropic
Z Z
1 = ~ ) =
d ( d ~ = J(~
~ r , E) (331)
4 4
2 = (~r , E) (332)
Z
3 = 4 d Y 00 ( ~ )Ylk ( ~ ) = 4l0 k0 , 400 = (333)
Z4
4 = (E) dE 0 f (E 0 )(~r , E 0 ) (334)
5 = S0 (~r , E) (335)
116
Lulu Li 22.211 Class Notes: March 14, 2012
Z
where weve defined the scalar current J(~r , E) = ~ (~r , E,
d ~ ), and scalar flux (~r , E) =
Z
~ ). We drop the 4 on the bottom of term
d(~r , E, 4 because now every term is in scalar flux.
(336)
This balance equation is basically equalling the source and the sink. Notice this equation is exact so far.
The problem is this is one quation with two unknowns , J. Thus we need to approximate J.
Basically going from transport to diffusion theory we drop the dependency. Thus we define the scalar flux
(~r , E) and the net current J~ (~r , E):
Z Z
d(~r , E, ~ ) = (~r , E) d ~ (~r , E,
~ ) = J~ (~r , E) (337)
4 4
The only reason we brought up the transport theory is that we need to set the boundary condition right.
117
Lulu Li 22.211 Class Notes: March 14, 2012
~ ) 1 (~r , E) + 3
h i
(~r , E, ~ J~ (~r , E) (339)
4
1. We start from transport equation again, multiple by and integrate over all angles:
5
6
zZ }| { zZ }| {
~
d ~ (~r , E,
~ ) +t (~r , E) ~ ~
d (~r , E, ) (340)
4 4
7
8
zZ }| { Z
L l
z
Z }| {
XX X
= ~ Ylk (
d ~) 0 i 0 0
dE s1 (~r , E E)lk (~r , E ) + ~ S(~r , E,
d ~)
i l=0 k=l 4 4
(341)
~ ) = 1 (~r , E) + 3
h i
(~r , E, ~ J~ (~r , E) (342)
4
and recall four identities,
Z Z
d = 4, ~ =0
d (343)
4 4
Z Z
~
d ~ = 4 A,
~ A ~ ~
d ~
~ =0 (344)
4 3 4
118
Lulu Li 22.211 Class Notes: March 14, 2012
We make the in-scattering approximation (may not be a good approximation!) that basically assume at low
absorption, the in-scattering (E 0 E) is equivalent to/would balance with the out-scattering (E E 0 ), notice
we can pull out the J term now as well,
Z Z
dE 0 s1 (~r , E E 0 )J(~r , E) = J(~r , E) dE 0 s1 (~r , E E 0 ) = J(~r , E)0 s0 (~r , E) (351)
1
In Ficks laws land, we define D = , then the flux is just J = D.
3tr
Alternative approach: start from the net current equation, assume the source is isotropic, then the source term
goes away:
1 XZ
~
(~r , E) + t (~r , E) J (~r , E) = dE 0 is1 (~r , E 0 E) J~ (~r , E 0 ) (354)
3 i E 0
Next we can either assume scattering to be isotropic in the Lab system, or isotropic in the Center of Mass
system.
1. Assume the scattering is isotropic in the Lab system, then the scattering kernel goes away:
1
(~r , E) + t (~r , E) J~ (~r , E) = 0 (355)
3
That is,
1
J~ (~r , E) = D(~r , E)(~r , E) D(~r , E) = (356)
3t (~r , E)
1
This is the first way to calculate diffusion coefficient, that is, D = 3 t
. Notice the assumption of isotropic
scattering in Lab system is not very good in a typical PWR because hydrogen scattering dominates, and
hydrogen scatters isotropically in the COM system, not in the lab system.
2. We can also start from the transport correction P0 approximation:
1
D(~r , E) = (357)
3tr (~r , E)
119
Lulu Li 22.211 Class Notes: March 14, 2012
The transport cross section in the transport-corrected Po approximation for isotope i is,
Z 1
i
s1 (E E 0 ) = 2 ds si (E E 0 , s )P1 (s ) (359)
1
Z 1
= ds si (E E 0 )(s s (E, E 0 ))s (360)
1
= s (E 0 , E)si (E E 0 ) (361)
Z
i
s1 (E) = dE 0 s (E, E 0 )si (E E 0 ) (362)
E0
= (E)si (E)
i
(363)
i
tr (E) = ti (E) i
s1 (E) (364)
= ti (E) (E)si (E)
i
(365)
Under the assumption of isotropic scattering in COM, the mean value of cosine theta over the interval 1, 1
2
is zero; the mean value of the cosine psi over the interval 1, 1 is 3A , hence,
i 2 i
tr (E) = ti (E) (E) (366)
3Ai s
How do we collapse D?
(a) Assuming isotropic scattering in lab system, we can calculate the flux-weighted total cross section,
X
1 t
D= , t = X (367)
3t
The problem with this method is that flux collapse under-predict transport.
(c) Most accurate: flux-weighted diffusion coefficient,
X 1
3tr
Dg == X (369)
More advanced methods include the out-scatter and in-scatter transport-corrected cross sections:
1. Out-scatter transport-corrected cross section for group g is just total cross section minus the total P1 scat-
tering rate from group g to all other group g 0 divided by P1 flux in group g:
120
Lulu Li 22.211 Class Notes: March 14, 2012
G
X 0
gg g
s1
G
g g g 0 =1 g
X gg 0
trout = tot = tot s1 (371)
g
g 0 =1
G
0 0
g g g
X
s1
g g g 0 =1
trout = tot (372)
g
This method requires knowledge of an approximate P1 flux (leakage) spectrum. C5 uses this method by
solving the Pn equations for a uniform fission spectrum source in a homogeneous slab of water of infinite
thickness.
121
Lulu Li 22.211 Class Notes: March 14, 2012
(0) = 0 (373)
Integrating over all angles in half space, that is incoming partial current is zero,
J (~r i , E) = 0 (375)
1 1 J 1
J (~r , E) = (~r , E) Jn (~r , E) = 0, = (376)
4 2 2
where D is the property of the material inside, and tr is transport mean free path. Notice that the
exact coefficient in dextrap may be different. Take into account that thermal neutron mean free path is
on the order of 1cm, so dextrap is even less than 1 cm, which means that we dont really care about it
for LWR cores.
Note: in diffusion theory, we approximate the flux as linearly dependent on D, thus in diffusion land we
can define an extropolated distrance at which the flux goes to zero. In reality the flux does not actually
get to zero, as it would just exponentially decay outside the core and approaches zero (and argulably never
reaches exactly zero).
3. Reflective BC: net current at surface is equal to 0.
4. Albedo boundary condition is the measurement of how much flux is reflected back:
J (~r i , E)
= (379)
J + (~r i , E)
122
Lulu Li 22.211 Class Notes: March 14, 2012
(~r ~ r+
i , E, ) = (~
~
i , E, ) (380)
In diffusion theory, the above is written as Continuity of scalar flux, from integrating angular flux over all
angles d,
(~r r+
i , E) = (~ i , E) (381)
2. Continuity of normal component of current from multiplying by omega and integrating over angle,
Z Z
~ ~ ~ (~r + ~
~n (~r i , E, ) d = ~n i , E, ) d (382)
4 4
J(~r r+
i , E) = J(~ i , E) (383)
~n D(~r
i , E)(~r
i , E) = ~n D(~r + r+
i , E)(~ i , E) (384)
(~r i , E) D(~r i , E)
= (385)
(~r +
i , E) D(~r +
i , E)
123
Lulu Li 22.211 Class Notes: March 19, 2012
8.8 Summary
Diffusion equation assumes only linear component of angular dependency (that is, isotropy). So anytime there
is any distortion on the angular phase (e.g., strong absorber, vacuum boundry condition),
Diffusion theory breaks down: [FIXME]
1. Near interfaces. Example: MOX and UO2 fuel boundary. MOX has a stronger thermal absorption cross
section (thus the thermal spectrum is more depressed by a factor of 4), the neutron experiences a big change
at the interface.
2. Near strong absorbers. Example: control rods.
124
Lulu Li 22.211 Class Notes: March 19, 2012
d2 (x)
2 =0 (386)
dx2 L
Assume that flux is in the form of (x) = C exp(kx), substitute into the differential equation, and it turns out
k = L1 . Thus
x x
(x) = C1 e L + C2 e L (387)
125
Lulu Li 22.211 Class Notes: March 19, 2012
BCs:
lim (x) = 0, (x) = Bex/L ;
x
S0 d
lim J(r) = . Equate that with J = D = BDex/L , we get B = S0 L
2D .
x0 2 dx
S0 L x/L
(x) = e (390)
2D
For the total geometry, we can simply place an absolute sign on x:
S0 L |x|/L
(x) = e (391)
2D
If the source is placed at x = b, the solution becomes,
S0 L |xb|/L
(x) = e (392)
2D
126
Lulu Li 22.211 Class Notes: March 19, 2012
d2 a S0
2
(x) = (393)
dx D D
We cannot play the trick in the infinite medium anymore to avoid solving for a homogeneous and a particular
solution:
The homogeneous solution is in the form of exponentials as we have solved for multiply times:
x x
H (x) = C1 e L + C2 e L (395)
Consider the particular solution has to be dependent on the source but independent of the space; that is,
0
z}|{
d2 a S0 S0
P = P = (396)
dx2 D D a
Putting the homogeneous and particular solution together, and consider vacuum BCs on the extrapolated bound-
aries
a = (a + 2D):
(
a) = 0.
(
a) = 0.
We get to,
1 S0
C1 = C2 = a
a
(397)
e +
L e L a
127
Lulu Li 22.211 Class Notes: March 21, 2012
d2 2 d 1 S0
2
+ 2 = (~r ) (399)
dr r dr L D
We look for the homogeneous solution, and set w = r (this is an important trick because it would make a
spherical case identical to the plane case),
d2 2 d 1 d2 w 1
2
+ 2 = 0, 2
2w = 0 (400)
dr r dr L dr L
The solution is,
er/L er/L
w = Aer/L + Ber/L , (r) = A +B (401)
r r
BCs:
lim = 0, A = 0.
r
lim 4r2 J(r) = S0 (this is also the particular solution), that is,
r0
d 1 1 S0
J = D = DBer/L + lim 4r2 J(r) = 4DB = S0 B= (402)
dr r2 rL r0 4D
S0 er/L
(r) = (403)
4D r
Interpretations:
2. Extreme condition 2: pure absorber. We can easily solve the transport kernel:
0
S(r0 )ea |rr |
Z
S0
(r) = 0 2
= (405)
V0 4|r r | 4r2
3. Diffusion results in a different shape for the decrease in neutron population as in Figure 41.
128
Lulu Li 22.211 Class Notes: March 21, 2012
129
Lulu Li 22.211 Class Notes: March 21, 2012
a f S
2 (~r ) + 1 (~r ) = (409)
D a D
= C1 er + C2 er (410)
S0
we know that the particular solution is p = , then
1 af a
C1 er C2 er S0
= + + (411)
r r
1 af a
f
(b) For k > 1 and a 1 > 0, then
d2 a
f
+ 1 (r) = 0 (412)
dr2 D a
that is, (r) = C1 cos(Bm r) + C2 sin(Bm r). Then the solution form is,
C1 C2 S0
(r) = cos(Bm r) + sin(Bm r) (413)
r r f a
BCs:
sin(Bm r) cos(Bm r)
(0) has to be finite. lim = Bm , lim = thus C1 = 0.
r0 r r0 r
= 0. We fine C2 .
(R)
130
Lulu Li 22.211 Class Notes: March 21, 2012
Comments:
Subcritical reactor: only has a steady state solution when there is a source.
Critical reactor: only has a steady state solution when there is no source.
The fact that we cannot solve for C2 implies that reactor can go critical at any power as long as we do
not have any feedbacks.
2
Bm = Bg means critical; Bm < Bg means subcritical; Bm > Bg means super-critical. Bm =
f a 2
2
D , Bg = R .
The bigger the core is, the more oscillation there is.
4. Fundamental mode of Helmholtz equation: we start with a bare homogeneous reactors, notice geometric
buckling is a measurement of the flux curvature
2 (r)
2 (r) + Bg2 (r) = 0 Bg2 = (416)
(r)
131
Lulu Li 22.211 Class Notes: March 21, 2012
2
(c) f > a , we can solve for the critical dimension using Bm = Bg2 (if R < Rcritical then sub-
critical etc). Alternatively we can perform the fundamental mode of Helmoltz equation and get
f
k= . Should have no source to be steady state critical.
DBg2 + a
1 S 1 f a
2 (~r ) + Bm
2
(~r ) = 2 (~r ) 2
(~r ) = 2
Bm = 2
= (420)
L D L D
132
Lulu Li 22.211 Class Notes: March 21, 2012
d2 1
D 2 + a (x) = f (x) (422)
dx k
2
1
DC1 cos x + a C1 x = f C1 x (423)
L
L
L k1
L
f
k1 = 2 (424)
D L
+ a
2
which agrees with Eq. 419, provided that Bg2 =
L
for a slab. Further more, the secondary mode yields an
eigenvalue of,
f
k2 = 2 (425)
2
D L + a
DR is a measure of the stability. The closer it is to 1 (the larger the system is), the less stable it is (the longer it
oscillates). It governs left-right/top-bottom oscillation.
A rule of thumb about k: in real reactor application, there is no source, so the balance equation is,
D2 + a = f (427)
and the above equation describes a critical system (because in real reactor otherwise either the flux would die out
or it would explode and we would have no steady state solution). However in academic reason we are interested
in the subcritical and the supercritical condition as well, so we add in an artificial k on the bottom of the fission
term which would make the equation critical:
f
D2 + a = (428)
k
Notice the k here is just a measurement of criticality of the system; it is not physical in the sense that a physically
existing system can only have k = 1.
133
Lulu Li 22.211 Class Notes: March 21, 2012
d2
D + a (x) = f (x) + S(x) (429)
dx2
d2 2 S(x)
+ Bm (x) = (430)
dx2 D
f a
Given a > f , B 2 = D < 0, we re-write B 2 = |B|2 ,
d2
|B|2 (x) = S(x) (431)
dx2
The general/homogeneous solution is,
134
Lulu Li 22.211 Class Notes: March 21, 2012
2 f a k 1 2
Bm = = < 0, Bm |Bm |2 (434)
D L2
1. For the homogeneous solution, we assume there is no source,
d2
|Bm |2 = 0, (x) = A cosh(|Bm |x) + B sinh(|Bm |x) (435)
dx2
4. Notice in both finite and infinite case, the fluxes have convex shapes; the finite curve is below the infinite
curve though.
135
Lulu Li 22.211 Class Notes: March 21, 2012
d2 2 2 f a k 1
+ Bm = 0, Bm = = >0 (x) = A cos(Bm x) + B sin(Bm x) (438)
dx2 D L2
S0
BC1: (H/2) = 0; BC2: J(0) = 2 . We can solve for the coefficients, and place absolute sign to represent the
whole geometry:
S0 H
(x) = Bm H
sin Bm |x| (439)
2DBm cos 2
2
Interpretations:
1. Because k > 1, the flux shape is concave, that is, increasing negative slopes in the positive domain. In
Fig. 43, flux is convex when k < 1 and concave when k > 1, until we hit the cosine shape when
keff = 1, k = 1 + B 2 L2 from
f k
keff a keff 1
B2 = = k = keff + B 2 L2 (440)
D L2
2. If H is increased to the critical dimension, that is, H Bm , then (0) ; that is, if the reactor is
critical, the flux at the source is infinite. That is, there is no steady-state solution for flux at the source site.
In fact, if we place a fission detector at x = 0 and measure source multiplication,
Nd f,d Vd (0) f BH
Ms (0) = tan (441)
S0 2DB 2
Or the inverse source multiplication factor,
1 1
0 (442)
Ms (0) tan BH
2
3. To reach criticality, we can add fuel elements, withdraw control rods, and dilute boron. 1/M is useful for
estimating criticality as shown in Figure 43.
136
Lulu Li 22.211 Class Notes: March 21, 2012
10.6 Summary
1. Super-positioning of sources: if we are given a random source that is the sum of a couple of common
forms of sources, we can super-position the flux from each of the common sources. This method works in
any non-multiplying medium, and in multiplying medium in subcritical condition10 . For instance, we can
super-position point sources to get a line source.
2. Diffusion theory is valid:
3. k is important11 :
k = 1: flux is straight line;
k < 1 means B 2 < 0, flux is sinh, cosh which is convex;
k > 1 means B 2 > 0, flux is sin, cos which is concave.
137
Lulu Li 22.211 Class Notes: March 21, 2012
2 (~r ) + Bm
2
(~r ) = 0 (446)
2 2 (~r )
Bm = (447)
(~r )
Notice,
2 2 ( ~
r)
1. Since Bm is a constant, that is to say ( ~
r) is a constant, that is to say (~r ) has a constant curvature.
2. Material buckling:
f
2 keff a
Bm = (448)
D
which matches geometry buckling,
f
keff = (449)
DBg2 + a
where DBg2 is the leakage per unit volume per unit flux. Notice keff does not depend on volume or flux.
3. Critical buckling: when a system is critical, the material buckling is uniquely determined by cross sections,
f
2 f a a 1 k 1
Bm = = D
= (450)
D a
M2
138
Lulu Li 22.211 Class Notes: March 21, 2012
d2
Slab dx2 + B 2 (x) = 0
d2
Sphere dr 2 + 2r d 2
dr + B (r) = 0
d2
Infinite Cylinder dr 2 + 1r d 2
dr + B (r) = 0
d2 1 d d2
Finite Cylinder dr 2 + r dr + dz2 + B 2 (r, z) = 0
d2 2 2
Cartesian dx2 + ddy2 + ddz2 + B 2 (x, y, z) = 0
4. Solutions exist only for certain values of the buckling such that the flux is everywhere positive and vanishing
on outer (or extrapolated) surfaces; we define these unique values as geometrical buckling Bg2 . The
allowable values of geometrical buckling that satisfy the boundary conditions are uniquely determined
reactor geometry.
5. For a criticality problem, we can solve find the B 2 such that the system is critical. Hence giving geometry,
only certain materials would make the system critical; given material, only certain geometries would make
the system critical. That is, given a reactor, it would be critical if and only if the material and the geometry
satisfy,
Bg2 = Bm
2
(452)
6. Solution only exists when k > 1 (exception: if there is external source, k can be less than 1 and the
system is still critical).
139
Lulu Li 22.211 Class Notes: March 21, 2012
d2
+ B 2 (x) = 0, (x) = A cos(Bx) + C sin(Bx) (453)
dx2
L L0
BCs: (L/2) = 0, 2 = 2 + 0.711tr . Two equations two unknowns,
cos(BL/2) sin(BL/2) A
=0 (454)
cos(BL/2) sin(BL/2) C
Set the determinant to be zero, we get 2 cos(BL/2) sin(BL/2) = 0. There are two possibilities:
n An cos(Bn x) n = 1, 3, 5,
Bn = (x) = (455)
L An sin(Bn x) n = 2, 4, 6,
140
Lulu Li 22.211 Class Notes: March 21, 2012
2 2 2
+ 2 + 2 + B2 = 0 (458)
x2 y z
Assume separation of variables,
d2 X d2 Y d2 Z
+ Bx2 X = 0, + By2 Y = 0, + Bz2 Z = 0 (460)
dx2 dy 2 dz 2
The solution is,
2 2 2
x y z
= A cos cos cos , B2 = + + (461)
Lx Ly Lz a b c
141
Lulu Li 22.211 Class Notes: March 21, 2012
2
1 d
r + 2 + B2 = 0 (462)
r dr r z
We plut the sepration of variables into the Helmholtz Equation, break B 2 = Br2 + Bz2 , and we get two equations
one for each direction,
1 d2 R 1 dR
+ + Br2 = 0 (464)
R dr2 r dr
d2 Z
+ Bz2 Z = 0 (465)
dz 2
In r direction, we can consider an infinite cylinder,
2.405
R(r) = A1 J0 (Br r), Br = (466)
R
In z direction, we can consider an infinite slab,
Z(z) = A2 cos(Bz z), Bz = (467)
H
We combine the two directions,
2 2
2.405 z
2 2.405
(r, z) = AJ0 r cos , B = Br2 + Bz2 = + (468)
R H R H
Problems with less than one direction that is non-homogeneous can be solved this way with separation of variables.
142
Lulu Li 22.211 Class Notes: March 21, 2012
1
Dc 2 c + ca c = cf c , Dr 2 r + ra r = 0 (469)
keff
k
keff 1 1 r
B2 = , 2 = = ar = Bm,r
2
(470)
L2c 2
Lr D
2 c + B 2 c = 0, 2 r 2 r = 0 (471)
The solutions are in the forms of,
c (x) = C1 cos(Bx) + C2 sin(Bx), r (x) = C3 cosh(x) + C4 sinh(x) (472)
BC1: reflective boudnary condition at x = 0, BC2: zero flux at reflector surface,
J c (0) = 0 C2 = 0, (H 0 /2) = 0 C3 = 0 (473)
We have:
c (x) = C1 cos(Bx), r (x) = C40 sinh[(H 0 /2 x)] (474)
Apply two interface conditions:
c (H/2) = r (H/2), J c (H/2) = J r (H/2) (475)
which can be written in the matrix form,
cos(BH/2) sinh( ) C1
=0 (476)
Dc B sin(BH/2) Dr cosh( ) C40
H 0 H
We define the reflector thickness = 2 , then the coefficients can be solved,
cos(BH/2)
C1 = c (0), C40 = C1 (477)
sinh( )
which gives us the criticality condition,
c BH
D B tan = Dr coth( ) (478)
2
Interpretation:
143
Lulu Li 22.211 Class Notes: March 21, 2012
That is, as H , . For large , like > 3, coth( ) 1, the smallest H to reach criticality is,
r
2 1 D
H= tan (481)
Bm D c Bm
h
i 1 1 1
tan tan1 x = cot[tan1 x] = = , tan1 x = tan1 (483)
2 tan[tan1 x] x 2 x
c
1 D Bm
s= tan1 tanh( ) (484)
Bm Dr
144
Lulu Li 22.211 Class Notes: April 2, 2012
11.6 Summary
Observations:
The lowest node is the only one remains after the source is gone.
For any positive value of materials buckling, there is a unique critical size for each reactor geometry.
Know how to get the critical buckling for different geometries.
Seperation of variables works as long as there is no one more than one direction that is heterogeneous.
Example: know how to find height-to-diameter to minimize leakage. The optimal cylinder has a H/D of
around 0.92.
145
Lulu Li 22.211 Class Notes: April 2, 2012
2. Define group terms. We define the yet unknown group fluxes by the integrals of the flux over individual
groups. Be careful about the units: g (~r ) is the real flux, (~r , E) is really the flux density. Similarly the
reaction rate is actually reaction rate density. This implementation does not have volume term in it because
we are considering an infinite system.
Z Eg1
g (~r ) = (~r , E) dE (487)
Eg
The group cross sections are defined such that the reaction rates in each interval are preserved,
R Eg1 i
1
Z Eg1
E
x (~r , E)(~r , E) dE
i
xg (~r ) = xi (~r , E)(~r , E) dE = g R Eg1 (488)
g (~r ) Eg (~r , E) dE
Eg
Group constants are typically determined using neutron spectra obtained by solving local problems with
approximate boundary conditions, like our MC code. Group-wise cross sections cannot match the details
of Monte Carlo spectrum calculation, so it is important that detailed resonance effects are contained in the
multi-group cross sections when produced from some detailed spectral calculations. For instance, we can
approximate (E) using (E) for fast energies, using narrow resonances for intermediate energies, and
Maxwellian distribution for the thermal energies.
