Multiscale modeling of fission gas bubble evolution

in UO2 under nominal operating conditions

Institution Principal Investigator Additional Personnel
UTK* Brian Wirth* Sophie Blondel, Bamin Khomami, Aaron Kohnert
ANL Barry Smith Shashi Aithal (FASTMath)
(FASTMath)
INL Giovanni Pastore
LANL Blas Uberuaga** Li-Ta “Ollie” Lo (SDAV), David Andersson, Danny Perez,
Wathsala Widanagamaachchi (U Utah)
ORNL David Bernholdt** Valmor de Almeida Jeremy Meredith (SDAV), Phil Roth
(SUPER)
PNNL Rick Kurtz Ken Roche, Wahyu Setyawan
In partnership with:

Project web site:

https://collab.cels.anl.gov/display/FGS
Presented at OECD/NEA Workshop on Nuclear
* bdwirth@utk.edu Fuel Modeling to support safety &performance
enhancement for water cooled reactors

8 March 2017, Paris, France #

This work was supported by the U.S. Department of Energy, Office of Advanced Scientific Computing Research, and
the Office of Nuclear Energy, in addition to the Nuclear Energy Advanced Modeling & Simulation (NEAMS)
program.
 Introduction: the long-standing fission gas problem
Fission reactions
Intra-granular
Generation of fission gases (Xe, Kr) bubble swelling
Trapping
Intra-granular gas Intra-granular gas
(single atoms) (bubbles)
Resolution
Diffusion
Diffusion

Grain-boundary gas

Grain-boundary
bubble swelling

Saturation / Micro-cracking

Fission
gas release
* G. Pastore (INL) – micographs from White, Corcoran and Barnes, Report R&T/NG/EXT/REP/02060/02 (2006).
 Introduction: the long-standing fission gas problem
Intra-granular
§  Fission gas located:
–  Mobile single gas atoms
–  Intra-granular bubbles
–  Inter-granular bubbles
Inter-granular

§  Gas release driven by inter- Release to

granular bubble interconnection plenum
∂c
= D∇2c − gc + b′m+ β!
∂t
Diffusion Creation

Absorption Re-solution

(
§  Effective diffusion rate: D' = Db' b' + g )
* G. Pastore (INL) and D. Andersson (LANL)
 Fission gas bubble observations

*Cornell, Speight and Masters, JNM 30 (1969) 170-178. *Baker, JNM 66 (1977) 283-291.
 tate. In fact, in larger bubbles above several nm in the
3.1. Bubble morphology high burnup fuels of 44 and 83 GWd/t, solid fission
products precipitate, which are presumably composed
Figs. la-lc
base-irradiated
Fission gas bubble observations
show bright-field TEM images for the
fuels with various burnups of 23, 44
of molybdenum, technetium, ruthenium, rhodium and
palladium.

23GWd/t 44GWdA 83GWdA

Fig. 1. Bright-field electron micrographs of base-irradiated fuels: (a) 23 GWd/t; (b) 44 GWd/t; (c) 83 GWd/t.

*Kashibe, Une and Nogita, JNM 206 (1993) 22-34.

 Fission gas challenges

*White, Corcoran and Barnes, Report R&T/NG/EXT/REP/02060/02 (2006).

 Fission gas challenges
• Current, engineering scale models do not do a good job of predicting fission gas
release and gas bubble contributions to swelling but are of paramount importance –
well recognized problem in fuel performance modeling:

- Lösönen* states in recent review: “In transients in particular, when

pronounced bubble coarsening takes place, the release rate depends
strongly on the development of the characteristics of the bubble
population”, and “a mechanistic approach in modeling the bubble
coarsening process should be applied.”

- Pastore** performed an uncertainty quantification assessment in BISON

and concluded “a better characterization of the parameters through
through experimental and theoretical research may reduce the uncertainty
in fission gas behavior calculations and in the multiple related aspects of
fuel performance analysis” and further noted that there is at least a factor
of 2X uncertainty in current engineering predictions of fission gas release

*Lösönen, JNM 280 (2000) 56-72. **Pastore, Swiler, Hales, Novascone, Perez, et al., JNM 456 (2015) 398-408.
Our vision of multiscale modeling
Initial focus on unresolved
questions:
-  Large-scale atomistic
calculations to improve
understanding of
fission gas behavior
(migration, bubble
evolution including re-
solution)

-  Developing &
applying Xolotl cluster
dynamics model to
spatially resolve fission
gas bubble populations
 Xe diffusion mechanisms

Current empirical model:

Total: Dxe = D1 + D2 + D3

Intrinsic: D1 = 7.6⋅10−10 × exp( −3.04 /kBT ) ⎡⎣m2 /s ⎤⎦

(
Irr. Enhanced: D2 = 4 ×1.4⋅10 × F! exp −1.2/kBT
−25
) ⎡m2 /s ⎤
⎣ ⎦

Athermal: D3 = 4 × 2⋅10−40 × F! ⎡⎣m2 /s ⎤⎦

§  Empirical relationships.

§  The mechanisms for D1, D2, and D3 are not fully
understood, which complicates development of
predictive models.
§  D1 and D2 driven by vacancy population, similar to U
behavior. J. A. Turnbull et al., JNM 107, 168 (1982)

§  D3 is believed to be caused directly by damage.

Goal: Calculate D1 and D2 fission gas diffusion through simulation

using point defect dynamics and D3 by direct MD simulations.
 Governing reaction-diffusion equations

•  D1 and D2 calculated by combining empirical potentials and

DFT calculations to parameterize reaction-diffusion equations
formulated to be consistent with phase-field models, e.g.:

Diffusion (XeU2O):

Cluster formation (XeU2O):

Reactions with interstitials (XeU2O):

Production and reaction with

sinks/sources (VU):

Chemical potential
differences, free energies Slide 10

and mobilities from DFT/MD:

 Point defect evolution in irradiated UO2

§  Initial point defect dynamics model

–  Uranium vacancies (mono- and di-)
–  Uranium interstitials
–  Equilibrium oxygen (stoichiometric)
–  Xenon residing in uranium single
vacancy(+oxygen) and diffusing as a di- 20

vacancy 15 Miekeley and Felix

–  Damage source term (uranium 10

interstitials, vacancies) Turnbull

ln(D [nm /s])

2
5
–  Sinks (static bubble population)
0
§  Xe/Vacancy cluster dominates low
temperature diffusion -5 Irradiation
No irradiaition
Intrinsic analytical model
-10 XeU2O cluster

300 400 500 600 700 800x10

-6

1/T [1/K]

Xe+2V diffusion behavior does M. R. Tonks, et al., Comput. Mater. Sci. 51 20 (2012)
D. A. Andersson, et al., JNM 451, 225 (2014)
not capture experiments D. A. Andersson et al., Phys. Rev. 84, 054105 (2011)
D. A. Andersson et al., JNM 462, 15 (2015) Slide 11
 Migration properties as function of cluster size
Preliminary estimates from statics calculations of the
•  Preliminary results migration barrier for different vacancy clusters in UO2.
m=4
indicate that the 7

mobility increases

Migration barrier (eV)

6
m=2
for increasing cluster 5 m=3
m=8

size. Xe
4 VU
•  The stable XeU6O8 VO
3
cluster moves by
binding two 2
U2O2 U3O4 U4O7 U6O8-U2O2
Vacancy cluster
additional vacancies.