3. Integrate steady-state diffusion equation over each of the g energy groups:
leakage/diffusion term
1 total interaction term
2 source term
3
z }| { z }| { z }| {
Z Eg1 Z Eg1 Z Eg1
dE D(~r , E)(~r , E) + dEt (~r , E)(~r , E) = dES(~r , E) (489)
Eg Eg Eg
146
Lulu Li 22.211 Class Notes: April 2, 2012
The group diffusion coefficient Dg should be a tensor (a 3x3 symmetric matrix, called the Eddington
tensor); but in application, we assume that Dg is the same for all directions. That is, the tensor is
approximated by a diagonal matrix (or call it current-weighted-diffusion-coefficient):
R Eg1
Eg
dED(~r , E) u (~r , E)
u
Dg (~r ) =
, u = x, y, z (492)
u g
It is not a great definition because for an infinite system u = 0 which would give us 0 over 0. A
more accurate approach is to let Dg be the same in the x-y plane, but different in axial direction.
(b) Total interaction term:
Z Eg1 Z Eg1
2 = dEt (~r , E)(~r , E) = tg (~r ) dE(~r , E) = tg (~r )g (~r ) (493)
Eg Eg
(d) Fission source term (this formation has only one fissioning species):
Z Eg1 Z G
X
4 = dE(E) dE 0 f (~r , E 0 )(~r , E 0 ) = g f g (~r )g (~r ) (497)
Eg E0 g 0 =1
where
Z Eg1 Z Eg1
1
g = dE(E), fi g (~r ) = (E)fi (E)(~r , E) dE (498)
Eg g (~r ) Eg
G Z
X Eg1 Z Eg0 1
= dE dE 0 s (~r , E 0 E)(~r , E 0 ) (501)
g 0 =1 Eg Eg0
G
X
= sg0 g (~r )g0 (~r ) (502)
g 0 =1
where
Z Eg1 Z Eg0 1
1
i
sg 0 g (~
r) = dE dE 0 si (E 0 E)(~r , E 0 ) (503)
g0 (~r ) Eg Eg0
147
Lulu Li 22.211 Class Notes: April 2, 2012
Cancelling the within group scattering cross section from both sides and defining the group-wise net re-
moval cross section,
X
rg = tg sgg = ag + sgg0 (505)
g 0 =1,g 0 6=g
G
X G
X
Dg (~r )g (~r ) + rg (~r )g (~r ) = g f g0 (~r )g0 (~r ) + sg0 g (~r )g0 (~r ) + Sg (~r )
g 0 =1 g 0 =1,g 0 6=g
(506)
Notes:
A note about the summation sign: the loss terms do not need to be summed over energy groups because
once a neutron leaves group g it does not matter where it goes, whereas the gain terms need to be summed
over energy groups to take into account all neutrons contributing from every other energy group.
A note about the sign of the leakage term: the absolute value of leakage always have a negative sign.
Typically RHS > 0 (we have either fission, scattering, or source), then the curve is concave down. We
can either think about it in terms of second derivative, which is negative; or we can think the slope
is decreasing, which means that 2 < 0. The heat transfer analogy is that when there are sources
everywhere in the slab, there would be heat transfer away at every point.
In the extreme case RHS < 0 , and the system is losing neutrons due to absorption or scattering out,
then the flux would be concave up, meaning 2 > 0. The heat transfer analygo is that when there
are sinks everywhere in the slab, there would be heat transfer inwards at every point.
When RHS = 0, the flux would be a straight line, = constant, 2 = 0, there is no leakage/dif-
fusion at every point, that is, there is no net neutrons leaving any point, the neutron diffusion breakes
even at every point. The heat transfer analogy is that there is no heat leaving or entering any point.
148
Lulu Li 22.211 Class Notes: April 2, 2012
D1 2 + r1
1
D2 2 + r2 2
(507)
. .. ..
.
2
DG + rG G
1 f 1 1 f 2
sG1 + 1 k f G
k s21 + k 1
2 f 2
s12 + 2 k f 1 sG2 + 2 k f G 2
k
= (508)
.. .. ..
. . .
G f 1 1 f 2 1 f G G
s1G + k s21 + k k
Notes: we have the k term in the equations to keep the fission source general. In the absence of spatial dependency
(for instance, in this formulation weve only considered energy dependency so far and assume spatial homoge-
nization), notice the RHS matrix has order 1 (this can be seen from that every row vector is a constant times the 1st
row), leading to that there is only one eigenvalue, thus we do not have to worry about multiple eigenvalues. This
implies that the eigenvalue is a measurement of the spatial dependency. Once we add in the flux depending on
both space and energy, we cannot easily write each row as the 1st row times a constant anymore.
149
Lulu Li 22.211 Class Notes: April 2, 2012
2. Net current,
Z
J~ (~r , E) = ~ (~r , E,
~ ) d (511)
4
4. A common expansion for angular flux (I think this is second order) from Reuss:
Z
d = (519)
4
Z
d = J~
(520)
4
1 h J~
i
= +3 (521)
4
150
Lulu Li 22.211 Class Notes: April 2, 2012
= R + 2Dg )
Then the question is of course what do we do with Bg2 ? Bg2 depends on extrapolated distance (e.g., R
which depends on group. One way we can approximate is that we use D1 because fast group leakage dominates.
Then our leakage term is 2 2g = Dg Bg2 .
151
Lulu Li 22.211 Class Notes: April 4, 2012
152
Lulu Li 22.211 Class Notes: April 4, 2012
D1 2 1 + (a1 + s12 )1 = f 1 1 + f 2 2 + S1
(524)
D2 2 2 + a2 2 = s12 1 + S2
If we are in an infinite system that we can ignore the leakage term, then we can write out an eigenvalue
problem in matrix form immediately. More specifically, without leakage, 1 , 2 are flat, and we can
directly write out the ratio of the two:
2 s12
= (525)
1 a2
and the eigenvalue:
f 1 f 2 s12
k= + (526)
r1 a2 r1
Prof. Forget made a connection between the four-factor/sixe-factor formula and the 2 group theory:
s12 1 s12 f 1 f 2 f 1 a2
1. Four factor: p = = , k1 = 1 f1 = , k2 = 2 f2 = , = 1+ , k =
r1 1 r1 r1 a2 f 2 r1 12
pf .
1 D1 1
2. Six factor: keff = pf P1,N L P2,N L where P1,N L = , L2 = , P2,N L = , L2 =
1 + L21 Bg2 1 r1 1 + L22 B 2 2
D2
.
a2
However to keep it general, we need to approximate the leakage term by constructing a finite spatial mesh
in which cross sections are constant. Integrate Eq. 524 over for node n with width n ,
Z
d n d n
D dx + nr1 n1 n = nf1 n1 n + nf2 n2 n + S1 n (527)
dx 1 dx 1
Z
d n d n
D dx + na2 n2 n = ns12 n1 n + S2 n (528)
dx 2 dx 2
153
Lulu Li 22.211 Class Notes: April 4, 2012
where n+ is the right surface of node n, n is the left surface. Then the balance equations become,
d d
D1n n1 + D1n n1 + nr1 n1 n = nf1 n1 n + nf2 n2 n + S1 n
(530)
dx n+ dx n
d n n d n
D2n 2 + na2 n2 n = ns12 n1 n + S2 n
2 + D2 (531)
dx n+ dx n
Sg ng
d n
Jn+ = Dgn g = Dgn (532)
dx n+ /2
n+1 Sg
d n+1 g
Jn+1 = Dgn+1 g = Dgn+1 (533)
dx n /2
By continuity of the net current, we know the net current at the surface from the left cell (Jn+ ) and from the right
cell (Jn+1 ) should be the same,
Sg ng n+1
g Sg
Dgn = Dgn+1 (534)
/2 /2
154
Lulu Li 22.211 Class Notes: April 4, 2012
Now we consider three neighboring cells with the same width, and following the previous analogy, we can get
the net currents in cell n:
n1,n (n2 n1 ) D
D n,n+1 (n+1 n2 ) + na2 n2 n = ns12 n1 n + S2 n (539)
2 2 2 2
n1,n n1
D 1
n,n+1 n+1
D 1 + [n n n1,n + D
r1 + D1
n,n+1 ]n n n n n n n
1 = f 1 1 + f 2 2 + S1
n
(540)
1 1 1
n1,n n1
D 2
n,n+1 n+1
D 2 + [n n n1.n + D
a2 + D2
n,n+1 ]n n n n
2 = s12 1 + S2
n
(541)
2 2 2
13 Kord uses the notation Jn+ , Jn , which in this content are actually net currents not partial currents!!!
155
Lulu Li 22.211 Class Notes: April 4, 2012
2. Zero flux: flux at the RHS of node N is zero. We can either think of the flux as keep going to be N
g in
the imaginary N + 1 node14 , or calculate the slope by doing N
g /(/2),
N N
g g 2DgN N
JgN = DgN = (542)
N N g
3. Zero incoming flux, zero incoming current, vacuum BC: there is no partial current coming in from RHS,
1 1
JN = JN = 0 (543)
4 2
which implies,
2DgN
sg sg N
g 2DgN N
1 N
JgN = = DgN = N 2D N g (544)
2 /2 1
+ g
2
" #
2DgN 1
= N 4DgN
N
g (545)
1+ N
2DgN N
g
JgN = (546)
N + C DgN
Alternatively, we can also take Eq. 545 and make the DgN term into the diffusion coefficient of the imaginary
N + 1 node:
N
2D g 1
JgN = N N
g (547)
4DgN +1
1+ N
14 we have to interpolate the node center value from the node edge value because in linear difference equation the center value is the averaged
156
Lulu Li 22.211 Class Notes: April 4, 2012
The two matrix are plotted in Figure 47. Notice there are missing points in A. Explaination: Block matrix: the
L + D + U term should only be doing in-group stuff, and the missing points would have been throwing group 1
stuff into group 2 when it is not supposed to do it.
In this method, we guess keff , and given a source we can solve for the flux. We can either use Matlab to solve the
Ax = b system, or we use Gaussian elimination/forward elimination backward substitution by hand. The basic
157
Lulu Li 22.211 Class Notes: April 4, 2012
Figure 48: Power Iteration Convergence Rate (left) and Dominance Ratio (right)
idea is to subtract ai,i1 /ai1,i1 times the (i1)th equation from the ith equation to eliminate the ai,i1 element
in the ith equation. The modified ith equation is then divided by ai,i . This process is repeated successively for
i = 1 through i = I 1.
We never solve real questions this way, because the flux approaches the right shape (as keff keff correct ),
but once keff > keff correct , the flux can be entirely inverted to become negative (recall that the 1/M curve goes
to 0 when we become critical, implying that M . Having a negative flux is fine because the flux solved can
be arbitrarily scaled). If we have a super-critical problem, then the only way to get criticality with an additional
source is through negative flux. A reactor is like an amplifier; the closer it is to criticality the more amplification
it is; when critical, the amplification is infinity.
Notice that the eigenvectors are arbitrarily normalized; hence it is fine if the flux is negative.
14.5 Method 3+: Power Iteration with Matlabs GMRES Numerical Invertion
With no pre-conditioner, GMRES is actually longer than Method 3.
158
Lulu Li 22.211 Class Notes: April 4, 2012
14.6 Method 4: Power Iteration with Point-Jacobi Iterative Flux Matrix Inversion
In a steady-state problem, it does not matter whether we fully converge our flux iteration.
14.7 Method 5: Power Iteration with Gauss-Seidal Iterative Flux Matrix Inversion
The outter loop is almost the same as Point-Jacobi,
A=L+D+U (563)
1
(L + D)m+1 = M i U m (564)
keff
159
Lulu Li 22.211 Class Notes: April 4, 2012
Notice these Point-Jacobi and Gauss-Seidal only work with 2D problem that we can easily split a matrix into L,
D, and U. 3D and above problems would be hard to solve using these two methods.
2. The matrix representation for higher dimensions looks the same as in 1D, except with one additional pair
of off-diagonal term for each additional dimension as in Figure 50
3. The size of the matrix increases as the dimensions increases. For instance, if we have 10 meshes per
dimension, then 1D matrix is 10x10, 2D matrix is 100x100, 3D matrix is 1000x1000.
160
Lulu Li 22.211 Class Notes: April 4, 2012
|1 | > |2 | |3 | (566)
then |1 | is the spectral radius of the iteration matrix, and every mode has a dominance ratio,
n
drn = (567)
1
and power iteration kills of the lowest dominance ratio modes. The last remaining mode is the fundamental mode
dr = 12 . If |1 | 1, the ietration scheme is unstable and it would not converge15 . Convergence of the power
method is slow when dr is close to unity; in fact in most numerical methods, convergence rate = 1 - dominance
ratio. Knowing the dominance ratio, we can estimate the number of iterations needed for convergence.
Dominance ratio measures the spatial decoupling. It depends on:
1. Symmetric mode. If the initial guess is symmetric, the solution is symmetric, and the method used is
symmetric, then only the symmetric mode shows up in the dominance ratio, and the asymmetric mode is
hidden. As shown in Figure 51, the symmetric guess may not display the highest dr convergence (0.99 in
this case), instead it display a symmetric mode (0.975 in this case). Whereas a random guess would excited
all modes and show all the dr convergence (though the lower modes die too fast to be observed; we can
only see the last two dominance ratio, 0.97 and 0.99), and the dominant mode has an asymmetric shape.
Figure 51: Dominance Ratio With An Symmetric Guess (left) or Random Guess (right)
2. Core size: As seen in Figure 52, as core size increases, dominance ratio increases, and it takes longer to
converge.
3. Decoupling of radial zones from inserting control rod, having asymmetric core loadings, or having xenon
distributions. For instance, in BWRs if we insert control rods to make the core into two halves, then it
would take a lot longer to converge.
4. Decoupling of axial zones form partially inserting control rods, axial fuel enrichment zoning, and axial
burnable absorber loading.
Dominance ratio should be a physical property, so it should not depend on mesh size.
15 See Lewis Computational Methods of Neutron Transport Appendixe C for more details
161
Lulu Li 22.211 Class Notes: April 4, 2012
162
Lulu Li 22.211 Class Notes: April 9, 2012
14.10 Summary
Remember for real 3D problem,
Matrix inversion if of order N 3 , so in real applications no matrix inversion;
Finding all eigenvalues is at least order N 2 ;
Iterative inversion must be order N to be practical for large problems;
163
Lulu Li 22.211 Class Notes: April 9, 2012
1. Number: almost all fissions produce precisely two fission produces per fission;
2. Unstability: most fission products are highly unstable and radially decay to other nuclides;
3. Thermal spectrum: some fission produces have large thermal absorption cross sections, which are important
for thermal spectrum reactors as in Figure 55;
4. Fast spectrum: fast reactors are much less sensitive to fission products, because thermal spectrum does not
matter.
5. Distribution: contributing factors are: which species are fissioned (U235, Pu239, etc), the incident fission
neutron energy, random statistical fluctuations nuclide breakup.
6. Independent fission yield is a different concept from the cumulative fission yield: the former only accounts
for fission yield, whereas the latter accounts for generation of the isotope from independent yield as well as
all the precursor isotopes that decay into a specific isotope.
7. Meda-stable state: an isotope may not directly decay into a stable state; it may branch into a meda-stable
state. For our purpose, we are going to take the simplified approach.
As an example, we look at the Xenon-135 Chain in Figure 54. Typically very rapid transitions are not followed
in details when modeling reactor behavior. For instance, because Tes half-life is 19s, we can assume that 135I is
producted instantenously. Most of the chains are similar and are dominated by beta decays until they are stable.
164
Lulu Li 22.211 Class Notes: April 9, 2012
165
Lulu Li 22.211 Class Notes: April 9, 2012
dN
+ A(t)N (t) = Y (568)
dt R
IF = e A(t) dt = eAt (569)
dN
eAt + eAt A(t)N (t) = eAt Y (570)
dt
d At
[e N (t)] = eAt Y (571)
dt
eAt N (t) = A1 eAt Y + C (572)
N0 = A1 Y + C C = N0 A1 Y (573)
We can also write the above differential equation into the matrix form as in Figure 56. Notice we asume fission
yields, cross sections, and fluxes are constant over the time interval tn1 , tn ,
166
Lulu Li 22.211 Class Notes: April 9, 2012
167
Lulu Li 22.211 Class Notes: April 9, 2012
15.3 Iodine/Xenon
Xe-135 has a thermal absorption cross section of 2.6 106 barns.
Major source: Iodine decay; 135I 135
Xe which happens with a half-life of 6.6 hours.
Major sink: burnup. Xe has a huge absorption xs (106 barns), so 135
Xe + n 136
Xe. Xe can also beta
decay again with a half-life of 9.1 hours.
Xenon peaking happens because after shutdown, the major sink is removed, but the major source remains, hence
Xenon peaks until the Iodine depletes. Reactors must be designed with enough fuel to offset the effect of Xenon.
At the end of core life, there may not be enough reactivity to override peak Xenon; such reactors are called
xenon-precluded. If a scram occurs, a restart may not be possible for 30-40 hours. See Stacys Example 6.2
(p.214) for an example of xenon reactivity worth and illustrating diagrams.
168
Lulu Li 22.211 Class Notes: April 9, 2012
15.4 Promethium/Samarium
Samarium is another fission poisoner. It has one source: decay of Promethium; it has one sink: burnup.
170
Lulu Li 22.211 Class Notes: April 9, 2012
171
Lulu Li 22.211 Class Notes: April 9, 2012
172
Lulu Li 22.211 Class Notes: April 11, 2012
Benchmark: Nd is a burnup marker term b/c of its low absorption cross section. Each fission produces a fixed
amount of Nd (Pu and U produces slightly different amount of Nd). Use Nd number density and the yield, you
can back out the burnups.
Gd156 and Gd157 are important for startups when using old rods (15 years old). Once operating, we dont
really care about Gd that much.
173
Lulu Li 22.211 Class Notes: April 11, 2012
16 Fuel Depletion
For fuel assembly depletion, the reactivity of fuel changes dramatically with burnup as in Fig. 58. Typically fresh
assemblies would include burnable poisons to control initial reactivity. Notice the rapid drop right after reactor
start-up it is due to Xe buildup.
1. FP absorbe neutrons and reduce reactivitiy of the fuel significantly, often 10-20%;
2. Fission cross section: isotopes have threshold energies, producing step-shape near higher energies (around
10 MeV)17 . Remember Pu is very fissile material. Am242 is a funny one although its an even isotope, its
fission cross section is large.
3. Simple actinide nuclide transmutation model as in Figure 59: red arrows designate the actinide chains we
model (chains end at isotopes that decay quickly compared with the phenomena we are trying to model);
blue arrows designate beta decay.
4. The higher you go up the chain (e.g., into the MA region), the worse the absorption would be.
5. Branching ratio is a function of energy. Example: 241Am can produce 242Am with a half-life of 16 hrs, and
the meta-stable 242mAm with a half-life of 141 years. 242Ams thermal branching ratio is about 10% that of
242m
Am.
17 Both even and odd isotopes have threshold energies, but odd isotopes have higher resonance fission cross section, hence even isotopes
174
Lulu Li 22.211 Class Notes: April 11, 2012
175
Lulu Li 22.211 Class Notes: April 11, 2012
Actinides matrix form has no fission yield term so the solution is just [N (tn )] = e[A]tn [N (tn1 )] ;
FP matrix is decoupled; actinide matrix is not.
More specifically, we focus on the actinides matrix which is illustrated in both Fig. 62 and Fig. 63.
176
Lulu Li 22.211 Class Notes: April 11, 2012
is a lot harder to invert. Another simplification we made is to ignore all meta-stable states and only have
the ground states, hence ignoring branching.
One side-note here is that Bateman equations can be very stiff. For a system of differential equation
N10
1 0 0 N1
N20 = 11 2 0 N2 (577)
N30 21 2 0 N3
1 0 0
11
[] = Diag[e1 t , e2 t , 1] [V ] = 1 0 (578)
2 1
21 2
2 1 1 1
So if 1 = 2 (which is possible when absorption terms are present), numerical instability could happen
depends on the flux level!
177
Lulu Li 22.211 Class Notes: April 11, 2012
6. (remember) Np has a half-life of 2 and 2.5 days; Nps cancels out Am etc, and they have the same half-life,
so people use to ignore them.
7. Production with Pu;
8. Am244: takes a certain burnup to come up. Even though Am is a small concentration, we still care about
them.
Fig. 65 shows that about 40-50% of power would be produced from Pu instead of U by the time we shut down the
reactor.
178
Lulu Li 22.211 Class Notes: April 11, 2012
179
Lulu Li 22.211 Class Notes: April 11, 2012
180
Lulu Li 22.211 Class Notes: April 11, 2012
Fig. 66 shows that power level does not affect number density vs. BU plot. BU is a better gage of depletion
effect than time.
5. EFPHs = Effective Full Power Hours; EFPD = Effective Full Power Days.
181
Lulu Li 22.211 Class Notes: April 11, 2012
3. Fuel Temperature Depletion History Effects (Fig. 69): initially, the instantenous temperature effect is almost
independent of depletion; as the burnup increases, something happens. The moral of the story is, we not
only need to know the burnup but also the temperature.
4. Gadolinium burnable absorber (Fig. 70):
Gadolinium has a huge thermal absorption cross section. It is almost as high as Zn, but we cannot use
Zinon because it would decay;
What we do with Gadolinium is to replace part of the fuel with Gd and reactivity would be flatter with
respect to time.
Gad depletes/burns in layers around the fuel pin from outside to inside. ]
Hold-down is the difference between the reactivity at the BOL with and without the reactivity control
mechanism.
Gad residual: there is a little bit loss of reactivity due to Gad.
182
Lulu Li 22.211 Class Notes: April 11, 2012
183
Lulu Li 22.211 Class Notes: April 11, 2012
184
Lulu Li 22.211 Class Notes: April 11, 2012
185
Lulu Li 22.211 Class Notes: April 11, 2012
data set is the single most important things in determine the accuracy of your codes.
U235 has a pretty large error, but we dont care because there is so little of it;
Accuracy tends to decrease the further up the decay chains.
Measurements also have an intrinsic uncertainty; do not rely on single measurement campaigns; sys-
tematic errors in measurements are common.
In performing actinide burner analysis, when we place the actinide at the end of one analysis into
another one, the error builds up.
186
Lulu Li 22.211 Class Notes: April 11, 2012
187
Lulu Li 22.211 Class Notes: April 11, 2012
5. Multi-recycling: we cannot recycle too many times. Reason: after 1st cycle the fissile content is 64.8%,
after 2nd cycle fissile is only 51.1% and requires a higher Pu enrichment; and if Pu enrichment is pass xxx,
then the void coefficient would become positive.
6. DUPIC cycle (used in CANDU): it is an entirely mechanical process during which we grind up everything,
the gas comes out, and re-construct the grinds into fuel which contains about 0.8% U235 and 0.8% Pu239,
which is about twice the excess reactivity of natural uranium and is more than enough for a CANDU.
7. Fast reactor transmutation: at high energy, all MA are fissionable (also high absorption cross section which
is why we dont expect to get energy from MA), so we can transmutate MA.
8. Safety concerns: See Fig. 72.
188
Lulu Li 22.211 Class Notes: April 11, 2012
189
Lulu Li 22.211 Class Notes: April 11, 2012
(f) SFP loading curves are developed where the y-axis is Burnup, x-axis is initial enrichment, and lines
were drawn to illustrate limitations.
2. Interim Storage: CLAB (Central Interim Storage Facility for Spent Fuel) is an interim storage facility for
all spent fuel from Swedish nuclear reactors. It was opened in 1950 and was expanded in the 2000s to its
current capacity. A Swedish reactor generates between 15 to 25 tonnes per year. It has also been used for
performing calorimeter measurements of full-length spent fuel assemblies.
3. On-site dry cast storage. These are dry casks built on-site. Under the current consideration, all sites are
going to get dry cast storage eventually. It consists of steel canisters (one canister holds 36 assemblies),
storage casts, and transport casts. Loading 36 assemblies into a canister takes about 1.5 days. Each fuel
transport operation (placing canister in cast, two welded steel lids, take out water and pump in He) takes
about 8 days to complete. At Yankee Rowe, the inner cast is a transport cast that can be taken out and
transport to a permanent site later on.
190
Lulu Li 22.211 Class Notes: April 11, 2012
2. UO2 is used in LWRs regardless of its low heat transfer property because it does not interact with water
violently. UO2 has a relatively low thermal expand, so the gap between the fuel and the clad can be relatively
small.
3. Fast reactors use metallic fuel because a) better heat transfer, b) coolant is not water so no worry for
chemical reaction, c) metallic fuel thermal expansion is large, which replaces Doppler effect(recall Doppler
is due to U238 ). Also because metallic fuel expands more, we build fuel pellets with more porosity to
limit the amount of volume expansion.
4. Grid straps: egg crates to place pins in them. Pin starts vibrating and may cause fuel failure.
5. To flattern power peaking:
(a) Core loading.
(b) Poisons: WABA, IFBA ZrB2 , Gd and other Lanthanum.
(c) Control rod sequence.
(d) Reflector.
(e) Core enrichment.
(f) Flow orificing: change flow pattern at different assemblies. It is not done in LWRs that often, more
common in SFRs to control temperature and flux spectrum.
6. Thermal transients in fuel:
dTf 1
MF Cp,f = P (t) (Tf (t) Tc (t)) (579)
dt Rf
where Rf captures all the thermal resistance. Assume steady state, then
1
P (t) = (Tf (t) Tc (t)) (580)
Rf
If we assume a fast transient with no cooling (Rf ), then we get adiabatic heat rate:
dTf
Mf Cp,f = P (t) (581)
dt
191
Lulu Li 22.211 Class Notes: April 11, 2012
dTc 1
MC CpC = (Tf (t) Tc (t)) 2m
c Cpc (Tc (t) Ti ) (582)
dt Rf
where m= mass flow of coolant, and the factor of 2 is there because the Tc (t) in the equation is assumed to
be the average.
1
Pc (Tc Ti ) =
Steady state: 2mC (Tf Tc ) = P . That is, reactor power is determined from flow
Rf
and temperature measurements.
8. We can relate the above two equations with the PKE using P (t). That is, how do we relate temperature
change to reactivity change.
k1 dk dk
(a) Set-up: consider = , d = 2 . If we assume k 1, then d = = d(ln k). There is
k k k
benefit in writing k in log form as now we can separate multiplication into additions.
1
(b) Consider keff = k PN L , where PN L = L2 Bg2 , we can write,
dk dPN L
d(ln k) = d(ln k ) + d(ln PN L ) = + (583)
k PN L
We can do:
!
dPN L L2 Bg2 dL2 dBg2
= + (584)
PN L 1 + L2 Bg2 L2 Bg2
(c) Recall
f
z }| !{ z }| { z }| { z}|{
2 F
f 1 a2 + D2 Bg s12 f 2 F a2
k = 1+ (585)
f 2 12 R1 F a2 a2
dk 1 d k 1 d k
d = = + (586)
k k dTf k dTc
1 dk 1 d 1 d 1 df 1 d
x = = + + + (587)
k dTx dTx dTx f dTx dTx
192
Lulu Li 22.211 Class Notes: April 11, 2012
1 dkm 1 dk dNm
m = (589)
k dTm k dNm dTm
1 dNm 1 dk
m = m = m Nm (590)
Nm dTm k dNm
The expansion coefficient of sodium and salt is very large; it will change density, cross section, and
push some coolant outside thus changing the moderator-to-fuel ratio.
For our case, two terms depend on the fuel and do not change much: , . would go up a little
bit, as s12 = Nm s12 as Nm . Most of the effect is felt by resonance escape probability p and
thermalization factor f .
s12 Nm H,s12
In the resonance escape probability p, p = = as Nm . The cross sections are
R1 R1
considered flat, and in the R1 = a + s12 in which a dominates. The real formulation should be,
0
z }| {
Nm s12,m + NF s12,F
p= s12
(591)
z }| {
R,m +R1,F
193
Lulu Li 22.211 Class Notes: April 11, 2012
11. Fast reactor considerations: we pack fuels more tightly, m > 0, f < 0, we use leakage to control
reactivity:
!
dPN L L2 Bg2 dL2 dBg2
= + (592)
PN L 1 + L2 Bg2 L2 Bg2
Core flowering: fuel expands at different rate, and it would expand more on the top, creating a negative
coefficient.
As temperature , number of neutrons generated per fission event producing a positive feedback .
This is another positive temperature coefficient that LWRs do not have to worry about.
12. Power coefficient: at the end of the day, we really care about the power coefficient that combines fuel
temperature effect and moderator temperature coefficient. We want the power coefficient to be negative.
d 1 k 1 dk dTf 1 dk dTc
P = = = + (593)
dP k d k dTf dP k dTc dP
The one reactor built that has a positive power coefficient unexpected (designed to be -0.12mk/MW, mea-
sured at 0.28 mk/MW had to be shutdown) is the MAPLEs medical reactor built around 2000. Fuel expands
by less than 1mm but because it is a small reactor it pushes the power coefficient to be positive.
Key points:
(a) Temperature defect: when we start up the reactor from CZP to HZP, we experience a negative reactiv-
ity due to resonances absorption increase:
Z Ti
DT = (m + f ) dT (594)
Troom
194
Lulu Li 22.211 Class Notes: April 11, 2012
Z Tf Z Tc
DP = f dTf + c dTc (595)
Ti Ti
where Tc Ti is fairly small ( 20 C), whereas the power defect is dominated by the fuel materials.
(c) One way we can use power defect: if for some reason we need to stretch out the cycle length for some,
we can do coast-down by lowering the power, and by power defect, the reactivity would increase, thus
we can operator for a longer period of time before reaching subcritical.
(d) Excess reactivity: value of if all movable rods and poisons are removed from the core. We need
excess reactivity to compensate for the temperature defect, power detect and shutdown margin.
13. Reactor transients: if we assume initially critical core, m , f < 0. With a reactority insection (t), we
can write
2. Neutron & gamma spectra: for instance, it is useful to know neutron and gamma energy release during
shipping and storage.
3. Cask design: heat source, heat conduction, gamma/neutron transport.
195
Lulu Li 22.211 Class Notes: April 12, 2012
From there we use the Ficks law, J~ = D to get the leakage term D2 .
196
Lulu Li 22.211 Class Notes: April 12, 2012
1 2
D= , tr = t s1 = t s (599)
3tr 3A
Effective down-scattering cross section (if assume up-scattering is zero, effective down-scattering
is real down-scattering):
1 1 J 1
J (~r , E) = (~r , E) Jn (~r , E) = 0, = (603)
4 2 2
where D is the property of the material inside, and tr is transport mean free path. The
coefficient before tr can be 0.711 in other formulism.
(c) Collapsing coefficients and D:
X X
D
= X D = X (606)
197
Lulu Li 22.211 Class Notes: April 12, 2012
198
Lulu Li 22.211 Class Notes: April 12, 2012
2. Fission yields.
3. Independent fission yield vs. cumulative fission yield (12 Exam 2 #4): independent yield is the yield of
a specific isotope directly from fission and nothing else. Cumulative yield includes the independent yield
PLUS all the precursor isotopes that decay into a specific isotope.
199
Lulu Li 22.211 Class Notes: April 12, 2012
5. FP balance equation:
dN R
6. Analytic solution first order ODE + A(t)N (t) = Y , know to do IF = e A(t) dt = eAt ,
dt
dN
eAt + eAt A(t)N (t) = eAt Y (618)
dt
d At
[e N (t)] = eAt Y (619)
dt
eAt N (t) = A1 eAt Y + C (620)
N0 = A1 Y + C C = N0 A1 Y (621)
See Figure 56 for how to construct matrix form, assuming that fission yields, cross sections, and fluxes are
constant over the time interval tn1 , tn ,
200
Lulu Li 22.211 Class Notes: April 12, 2012
201
Lulu Li 22.211 Class Notes: April 12, 2012
10. I/Xe, Pm/Sm reactivity effects: Xe and Sm both are fission poisoners18 .
Major source of Xe: Iodine decay; Source of Sm: Pm decay;
Major sink of Xe: burnup; Sink of Sm: burnup.
Xenon peaking happens because after shutdown, the major sink is removed, but the major source remains,
hence Xenon peaks until the Iodine depletes.
11. In a LWR, equilibrium Xe accounts for about -4 (thats negative 2500 2800 pcm) of reactivity during
operation and does not depend on flux level. On shutdown, additional negative reactivity is added which
depends on flux level.
12. In a LWR, Sm-149 builds up at shutdown and does not decay. It accounts for about during operation,
and 0.5 at shutdown. This becomes a much bigger issue for high flux reactors.
13. Xe-precluded cores: reactors must be designed with enough fuel to offset the effect of Xe; otherwise at the
end of core life, there may not be enough reactivity to override peak Xe; such reactors are called xenon-
precluded. If a scram occurs, a restart may not be possible for 30-40 hrs. See Stacys Example 6.2 (p.214)
for an example of xenon reactivity worth and illustrating diagrams.
14. Spatial Xe oscillations: if a core has a flux gradient, then regions with low flux would have a higher Xe
concentration (due to less burnup of Xe), thus a even reduced flux, until the I concentration becomes low
(Ieq ), and Xe would first slowing decrease then suddently decrease thus increasing the flux (see Fig in
class notes). The point is, with a feedback due to flux changing, Xe concentration and flux can oscillate.
15. Xe chasing is not right, because Xe concentration is 9 hrs behind what is actually going on in the core.
16. Production tool: Nd is a BU marker term b/c each FP produces a fixed amount of Nd. Thus using Nd
number density and the yield we can back out the BU.
17. Old rods (15 yrs) have 156Gd and 157
Gd which would make startup hard.
202
Lulu Li 22.211 Class Notes: April 12, 2012
204
Lulu Li 22.211 Class Notes: April 12, 2012
18. Actinide chain equations: similar to FPs except there is no direction production by fission anymore (so no
Y on the RHS) so the solution is just [N (tn )] = e[A]tn [N (tn1 )], and the matrix is more coupled than
FPs.
205
Lulu Li 22.211 Class Notes: April 12, 2012
19. Zooming in the matrix, there are four blocks for 4 matrixes,
The decay terms sit outside of the blocks; they couple the blocks.
The capture terms are one line off the diagonal term (NP238 decays so quickly that its capture cross
section is zero and is not shown throug the spy function in matlab);
We dont have above the diagonal term because we ignore the n2n reaction and the n reaction. The
reason we ignore them is that having a sub-diagonal matrix we can invert analytically; whereas having
a full matrix is a lot harder to invert. Another simplification we made is to ignore all meta-stable states
and only have the ground states, hence ignoring branching.
206
Lulu Li 22.211 Class Notes: April 12, 2012
20. Know some solutions of the actinide chain equations. E.g., U238 barely changes; U235 decreases b/c no
production; U236 increases b/c source is U235 plus n and no sink; U239 is constant b/c capture from U238
is constant, and U239 decays so quickly (23 mins) that it is at equilibrium the whole time.
21. Know the fission fraction vs. BU plot: by the EOL about 40-50% power was produced from Pu instead of
U.
22. BU is a good measurement, because number density vs. BU is independent of power level/time. BU units:
(a) FIFA = fission per initial fissile atoms;
(b) FIMA = fission per initial (heavy) metal atoms;
(c) Atom percent (A%) = FIMA * 100;
(d) Burnup: GWd/T = MWd/kg = thermal power per weight of heavy metal.
Advantages: we know the reactor power and initial fuel loading (almost).
Disadvantages: energy released is not a very clean term; we dont really care about neutrino
energy, gamma energy released from capture of neutrons; energy deposited in fuel assembly B
from fissions in assembly A etc. In LWR, it is a good approximation to assume that energy is
depositied locally where the fission happens; in MITR it is a bad approximation, fission product
energy (deposite locally), gamma heat energy (not necessarily deposited locally), etc.
(e) EFPHs = Effective Full Power Hours; EFPD = Effective Full Power Days.
23. History effects:
Burnable poisons: harden the spectrum. Remove 20 WABA at 20 MWd/kg generates a positive 350
pcm reactivity jump.
Fuel temperature depletion history effect: initially temperature independent of depletion; as BU ,
histories are approximately linear in BU.
Gd burnable poisons: burnup rates dependent on Gd concentration; residual is proportional to w% of
Gd.
207
Lulu Li 22.211 Class Notes: April 12, 2012
leakage/diffusion term
1 total interaction term
2 source term
3
z }| { z }| { z }| {
Z Eg1 Z Eg1 Z Eg1
dE D(~r , E)(~r , E) + dEt (~r , E)(~r , E) = dES(~r , E) (625)
Eg Eg Eg
Cancel the within group scattering cross section and define the group-wise removal cross section, we get
the final form,
G
X G
X
Dg (~r )g (~r ) + rg (~r )g (~r ) = g f g0 (~r ) (~r ) +
g0 sg0 g (~r )g0 (~r ) + Sg (~r )
g 0 =1 g 0 =1,g 0 6=g
(627)
1 1 1 D 1 1 1 D
J+ = + Jn = , J = Jn = + (628)
4 2 4 2 4 2 4 2
1
= 2[J + + J ], +
Jn = J J = = D (629)
3tr
Albedo coefficient, the coefficient of reflection, is the measurement of how much flux is reflected back:
208
Lulu Li 22.211 Class Notes: April 12, 2012
)= 1 h J~ (~r , E)
i
(~r , E, (~r , E) + 3 (632)
4
We can get the net current equation,
Z
1 X
(~r , E) + t (~r , E) J~ (~r , E) = dE 0 is1 (~r , E 0 E) J~ (~r , E 0 ) + S
~1 (~r , E) (633)
3 i E0
3. From net current equation, assume no source and isotropic scattering in lab system, we can approximate
the transport equation as,
2 i
itr (E) = it (E) (E) (634)
3Ai s
1
Then we can solve for diffusion coefficient D = 3tr .
6. Helmoltz Equation: from one-group diffusion equation, we assume spatial constant cross section and use
buckling term,
209
Lulu Li 22.211 Class Notes: April 12, 2012
d2
+ B 2 (x) = 0, (x) = A cos(Bx) + C sin(Bx) (638)
dx2
L L0
BCs: (L/2) = 0, where 2 = + 0.711tr . Two equations two unknowns,
2
cos(BL/2) sin(BL/2) A
=0 (639)
cos(BL/2) sin(BL/2) C
Set the determinant to be zero, we get 2 cos(BL/2) sin(BL/2) = 0. There are two possibilities:
n An cos(Bn x) n = 1, 3, 5,
Bn = (x) = (640)
L An sin(Bn x) n = 2, 4, 6,
2 2 2
+ 2 + 2 + B2 = 0 (643)
x2 y z
Assume separation of variables,
d2 X d2 Y d2 Z
+ Bx2 X = 0, + By2 Y = 0, + Bz2 Z = 0 (645)
dx2 dy 2 dz 2
The solution is,
2 2 2
x y z
= A cos cos cos , B2 = + + (646)
Lx Ly Lz a b c
(c) Consider a finite cylinder from H/2 to H/2 with radius R. We assume azmimuthal symmetry of
Helmholtz Equation,
2
1 d
r + 2 + B2 = 0 (647)
r dr r z
Assume separation of variables,
210
Lulu Li 22.211 Class Notes: April 12, 2012
We plut the sepration of variables into the Helmholtz Equation, break B 2 = Br2 + Bz2 , and we get two
equations one for each direction,
1 d2 R 1 dR
+ + Br2 = 0 (649)
R dr2 r dr
d2 Z
+ Bz2 Z = 0 (650)
dz 2
In r direction, we can consider an infinite cylinder,
2.405
R(r) = A1 J0 (Br r), Br = (651)
R
In z direction, we can consider an infinite slab,
Z(z) = A2 cos(Bz z), Bz = (652)
H
We combine the two directions,
2 2
2.405 z
2 2.405
(r, z) = AJ0 r cos , B = Br2 + Bz2 = + (653)
R H R H
Problems with less than one direction that is non-homogeneous can be solved this way with separation
of variables.
(d) Summary for a couple of geometries
(e) Find the height-to-diameter that would minimize the leakage from a fixed-volume cylinder. To mini-
2
mize leakage, we need to maximize Bm hence Bg2 . For a finite cylinder, we know
2 2
2.405
Bg2 = + (654)
R H
dBg2
and we solve dr = 0. The optimal H/D ratio comes out to be 0.9236.
211
Lulu Li 22.211 Class Notes: April 12, 2012
d2
D a = 0 (655)
dx2
The general solution is in the form of,
1 2D/L
= (659)
1 + 2D/L
(b) Consider a subcritical multiplying medium, slab geometry from L/2 to L/2, and with a source19 .
d2
D + a (x) = f (x) + S(x) (660)
dx2
2
d 2 S(x)
2
+ Bm (x) = (661)
dx D
f a
Given a > f , B 2 = D < 0, we re-write B 2 = |B|2 ,
d2
|B|2 (x) = S(x) (662)
dx2
The general/homogeneous solution is,
212
Lulu Li 22.211 Class Notes: April 12, 2012
S0 S0 er/L
B= , (r) = (669)
4D 4D r
(d) Consider an infinite non-multiplying medium with a line source along the z-axis. Use cylindrical
coordinates, only is involved:
d2
1 d 1 S
D + 2 = (670)
d2 d L D
The solution is in terms of Bessel functions,
d
J = D = BDex/L (674)
dx
S0 L x/L
(x) = e (675)
2D
For the total geometry, we can simply place an absolute sign on x:
S0 L |x|/L
(x) = e (676)
2D
If the source is placed at x = b, the solution becomes,
S0 L |xb|/L
(x) = e (677)
2D
213
Lulu Li 22.211 Class Notes: April 12, 2012
(f) Consider a finite multiplying medium with k > 1, it is a slab extending from H/2 to H/2, it is
subcritical with leakage, BC are from the current at center and the zero flux at the H/2 boundary.
d2 2 2 f a k 1
+ Bm = 0, Bm = = >0 (678)
dx2 D L2
(g) Consider a finite multiplying medium with k < 1, a slab geometry from 0 to H, and a plane source
at 0,
2 f a k 1 2
Bm = = < 0, Bm |Bm |2 (681)
D L2
The homogeneous solution is,
d2
|Bm |2 = 0, (x) = A cosh(|Bm |x) + B sinh(|Bm |x) (682)
dx2
BCs: (0) = 0 , (H) = 0, we can solve for the coefficients,
cosh(|Bm |x)
(x) = 0 cosh(|Bm |x) sinh(|Bm |x) (683)
sinh(|Bm |x)
214
Lulu Li 22.211 Class Notes: April 12, 2012
1
Dc 2 c + ca c = cf c , Dr 2 r + ra r = 0 (684)
keff
k
keff 1 1 r
B2 = , 2 = = ar = Bm,r
2
(685)
L2c 2
Lr D
2 c + B 2 c = 0, 2 r 2 r = 0 (686)
We have:
cos(BH/2)
C1 = c (0), C40 = C1 (692)
sinh( )
which gives us the criticality condition,
BH
Dc B tan = Dr coth( ) (693)
2
Interpretation:
215
Lulu Li 22.211 Class Notes: April 12, 2012
That is, as H , . For large , like > 3, coth( ) 1, the smallest H to reach
criticality is,
r
2 1 D
H= tan (696)
Bm Dc Bm
h
1
i
1 1 1 1
tan tan x = cot[tan x] = 1 = , tan1 x = tan1
2 tan[tan x] x 2 x
(698)
c
1 D Bm
s= tan1 tanh( ) (699)
Bm Dr
216
Lulu Li 22.211 Class Notes: April 12, 2012
2. Balance equation:
3. Interface flux:
4. BC:
Ng g
N
2DgN N
Zero Flux BC: JgN = DgN N
= N
g (703)
" #
2DgN 1 2DgN N
g
Zero Incoming Flux BC: JgN = 4DgN
N
g = (704)
N 1 + N N + 4D N
g
5. Matrix form:
[L1 + D1 + U1 ] [0] 1 [M1 ] [M2 ] 1 S1
= + (706)
[T2 ] [L2 + D2 + U2 ] 2 [0] [0] 2 S2
217
Lulu Li 22.211 Class Notes: April 12, 2012
9. Dominance ratio estimation: given an eigenvalue problem, if we specify the solved eigenvalues to be:
|1 | > |2 | |3 | (711)
then |1 | is the spectral radius of the iteration matrix, and every mode has a dominance ratio,
n
drn = (712)
1
and power iteration kills the lowest dominance ratio modes. The last remaining mode is the fundamental
mode dr = 21 .
k2
12 Exam 2 #5, 13 S904 #4: What is dominance ratio for a reactor core? Answer: DR = k1 = the ratio
of the two largest eigenvalues of the equation:
1
M = F M 1 F = k (713)
k
218
Lulu Li 22.211 Class Notes: April 12, 2012
12 Exam 2 #19: continue from #18, estimate how many more iterations will it take to converge the
peak fission source to a fractional change of 106 between successive iterations?
Answer: the fractional change from iteration 10013 to 10014 is 0.005. We know that error reduction
is proportional to DR:
6
10
ln 0.005
106 = 0.005DRn DR = = 1708 (714)
ln DR
Prof. Smith noted that do not assume linear, as all convergence is exponential at the end.
13 S904 Exam #5b: physical meaning of spectral radius?
Answer: spectral radius |1 | is the max of all the absolute value of eigenvalues of a matrix. It is defined
for a particular iterative method, and has impact on how iterative methods converge. For instance, if
|1 | 1, the iteration scheme is unstable and it would not converge.
Convergence of the power method is slow when d is close to unity; in fact in most numerical methods,
convergence rate = 1 - dominance ratio.
10. Order of operations (not included in official list): Matrix inversion is of N 3 , finding all eigenvalues is at
least N 2 , iterative inversion must be order N to be practical, multi-level iteration is a necessity.
219
Lulu Li 22.211 Class Notes: April 12, 2012
G G
X 1X
Jg (x) + t,g g (x) = s,kg k (x) + f,kg k (x) (715)
k
k=1 k=1
G G
dJg X 1X
+ t,g g (x) = s,kg k (x) + f,kg k (x) (716)
dx k
k=1 k=1
Then we can discritize the domain into the following (insert image).
We make assumptions,
All groups are spatially constant of a computational cell;
Average flux equals flux at the center of a cell;
d
First order approximation on dx (linear that is);
Integrate over an arbitrary cell,
Z i+1/2 Z 1/2 Z i+1/2 G Z G
i+1/2 X
dJg X 1
dx + rg g (x) dx s,kg k (x) = f,kg k (x) (718)
i1/2 dx i1/2 i1/2 k=1,k6=g k i1/2 k=1
220
Lulu Li 22.211 Class Notes: April 12, 2012
2. First we find the exact expression for the volume-averaged flux in each cell:
R i+1/2
i i1/2
g (x) dx
g = (720)
xi
Then we make a second-order approximation,
Z i+1/2
i
g = ig = g (x) dx = ig xi (721)
i1/2
Then we have,
G G
X 1X i
Jgi+1/2 Jgi1/2 + irg ig xi is,kg ik xi = f,kg ik xi (722)
k
k=1,k6=g k=1
i+1/2 i1/2 G G
Jg Jg X 1X i
+ irg ig is,kg ik = f,kg ik (723)
xi k
k=1,k6=g k=1
Notice here if we were to expand 1D into 3D, we need to add two more leakage terms making a total of six net
current terms.
221
Lulu Li 22.211 Class Notes: April 12, 2012
We can write the net current at the boundary two different ways,
i+1/2
!
ig
i+1/2 i d
i
g
Jg = Dg Dg (724)
dx i+1/2 xi /2
i+1/2
!
i+1 g
d g
Jgi+1/2 = Dgi+1 Dgi+1 (725)
dx i+1/2 xi+1 /2
and set the above two expressions to be the same, we can hence solve for the interface current,
i+1/2 i+1/2
! !
i
g ig i+1
i+1
g g
Dg = Dg (726)
xi /2 xi+1 /2
Dgi xi+1 ig + Dgi+1 xi i+1
g
i+1/2
g = (727)
Dgi xi+1 + Dgi+1 xi
Plug the interface flux into either expression for the interface net current, we get,
!
i+1/2
2Dgi Dgi xi+1 ig + Dgi+1 xi i+1 g i
Jg = g (728)
xi Dgi xi+1 + Dgi+1 xi
!