Xe+Uv8Ov10 Vacancy drives

intermediate diffusion (D2)
 D3: Fission gas migration under irradiation
Left axis diffusivity per fission rate Temperature dependence of
Right axis typical diffusivity in reactor the diffusivity

Diffusion for cation, O, Xe and Kr in UO2, ThO2 and PuO2. Near-

athermal mechanism and within scatter of experiment. Ratio of 10:1
for ratio of electronic to ballistic contribution. Little difference
between actinide oxides.
M.W.D. Cooper et al. J. Nucl. Mater. 481 125-133 (2016).
Fission gas bubble re-solution mechanisms
 MD simulations* of R=0.8 nm (hetero re-solution)

Average number and standard deviation of re-solved Xe, over-pressurized bubbles

10 keV/nm 20 keV/nm
40 40
Number of resolved Xe

Number of resolved Xe
70211 31
30 30 29
2.2 ' 2.7749
26
25
23
20 20

10 10
7 23 25 31 26 29
26.8 ' 3.1937
2
1
0 0 0
0 10 20 30 40 0 10 20 30 40
Time (ps) Time (ps)

* W. D. Cooper, et al., J. Phys.: Condens. Matt. 28, 405401 (2016).

 MD simulations* of R=0.8 nm (hetero re-solution)

Average number and standard deviation of re-solved Xe, equilibrium bubbles

35 35
10 keV/nm 20 keV/nm
30 30
Number of resolved Xe

Number of resolved Xe
25 25 26
14342 23
20 2.8 ' 1.3038 20 19
18
15 15 16

10 10

5 5 26 18 23 16 19
4
3
2 20.4 ' 4.0373
0 1 0
0 10 20 30 40 0 10 20 30 40
Time (ps) Time (ps)

* W. D. Cooper, et al., J. Phys.: Condens. Matt. 28, 405401 (2016).

 MD simulations* of bubble re-solution (hetero)
N = Number of Xe in a bubble
M = Number of re-solved Xe
30
Bubble 10 keV/nm 16 keV/nm 20 keV/nm R0.6N10
R0.6N15

Number of resolved Xe, M

25
R0.6N10 1.2 ± 1.6 5.4 ± 2.3 6.2 ± 1.3 R0.6N20
R0.8N24
R0.6N15 0.8 ± 0.8 8.0 ± 1.7 9.4 ± 1.5 20 R0.8N36
R0.8N48
R0.6N20 2.6 ± 2.2 10.0 ± 2.0 13.6 ± 1.3 15

10
R0.8N24 3.2 ± 1.6 10.2 ± 2.8 12.2 ± 2.3
5
R0.8N36 2.8 ± 1.3 14.6 ± 2.3 20.4 ± 4.0
0
R0.8N48 2.2 ± 2.8 18.2 ± 5.2 26.8 ± 3.2 10 12 14 16 18 20
Effective dE/dx (keV/nm)
M~2 Né, Né, Mé
Mé
le 3
mber of gas atoms resolved into the crystal lattice as a function of PKA direction and energy from a 2.5 nm radius bubble of Xe atoms.

10 keV 20 keV 50 keV

[1 1 0] MD studies of fission gas bubble re-solution
[2 1 0] [5 6 1] [1 1 0] [2 1 0] [5 6 1] [1 1 0] [2 1 0] [5 6 1]
# Xe at.
500 0 0 1 0 2 1 0 0 0
1000 Comparison
0 0 with data
0 by Govers
0 et al1 [2] with Morelon
3 potentials
1 0[3]. 2
1500 0 2 0 0 0 2 0 2 2
Our data with Cooper potentials [1].

7. Number of re-solved atoms as a function of thermal spike energy, plotted for various bubble sizes and densities. The regression line corresponds to Eq. (2). In a few
es, clustering of some Xe atoms lead to the creation of a secondary small bubble.

[1] W. D. Cooper, et al., J. Phys.: Condens. Matt. 28, 405401 (2016).

gh (dE/dx)[2]
effK. Govers,
; these et al.,counted
were J. Nuc. Mater. 420, 282 atoms
as re-solved (2012). as well. At The series of simulations discussed in Section 5 indicate that the
[3] N. D. Morelon,
ry rare occasions et al.,ofPhil.
a splitting theMag. 83, 1533
initial (2003).
bubble into two large smallest clusters of Xe atoms can be fully destroyed but above a
ore than 20 atoms) fragments could be observed; the atoms they threshold size, only partial re-solution will take place. The number
ntain were then not counted as re-solved atoms. Interestingly, the of re-solved atoms per bubble (Nre!sol.) depends critically on the
mber of re-solved atoms does not depend on bubble size or gas actual energy transferred to the lattice, (dE/dx) , but it is indepen-
le 3
mber of gas atoms resolved into the crystal lattice as a function of PKA direction and energy from a 2.5 nm radius bubble of Xe atoms.