2Dgi Dgi xi+1 ig + Dgi+1 xi i+1
g Dgi xi+1 ig Dgi+1 xi ig
= (729)
xi Dgi xi+1 + Dgi+1 xi
!
2Dgi Dgi+1 xi i+1
g Dgi+1 xi ig
= (730)
xi Dgi xi+1 + Dgi+1 xi
2Dgi Dgi+1
Jgi+1/2 = (i+1
g
gi+1/2 (i+1
ig ) = D g ig ) (731)
Dgi xi+1 + Dgi+1 xi
Similarly, we can construct the net current on the left hand side,
2Dgi1 Dgi
Jgi1/2 = (ig i1 i1/2 (ig i1
g ) = Dg g ) (732)
Dgi1 xi + Dgi xi1
222
Lulu Li 22.211 Class Notes: April 12, 2012
gi+1/2 gi1/2 G G
D D X 1X i
(i+1
g ig ) + (ig gi1 ) + irg ig is,kg ik = f,kg ik (733)
xi xi k
k=1,k6=g k=1
Rearranging,
! G G
gi+1/2
D gi+1/2
D gi1/2
D gi1/2
D X 1X i
i+1 + + + irg ig i1 is,kg ik = f,kg ik
xi g xi xi xi g k
k=1,k6=g k=1
(734)
1
[M ][] = [F ][] (735)
k
The generalized eigenvalue problem is F = kM . If using solver, solving the generalized eigenvalue
problem is faster than the eigenvalue problem.
The eigenvalue problem is M 1 F = k. Notice we need to re-arrange the matrix form to the eigenvalue
form to get that k is the eigenvalue. Also keep in mind that [F ] can be singular, so we try to avoid inverting
[F ] whenever possible.
Keep in mind it is always column impact rows. The above expression is true all the time; with different BC, the
terms) would change20 . Next we will investigate the different boundary conditions.
coupling coefficients (the D
1. Zero Flux BC:
0
z }| {
I+1/2 Ig 2DgI I
d I g I+1/2 I
JgI+1/2 = DgI
Dg = =D (736)
dx I+1/2 xI /2 xI g g g
JgI+1/2 = 0 (737)
Jg = (Jg+ Jg ) ~n (738)
223
Lulu Li 22.211 Class Notes: April 12, 2012
I+1/2 I+1/2
that is, Jg = 21 g .
I+1/2
!
g Ig
JgI+1/2 = DgI (740)
xI /2
2DgI I+1/2
= ( Ig ) (741)
xi g
2DgI
= (2JgI+1/2 Ig ) (742)
xi
2DgI
JgI+1/2 = I (743)
xI + 4DgI g
=D gI+1/2 Ig (744)
224
Lulu Li 22.211 Class Notes: April 12, 2012
2. Hyperbolic functions:
ex ex ex + ex
sinh(x) = cosh(x) = (748)
2 2
sinh(x) = i sin(ix) cosh(x) = cos(ix) (749)
x
e = cosh(x) sinh(x) (750)
225
Lulu Li 22.211 Class Notes: April 12, 2012
(a1 + s12 )1 = k1 (f 1 1 + f 2 2 )
(751)
a2 2 = s12 1
f 1 + f 2 s12
a2
k = (752)
a1 + s12
2 s12
(b) 1 = a2 .
X X
D
(c) To preserve reaction rates, = X To preserve leakage rates, D = X .
(d) Us the f , a , we can calculate
f
k = (753)
a
The k calculated from effective one-group cross sections is the same as the one calculated from
two-group cross sections, because the one-group data is calculated preserving reaction rates.
226
Lulu Li 22.211 Class Notes: April 12, 2012
(f) unreflected means that there are leakages now, energy-independent means that B 2 is the same for
group 1 and 2,
1
(D1 B 2 + a1 + s12 )1 = keff
(f 1 1 + f 2 2 )
2 (755)
(D2 B + a2 )2 = s12 1
1 D2 B 2 + a2
= (756)
2 s12
It is different from part b because the finite unreflected geometry has leakage terms.
(h) The k using one-group cross sections collapsed with the critically-buckled spectrum from part g
( 21 ),
f 1 12 + f 2
k = (757)
a1 12 + a2
This k is smaller then the one in part d because most fission comes from thermal spectrum, and a
lot of it leaks out. Notice is is good to double check to make sure that the calculated k result is no
larger than .
227
Lulu Li 22.211 Class Notes: April 12, 2012
(i) Though this appears to be a multi-region problem. But since the reflector is made of absorbium whose
absorption cross section is infinite for both energy groups, the problem is effectively a one-group
problem, including the core with length L and the approximated extropolated distance of 2D on each
side. Hence we solve for,
2
B2 = , L = L0 4D (758)
L0
(j) There are two approaches. Approach 1:
Ff
keff = 1 = (759)
DB 2 + NF aF + NP aP
2
Approach 2: we solve for the unknown a that would make Bm = Bg2 ; then solve for the concentra-
tion.
228
Lulu Li 22.211 Class Notes: April 12, 2012
2. (12 Final #23) Compute the finite-medium energy-independent critical buckling and fast-to-thermal flux
ratio for Fuel 1.
Answer: This is essentially #10 in Exam 2. We start by writing the 2-group diffusion equations with B 2
being the same for the two groups,
D1 B 2 1 + (a1 + s,12 )1 = f 1 1 + f 2 2 (760)
2
D2 B 2 + a2 2 = s,12 1 (761)
From Eq. 761, we can write,
1 D2 B 2 + a2
= (762)
2 s,12
Plug into Eq. 760 and divide every term by 2 we get,
D2 B 2 + a2
(D1 B 2 + a1 + s,12 f 1 ) = f 2 (763)
s,12
Solving this quadratic equation, we get B 2 = 0.008876 . Then plug into Eq. 762, we get,
1 D2 B 2 + a2
= = 4.31 (764)
2 s,12
229
Lulu Li 22.211 Class Notes: April 12, 2012
3. (12 Final #24) Using the infinite medium spectrum from #23, compute the effective one-group cros sections
for Fuel 1 by preserving reaction and leakage rates. X
Answer: Recall collapsing cross sections is essentially = X . Thus we do,
1 f 1 12 + 2 f 2
f = 1
= 0.03515 (765)
2 +1
a1 21 + a2
a = 1
= 0.02384 (766)
2 + 1
1
D1 1 + D2
D= 1
= 1.27387 (767)
2 + 1
4. (12 Final #25) Using one-group cross sections from #24, what is the one-group, critical radius of a bare
square cylinder (i.e., height = diameter) using diffusion theory extrapolation distances for all surfaces?
Answer: There are two slightly different ways to do it.
Using collapsed one-group cross sections, we can write the material buckling as,
2 f a
Bmaterial = = 0.008876 (768)
D
The geometrical buckling with extrapolated distance 2 D is,
2 2
2 2.405
Bgeometrical = + (769)
R+2D 2R + 4 D
Setting the two bucklings to be equal, we can solve for the critical radius to be R = 27.94cm .
Alternatively, we can recognize that we have already solved for B 2 in #23 and get 0.008876 (which
agrees with the previous method using collapsed one-group cross sections, because we are preserving
leakage by collapsing D ).
230
Lulu Li 22.211 Class Notes: April 12, 2012
Taking advantage of the square cylinder geometry and the assumption of one-group, if we define
R0 = R + 2 D where R is the critical radius we are actually solving for, R0 is the distance including
the extrapolated distance, we can easily solving for R0 through
2
2 2.405 2
B = + (770)
R0 2R0
2 2 f 2 s12
D1 B1,R + D1 B1,H + a1 + s12 = f 1 + 2 2 (775)
D2 B2,R + D2 B2,H + a2
We obtained a one equation one unknown (R), thus can solve for the critical radius R = 27.668cm .
6. (12 Final #27) Los Alamos scientists have created a system to monitor sub-criticality of spent fuel pools by
comparing detector responses from two identical point sources, one source/detector pair placed in the pool
and another source/detector pair placed in the fuel, as depicted in the following figure. What is the required
distance, L, between the water pool source and detector such that the signal will be exactly equal to the fuel
source and detector signal as the k of the fuel reaches the regulatory limit of 0.95?
Answer: Recall in lecture we solved the point source in an infinite non-multiplying medium problem to
S0 er/L
have (r) = . This problem is essentially the same set-up, thus we can write,
4D r
q F
|B|r F
a f
S0 e
D F = 0.0707 Fuel region
(r) = , where |B| = q (776)
4D r P
a
= 0.1414 Pool region
DP
where F F
f = k a . A quick review of how we obtained the above flux expression,
231
Lulu Li 22.211 Class Notes: April 12, 2012
S0 e|B|r
(r) = (779)
4D r
Coming back to Eq. 776, the question is basically asking for,
S0 e0.1414r S0 e0.0707200
= (780)
4D r 4D 200
Thus r = 104.58cm .
7. (12 Exam 2 #10) What is the energy-independent critical material buckling and fast-to-thermal flux ratio
for Fuel 1 material?
Answer: write out the two group diffusion equations with D1 B 2 , D2 B 2 and solve for B 2 . Notice the hint
of energy-independent critical material buckling, implying that the two groups share the same B 2 .
8. (12 Exam 2 #11) Los Alamos scientists have created a perfect neutron reflecting material, Reflectium. If
a homogeneous hemisphere is placed flat side down on an infinitely large table of Reflectium, what is the
critical radius of the hemisphere?
Answer: Dont forget extrapolation distance!
2
2 =R+d
B = R (781)
R
9. (12 Exam 2 #12) I.M. Lazy wants to analyze a finite slab reactor of Fuel 1 with infinitely thick pure water
reflector on both sides. I.M. decides to perform a two-group zero-leakage infinite-medium spectrum calcu-
lation for Fuel-1 material and collapse one-group cross sections and diffusion coefficients for both Fuel-1
and Reflector material using that spectrum. What critical thickness will he compute with his collapsed 1G
cross section data?
Answer:
232
Lulu Li 22.211 Class Notes: April 25, 2012
233
Lulu Li 22.211 Class Notes: April 25, 2012
Prompt neutrons: more than 99% of all neutrons are emitted in < 1010 s after the fission (prompt neutron
lifetime in a PWR is about 2 105 sec).
Delayed neutrons:
1. Importance. Assume no delayed neutrons, a reactor was operated at 1W, a control rod was moved to produce
an excess reactivity of 0.0005k, what would the power be 1s later?
5 5
P (t) = P0 (1.0005)t/(210 )
= 1W (1.0005)1/(210 )
= 70, 000M W (782)
This reactor would be virtually impossible to control! Fortunately, delayed netrons exist and the reactor
time constant depend on more than just prompt neutron lifetime.
3. Measurement: delayed neutrons can be measured by counting neutrons emission after a pulsed irradiation
of a pure U235 foil.
Burst measurement: the amount of prompt neutrons.
Saturated measurement: the total amount of neutrons.
Measurement vs. evaluation: the data are recorded from measurements, but upon different evaluations
the final data would be different.
4. Neutron Precursor data. Delayed neutrons are emitted through the decay of fission products.
Br-87 is a dominant FP that emits delayed neutrons (longest half-life).
Delayed neutron precursors are embedded in fission species delayed yields.
234
Lulu Li 22.211 Class Notes: April 25, 2012
Delayed neutron fraction : absolute yield (# of delayed neutrons per fission) divided by (total
fission neutrons) = percentage of neutrons that are delayed.
We talk about effective once it is averaged over all isotopes and flux.
changes as a function of time through breeding of Pu-239 in U-239 cycle, though the variations in
BU are on a much larger time scale than usual range of application of PKEs so typically we ignore the
change in for short transients.
6 group yields and decay constants depend on fissioning species, as in Table 17.
U238 produce 4% delayed neutron per fission, more than U235s 1.7% delayed neutron per fission,
making U238 a very important isotope when it comes to delayed neutrons. Fast group absolute yield
for U238 is 0.0412, U235 is 0.0165, U239 is 0.0063 as in Table 18. Keep in mind: model the term
technically we need to know the delayed neutron spectrum etc.
235
Lulu Li 22.211 Class Notes: April 25, 2012
all isotopes have the same half-life for the same group; but different isotopes would still have different
delayed neutron fraction. Remember the terms circled in blue in Table 19 (the delayed neutron frac-
tion listed in this table might be in % unit, because U235 thermal fission fraction should be 0.0067.).
Delayed energy group: neutron spectra have similar shapes for all delayed groups. However different
energy groups have different mean energies.
Summary: Delayed neutron spectra vary only slightly for different fissioning nuclides; but the
spectra depends significantly on the delayed neutron group (in real production tools, there is a
separate for each energy group).
236
Lulu Li 22.211 Class Notes: April 25, 2012
237
Lulu Li 22.211 Class Notes: April 25, 2012
In steady-state transport or diffusion equation, we do not treat delayed neutrons directly. Instead the fission
emission spectrum must be properly weighted with prompt and delayed contributions.
A chicken-and-egg problem: we need fission rates to compute (E), and we need (E) to get fission rates,
so the real way to solve the balance equation is to iterate, though this is not how it is done normally.
Next we will start from the time-dependent neutron diffusion equation and go through various steps to arrive
at the generalized PKE.
1. The time-dependent neutron diffusion equation: now delayed neutrons must be treated explicitly. j =
X j
i is the delayed fission fraction (0.66% for instance); we use i is for isotropes, j is for energy groups.
i
1
(~r , E, t) = D(~r , E, t)(~r , E, t) t (~r , E, t)(~r , E, t)
t v
Z
+ s (~r , E 0 E, t)(~r , E 0 , t) dE 0
0
X Z
+ jp (E)(1 j ) jf (~r , E 0 , t)(~r , E 0 , t) dE 0
j 0
X
+ id (E)i Ci (~r , t) + Q(~r , E, t) (784)
i
Z
X j
Ci (~r , t) = i jf (~r , E 0 , t)(~r , E 0 , t) dE 0 i Ci (~r , t) (785)
t j 0
2. Assume that flux can be separated into a space/energy term and a time-dependent term:
where S(~r , E) is the Shape function, T (t) is the amplitude function (remember this approximation may
not always be valid, and certainly in some cases it would not be a good approximation). Plug it into the
238
Lulu Li 22.211 Class Notes: April 25, 2012
above expressions.
(
1
S(~r , E)T (t) = D(~r , E, t)S(~r , E) t (~r , E, t)S(~r , E)
t v
Z
+ s (~r , E 0 E, t)S(~r , E 0 ) dE 0
0
!Z
X X
j
+ j j
p (E)(1 ) + i
d (E)i jf (~r , E 0 , t)S(~r , E 0 ) dE 0
j i 0
)
X Z
jp (E) j jf (~r 0 0
, E , t)S(~r , E ) dE 0
T (t)
j 0
X
+ id (E)i Ci (~r , t) + Q(~r , E, t) (787)
i
Z
X
Ci (~r , t) = ij jf (~r , E 0 , t)S(~r , E 0 ) dE 0 T (t) i Ci (~r , t) (788)
t
j 0
3. Integrate over all space and energy and normalize to dE dr v1 S(~r , E) = 1.0,
R R
Z Z (
d
[T (t)] = dE dr D(~r , E, t)S(~r , E) t (~r , E, t)S(~r , E)
dt
Z
+ s (~r , E 0 E, t)S(~r , E 0 ) dE 0
0
!Z
X X
j
+ j j
p (E)(1 ) + i
d (E)i jf (~r , E 0 , t)S(~r , E 0 ) dE 0
j i 0
)
X Z
jp (E) j jf (~r 0
, E , t)S(~r , E ) dE 0 0
T (t)
j 0
Z Z " #
X
+ dE dr id (E)i Ci (~r , t) + Q(~r , E, t) (789)
i
d
Z Z Z Z X Z
dE drCi (~r , t) = dE dr ij jf (~r , E 0 , t)S(~r , E 0 ) dE 0 T (t)
dt
j 0
Z Z
i dE drCi (~r , t) (790)
X Z
4. Simplify energy dependency. In Eq. 789, notice in the term jp (E) j jf (~r , E 0 , t)S(~r , E 0 ) dE 0 ,
j 0
239
Lulu Li 22.211 Class Notes: April 25, 2012
only jp (E) depends on energy, we can use jp (E) dE = 1 to get rid of the energy integral,
R
Z Z (
d
[T (t)] = dE dr D(~r , E, t)S(~r , E) t (~r , E, t)S(~r , E)
dt
Z
+ s (~r , E 0 E, t)S(~r , E 0 ) dE 0
0
!Z )
X X
j j i j j 0 0 0
+ p (E)(1 ) + d (E)i f (~r , E , t)S(~r , E ) dE T (t)
j i 0
( )
Z X Z
dr j jf (~r , E 0 , t)S(~r , E 0 ) dE 0 T (t)
j 0
Z Z " #
X
+ dE dr id (E)i Ci (~r , t) + Q(~r , E, t) (791)
i
R
In a similar fashion, notice Eq. 790 no term has a E dependency, so we can cross out the dE in every
term to get,
Z Z Z
d X
drCi (~r , t) = dr ij jf (~r , E 0 , t)S(~r , E 0 ) dE 0 T (t)
dt
j 0
Z
i drCi (~r , t) (792)
5. Assume no net current (leakage) out of the reactor model22 , thus D(~r , E, t)S(~r , E) 0. Assume
no external source of prompt or delayed neutrons, thus Q(~r , E, t) 0.
6. Assume prompt fission spectrum jp is independent of fissioning species j, that is, we drop the j depen-
dency on prompt terms jp , j ,
X
jp (1 j ) = p (1 ) (793)
j
7. Assume delayed fission yield j is independent of fissioning species j, that is, we drop the j dependency
on delayed terms,
X X j
j = i = i (794)
j j
8. Assume delayed fission spectrum id is independent of delayed neutron group i23 , that is, we drop the i
dependency on id ,
X X
id (E)i Ci = d (E) i Ci (795)
i i
22 Recall we used Gauss theory to make the volume integral into surface integral, thus creating zero leakage for certain BCs.
23 we saw from data two lectures ago that is slightly dependent on neutron group, but we dont care much about it for this context.
240
Lulu Li 22.211 Class Notes: April 25, 2012
Z Z Z Z
d 0 0 0
drCi (~r , t) = i dr f (~r , E , t)S(~r , E ) dE T (t) i drCi (~r , t) (798)
dt 0
R
10. We further assume that dEd (E) = 1, thus the last term simplifies, and two long terms cancel out,
Z Z Z Z
0 0 0
dE drd (E) i
f (~r , E , t)S(~r , E ) dE T (t) dr f (~r , E 0 t)S(~r , E 0 ) dE 0 T (t) = 0
0 0
11. Traditionally we weight the PKEs by adjoint flux, not by unity. We add because the prompt neutron
spectrum has energy dependency, but no for precursors because precursors emission is not energy de-
pendency, it is just a fission rate.24
Z Z Z
d
[T (t)] = dE dr (~r , E) t (~r , E, t)S(~r , E) + s (~r , E 0 E, t)S(~r , E 0 ) dE 0
dt 0
Z Z Z X
0 0 0
+ p (E)(1 (~r )) f (~r , E , t)S(~r , E ) dE T (t) + dE dr (~r , E)d (E) i Ci (~r , t)
0 i
(800)
241
Lulu Li 22.211 Class Notes: April 25, 2012
X Z 1
1. Multi-group PKEs: normalize drg (~r )
Sg (~r ) = 1.0, we basically add a g subscript
g
vg (~r , t)
R X
to every term applicable, and convert any dEA(E) to Ag ,
g
Z
d X
[Tg (t)] = drg (~r ) t,g (~r , t)Sg (~r )Tg (t) + s,g0 g (~r , t)Sg0 (~r )Tg0 (t)
dt
g0
X Z X
+ p,g (E)(1 (~r )) f,g0 (~r , t)Sg0 (~r )Tg0 (t) + dr (~r )d,g i Ci (~r , t)
g0 i
(801)
Z Z Z
d X
drCi (~r , t) = dri (~r ) f,g (~r , t)Sg (~r )Tg (t) i drCi (~r , t) (802)
dt g
Z
d
[T1 (t)] = dr1 (~r ) {r,1 (~r , t)S1 (~r )T1 (t) + [1 (~r )][f,1 (~r , t)S1 (~r )T1 (t) + f,2 (~r , t)S2 (~r )T2 (t)]}
dt
Z X
+ dr (~r ) i Ci (~r , t) (803)
i
Z
d
[T2 (t)] = dr2 (~r ) [a,2 (~r , t)S2 (~r )T2 (t) + s12 (~r , t)S1 (~r )T1 (t)] (804)
dt
Z Z Z
d
drCi (~r , t) = dri (~r ) [f,1 (~r , t)S1 (~r )T1 (t) + f,2 (~r , t)S2 (~r )T2 (t)] i drCi (~r , t)
dt
(805)
Interpretaion:
(a) Notice we include the adjoint flux as weighting factors. Technically the parameters in the previous
sections should include corresponding weighting factors as well.
(b) has no T (t) temporal dependency. only depends on 1 but not 2 , because all delayed neutrons
are born in group 1, hence we do not care about the neutron importance in group 2.
(c) is the prompt neutron life time, thus only the importance of group 1 neutrons matters.
(d) Keep in mind that the classical PKE expression that we would show in the next subsection looks
simple; it is how these parameters are defined that is involved.
242
Lulu Li 22.211 Class Notes: April 25, 2012
3. We can further write the 2-group PKEs in matrix form as in Table 20.
243
Lulu Li 22.211 Class Notes: April 25, 2012
A special word about reactivity: (t) does not describe how much neutrons there are at a specific time;
it describes how much neutrons there would be eventually.
Static comes from steady-state transport equation, and it indicates excess (or lack) of reactivity
present in the system.
keff 1.0
(t) = (811)
keff
Dynamic changes as a function of time and indicates changes from an initially stable system.
f i
(t) = (812)
f
The unit of can be confusing. See Table 21.
2. The delayed neutron fraction (notice although this is called delayed neutron fraction, there is jp (E) term
X
in it). If we ignore fissioning spectrum, i .
i
X j Z
(E) j jf (~r , E 0 , t)S(~r , E 0 ) dE 0
R R
dE dr d
j 0
i (t) = !Z (813)
X X
jp (E)(1 j
id (E)ij jf S 0
R R
dE dr )+ dE
j i
244
Lulu Li 22.211 Class Notes: April 25, 2012
dE dr v1 S(~r , E)
R
R
(t) = !Z (814)
R X X j
j j i
jf S dE 0
R
dE dr p (E)(1 ) + d (E)i
j i
1 criticality 1
(t) (815)
vf va
w, v 1 0
hw, (F M )0 i hw, Fd 0 i
(t) = = eff = (818)
hw, F 0 i hw, F 0 i hw, F 0 i
245
Lulu Li 22.211 Class Notes: April 25, 2012
d i (t)
Ci (t) = T (t) i Ci (t) (820)
dt (t)
Remember that the above classic PKEs are obtained using assumptions:
No external source.
Constant i , i .
The approximation (~r , E, t) = S(~r , E)T (t), which is not always valid.
There is no feedback (e.g., the reactor is operating at low enough flux, no feedback, coefficients are
independent of solution), if (t) is known, one can solve for reactor power as a function of time.
2. For a steady state solution at power level T0 , we know,
d i
Ci (0) = 0 Ci (0) = T0
dt i
d
T (0) = 0 (0) = 0 (821)
dt
3. The matrix form of the 1st order ODE is (assuming three precursor groups, can be easily extend for more
precursor groups),
(t)
T 1 2 3
d 1
C1
1
N =
C2 N (t) = 2
[N (t)] (822)
dt
2
C3 3
3
246
Lulu Li 22.211 Class Notes: April 25, 2012
0
P = P + C C = P C (823)
This is a homogeneous linear system that can be written in the matrix form:
0
P 0
= (824)
+ C 0
Non-trival solution exists if and only if the determinant of the matrix is 0, that is,
2 + ( + 0 ) 0 = 0 (825)
247
Lulu Li 22.211 Class Notes: April 25, 2012
2. 2s rod drop scram ( = 8). This is a common assumption to make for real application because rods
are dropping to replace the water. Again, power goes to about 10% when rod fully inserted, and power
ultimately decays with time constant of the longest lived precursor (55s).
248
Lulu Li 22.211 Class Notes: April 25, 2012
3. Instantaneous rod withdraw ( = 0.1). In this senario, there is only one delayed neutron precursor group,
and we withdraw rod instantly. The reactor wants to response immediately (hence the jump) but it does not
have enough delayed neutron to sustain that increase, hence the jump is not enough and it will increase for
the rest of the way, until ultimately power increases with time constant of the longest precursor group (55s).
See next section for the derivation of prompt jump estimation.
4. Instantaneous rod withdraw with 8 delayed groups. The intial part of the transient is slightly more compli-
cated in Fig. 84a than 1-group case in Fig. 83(single delayed group is very sensitive to the time-constant).
Fig. 84b shows that, if we wait long enough, the power becomes an asymptotic exponential rise. We reach
secular equilibrium between power and precursor rate that the precursor rates have the same shapes as the
power, altough the rates may be off by a factor.
Figure 84: PKE Example 4, Instant Rod Withdrawal With 8 Delayed Groups
249
Lulu Li 22.211 Class Notes: April 25, 2012
6. Regulating rod withdrawal/insertion. Senario: rod is moved in for 2s, hold for 20s, re-positioned in 2s.
Power does not come back to the original state, because we are solving for an eigenvalue problem,
and what the asymptotic power is depends on how to get there. In the withdrawal case, the precursor
builds up and is continuing to build up during the second phase, so the power ends up higher than initially25 .
7. Super-prompt criticality/rod ejection/failed housing. Senario: rod is moved in for 1s. The point is, power
ascension is very sensitive as . When > , reactivity does not have to wait for the delayed
neutrons anymore, changes can happen instantaneously. Thus if your data is more than 5% uncertain,
the answer would vary quite a lot.
250
Lulu Li 22.211 Class Notes: April 25, 2012
251
Lulu Li 22.211 Class Notes: April 25, 2012
4. Interpretation:
Prompt jump approximation only works when < ! The smaller is compared to , the more
accurate the prompt jump approximation is, because at small reactivity, we can assume the change
in prompt neutrons are negligible compare with delayed neutrons (or say precursors dominates), in
d
which case it is valid to set dt P (t) = 0.
If = 0, both P (t), C(t) increases linearly with respect to time.
If is slightly larger than 0, both P (t), C(t) explodes exponentially. Eg., given a reactivity change of
+
0.1, the prompt jump is T (0 T0
)
= 0.1 = 1.111.
252
Lulu Li 22.211 Class Notes: April 25, 2012
2. Small positive means a rapid increase in power. Small negative means a rapid decrease in power.
3. If period is constant, power varies exponentially:
4. For one delayed precursor gropu, the reactor period can be separated into two parts (prompt period and
stable period) which are the two exponential of the solution with different coefficient. Recall Fig. 80.
5. After the initial transient, there is a fixed relation between the stable period and reactivity:
(
<
stable = (841)
>
6. Typical values: when <1 mk, the period is larger than 1 min; when = 6 mk, 1 s.
253
Lulu Li 22.211 Class Notes: April 25, 2012
X
T (t) = T (t) + i Ci (t) (843)
i
i
Ci (t) = T (t) i Ci (t) (844)
From Eq. 844, we can write Ci (t) in terms of T (t),
i
Ci (t) = T (t) (845)
( + i )
X i i
T (t) = T (t) + T (t) (846)
i
( + i )
X i i
= + i (848)
i
+ i
That is,
X i
= + (849)
i
+ i
Implications:
Asymptotic periods 0 = 1 : we can only use Inhour equation for asymptotic period, that is, a stable
condition indicated by a constant period. Thus to use Inhour equation, we have to make the reactor stable
first. For any time dependent problem, we have to use the more general IK.
We can compute using inhour equation if we know i , , i .
A typically assumption is to set 0 because the prompt neutron lifetime is so small (with the exception
of super prompt in which case is huge). Another way to put it is that the frequency of the reactor is going
to be small.
254
Lulu Li 22.211 Class Notes: April 25, 2012
2. In-hour assumes constant i , where in reality Bete-effective can be changed by rod worth by 3%. For
instance, when we drive the control rod in the center of a core, we drive the flux towards the outter edge of
the core and thus change beta.
3. To perform an accurate measurement, we really need to perform calculation.
Fig. 88 demonstrate this process. The problem is, the entire process takes more than 1 hour for each rod (because
we have to carefully wait for the power to stablize after each step). This is too costly for real reactors, hence we
need to derive a more general expression, which leads us to the next section on IK.
255
Lulu Li 22.211 Class Notes: April 25, 2012
256
Lulu Li 22.211 Class Notes: April 25, 2012
18.9 Derivation of IK
Assume we know the amplitude function T (t) (that is, we are given a desired reactor power shape vs. time),
we can solve for the precursor function C(t); then plugging back in the PKEs, we can get (t), that is, how to
operator reactor to achieve the desired power shape.
1. Given T (t), we can solve for the precursor equations,
d i
Ci (t) = T (t) i C(t) (850)
dt
d i
Ci (t) + i Ci (t) = T (t) (851)
dt
diag[i ]
If we define matrix [A] = diag[i ], [B] = , [Y (t)] = [B]T (t), then the above equation forms a
matrix system for all i,
d
[C(t)] + [A][C(t)] = [B]T (t) = [Y (t)] (852)
dt
2. Multiply by IF [eAt ],
d
[eAt ] [C(t)] + [eAt ][A][C(t)] = [eAt ][Y (t)] (853)
dt
d
[eAt ][C(t)] = [eAt ][Y (t)]
(854)
dt
That is, we can solve for the time-varying concentrations of precursors for any desired reactor power shape
in time by applying this equation successively for discrete steps.
4. From the PKEs, we can solve for the reactivity in terms of C(t), T (t) by making finite-difference approxi-
mation for derivative term:
d (t) X
T (t) = T (t) + i Ci (t) + Q(t) (857)
dt i
d X
(t) = T (t) + i Ci (t) Q(t) (858)
T (t) dt T (t) i T (t)
Tn Tn1 X
n = + i Ci,n Qn (859)
Tn tn tn1 Tn i Tn
257
Lulu Li 22.211 Class Notes: April 25, 2012
18.10 Applications of IK
IK is what is used in production tools. We will discuss three examples.
Prompt neutron life time is so short that spatial distribution of flux changes rapidly, thus we can assume that
it stabalizes already at each rod drop.
Disadvantages: time concern; boration dilution increases cost with waste disposal.
Alternatives: we use General Inverse Kinetics to get around about it.
258
Lulu Li 22.211 Class Notes: April 30, 2012
In this example, we would scram the core, then measure power change and infer T (t), thus estimating the
reactivity.
The rough method uses Prompt Jumpt Approximation.
More accurately, we need to model all the neutron sources, including (, n) reactions etc.
In reality, there are noisy signals in measured power and reactivity data; we need to smooth them out with
a fitted function before we can apply IK with external neutron sources.
The point is that we can use IK instead of prompt jump approximation to gain precision.
This technique is not used in the US that much because it is harsh on your system and it requires work to bring
the system back to critical. But it is one of the best ways to demonstrate to the licensing committee that a reactor
is capable of scraming.
Alternatively, we can do sub-critical source multiplication, that is, IK with a source in the reactor. This way
we dont even need to start with a critical reactor. This technique has only been used in experimental reactors but
not so much in power reactors because it requires detailed knowledge about the source distribution.
Note: we should be careful about power level. As shown in Fig. 92, power can change from a few watts to
20,000 MW in less than a second! In order to make IK valid, we have to bring power down to make sure there
is no thermal feedback. So it is a balance between getting a good signal and making sure there is no thermal
feedback, thus people end up running the reactor in the intermediate range.
259
Lulu Li 22.211 Class Notes: April 30, 2012
260
Lulu Li 22.211 Class Notes: April 30, 2012
2. If we perform a rod ejection at hot zero power, the magnitude of flux/power can change from a few Watt to
20,000 MW in less than 1s! Also the fuel enthalpy at failure is dependent on fuel BU (the higher the BU,
the easier it is for the fuel to fail) as in Fig. 94. Thus we need to model control rods somewhat faithfully to
get a reasonable safety analysis.
3. The importance of modeling feedback can be illustrated in Fig. 95. Without feedback, the reactor wants to
be a bomb and the power shoots up with an asymptotic period. With Doppler feedback, both the reactivity
and the power decreases before the rod is even fully inserted. When correctly designed, reactors use feed-
261
Lulu Li 22.211 Class Notes: April 30, 2012
back effects to keep them from acting like bombs26 . The feedback is especially strong when > 1$. A
second peak is the characteristic of longer rod change.
Figure 95: 1.5$ RIA Without Feedback (top row) and With Doppler Feedback (bottom row)
26 There is no Doppler feedback for fast reactors, so fast reactors have to do all kind of tricks to get a negative feedback. Also fast reactors
have positive sodium void worth, because as energy deposits in the sodium, voids are generated, pushing sodium away, generating a positive
sodium void worth
262
Lulu Li 22.211 Class Notes: April 30, 2012
19.2.1 Assumptions
To start, we make three assumptions27 ,
1. If , we can ignore the delayed neutrons, hence the precursor equation in PKEs goes away, and we are
left with the power distribution P (t) with no precursor term28 .
d (t) X (t)
P (t) = P (t) + i Ci (t) P (t) (860)
dt i
It is fair to ignore the precursors, because as in Fig. 96 top row, we can see that the power with precursor is
small enough that the F-N model provides a good approximation; the bottom row images are plotted on a
log-log scale to show the small difference made by the precursors.
27 The historical reason for these assumptions is that the model was developed for weapon use, hence and rapid transient.
28 we use P (t) instead of T (t) for the amplitude function now to avoid the confusion with temperature T that we would use repeatively in
this section.
263
Lulu Li 22.211 Class Notes: April 30, 2012
3. Assume a Doppler feedback coefficient independent of temperature (recall that we calculate that PWRs
have a Doppler coefficient is abou -3 pcm/K),
0
(t) = rod (Tfuel Tfuel ) (862)
Eq. 866 can be used to find peak power behavior, that is, we set dTdP
fuel
= 0,
1h peak
i
Cp (rod ) Cp Tfuel =0 (867)
peak rod
Tfuel = (868)
Then we can solve for peak power as well,
" 2 #
peak 1 rod Cp rod
P = P0 + Cp (rod ) (869)
2
Cp (rod )2
P peak = P0 + (870)
2
Take-away Messages:
1. Peak temperature is independent of neutron lifetime and heat capacity Cp .
2. Peak power is proportional to heat capacity Cp and inversely proportional to prompt neutron lifetime
and Doppler feedback coefficient . A direct consequence is, fast reactors whose is orders of magnitude
smaller than LWRs are going to generate a lot of heat in a short amount of time.
3. These equations are derived with > . As , the peak temperature and power become very sensitive
to .
29 The time constant for heat to be removed from UO2 fuel is about 5 mins.
264
Lulu Li 22.211 Class Notes: April 30, 2012
dP (t) dP (t) d
= (871)
dt d dt
dP (t) dP (t) d (t) Cp
= = P (t) P (t) = [(t) ] (872)
d dt dt Cp
We integrate with repsect to and evaluate the constant of integration by using the step reactivity rod ,
Cp
((t) )2 + (rod )2
P (t) = P0 + (873)
2
Consider the transient terminated when P (t) returns to P0 , then
That is,
end 2(rod ) 0
Tfuel = + Tfuel (876)
265
Lulu Li 22.211 Class Notes: April 30, 2012
266
Lulu Li 22.211 Class Notes: April 30, 2012
d
Tfuel (t) = aP (t) b[Tfuel (t) Tcoolant (t)] (878)
dt
d
Tcoolant (t) = c[Tfuel (t) Tcoolant (t)] d[Tcoolant (t) Tinlet (t)] (879)
dt
Re-write in a matrix form,
d Tfuel (t) b b Tfuel (t) aPfuel (t)
+ = (880)
dt Tcoolant (t) c c+d Tcoolant (t) dTinlet (t)
Using integrating factors to solve the above first order ODE system (we solved the exact form of problem
d
dT C + AC = Y for the IK equations as in Section 18.9), we get (assuming zero initial power, zero initial
inlet temperature),
n+1 n
Tfuel b b n Tfuel
n+1 = exp t n
Tcoolant c c + d Tcoolant
1 n+1
b b b b b b aPfuel
+ exp nt exp nt n+1
c c + d c c + d c c + d dTinlet
(881)
We make use of that we know at full power, flow rate is 30 W/g, Cp = 300J/kg-s, = 3 pcm/K, Vcoolant =
0 0 0
2m/s, Tinlet = 540K, Tcoolant = 560K, Tfuel = 900K, core height is 4m.
Recall if we do not have heat transfer (axial or radial), then the fuel temperature/power would (linearly)
increasing as weve demonstrated before.
267
Lulu Li 22.211 Class Notes: April 30, 2012
For the TH part, recall that we have derived the analytical solution Eq. 881.
268
Lulu Li 22.211 Class Notes: April 30, 2012
Take-away:
269
Lulu Li 22.211 Class Notes: April 30, 2012
270
Lulu Li 22.211 Class Notes: April 30, 2012
Figure 101: Augmented Matrix Equations for 1st Order Coupled ODE
Then we can use our familar integrating factor method to solve this first order ODE:
1
[N ]n+1 = exp[A]m [N ]n + h[A]m i h[1] exp[A]m i [S]m (886)
271
Lulu Li 22.211 Class Notes: April 30, 2012
This solution requires no synchronization iteration and allows very large time steps.
There are two approaches here:
1. Replace with its old time step value.
2. Alternativley, we can do a synchronization step. Effectively we are doing JFK in the sense that derivative of
a term with two variables is just the derivative of one times the value of the 2nd term plus the derivative of
the 2nd term times the 1st term. We ignore the 2nd term of Jacobian, and as shown later with a reasonable
fine time step, we do not even need synchronization as the 2nd term of the Jacobian is so small.
Take-away:
Compare augmented solution with operator splitting, even with no synchronization, the augmented solution
is more accurate than operator splitting.
Time step is far less important than the order of the operator. In the 0.5 dollar ramp insertion case, direct
solution provides a factor of 10 more accuracy.
272
Lulu Li 22.211 Class Notes: April 30, 2012
2. Ramp 0.5 reactivity. Undershoots are very common in negative reactivity transients. The slower ramp-
ing results in smaller , P because slow transients are in pseudo-equilibrium, and there is enough time
for the feedback. That is, The smaller rate of reactivity change is, the smaller the overshoot/undershoot
is. The final/asymptotic results are a different story, as they only depend on (system always find stable
points) and are independent of ddt .
273
Lulu Li 22.211 Class Notes: April 30, 2012
3. 1.5 RIA from HZP(eject rod in 0.1s, scram delay = 0.5s, scram time = 1s, follow to 3s). Nominal rated
power increases by about 100 times in 0.2s. The power and temperature turns around due to Doppler
feedback. Temperature peaks around 800K and ends around 760K:
4. 1.5 RIA from HFP (exact same condition as last one except HFP this time). Power increases by about
70 times. Because we start at a higher power region, we get feedback more quickly, thus peak power is
lower compare with the HZP case. T is about the same as the HZP (initial temperature is higher, peak
fuel temperature around 1200K) because the are the same for the two cases and our Doppler feedback
coefficient is linear.
274
Lulu Li 22.211 Class Notes: April 30, 2012
5. 1.5 RIA from HZP (same condition as the previous HZP RIA case, but eject rod in 1s instead of 0.1s).
Power only goes up by 4 times in 1s. Again Doppler feedback turns the power around, and temperature
peaks around 800K which is the same as the fast ejection case.
Figure 107: 1.5 RIA from HZP with Feedback, Slower Ejection
6. 1.5 RIA from HFP (same condition as the previous HFP RIA case, but eject rod in 1s instead of 0.1s). The
ejection is slow enough that the reactor is at equilibrium the whole time. The power peaks by a factor of 5
in 0.2s. Peak fuel temperature is around 1200K which is again the same as the fast ejection case.
Figure 108: 1.5 RIA from HFP with Feedback, Slower Ejection
275
Lulu Li 22.211 Class Notes: April 30, 2012
7. 1.5 bank withdrawal RIA from HZP (eject in 100s). The ejection is so slow that the reactor is at equilib-
rium the whole time and the system never get super prompt critical (Doppler stops power ascension). The
system reaches its nominal rated power in 100s.
Figure 109: 1.5 RIA from HZP with Feedback, Slowest Ejection
8. 1.5 bank withdrawal RIA from HFP (eject in 100s). Doppler reactivity is continually synchronized with
rod reactivity (hence the straight line looking?). The system reaches twice its nominal rated power in 100s.
Figure 110: 1.5 RIA from HFP with Feedback, Slowest Ejection
276
Lulu Li 22.211 Class Notes: April 30, 2012
9. 1.5 bank withdrawal RIA from HFP (eject in 100s, no scram). Doppler reactivity is continually synchro-
nized with rod reactivity (hence the straight line looking?). The ejection is slow enough that if we can take
away the scram, and the system reaches a new equilibrium at twice its nominal rated power in 100s.
Figure 111: 1.5 RIA from HFP with Feedback, Slowest Ejection, No Scram
Take-away: we dont really care about the temperal difference/integration, because the asymptotic temperature
only depends on the reactivity change, not on the time dependent behavior.
277
Lulu Li 22.211 Class Notes: May 2, 2012
19.5 Summary
1. PKEs assume you are solving for the core average properties; PKEs assume flux spatial shape is constant.
Thus PKEs are not very precise for large spatial flux changes.
2. Peak temperature are proportionally larger than core average properties.
3. If we really want spatial dynamics, we need to do 3D spatial dynamics to get correct predictions to compli-
cated problems.
This thermal hydraulics model we have is very crude. For one thing, it has a huge diffusive property, that is, it
would not predict any thermal shock behavior.
As in Fig. 112, about 3% of the energy deposited into the coolant (about 2.5 2 MeV =5 MeV worth of
neutrons from 200 MeV just from neutron slowing down that is deposited into the coolant), hence the coolant
temperature would change slightly as well. Net reactivity comes back to zero. Remember PKEs are not precise
for large spatial flux changes.
278
Lulu Li 22.211 Class Notes: May 2, 2012
20.1.1 Challenges
There are a couple levels of complexity for performing 3D core analysis,
One reactor at one state point: couple core neutronics/fuel heat conduction/coolant hydraulics/structural
mechanics for 60,000 fuel pins and 100 axial levels to evaluate thermal margins;
Repeat for 50 depletion states and track 300 isotopes;
Repeat for 10,000s limiting conditional simulations and 100s of transient accident simulations for safety
analysis;
Repeat for 1000s of startup and maneuver simulations;
Repeat for millions of cycle depletion simulation for designing loading design and optimization;
For operator training simulator in real time, 24/7 operation at 5-10 Hz.
The challenges for 3D core calculation include,
Predict pin power, axial shapes of pin power: as in Figure 113, the spacers depress the flux. Though even
without the spacers, the axial shape would still be more towards the bottom compare with a cosine shape
due to the temperature coefficients;
Predict core reactivity with core burnup: as in Figure 113, burnable poisons deplete as burnup increases;
Predict in-core detector response;
Predict control rod worths and temperature coefficients.
279
Lulu Li 22.211 Class Notes: May 2, 2012
The best parallel transport grind time currently comes from Texas A&M PDT code is 300 nanosecond/un-
known. On top of that, assume we perform 20 fission source iterations, we are looking at 100,000 core hours.
Not to mension that we have no inlcude thermal-hydraulic feedback, cross section evaluations, boron search
to criticality, equilibrium xenon, and control rod search to criticality.
2. Pin-by-pin diffusion with homogenized pin cells: the number of meshes we need is approximately
Thus we reduce the core hours from 100,000 to 1 core hour. The drawback is that pin homogenization causes
a loss of accuracy around strong absorbers. This is now a tractable problem, but not used for production.
3. Nodal method: this is what production codes typically use. There is a huge push towards nodal method,
because nodal methods take a mesh size of 20cm to be accurate, whereas Finite Difference method requires
a mesh size of about 1 cm. In terms of the amount of errors, using CMFD on a reactor application would
yield somewhere between 5-10% error.
Multilevel Approach for LWRs (the reason we emphasize this approach is for LWRs is because the spatial
mesh should agree with mean free path which depends on energies):
1. Spectral code to get from 1000s of energy groups to 100s of energy groups.
280
Lulu Li 22.211 Class Notes: May 2, 2012
D
where the migration coefficients Mp2 = pp , g is an adjustable parameter, and the coupling coefficients
q
Mp2 Mp2 X
are wpq = (1 g) + g 2 , wp p = 1 wpq . There is not much physics in this model,
2h h 6
q=1
because the coefficients are derived assuming two infinite half planes.
3. 70s-80s advanced nodal diffusion methods.
(a) Goals:
Directly represent baffle/reflector, instead of like previous codes that just use screw-driver to
account for the albedos;
No user-adjusted parameters, should be all physics;
Full 3D thermal-hydraulic feedback by assemblies;
Recover individual pin powers;
(b) In Germany: KWU became Siemens who became Areva, and they have two version of the nodal
diffusion method:
NEM: Wagner, Koebke, Winter et. al.
CUBOX/QUABOX: Finnemann.
(c) At MIT:
Prof. Kent Hansen was working on FEM (finite element method), which turned out to be limited
because the methods are more complex compared with nodal methods.
Prof. Allan Henrys group did ANM-1D analytic, NEM variants, ANM-FLT, and QUANDRY.
(d) At University of Illinois, Prof. Jack Dornings group has NGFM.
4. Modern advance nodal codes: there are numerous of them, and they should all be the same. Specific code
names are listed in Lec 19 in 22.212.
281
Lulu Li 22.211 Class Notes: May 2, 2012
2D+ D
J = (+ ) (891)
D h+ + D+ h
+
where nodal methods is no better than the D2D D
h+ +D + h .
Next we derive the nodal balance equations.
1. We start from 3D steady-state multigroup neutron diffusion equation,
G G
1 X X
J~ g (~r , E) + rg g (~r , E) = g f g0 g0 (~r , E) + sg0 g g0 (~r , E) (892)
keff 0 0
g =1 g =1
2. Volume average of diffusion equation for a node: we integrate every term over the node volume then divide
by volume. The volume averaged flux becomes,
Z hx Z hy Z hz
1
= (x, y, z) dx dy dz (894)
hx hy hz 0 0 0
3. The leakage term becomes (after integrating the divergence term using Gauss Theorem),
Z hx Z hy Z hz Z hy Z hz
~
J g dx dy dz = (Jx (hx , y, z) Jx (0, y, z)) dy dz (895)
0 0 0 0 0
Z hx Z hz
+ (Jy (x, hy , z) Jy (x, 0, z)) dx dz (896)
0 0
Z hx Z hy
+ (Jz (x, y, hz ) Jz (x, y, 0)) dx dy (897)
0 0
Notice,
282
Lulu Li 22.211 Class Notes: May 2, 2012
283
Lulu Li 22.211 Class Notes: May 2, 2012
284
Lulu Li 22.211 Class Notes: May 2, 2012
2. The RHS becomes not interesting after a while because it is just the average flux times cross sections.
3. Define dimensionless independent variables,
x y z
x = , y = , z = (902)
hx hy hz
and we can transform the integration and the derivative operator as well,
du d 1 d
= du , = , u = x, y, z (903)
hu du hu du
4. Simplify the averaging with dimensionless variables (l is left, r is right): we basically say the net current
on the left plane can be found by integrating the 3D net current at x = 0 over the left plane.
Z 1Z 1 Z 1Z 1
Jxgl = Jx (0, y , z ) dy dz , Jgxr = Jx (1, y , z ) dy dz (904)
0 0 0 0
Z 1 Z 1 Z 1
= (x , y , z ) dx dy dz (905)
0 0 0
285
Lulu Li 22.211 Class Notes: May 2, 2012
6. We look at the leakage term after the transverse integration, notice that at the end of the day terms only
depend on x direction becuse we integrate away the corerwponding y and z directions:
Z 1Z 1 Z 1Z 1
1 Jx 1 Jy 1 Jz
J~ dy dz = + + dy dz (907)
0 0 0 0 hx x hy dy hz dz
Z 1Z 1
Dg 2
1 Jy 1 Jz
= 2 + + dy dz (908)
0 0 hx x2 hy dy hz dz
Z 1Z 1
Dg 2
= 2 (x , y , z ) dy dz (909)
hx x2 0 0
Z 1 Z 1
1 1
+ (Jy (x , 1, z ) Jy (x , 0, z )) dz + (Jz (x , y , 1) Jz (x , y , 0)) dy (910)
hy 0 hz 0
Dg x (x ) J yr (x ) J yl (x ) J zr (x ) J zl (x )
= + + (911)
h2x x2 hy hz
where we defined the plane-averaged 1D flux, and it does not have mesh spacing on the bottom because
we defined dimentionless terms:
Z 1Z 1
x = (x , y , z ) dy dz (912)
0 0
Z 1 Z 1
J zr (x ) = Jz (x , y , 1) dy , J zl (x ) = Jz (x , y , 0) dy (914)
0 0
7. Collect all the terms, we get the transverse integration of 3D diffusion equation,
G G
X J gur (x ) J gul (x ) D d2 1 X X
2 (x ) + rg ( x ) = g f g 0 g (x ) + sg0 g g (x )
u=y,z
hu hx dx2 gx gx
keff 0 0 g =1 g =1
(915)
If we define a transverse leakage term (the source that comes in the plane) and move it to the RHS,
1
Lgu (x ) = J gur (x ) J gul (x ) , u = y, z (916)
hu
and also define the diffusion equivalent group constant,
D
xDg = (917)
h2x
286
Lulu Li 22.211 Class Notes: May 2, 2012
We get the transverse integrated 1D diffusion equation, with two additional terms that describe the leakage
in the other two directions. If we know the two leakage terms exactly, then we can solve the 1D diffusion
equation exactly.
G G
d2 1 X X
xDg gx (x ) + rg gx ( x ) = g f g 0 0 (x ) +
g x sg0 g g0 x (x ) Lgy (x ) Lgz (x )
dx2 keff 0 0
g =1 g =1
(918)
8. Repeat for the other two directions, we get a set of 3 directional 1D diffusion equations.
G G
d2 1 X X
xDg (x ) + rg gx (x ) = g f g g0 x (x ) +
0 sg0 g g0 x (x ) Lgy (x ) Lgz (x )
dx2 gx keff 0 0
g =1 g =1
G G
d2 1 X X
yDg (y ) + rg ( y ) = g fg 0 g0 y (y ) + sg0 g g0 y (y ) Lgz (y ) Lgx (y )
dy2 gy gy
keff 0 0
g =1 g =1
G G
d2 1 X X
zDg ( z ) + rg (z ) = g fg 0 0 ( z ) + sg0 g g0 z (z ) Lgx (z ) Lgy (z )
dz2 gz gz
keff 0
g z
0
g =1 g =1
9. Interpretations:
We turn a 3D partial differential equation into three 1D ordinary differential equation that are coupled
through average transverse leakage term.
It is exact if the transverse leakage shape is known.
Why nodal methods work? As the node becomes bigger and bigger, it is less likely that a random
neutron to stream right through the node (thus contributing to the J~ ). That is, J~ streaming out
the node from the right side is insensitive to the transverse leakage term from the top and the
bottom. Thus we can decouple leakage into three dimensions at large meshes.
10. Approximation On Transverse Leakage: Quadratic Approximation in Each Node (Finnemann, KWU).
From observation, flux is insensitive to the transverse leakage shape, so we are going to perform a quan-
drature polynomial with 2nd order polynomials of the leakage term, and iteratively update them. Quadratic
approximation in each node,
L() = L + l1 P1 () + l2 P2 () (919)
We apply average TL conservation scheme to determine l1 , l2 . We use three node average transver leakages
(the values of each node and its two adjacent nodes), and impose constraint of conserving the averages of
two adjacent nodes as in Fig. 116. Notice the parabola defined by the quadratic polynomial is assumed to
extend over three nodes, but is only used in the middle node to represent the transverse leakage shape. Alter-
native approach: Finnemann (1977) uses continuity on transver leakage and its first derivative to determine
the quadratic coefficients.
About how we handle the rest of this method, we present three methods in the following sections.
287
Lulu Li 22.211 Class Notes: May 2, 2012
288
Lulu Li 22.211 Class Notes: May 2, 2012
P0 () = 1 (921)
P1 () = 2 1 (922)
P2 () = 6(1 ) 1 (923)
P3 () = 6(1 )(2 1) (924)
2
P4 () = 6(1 )(5 5 + 1) (925)
Notice these basis functions are not orthogonal, and integration from 0 to 1 would result in zero which is
why we choose these basis (we should use these basis for 22.212 HW9). Keep in mind that our transverse
leakage is 2nd order (because the result is not very sensitive to the expansion of the transverse leakage).
2. BCs: For each node at each energy group, we have 5 unknown coefficients, and 3 constrains:
1 from node average fluxes, or say nodal balance.
2 incoming current boundary conditions.
Hence we need 2 additional constraints because polynomials can not satisfy the differential equations ex-
actly.
3. To find the next 2 constrains, we use weighted residual method for each group in the 1D diffusion equation,
Z 1 Z 1
d 1
w() D 2 () + r () d = w() () + S() L() d (926)
0 d 0 keff
The closure relationship basically forces the P1 , P2 integration to go to zero. Forcing the 1st moment of
diffusion equation to go to zero gives us,
Z 1
d 5q1 + 3q3 5a1 r
P1 () D 2 () + r () Q() d = 0, a3 = (927)
0 d 3(60D + r )
Notice how simple the expressions for a3 , a4 come out to be. Notice a3 only depends on the odd nodes;
a4 only depends on the even nodes. Notice we dont know a1 , a2 . But we get two pieces of information
from the spatial moments. That gives us in total 2 spatial moments 2 nodes 2 groups = 8 constraints.
Together thats 16 constraints for 16 unknowns.
4. The choice of the weighted residual equations is flexible the results are not very sensitive to the weights.
For instance, we can choose the same expansion coefficients for Pi .
5. Two-node 2-group NEM is different from finite-difference methods in terms of:
289
Lulu Li 22.211 Class Notes: May 2, 2012
Flux shapes in each group affects coupling equations for other group;
The coupling equations depend on transverse leakages;
Final form of the equations can be made similar to 3D finite difference.
There has been more paper on how to find other forms of weights, but it turns out Finnemanns weights were good
enough. Each energy group can be solved independently.
290
Lulu Li 22.211 Class Notes: May 2, 2012
D1 B 2 + r1 keff 1 1
1
H
f 1 keff f 2 0
= (933)
12 2
D2 B + r2 2H 0
For nontrivial solution, we set the determinant to zero, and re-write the expression in terms of B 2 :
2b
z }| c
f 1
!{ z
}|
{
r1 r2 keff k r1 r2
(B 2 )2 + + 2
B + 1 =0 (934)
D1 D2 D1 keff D1 D2
The roots of characteristic equations are the eigen-buckling; more specifically, one of them is the funda-
mental mode (this is the fast group), and the other is the harmonic mode (this is the thermal group)30 ,
r
2 c > 0 k > keff
Fundamental Mode B1 = b 1 + 1 2 = (935)
b < 0 k < keff
r
c
Harmonic Mode B22 = b 1 1 2 < 0 (936)
b
The homogeneous solution for group g is:
H ag1 sin(B1 x) + ag2 cos(B1 x) k > keff
Fundamental Mode g1 (x) = (937)
ag1 sinh(B1 x) + ag2 cosh(B1 x) k < keff
H
Second-Harmonic Mode g2 = ag3 sinh(B2 x) + ag4 cosh(B2 x) (938)
where k is for this specific node (depends on material properties), whereas keff is for the entire reactor
(depends on geometry). The combined homogeneous solution shows that group 1 and group 2 equations
are linearly dependent,
H
1 (x) a11 sin(B1 x) + a12 cos(B1 x) r1 r2 a21 sin(B1 x) + a22 cos(B1 x)
= =
2H (x) a21 sinh(B2 x) + a22 cosh(B2 x) 1 1 a23 sinh(B2 x) + a24 cosh(B2 x)
(939)
b b2 4ac
30 recall quadratic roots for ax2 + bx + c = 0 are x = 2a
291
Lulu Li 22.211 Class Notes: May 2, 2012
4. Next we find the particular solution, which is determined solely by quadratic transverse leakage,
where
b10 + 6D1 c12 /x2
c1p 1 b1p c10 1
=A , p = 1, 2. =A . (942)
c2p b2p c20 b20 + 6D2 c22 /x2
292
Lulu Li 22.211 Class Notes: May 2, 2012
d2
D2 2 (x) + r2 2 (x) = s12 1 (x) Ly2 (x) = Q2 (x) (944)
dx2
We end up with exponential homogeneous solution and polynomial particular solution. The reason we
like this method is that these sinh, cosh terms provide huge gradient changes.
4. We differentiate to get expression for net current at the interface:
d
Jg (x) = Dg g (x) (947)
dx
We use the continuity of net currents/flux to get analytic expression for coupling.
5. Summary. SANM is advantage for two reasons:
(a) As in Fig. 117 the NEM predicts the center flux (where the instruments would be) wrong.
(b) Weighted residual has the advantage of capable of incorporating cross sections spatial dependence.
293
Lulu Li 22.211 Class Notes: May 2, 2012
294
Lulu Li 22.211 Class Notes: May 2, 2012
4. We solve the 3D balance equation by fission source and flux/leakage iteration, similar to finite difference,
but leakate term has group-to-group coupling.
5. Update keff and return to 2 until converged.
Next the non-linear iterative solution sequence was developed,
1. Guess keff , solve the standard 3D finite difference equations.
2. Evaluate all interface currents from 2-node equations,
" + # " # X c11
J1 a11 a12 1 b11 b12 1 c12 Lm1
= + + (950)
J2 a21 a22 2 b21 b22 2 c21 c22 Lm2
m=1,4
Its almost finite difference, but higher order in the sense that we say if flux has already converged, what
would the current be.
3. Redefine equivalent coupling terms to preserve nodal solution for current. Notice the coefficients , , ,
are non-linear functions of the solution.
" + # " #
J1 0 1 0 1
= + (951)
J2 0 2 0 2
This is almost CMFD. As we solve the problem, we will slowly find out the 4 coefficients and if the solution
converges it should be identical to the original matrix problem.
4. Set up global matrix equation,
1
[A][[ ]] = [M ][[ ]] (952)
keff
5. Solve 3D balance equations by fission source and flux iteration (same as finite-difference, and no group-to-
group spatial coupling term).
6. Update keff and return to 2 until converged.
This method is almost the same as CMFD, but with slightly different coefficients , , , . The point is, whether
we do any one of the three nodal methods, we get a slightly different net current, the rest of the iteration looks the
same. Mechanically, it is the same as the two-node two-group finite difference solution that weve already done.
295
Lulu Li 22.211 Class Notes: May 2, 2012
4. SANM is the most practical for LWR applications, because it has simple algebra, is multi-group applicable,
and is comparable in accuracy to ANM.
5. Non-linear solution method (eg, iterative) can be easily applied to all three methods.
296
Lulu Li 22.211 Class Notes: May 7, 2012
21 Homogenization Methods
We continue our discussion on nodal methods, and would cover spatial dependency of cross sections and ho-
mogenization methods. This lecture follows closely on Smiths Assembly Homogenization Techniques for Light
Water Reactor Analysis (Prog. in Nulear Energy, 1986).
Again, we set the 1st moment of neutron diffusion equation to zero, hence deriving an expression for a3 ,
Z 1
d2
5q1 + 3q3 5a1 r +
P1 () D 2 () + r ()() Q() d = 0, a3 = (954)
0 d 3(60D + r ) +
Similarly we set the 2nd moment of neutron diffusion equaion to zero, hence deriving an expression for a4 ,
Z 1
d2
7q2 + 3q4 + 7a2 r +
P2 () D 2 () + r ()() Q() d = 0, a4 = (955)
0 d 420(D + 3r ) +
Interpretations: flux distribution changes after we add in the spatial dependency of cross sections; but
mechanically there is nothing tricky: we are just adding more terms, designated by , to the expressions
in the above two equations.
3. KWU depletion benchmark using BOC-2 Powers: without space-dependent xs, the error on keff is about
12pcm (on the edge); with space-dependent xs, the error on keff drops to about 4pcm (on the edge). The
error on the interior is always smaller. Basically if we do not model the space-dependent xs, then our
assembly is solved using zero current BC and is not accurate when the location and/or the orientation of the
assembly is changed. Alternatively we can further sub-divide the node to get a reduced error as well.
297
Lulu Li 22.211 Class Notes: May 7, 2012
and analogous equation for the homogenized model using homogenized parameters,
G
~g (~r ) +
tg (~r )g (~r ) =
X 1 gg0 (~r ) g0 (~r )
J Mgg0 (~r ) + (957)
0
keff
g =1
2. We need to decide what terms to preserve. For instance, if we preserve the scalar flux, reaction rates (for
every reaction type and every energy = t, gg 0 etc) and the leakage term, keff will be preserved.
R
g g dr
Z Z
g g dr = g g dr g = VRi
i
(958)
Vi Vi dr
Vi g
S k J~ g dS
Z Z R
Jg dS = ~
J g dS i
Jg = Dg g , where Dg = R i (959)
Sik Sik Sk
g dS
i
3. What happens is that we dont really know the true heterogeneous flux, so we further approximate hetero-
geneous flux using results from assembly results.
Z Z
g dr = Ag dr (960)
k J~ g dS
R
gi = R Si
D (961)
Sk
Ag dS
i
Using the assembly results give us AXS (assembly flux-weighted cross section).
4. AXS (assembly flux-weighted cross sections): in a HAFAS BWR benchmark case, notice that the error
(-0.44% for keff , 5.5% for average assembly power) is somewhat independent of the mesh size (1x1 vs.
3x3) and angular expansion (P1 vs. P3), which suggests that the error comes from homogenization of
the cross sections .
5. RXS (reference cross sections): the errors mentioned above do not go away neither if we use perfect cross
sections (RXS): -0.34% for keff , 4.1% for average assembly power. As we will examine in the next section,
the errors have to do with the discontinuity of scalar flux.
6. Flux discontinuity: since the homogenized flux distribution in each node is affected by the diffusion coef-
ficients, and the choice of flux weighted diffusion coefficients is in a sense entirely arbitrary, the interface
fluxes can be different as in Figure 118(c). As a direct result, the homogenized flux in nodes i, i + 1 when
the two-node homogenized diffusion problem is solved with BCs Ji , Ji+1 +
and continuity of flux and cur-
rent interface conditions would yield Figure 118(d), which is different from Figure 118(c). That is, the
homogeneous flux is not continuous across the surface anymore.
298
Lulu Li 22.211 Class Notes: May 7, 2012
7. Heterogeneity factor(HF): Koebke introduced the Heterogeneity Factor(Lugano, 1978), which states,
i fi+ =
i+1 fi+1 (962)
where the equivalence factors are defined as (and computed from reference solution),
+ i+1
fi+ = i
, fi+1 = (963)
+
i
i+1
which says that the heterogeneous flux is continuous across the interface and that there exists a direct
relationship between the heterogeneous and homogenized surface fluxes. When the homogenized two-
node problem is solved, the homogenized flux is made discontinuous (by a factor of fi+ /fi+1 ) and the
homogenized flux distribution will be the same as that in Fig. 118(c), which results in the preservation
of interface current / net current by adding a degree of freedom which essentially relaxes boundary
conditions.
8. Nodal equivalence theory and discontinuity factor(DF): Koebkes method of constraining the diffusion co-
efficients such that the heterogeneity factors are identical on both surfaces of a node requires an iterative
299
Lulu Li 22.211 Class Notes: May 7, 2012
method to determine diffusion coefficients. A variation of Koebkes method takes advantage of the fact that
exact heterogeneity factors can be defined from Eq. 962 for any value of the diffusion coefficient. Note
that unless the diffusion coefficients are found iteratively, the values of the heterogeneity factors on the
opposite faces of a node will be different. These two factors are referred to as discontinuity factor(DF) to
distinguish them from heterogeneity factors,
u
ugi,j (ul ) u+
ugi,j (ul+1 )
fgi,j = , fgi,j = (964)
u (ul )
gi,j u (ul+1 )
gi,j
where ul , ul+1 are the lower and upper u-direction boundaries of node i, j.
9. Application of DF: Koebke modified the NEM code, demonstrating that heterogeneous reference solutions
could be reproduced by using heterogeneity factors; Smith modifed the QUANDRY code, demonstrating
that even CMFD method could reproduce heterogeneous reference solution using discontinuity factors. The
updated nodal interface current on the upper u-direction surface of node i, j in the CMFD approximation
is,
300
Lulu Li 22.211 Class Notes: May 7, 2012
301
Lulu Li 22.211 Class Notes: May 7, 2012
Collapse baffle cross sections into two groups to assure that finite difference diffusion calculation would
reproduce the transport results;
Use flux, current distribution and reaction rates directly to define homogenized cross sections and DFs,
which accurately model the baffle and reflector.
The inherent advantage of HFs or DFs is that they are chosen in such a way that the nodal model would reproduce
the net currents at the core/baffle interface and the net reaction rates in both the fuel assembly and the homogenized
baffle/reflector without explicitly representing the baffle. It turns out that for reflector purpose, the fast neutron
leakage makes up for 90% of the leakage, and it is very insensitive to enrichment, burnup etc.
Application and Limitation of DFs:
Other geometries: Assembly/Reflector DFs also work in hexagonal geometry.
Other reactor types: DFs may not work because graphite or gas or fast reactor the mean free path is long.
DFs work well mostly for thermal LWRs.
Other applications of homogenization theory:
Fine mesh data generation for 3D/1D axial models. Basically we homogenize axial fuel asembly, and
depens on how we define the node, the DFs come out to be different.
Nonlinear acceleration of fine-mesh transport methods.
Nonlinear acceleration of Monte Carlo.
302
Lulu Li 22.211 Class Notes: May 7, 2012
21.5 Example
Fast flux peaks in fuel; thermal flux peaks in moderator.
Energy-in on every term, energ-out for the terms on the RHS. Working with a be careful about the (n,2n)
reactions. That is,
t = s + a + n,2n + n,3n + s s = s + 2n,2n + 3n,3n + (967)
Loop over all the surfaces, solve a fixed-source diffusion equation for each region (that is, given the current
on the boundary, we can solve each region separately). To get an accurate solution, we need to control D.
If we use the heterogeneous reference solution divided by single assembly (SA) solution, we get sort of like
the reference energy shape. It is almost the same as what we get out of the diffusion theory, except at the
interface, which has two reasons: 1) diffusion coefficient is off in diffusion theory; 2) we assume we can
separate form function and background behavior.
Full core diffusion eigenvalue solver: we need to preserve both leakage rate (DFs) as well as reaction rates
(XSs).
Apply DFs to spatial discretization method: puts f terms in there. Only the ratio matters. f s are different
for the two surfaces of the same node.
Diffusion coefficient: spatial first then energy to treat spatial void.
1. Spatial: we first do calculate resolve the spatial dependency of diffusion coefficient, if there is a void,
we average out the diffusion coefficient.
R 1
2. Energy: then we resolve the energy dependency by performing R3tr . In Monte Carlo, we cannot
really perform this in fine energy space, because for materials like hydrogen we quickly run into
tr 0 for hydrogens. Instead, we do a two-step process:
First flux collapse tr , get an approximate of Dg in fine groups.
Then flux collapse Dg for few groups.
It is not important for preserving keff what diffusion coefficient you use. Because keff only captures the
average behavior as it is integrated over all space.
Reconstruction: need to re-normalize the convolution flux, because the integral of the product is not the
same as the product of the integrals.
Subtle point: net down-scattering can be more accurate for certain single assembly calculation.
Recap: this DF is the physical DF that preserves the leakage in assembly calculation; whereas as the D
that we talked about before is a pure mathematical concept that we come up with to do non-linear iteration
to preserve physics between the high-order and low-order calculation. We need both to get it right. For
terms to force our low order finite difference model to match
instance, the f terms going into computing D
the higher order model. CMFD is effectively the physics MG.
On-the fly Re-homogenization: can generate the integrated form functions ahead of the time, and construct
the higher-order correction terms on the fly onece we resolve the coefficients on each basis.
Exact homogenization parameters depend on reference solution (which we do not have), and depend on
details of the nodal spatial model slightly. Homogenization for transport methods is not as simple as for
diffusion methods, because now we have to have DFs for the higher moments as well. Noweverday every-
one uses AXS/ADF (assembly cross section and assembly discontinuity factors) in production tools. The
advantage of the ADFs are that they are almost free as they are just edits; we do need to tally them as
functions of temperature, etc.
303
Lulu Li 22.211 Class Notes: May 9, 2012
2. Homogenization is required any time spatial reduction is employed, be it at the pin-cell level, assembly
level, assembly clusters (BWRs);
3. Exact homogenization parameters depend on reference solutions, which are seldome available and certainly
self-defeating;
4. Exact homogenization parameters depend slightly on details of the nodal spatial model;
5. Homogenization for transport methods is not as simple as for diffusion methods;
6. Methods more accurate than simple AXS/ADFs are very desirable to further improve accuracy.
304
Lulu Li 22.211 Class Notes: May 9, 2012
22 Reconstruction/De-homogenization Methods
In this lecture we consider given a homogenized solution, how to we re-construct the heterogeneous solution.
Three references: Scott Palmtags MIT PhD thesis Advanced Nodal Methods for MOX Fuel Analysis for good
detailed math, Ken Rempse NSE paper SIMULATE-3 Pin Power Reconstruction: Methology and Benchmark-
ing, Qunlei Jiangs NCS MS thesis for detailed plots of complex terms Intra-nodal Study for the Mixed LEU-
MOX Cores.
het hom
g (x, y) = g (x, y)SA
g (x, y) (968)
Keep in mind that, as the example in Fig. 120, the heterogeneous flux should be continuous acorss the material
boundary now,
SA,U O2
(x)ghom,U O2 (x)x=0 = SA,M OX
(x)hom,M OX
g g g (x)x=0 (969)
Extensions:
305
Lulu Li 22.211 Class Notes: May 9, 2012
The above equation is for flux, we reconstruct pin power too, in which case we need to know the pin-wise
fission cross section form functions as well as the group-wise flux form functions. This would be a lot of
data considering that form function, like cross sections and discontinuity factors have to be built into giant
libraries vs. burnup, temperature, density etc).
Consequently, most often powers are reconstructed be approximated by a superposition of homogeneous nodal
powers and lattice pin powers.
where
For detailed pin power reconstruction, one needs 2D (x, y) shapes of fluxes which are not directly available.
Nodal methods only provide flux shape information for the 1D shape in each direction. Cross section shapes are
also only 1D averages.
FIXME: If we want to do pin power reconstruction in our current frame-work, we need to store four form
functions, two for flux and two for cross sections. Also be careful about fission rates, power and heat conducted.
306
Lulu Li 22.211 Class Notes: May 9, 2012
2. Using 5th order polynomial, there are 25 unknown coefficients in the fast group (amn , where m, n each
goes from 0 to 4). For the thermal group, because of the polonomial and the hyperbolic functions, there
appear to be 50 unknowns bmn , cmn ; but it is really 25, because the other 25 for the specific colution can
be solved from the fast group.
3. We have 9 constrains provided by the nodal solution,
307
Lulu Li 22.211 Class Notes: May 9, 2012
308
Lulu Li 22.211 Class Notes: May 9, 2012
Since the expansion has been performed about the corner point, the error in each estimator of the corner-point
flux can be determined to be second-order; for instance, one of the corners has an error like,
1 1 2
3
= +O (973)
4cp x y 4 xy 3u
Another important point is that the heterogeneous corner point flux has to be continuous. Thus we need corner
point ratios that are analogous to discontinuity factors that can be edited from lattice calculations. The corner
point form function cpg,i,f orm,f et is coming from the lattice code.
In the general case, the corner-point fluxes are determined by averaging the four estimates of the heteroge-
neous corner-point flux, cp is corner point,
1 h hom,cp f ct,cp i
het,cp
g = g,i,j g,i,j + hom,cp f ct,cp hom,cp f ct,cp hom,cp f ct,cp
g,i+1,j g,i+1,j + g,i,j+1 g,i,j+1 + g,i+1,j+1 g,i+1,j+1 (974)
4
In order to satisfy the continuity conditions of reconstructed fluxes, the best-estimate homogeneous corner-
point fluxes for each node are computed by,
het,cp
g
hom,cp
g,i,j = (975)
fg,i,j
ct,cp
The reconstructed corner point fluxes are then continuous. CP ratio is the corner point analogue of ADF
for assembly edges.
Surface constrains preserve flux shape. Each corner has 4 surfaces connected to it.
FIXME: this reconstruction method works for PWRs because the heterogeneous terms cancel out, whereas
this is not the case for BWRs. This leads to Studsviks separable corner point estimator:
309
Lulu Li 22.211 Class Notes: May 9, 2012
8. Integrate the heterogeneous pin powers over time (usually exposure) to get heterogeneous pin burnups.
Recent years there are trends to do this heterogeneous power separately in fast and thermal groups (in
whatever number of groups that we need).
9. Determine pin-wise thermal margins (DNBR, CPR, LHGR, etc).
For every axial node, we perform this reconstruction, and use an axial shape for getting the 3D distribution. The
benefit of doing this two-step process (first solve things homogeneously, then re-construct) helps with water holes
etc., because in the first
gamma smearing affects Gd pins significantly.
310
Lulu Li 22.211 Class Notes: May 9, 2012
single assembly,
B SA
D1 2 1 (x)
1 (x) 1
= B (976)
SA
1 a1 (x)1 (x) + 21 (x)1 (x)
311
Lulu Li 22.211 Class Notes: May 9, 2012
Sg (x) B (x) SA Pg
g (x)
= Cg
(977)
B
g (x) SA
2 (x)
where (x) = 1 (x) . As in Fig. 123, we can see the cross section after approximation corrections. To
measure power of a core, we pull out individual pins and measure the gamma deposited in it to get the
power profile of it. The uncertainty comes out to be about 1.5%, which is considered to be high precision.
Advanced topic: fixing Cross-sections On The Fly/Adhop Method.
This method is somewhat an empiracle one; hence it is not very robust. Nowadays, nodal methods just go
to higher order of energy groups.
2. Burnup effect on cross section: SH-1 etc. FIXME.
Sidenote: PWR tends to burn out the error because as the exposure increases, the faster spectrum
312
Lulu Li 22.211 Class Notes: May 9, 2012
313
Lulu Li 22.211 Class Notes: May 12, 2012
314
Lulu Li 22.211 Class Notes: May 12, 2012
h , f i = h, f i (980)
Consequently,
1 1
h , M i = 0 (981)
k k
The fundamental mode solution, which is the one we are interested in, has everywhere positive real and adjoint
fluxes, and since the fission operator is an everywhere positive operator, we find that the real and adjoint eigenvalus
must be identical,
k = k (982)
315
Lulu Li 22.211 Class Notes: May 12, 2012
0
z
}| {
1 1 1 k
A0 M0 0 + A0 M0 + A M 0 = 2 M0 0 + O( 2 ) (989)
k0 k0 k0 k0
k 1 1
2 M0 0 = A M 0 + A0 M0 + O( 2 ) (990)
k0 k0 k0
Multiply by and integrating over phase space,
1
z }|
{
k 1 1
h , M0 0 i = , A M 0 + , A0 M0 +O( 2 ) (991)
k02 k0 k0
316
Lulu Li 22.211 Class Notes: May 12, 2012
There are only second order errors in reactivity, because the term multiplying is zero, hence the
first order error vanishes in the reactivity expression.
Thus Eq. 993 is clearly much more accurate than the 1st order accurate expression we had before we
multiply by 0 and integrate,
k A k10 M 0 + A0 k10 M0
2 = + O( 2 ) (994)
k0 M0 0
k1 k 1 k0 1 k k
= 0 = = = + O(k 2 ) (995)
k k k0 kk0 k02
One obtains the first order perturbation (FOP) expression for reactivity,
D E
0 , 1
k0 M A 0
(996)
h0 , M0 0 i
Each reactivity perturbation can be computed without need for solving for perturbed spatial flux distribu-
tions.
317
Lulu Li 22.211 Class Notes: May 12, 2012
d
N (t) = Q (999)
dt
That is, for a constant external source Q (unit: neutrons/s), neutron polulation grows linearly in time with
Q,
, E) then
2. Now consider introducing Q neutrons: at time zero at position, energy, and direction (~r ,
d
N (t) = Q(t t0 ) N () = N0 + Q X (1001)
dt
Or
d , E)
N (t) = Q(t t0 ) N () = N0 + Q (~r , (1002)
dt
N () N0 , E)
= (~r , (1003)
Q
The adjoint flux is defined as the asymptotic increase in total neutron population of a critical reactor for a
neutron introduced in a phase space (position r, direction , and energy E). The adjoint is only defined for
what reaction that we are interested in: in a critical reactor, that is neutron population; in a subcritical system, like
the one in detector, then it is whatever purpose, like detector response, that we are interested in.
318
Lulu Li 22.211 Class Notes: May 14, 2012
G
X G
X
Dg (~r )g (~r ) + tg (~r )g (~r ) = g f g0 (~r )g0 (~r ) + sg0 g (~r )g0 (~r ) (1004)
g 0 =1 g 0 =1
G
X G
X
Dg (~r )g (~r ) + tg (~r )g (~r ) = f g (~r ) g0 g0 (~r ) + sgg0 (~r )g0 (~r ) (1005)
g 0 =1 g 0 =1
For two group, infinite medium case, with effective downscatter only,
a1 + 12 k1 f 1 k1 f 2
1 2 12
=0 = (1006)
12 a2 2 1 a2
1 1
2 k f 2
a1 + 12 k f 1 12 1
=0 = (1007)
1
k f2 a2 2 1 a2
where
12
f 1 + f 2
k = k = a2
(1008)
a1 + 12
The above expressions show that knowing cross sections, we know k , hence we know k = k , then we
can solve for 2 .
1
We are also interested in adjoint weighted reactivity expression as in Fig. 125. Recall that PKE assumes flux
shape does not change with respect with time. We use adjoint flux weighted PKE to account for the flux change.
is very sensitive to adjoint flux.
319
Lulu Li 22.211 Class Notes: May 14, 2012
320
Lulu Li 22.211 Class Notes: May 14, 2012
321
Lulu Li 22.211 Class Notes: May 14, 2012
322
Lulu Li 22.211 Class Notes: May 14, 2012
1. Meet the cycle length requirement provided by the finance people; typically 18-20 months. Assure core
criticality for desired cycle operation.
2. Maximize economy, for instance, BU.
3. Assure plant availability: minimize vessel fluence, minimize back end costs.
323
Lulu Li 22.211 Class Notes: May 14, 2012
1. Monte Carlo: 75 hours on 750 CPU to get 1% 95/95 accuracy for the Hoogenboom/Martin problem.
2. Discrete transport method (eg: SN method):
Lots of unknowns need to be solved from the energy groups, angles, axial mesh etc. For example,
we can estimate the number of unknowns for a LWR core: (100 mesh/pin)(300 pins/assembly) (200
assemblies/core) (100 axial mesh) (100 energy groups) (1000 angles) = 6 1013 unknowns.
Fine mesh transport takes roughly the same time with Monte Carlo. For instance, the best parallel
transport Grind time SN code is Texas A&Ms PDT code which takes 300 nanosecond/unknown.
Then assuming perfect scaling, the time required to solve a full core is, (6E13 unknowns)(3E-7
s/unknown)(20 fission source iterations/case) = 100,000 hours. Not to mention we havent consider
any feedback, cross section evaluations, boron search to criticality, equilibrium Xenon, control rod
searches to criticality etc.
Hence discrete transport method has never been used for full core calculation.
3. Lattice code and nodal code (see Fig. 127: we start with basic cross sections (for instance, from ENDF),
generate multi-group library, perform an unit-cell calculation, perform a lattice calculation, and perform a
whole core calculation. Basically we go towards coarser energy groups in exchange for more spatial
dimensions/details(1D to 2D to 3D).
324
Lulu Li 22.211 Class Notes: May 14, 2012
325
Lulu Li 22.211 Class Notes: May 14, 2012
2. Intermediate group pin spatial calculation (spacial domain, see Fig. 129). Lattice code solves for radial
distributions of nuclides and fluxes in each type of pin. For a lattice depletion code, we want to know is the
pin-wise isotropic inventories vs. BU. Lattice code produces data library as a function of fuel BU, coolant
density, fuel temperature, void, etc (so lattice code does not just solve for a steady state condition, it solves
for about 2000 cases).
3. Solve full assembly neutronics. Lattice code solves for pin-wise distribution of neutron fluxes (10-100
energy groups, homogenized pins). Thats about 5000 spatial regions per assembly; MOC would take
about 20 energy groups, 1000 angles, 0.05cm ray spacing. Then we compute keff .
326
Lulu Li 22.211 Class Notes: May 14, 2012
Comments:
One example of how lattice codes generate many lattices for possible uses: in PWRs, vs. BU (with no
burnable poisons) are just straight lines depending on enrichment as in Fig. 130. Thus to achieve desirable
long cycle and high BU, we need high enrichment, which is why almost all US fuel is around the 4-5%
enrichment level (currently fuels are made at 4.4% enrichment; US regulation requires less than 5.2% for
trasportation reason).
Figure 130: Lattice Code: PWR Assembly Reactivity vs. Burnup, Enrichment
327
Lulu Li 22.211 Class Notes: May 14, 2012
2. Ex-core detector: calibrate based on flux maps, used to monitor core flux (power level) and axial power
shape.
3. HZP rod worth measurements. We solve IKE (given power extracted from the ex-core detector, we can
back out (t) etc). Though when we are moving the rods, we cannot believe the point kinetics. Thats why
to perform a HZP rod worth measurement we have to wait: after a boron dilution, we move the rod in small
number of steps, then wait. This takes more than 1 hours/rod, thats why dynamic rod worth calculation
gets popular. Control rod worth has to be within 10%.
4. Dynamic rod worth: you pull rod out a little bit, then you drive your rod all the way in called banking (takes
2 minutes), then you pull the rod out again. A subtle point here is that the power does not start to go up
until the control rod is back in its criticality position (the power actually goes up a bit faster because of the
delayed neutrons). Caution: flux should be high enough that statistics is reliable, but cannot be too high
that the reactor goes critical. See Section 18.10.3 for more details.
Side note: Westinghouse designs have 228 notches in a control rod, CE has 100 notches.
328
Lulu Li 22.211 Class Notes: May 14, 2012
329
Lulu Li 22.211 Class Notes: May 14, 2012
After startup, it takes about 30 hours for the two isotopes to reach saturation.
After shutdown, Xe peaks in 9 hours (because Iodine decays into Xe, adding about 3000 pcm reactiv-
ity) and decays away in 60 hours.
If we perform a rapid startup after a scram, Xe would be too high, thus the flux may be too high, which
immediately burn out Xe and causing an overshoot.
Long term, a PWR core burns about 1000 pcm/month. So the 3000 pcm Xenon peak takes about 3
months to burn out.
Important concepts about Pm/Sm:
After startup, it takes about 600 hours for the two isotopes to reach saturation.
After shutdown, Sm peaks in 200 hours, but never decay away unlike Xe.
Following a refueling outage, Sm reactivity peaks by 200-300 pcm after shutdown, and then return to
equilibrium about 100 hours after restart.
See Section 15.3 and 15.4 for more details.
330
Lulu Li 22.211 Class Notes: May 14, 2012
Figure 132: Arevas Chart Of Cycle length, # Assemblies, Enrichment for PWRs
331
Lulu Li 22.211 Class Notes: May 14, 2012
1. Out/In Pattern. Notice the depletion terminates when boron becomes zero. Also the power peaking factor
as a function of time increases first and then decreases.
2. In/In/Out Pattern (low leakage): the center fuel is taken from ancient batch; new fuels are surrounded by
burned fuel; power at periphery is reduced.
332
Lulu Li 22.211 Class Notes: May 14, 2012
Equilibrium cycle: we repeat the same core shuffle pattern cycle after cycle until the cycle length stops chang-
ing as in Fig. 136. Desirable features we are looking for: very flat peak power, and power peaks moves from
assembly to assembly between cycles. In practice, it is also common to design a 2-3 cycle pattern and repeat the
pattern until equilibrium.
333
Lulu Li 22.211 Class Notes: May 14, 2012
334
Lulu Li 22.211 Class Notes: May 14, 2012
LOCA limit describes local behavior which is related to local power generated.
CPR is sensitive to spacers (because of dryout condition). When power increases or flow decreases, CPR
tells us what thermal margin we have.
MCPR is chosen from generic transient analysis. Core design is performed to satisfy those limits.
Linear heat generation rate limits.
335
Lulu Li 22.211 Class Notes: May 14, 2012
MAPLHGR tech spec are designed to cover LOCA (eg., peak cladding temperature).
1. Modern trend has been pushing the fuel limits (increasing the power density from 40 kW/L to 50 kW/L to
even higher as seen in Fig. 139. The issue with higher power density is that we need to install bigger pumps
and deal with more vibration issues.
2. Also 7x7 bundle are being pushed towards 11x11 bundle, which means linear heat generation rate decrease,
though CPR may or may not increase.
336
Lulu Li 22.211 Class Notes: May 14, 2012
3. Simple Control Cell Core (CCC) has been used by many annual cycles (common in Europe). The motivation
is, when CRs are pulled out, local peaking factor would increase, thus a CCC with 20% fresh fuel and low
BU would be easy for an annual cycle. Though this is not the case in the US as there are not enough
locations for CCC because we need to put about 37% fresh fuel to sustain a longer cycle length. Again in
CCC notice the check board pattern in which fresh fuels are surrounded by depleted fuel.
There is a need to have hot excess reactivity, though we want to keep the reactivity swing flatter. Eg., PWRs
have 20% k reactivity hold-down, whereas BWRs only have 2% k (the additional comes from Gd). Gd
is integrated with fuel, very self-shielding with huge a , and behave like onion skin burning31 . So the advantage
of Gd over IFBA is the Gd can control burning by density as in Fig. 142 whereas IFBA can only burn constantly.
We pick the Gd concentration and number of pins to flatten the reactivity vs. BU plot. Notice in Fig. 142,
337
Lulu Li 22.211 Class Notes: May 14, 2012
In one bundle, we can put multiple concentration of Gd. For instance, Sweden has done pellet-by-pellet
design where every other pellet is all Gd.
BWRs generate high void at top of the core which means that the bottom fuel is more burned, there are more
fuel left on the top, and the flow is reduced at the top. To compensate for the bottom-heavy flux, we consider:
1. Add water rods to keep more coolant in the top.
2. Add part length fuel rods to allow for more water at the top.
3. Have more absorber rods in the bottom of the bundle.
338
Lulu Li 22.211 Class Notes: May 14, 2012
1. PLRs make lattice more over-moderated when cold and lower reactivity (which is a shutdown margin im-
provement).
2. Deep inserted CR controls reactivity; shallowly inserted CR controls power peaking.
3. Towards the end of the cycle, we slowly remove all CRs, suddenly all the power rolls to the top, in which
case we can lose all CPR margin. Thus end-of-cycle condition can be more limiting for BWRs as in
Fig. 144.
Figure 143: BWRs Axial Power Shapes at BOC, MOC, and EOC
The last step is illustrated in Fig. 145, as we design the startup sequence (power, flow and rods) when the core
design is finished.
339
Lulu Li 22.211 Class Notes: May 14, 2012
340
Lulu Li 22.211 Class Notes: May 14, 2012
Maximize economics.
24.6.1 Fundamentals
The fundamental trade off is between economics (maximize energy) and minimize safety risks (minimize peak-
ing):
Maximize energy: maximize EOC reactivity, maximize cycle length, maximize fuel burnup.
Minimize peaking: minimize maximum enthalpy rise hot-channel factor FH , minimize heat flux hot-
channel factor.
1. Out-in: most burned fuel bundles are loaded on the inside. Result: reduced peaking factor (e.g., 1.404), but
bad economy (for instance, when an outer ring fresh fuel bundle has a kinf of 1.25, and the whole core has
a keff of 1, that implies that about 25% of the neutrons leak out).
2. In-out: most burned on the outside. good peaking factor, bad economics.
3. Ring of fire (or low leakage pattern, in/in/out pattern): checker board pattern of fresh and one cycle fuel,
with a ring of burned one near the out-side most. We get about 3 months additional operating time compared
with the traditional out-in pattern.
A note here is that a big difference between PWRs and BWRs is that:
PWRs: density is almost constant; a fuel shuffle affects the global problem and can produce flux tilting very
easily.
BWRs: because of the existence of void, effects tend to be local.
341
Lulu Li 22.211 Class Notes: May 14, 2012
BWRs typically do half core shuffling. There are four times as many bundles as in PWRs.
Multiple rod patterns are always possible, and rods must be moved for criticality.
Flow is a variable. It is common to drive pump speed to control reactivity (review load-follow). Because a
harder spectrum is beneficial for extending cycle length, typically we want to run the core with a low flow
rate for as long as we can, then slowly increasing the flow rate.
1. Boron search: at each cycle exposure, a search is performed to find the Boron concentration that achieves
a keff = 1.
2. F-Delta h: F is the power peaking factor, and h is the entropy increase. This is a good measurement
because it takes into account that not all energy generated from fission would be deposited locally. For
instance, gamma interactions in the structural material generate heat which is then picked up by the fluid
and reflected in h. The analog in BWR is the MCPR (Minimum Critical Power Ratio).
3. Equilibrium cycle calculation: assume the shuffling pattern (called the equilibrium cycle) stays identical
from cycle to cycle. The cycle length converges fairly quickly (e.g., in this case 5 cycles for a PWR).
Though in real life, the condition never stays perfectly identical from cycle to cycle, so small variation
exists. Typically a year before start-up, someone designs the core loading pattern, and this design finalizes
about 6 months before the actual start-up of the plant.
Right click gives option, middle click gives configurations (for instance, display maps left and right maps
is a useful ting to do). Use View-Map layout option on the top menu to change displayed terms.
Shuffle: octant shuffle would touch four assemblies to keep the octant symmetry. There might be warning
messages. For instance, PWR has odd number of assemblies along each side, so you cannot shuffle anyone
on the axis.
Assembly inventory (automatically opened window). First box is fresh fuel, second box is depleted fuel,
and third box contains depleted fuel that is
Click on a patter in the LP window, right click on the map to restore that pattern.
For this class, we optimize for maximum cycle length, F h < 1.60, MTC < 3 pcm/C. cycle length: 18.5
342
Lulu Li 22.211 Class Notes: May 14, 2012
2. Rods have to have limited burnup (say 30 MWd/kg) under control rod positions). The main reason is that
guide tube distorts when burnup is high, and the control rods may not go in. The secondary reason is that
we want the reasonable rod worth underneath the control rods.
3. We prefer quadrant/octant symmetry in the fresh fuel. Likely wise we can maintain batch symmetry in the
depleted fuel.
343
Lulu Li 22.211 Class Notes: May 14, 2012
This is an easy way to include constrains as penalties. A better approach would the Pareto optimal which is
not covered in this lecture (Ask stephano for questions).
Stochastic optimization schemes are only real options,
1. Simulated Annealing: The physical process of annealing means bringing metals to a high temperature and
let it cool slowly; if it cools slowly enough eventually it would get to its minimum energy.
2. Genetic Algorithm
344
Lulu Li 22.211 Class Notes: May 14, 2012
SA is simpler to understand. SA is harder to parallelize (because at each temperature you are making up a
L that depends on the previous steps).
GA is natural for true multiobjective optimization. GA is trivial to parallelize.
Both are heuristic and need some tuning. Both can be improved with hill-climbing heuristics (eg, greedy
exhaustive single binary swaps).
A subtle point: once you put the most burned cores on the outside-most ring, how the inside pattern does
not affect the cycle length anymore because the power on the outside is so small. But the peaking power is very
sensitive on the inner pattern.
345
Lulu Li 22.211 Class Notes: May 22, 2012
d (thermal/fast) (thermal/fast)
U-235 0.01668/0.01673 0.0067/0.0064
U-238 -/0.046 -/0.0164
Pu-239 0.0157/0.0152 0.0022/0.0020
346
Lulu Li 22.211 Class Notes: May 22, 2012
keff 1.0 f i
(t) = (t) = (1013)
keff f
2. PKE assumptions:
Flux is separable (~r , E, t) = S(~r , E)T (t) (S is the shape function, T is the amplitude function).
Classical PKE further assumes S(~r , E) to remain constant with respect to t.
Assume no leakage, no external source of neutrons, no feedback.
Average properties, by integrate over all space and energy and normalize.
Classical PKE further assumes constant i , i .
X
Minor: jp p , j j = independent of fissioning species j; id d independent of
j
delayed neutron energy group i; (~r ) aka no spatial dependency.
3. 2-group general PKE (dropped ~r for every term, and t for , T, C:
Z Z
d X
[T1 ] = dr1 {r,1 S1 T1 + (1 )(f,1 S1 T1 + f,2 S2 T2 )} + dr i Ci (1014)
dt i
Z
d
[T2 ] = dr2 [a,2 S2 T2 + s12 S1 T1 ] (1015)
dt
Z Z Z
d
drCi = dri [f,1 S1 T1 + f,2 S2 T2 ] i drCi (1016)
dt
Implications:
(t) does not describe how much neutrons there are at a specific time; it describes how much
neutrons there would be eventually.
has no T (t) temporal dependency.
only depends on 1 but not 2 , because all delayed neutrons are born in group 1, hence we do not
care about the neutron importance in group 2.
Similarly is the prompt neutron life time, thus only the importance of group 1 neutrons matters.
347
Lulu Li 22.211 Class Notes: May 22, 2012
4. Classical PKE:
X
i
dT i
X d i
= T+ i Ci + Q Ci = T i Ci (1017)
dt i
dt
PKE parameters:
w, v 1 0
hw, (F M )0 i hw, Fd 0 i
(t) = = eff = (1018)
hw, F 0 i hw, F 0 i hw, F 0 i
(a) Reactivity , where the top is the diffusion equation (or say the net production), and the bottom is
f a (1 + M 2 Bg2 ) keff 1.0
fission rate if all neutrons show up instantaneously. (t) = = ,
f keff
f
a
where keff = 1+M 2 Bg2 .
( !Z
0
X j j
X i j j
r , E 0E, t)S dE 0 +
R R R
dE dr DS t S + s ( ~ p (E)(1 ) + d (E)i f S dE
j i
(t) = !Z (1019)
0
R R X j j
X i j j
dE dr p (E)(1 ) + d (E)i f S dE
j i
(b) The delayed neutron fraction (notice there is prompt neutron dependent jp (E) term in it). If ignore
X
fissioning spectrum, i .
i
X j Z
j
jf (~r , E 0 , t)S(~r , E 0 ) dE 0
R R
dE dr d (E)
j 0
i (t) = !Z (1020)
X X
jp (E)(1 j
id (E)ij jf S 0
R R
dE dr )+ dE
j i
(c) Prompt neutron lifetime, which is 1/v of shape divided by almost-instantenous-fission-rate. Know
105 for LWRs; smaller for SFR, and larger for CANDU.
dE dr v1 S(~r , E)
R R
1 criticality 1
(t) = !Z
R X X v f v a
jp (E)(1 j ) + id (E)ij jf S dE 0
R
dE dr
j i
R R R R
(d) The i-th precursor, Ci (t) = dE drCi (~r , t). The total external source, Q(t) = dE drQ(~r , t).
5. Steady state solution of classical PKE:
d i d
Ci (0) = 0 Ci (0) = T0 T (0) = 0 (0) = 0 (1021)
dt i dt
6. The matrix form of the 1st order ODE is (assuming three precursor groups, can be easily extend for more
precursor groups),
(t)
T 1 2 3
C1 d 1 1
N = N (t) = [N (t)] (1022)
C2 dt 2 2
C3 3
3
348
Lulu Li 22.211 Class Notes: May 22, 2012
d
T (t) 0 = T (t) + C(t) = T (0+ ) + T0 (1023)
dt
T (0+ )
= (1024)
T0
T1 0
More general, if initial reactivity is not exactly zero (but small), = .
T0 1
9. In-hour equation: assume solutions take the form of, T (t) Ci (t) et , plug into PKE. Basically if we
know i , , i we can find . Know is typically small except super prompt critical when is huge.
X i
= + (1025)
i
+ i
Issues:
A small reactivity range such that ignoring feedback is valid.
In-hour assumes constant i which is not very accurate for, e.g. we drive the control rod in the center
of a core.
Asymptotic periods T0 = 1 : we can only use Inhour equation for asymptotic period, that is, a stable
condition indicated by a constant period. Thus to use Inhour equation, we have to make the reactor
stable first. For any time dependent problem, we have to use the more general IK.
10. Inverse Kinetics Equation: 1st order ODE again. First given T (t) we solve for the precursor concentration,
using [A] = diag[i ], [Y (t)] = diag[
i]
T (t).
d X
(t) = T (t) + i Ci (t) Q(t) (1027)
T (t) dt T (t) i T (t)
Tn Tn1 X
n = + i Ci,n Qn (1028)
Tn tn tn1 Tn i Tn
349
Lulu Li 22.211 Class Notes: May 22, 2012
1 Cp 2
P (t) = P0 + Cp (rod )Tf uel (t) T (t) (1029)
2 f uel
dP
At peak power, we set dTf uel = 0,
rod Cp (rod )2
Tfpeak
uel = P peak = P0 + (1030)
2
dP
dP
(c) Second method to derive P (t): d = dt
d , integrate over ,
dt
Cp
((t) )2 + (rod )2
P (t) = P0 + (1031)
2
Assume the transient terminates when P (t) returns to P0 , then end = 2 rod ,
2(rod )
Tfend
uel = + Tf0uel (1032)
350
Lulu Li 22.211 Class Notes: May 22, 2012
d
Tfuel (t) = aP (t) b[Tfuel (t) Tcoolant (t)] (1033)
dt
d
Tcoolant (t) = c[Tfuel (t) Tcoolant (t)] d[Tcoolant (t) Tinlet (t)] (1034)
dt
(a) Time constants: recall the asymptotic power decay time constant is just that of the longest precursor.
(b) < insertion/withdrawal: power overshoot/undershoot is proportional to the reactivity change rate
(rate , overshoot/undershoot ), and the asymptotic value is independent of the reactivity change rate.
(c) > insertion/withdrawal: power and temperature would turn around due to Doppler feedback.
(d) Asymptotic temperature only depends on the reactivity change, not the change rate.
5. Problem with our neutronics model: not accurate for large spatial flux changes. Problem with our TH
model: very huge diffusive property, would not predict any thermal shock behavior.
351
Lulu Li 22.211 Class Notes: May 22, 2012
3. Real and adjoint diffusion equations: group diffusion and absorption operators are self-adjoint; group scat-
tering and fission operators are not.
G
X G
X
Dg g + tg g = g f g0 g0 + sg0 g g0 (1039)
g 0 =1 g 0 =1
G
X G
X
Dg g + tg g = f g
g0 g0 + sgg0 g0 (1040)
g 0 =1 g 0 =1
For two group, infinite medium case with effective downscatter only,
a1 + 12 k1 f 1 k1 f 2
1 2 12
=0 = (1041)
12 a2 2 1 a2
1 1
2 k f 2
a1 + 12 k f 1 12 1
=0 = (1042)
1
k f 2 a2 2 1 a2
where
12
f 1 + f 2
k = k = a2
(1043)
a1 + 12
In the above example, we effectively have M = 0, M = 0, and by definition, the adjoint of Hermitean
conjugate of a mtatrix is just complex-conjugating each of its element and then transposing,
T
a11 a12 a11 a21
a11 a12
M = = = (1044)
a21 a22 a21 a22 a12 a22
352
Lulu Li 22.211 Class Notes: May 22, 2012
4. Boundary conditions: , are any two functions satisfying the appropriate boundary and continuity con-
< 0, then the adjoint satisfying
ditions for the angular flux and adjoint flux. If (~r , , E) = 0 for all n
, E) = 0 for n
(~r , > 0.
5. Spectra:
The PWR adjoint spectrum: high in fast and thermal range b/c high probability of fissioning.
For a typical two group problem, the adjoint flux should have the same shape for the two groups,
except the magnitude is offset by 2 .
1
33
6. Adjoint fluxes in kinetics parameters
(a) Delayed spectra: 10 times that of the prompt for up to 0.1 MeV; then approahces prompt until smaller
than prompt spectrum for 1 MeV and above.
(b) I is the ratio of the adjoint weighted delayed chi to prompt chi, and it has a value of about 0.97.
33 In operating reactors, they measure the control rod worth; they are no better than the delayed spectra
353
Lulu Li 22.211 Class Notes: May 22, 2012
keff 1 f 1
(a) = , keff = , = .
keff a + DB 2 vf
(b) is prompt neutron lifetime, is decay constant.
(c) We use the inhour equation, and can ignore the term can go away because is so small,
X i X i
= + (1046)
i
+ i i
+ i
354
Lulu Li 22.211 Class Notes: May 22, 2012
= 0.1 = (1047)
+
where = ln 2
. Thus = 7.7 103 , T = 1
= 130s.
(d) We use prompt jump approximation,
T (0+ )
= = = 1.111 (1048)
T0 0.1
2. (12 Final #28) When a control rod worth 37 cents is removed from a critical reactor (with beta-effective of
0.007), what is the ratio of power after and before the reactivity insertion (e.g., the prompt jump)?
Answer: We use prompt jump approximation, = $0.37 = 0.37,
T (0+ )
= = = 1.59 (1049)
T0 0.37
3. (12 Final #29) If a fully-enriched, uranium-fueld, thermal-spectrum reactor has a stable period of 60 seconds
after a control rod is moved following long-term stable operation at power of 100 watts, how much reactivity
(in dollars) was inserted? (note: use the eight group delayed neutron data from Lecture 17, slide 21).
Answer: Given i , i etc, we can calculate: = 6.65 103 , = T1 = 1/60s1 ; = v 1
f
5
1.87 10 s. Plug everything in the In-hour equation,
8
X i
= + = 9.60 104 = $0.1444 (1050)
i=1
+ i
You can also omit the part as it is very small and get = 9.59 104 = $0.1443.
4. (12 Final #30) If a room temperature critical reactor undergoes a $5.0 reactivity excursion from near zero
power, what is the approximate final core-averaged fuel temperature (K)? Note, assume prompt neutron
lifetime of 105 sec, beta-effective of 0.007, specific heat of the fuel of 1 J/gm-K, and Doppler coefficient
of 3.0 pcm/C
Answer: Be careful about the units! First we convert all the units into k : = 4 = 0.028k, =
3 105 k. Thus,
2( ) 2 0.0028k
Tfend
uel = + Tf0uel = + 293 = 2159.67K (1051)
3 105 k
5. (12 Final #31) What is the fast-to-thermal adjoint flux ratio in an un-reflected critical slab reactor of Fuel 1
material?
Answer: Keep in mind that un-reflected suggests there are leakage terms. We have solved earlier in #23
that B 2 = 0.008876, which would remains the same for the forward problem and the adjoint problem. We
write the two-group diffusion equations for the adjoint problem,
1 s12
= = 0.6525 (1054)
2 D1 B 2 + a1 + s12 f 1
355
Lulu Li 22.211 Class Notes: May 22, 2012
1 D2 B 2 + a2
= = 0.6525 (1055)
2 f 2
Notice the fast-to-thermal adjoint flux ratio can be solved from either equation and they come out to be the
same.
356
Lulu Li 22.211 Class Notes: May 18, 2013
Z 1 Z 1
J zr (x ) = Jz (x , y , 1) dy , J zl (x ) = Jz (x , y , 0) dy (1060)
0 0
34 weighted residual is not on the final
357
Lulu Li 22.211 Class Notes: May 18, 2013
G G
d2 1 X X
xDg gx (x ) + rg gx ( x ) = g f g 0 0 (x ) +
g x sg0 g g0 x (x ) Lgy (x ) Lgz (x )
dx2 keff 0 0
g =1 g =1
G G
d2 1 X X
yDg gy (y ) + rg gy ( y ) = g f g 0 0 (y ) +
g y sg0 g g0 y (y ) Lgz (y ) Lgx (y )
dy2 keff 0 0
g =1 g =1
G G
d2 1 X X
zDg ( z ) + rg (z ) = g fg 0 g0 z ( z ) + sg0 g g0 z (z ) Lgx (z ) Lgy (z )
dz2 gz gz
keff 0 0
g =1 g =1
3. Flux distribution is not sensitive to transverse leakage shape, so we approximate transverse leakage with
quadratic shape (2nd order polynomial):
L() = L + l1 P1 () + l2 P2 () (1061)
1
7q2 + 3q4 + 7a2 r
Z
d
P2 () D 2 () + r () Q() d = 0, a4 = (1066)
0 d 420(D + 3r )
d2 1 (x) 1
D1 + r1 1 (x) (f 1 1 (x) + f 2 2 (x)) = L1 (x)
dx2 keff
d2 2 (x)
D2 + r2 2 (x) 12 1 (x) = L2 (x) (1067)
dx2
358
Lulu Li 22.211 Class Notes: May 18, 2013
We need both modes because B12 and B22 are different. In this case fast group uses fundamental mode,
thermal group uses harmonic mode, thus the solution is,
H
1 (x) a11 sin(B1 x) + a12 cos(B1 x) r1 r2 a21 sin(B1 x) + a22 cos(B1 x)
= =
2H (x) a21 sinh(B2 x) + a22 cosh(B2 x) 1 1 a23 sinh(B2 x) + a24 cosh(B2 x)
(1071)
where the fast-to-thermal flux ratio is defined as,
2
a11 a12 D2 Bm + r2
rm = = = (1072)
a21 a22 12
The particular solution is determined solely by transverse leakage.
(a) The advantage of this analytical method is that we get an expression relating the net current and the
flux, so the partial current never shows up.
(b) No need for weighted residual equations because we have exact solution to the 1D diffusion equation.
(c) The drawback is, all groups are solved simultaneously, so the generalization is very hard. But with
the help of modal expansion, now we can do ANM for any number of energy groups.
3. SAMN: uses NEM for fast group, transverse integrated 1D diffusion equation for the thermal group. Ap-
proximate source with 4th order Legendre polynomial. Analytical solution of the thermal group is,
4 s
X 2x g
g (x) = A sinh(g x) + B cosh(g x) + ci Pi , g = (1073)
i=0
h Dg
u
ugi,j (ul ) u+
ugi,j (ul+1 )
fgi,j = , fgi,j = (1074)
u (ul )
gi,j u (ul+1 )
gi,j
ADFs are the homogeneous analogy to the heterogeneous assembly calculation (which is calculating a
single-node problem with zero net current BC). ADFs are simply ratios of the surface-averaged fluxes to
the cell-averaged fluxes in the heterogeneous assembly calculation.
359
Lulu Li 22.211 Class Notes: May 18, 2013
2. Reflector modeling:
Using empiracal albedos to replace the baffle and reflector is bad because using flux-volume weighted
cross section distribute the strong absorption xs over the entire region.
We use DFs because they are less spatially sensitive than albedos. It is advantagous in that DFs are
chosen so that the nodal model wouold reproduce the net currents at the core/baffle interface and the
net reaction rates in both the assembly and the homogenized baffle without explicitly representing the
baffle.
Fast neutron leakage makes up for 90% of all leakage, so leakage is insensitive to enrichment, burnup
etc.
het hom
g (x, y) = g (x, y)SA
g (x, y) (1075)
Heterogeneous corner point flux has to be continuous. Thus we need corner point ratios that are
analogous to DFs that can be obtained from the lattice calculations.
In general, the corner-point fluxes are determined by averaging the four estimates of the heterogeneous
corner-point flux,
1 h hom,cp f ct,cp i
het,cp
g = g,i,j g,i,j + hom,cp f ct,cp
g,i+1,j g,i+1,j + hom,cp f ct,cp
g,i,j+1 g,i,j+1 + hom,cp
f ct,cp
g,i+1,j+1 g,i+1,j+1
4
(1077)
In order to satisfy the continuity conditions of reconstructed fluxes, the best-estimate homogeneous
corner-point fluxes for each node are computed by,
het,cp
g
hom,cp
g,i,j = (1078)
fg,i,j
ct,cp
The reconstructed corner point fluxes are then continuous. CP ratio is the corner point analogue of
ADF for assembly edges.
Side note: there are two entirely different problems. One is an eigenvalue problem, using homogenized cross
section, and output k, and the flux computed would be continuous. The other is a fixed-source problem, using
keff as input (hence J at interface), and generate discontinuous flux.
360
Lulu Li 22.211 Class Notes: May 18, 2013
361
Lulu Li 22.211 Class Notes: May 23, 2012
= nv J R = RI
n
cm2 1/cm cm2 s e cmn2 s reactions
cm3 s barns
5. Fast flux in hydrogen is around 1014 n/cm2 s, and on the order of 1012 n/cm2 s for thermal flux.
6. Average energies prompt neutrons are released: 2 MeV (peak prompt neutron energies: 0.7 MeV). Average
energies delayed neutrons are released: 0.4 MeV.
7. Core decay heat after 1 day is about 1% rated.
362
Lulu Li 22.211 Class Notes: May 23, 2012
(a) H has no resonance; it has the highest scattering xs in LWR, so we can ignore any other isotopics
neutron scattering.
(b) Na has a huge resonance in 23 keV, and more resonances at higher energies because it is a heavy
isotope.
(c) Near zero energy,
r
kT
(E 0) (1079)
AE
6. Given an unknown material type, all we care is to count the nucleus density of each material and look at its
xs.
363
Lulu Li 22.211 Class Notes: May 23, 2012
Ft (u) 10
PF F = F
= = 0.909 (1080)
t (u) + e 10 + 1
Thus the probability of escaping the cylinder is just 9.09%.
2. (12 Final #1-20)
(a) What are the average energies at which prompt and delayed fission neutrons are emitted?
Answer: 2 MeV for prompt neutrons, 0.4 MeV for delayed neutrons.
(b) What are the units of real and adjoint flux?
Answer: Real flux has a unit of neutrons per cm2 per second. Adjoint flux is typically unitless.
(c) What is the difference between true and effective downscatter in two-group cross sections?
Answer: True down-scattering cross section s,12 measures the probability of scattering from group
s,12 = s,12 s,21 2 is like combining
1 to group 2. Effective down-scattering cross section
1
flux-weighted up-scattering and down-scattering into one down-scattering term.
(d) What phenomenon is responsible for the 1/v tail of thermal scattering cross sections for isotopes with
constant elastic cross sections at higher energies?
Answer: Thermal motion/thermal vibration of the interaction material.
(e) What is the mean number of isotropic elastic scattering collisions with Livermorium, Lv293
116 , required
to slow neutrons from 1 MeV
to 1.0
2 eV?
A1 ln ln(E2 /E1 )
Answer: A = 293, = = 0.986, = 1 + = 6.8 103 , n = =
A+1 1
2028.6. Hence it requires 2029 collisions.
366
Lulu Li 22.211 Class Notes: May 23, 2012
(f) What is the average spacing (in eV) of U238 resolved capture resonances?
Answer: 25 eV.
(g) For Einsteinium Es253 in the energe range from 11.102 eV to 99.653 eV, what is the ratio of group
absorption cross section to resonance integral?
E2
Answer: Recall that RIeff and g are related through: RIeff = g ln , hence the ratio is,
E1
g 1
= = 0.456.
RIeff ln E2E1
(h) What is the definition and units of dilution cross section for resonance absorbers?
Answer: Dilusion cross section is defined as,
Nm rm
d = (1081)
Nr
where Nm , rm are the number density and cross section of the moderating material, and Nr is the
number density of the resonant material. d has the unit of cm2 .
(i) What is the approximate flux disadvantage factor at 0.5 eV for a 17x17 PWR lattice?
Answer: Notice flux disadvantage factor seem to be defined two ways:
The moderator flux over the fuel flux (as in Reuss book, e.g. Fig. 9.4) in which case the flux
disadvantage factor would be about 1.05 for standard fuel, 1.1 for MOX fuel.
The fuel flux over the moderator flux as in Kords lectures, in which case the flux disadvantage
factor would be around 0.9
Hence results within 0.9 to 1.1 receive credits for this problem.
(j) What is the approximate Doppler temperature coefficient in pcm/K in an LWR?
Answer: -3 pcm/K (it is important that the Doppler coefficient is negative), see slide 17, Lecture 10.
(k) What are the two primary sources for the production of Xe135 ?
Answer: Iodine decay (major source), fission/burnup (minor source).
(l) What is resonance escape probability for a material with a resonance integral of 100 barns, a dilution
cross section of 2000 barns, and a mean logarithmic energy decrement of 0.333?
Answer: Recall the expression for resonance escape probability:
RIeff 100b
p exp = exp = 0.861 (1082)
d 0.333 2000b
(m) What is Dancoff factor for unclad PWR fuel rods with radius 0.5cm in a voided uniform infinite lattice
with a pitch of 1.99 cm?
Answer: When the moderator is void, that is the opacity is zero, hence Dancoff factor is 1.0.
(n) What is the difference between delayed neutron fraction and delayed neutron yield?
Answer: Absolute delayed neutron yield is the number of delayed neutrons per fission. Relative
delayed neutron yield is the number of delayed neutrons per fission for an isotope divided by number
of delayed neutrons per fission for all isotopes. Delayed neutron fraction is the absolute yield divided
by .
(o) What are the approximate delayed neutron fractions for U235 , U238 , and Pu239 ?
Answer: U235 : 0.00665. U238 : 0.01650. Pu239 : 0.00225.
(p) What is an approximate expression for the extrapolation distance in diffusion theory for one-group
bare homogeneous reactor?
2 2
Answer: = tr = 2D.
3tr 3
367
Lulu Li 22.211 Class Notes: May 23, 2012
(q) What is an expression for the mean cosine of the scattering angle in the lab system for an isotope with
atomic mass A, if scattering is isotropic in the CM system?
2
Answer: In 3D it would be cos = 0, cos = 3A .
(r) For dense matrices of size N , what is the order of numerical operations required to find the full matrix
inverse, as N becomes large?
Answer: N 3 .
(s) If an infinite-repeating lattice calculation that is used to produce flux form functions and discontinuity
factors has assembly-averaged flux of 0.9 1013 and an surface-averaged flux of 1.1 1013 , what is
the assembly discontinuity factor (ADF)?
surface-averaged flux 1.1 1013
Answer: Recall that ADF = = = 1.22.
asembly-averaged flux 0.9 1013
(t) If two neighboring nodes have ADFs of 1.30 and 1.25 in the thermal group, what is the percentage
discontinuity in homogeneous thermal flux at the shared interface when the nodal diffusion equations
are solved using the NEM method?
HOM
n+1 fn
Answer: Recall fn HOM n = HETn = HET HOM
n+1 = fn+1 n+1 , then = . Thus the
HOM
n fn+1
1.30 1.25
percentage difference in the homogeneous thermal flux is, = 3.9%.
(1.30 + 1.25)/2
3. (12 Exam 2 #3) What are the approximate values of Xe-135 and Sm-149 capture cross sections at 0.1 eV
and temperature of 300K?
Answer: Xe-135: 2 106 b. Sm-149: 1.15 105 b.
4. (12 Exam 2 #4) Explain the difference between independent and cumulative fission yields for a given
nuclide.
Answer: The independent yield is the yield of a specific isotope directly from fission and nothing else. The
cumulative yield includes the independent yield plus all of the precursor isotopes that decay into a specific
isotope.
5. (12 Exam 2 #5) Definition of dominance ratio for a critical reactor?
368
Lulu Li 22.211 Class Notes: February 6, 2013
Typically what we do is to write it in a matrix form and solve for a coupled system. But even better, we can
define the effective removal rate s12 , and re-write the two-group balance equation:
f 1 f 2
1 a1 1 s12 1 + 2 = 0 (1085)
k k
s12 1 a2 2 = 0 (1086)
1
Then we can solve for 2
from the second equation in terms of cross section, plug in the first equation, and
get k from there. Notice that we only know the relative magnitude of and 2 .
4. Know Two-group diffusion model: group 1 is the fast group larger than 0.625 eV, and group 2 is the thermal
group.
(a) D1 = 1.5, D2 = 0.5.
(b) Total fission source: 1 = 1.0, 2 = 0.0 implies that the fission source in thermal group is zero,
X
Sf (~r ) = g f g (~r )g (~r ) (1087)
g
(c) Scattering source: define effective down-scatter, so up-scattering is zero s21 (~r ) = 0.
(d) Final Two-Group Diffusion Equations:
D2 2 + a2 2 = s12 1 + S2 (1089)
5. Know the one-group fundamental mode eigenvalues and eigenvectors as in Table 14.
2
Slab L2 , L2 (x) = A cos x
L B2 = L
sin( r ) 2
Sphere [0, R] (r) = A r R B2 = R
2.405 2
(r) = AJ0 2.405r
Infinite cylinder [0, R] R B2 = R
2.405 2 2
Finite cylinder r [0, R], z H H
(r, z) = AJ0 2.405r cos z
2, 2 R H B2 = R + H
2 2 2
y
Parallelepiped L2i , L2i x x
+ Ly
(x) = A cos L x
cos L y
cos L z
B2 = Lx + Lz
369
Lulu Li 22.211 Class Notes: February 6, 2013
370
Lulu Li 22.211 Class Notes: February 6, 2013
371
Lulu Li 22.211 Class Notes: February 6, 2013
(e) If = 200 milli-beta of reactivity is introduced into a critical reactor, what is approximate magni-
tude of the prompt drop in core power?
Answer:
(f) The Fuchs-Nordheim prompt reactivity excursion model predicts that the fuel temperature at the peak
of the power excursion can be expressed in terms of the Doppler feedback coefficient T = +
T 0 . What is the Fuchs-Nordheim expression for fuel temperature at the end of the prompt reactivity
excursion?
Answer:
(g) If an LWR is on a positive 10s asymptotic period, what is the approximate core reactivity in dollars
(using a one delayed neutron group model with a precursor half-life of 10s)?
Answer:
6. Numerical Methods in Reactor Analysis (5%): an engineer sets up a 2-group, finite-difference, diffusion
model of a 1D reactor that has the following cross sections?
Answer:
372