10 keV 20 keV 50 keV

[1 1 0] MD studies of fission gas bubble re-solution
[2 1 0] [5 6 1] [1 1 0] [2 1 0] [5 6 1] [1 1 0] [2 1 0] [5 6 1]
# Xe at.
500 0 0 1 0 2 1 0 0 0
1000 Comparison
0 0 with data
0 by Govers
0 et al1 [2] with Morelon
3 potentials
1 0[3]. 2
1500 0 2 0 0 0 2 0 2 2
Our data with Cooper potentials [1].

100
R0.6N10
R0.6N15
Number of resolved Xe, M

80 R0.6N20
R0.8N24
R0.8N36
60 R0.8N48

40

20 Our
Data

0
0 10 20 30 40
Effective dE/dx (keV/nm)
7. Number of re-solved atoms as a function of thermal spike energy, plotted for various bubble sizes and densities. The regression line corresponds to Eq. (2). In a few
es, clustering of some Xe atoms lead to the creation of a secondary small bubble.

[1] W. D. Cooper, et al., J. Phys.: Condens. Matt. 28, 405401 (2016).

gh (dE/dx)[2]
effK. Govers,
; these et al.,counted
were J. Nuc. Mater. 420, 282 atoms
as re-solved (2012). as well. At The series of simulations discussed in Section 5 indicate that the
[3] N. D. Morelon,
ry rare occasions et al.,ofPhil.
a splitting theMag. 83, 1533
initial (2003).
bubble into two large smallest clusters of Xe atoms can be fully destroyed but above a
ore than 20 atoms) fragments could be observed; the atoms they threshold size, only partial re-solution will take place. The number
ntain were then not counted as re-solved atoms. Interestingly, the of re-solved atoms per bubble (Nre!sol.) depends critically on the
mber of re-solved atoms does not depend on bubble size or gas actual energy transferred to the lattice, (dE/dx) , but it is indepen-
 Xolotl*
• Xolotl (SHO-lottle) is the Aztec god of lightning and death
• Developed from ‘scratch’ for the SciDAC project, designed for HPC
(current and emerging architectures – multicore, multicore+accelerator)
to solve advection – reaction – diffusion cluster dynamics problems
within spatially-resolved continuum domain (C++ with MPI and
independent modules for physics, solvers and data management)
• 2D and 3D recently implemented
• Model considers continuum concentration of He, vacancies, interstitials and mixed
clusters at spatial grid points, solving the coupled advection-reaction-diffusion equations

* Available at http://sourceforge.net/projects/xolotl-psi/
 Using MD to ‘train’ continuum scale (Xolotl)
• 100 eV He implantation into W (plasma surface interactions)

MD simulations – Γ ~ 4E25 m-2s-1 Xolotl simulations

 Using MD to ‘train’ continuum scale (Xolotl)

• 100 He implantation into W (plasma surface interactions)

MDsimulations
MD simulations––ΓΓ~~4E25
5E27mm-2-2ss-1-1 Xolotl simulations
Using MD to ‘train’ continuum scale (Xolotl)

Xolotl simulations
&'&&-*#8 1)! $0*%* *+"#$ -1%9* 1%*1# <(,;-*$*-. 2%** (2
&'&&-*# =>A8@IB6 S0* $%1)#"$"()# $( $0"# 0*$*%(9*)*('#
&'&&-* !"#$%"&'$"() (<<'% 1$ -(D*% $*,;*%1$'%*# D"$0
Fission gas evolution modeling
")<%*1#")9 &'%):';6 S0* #,1--*#$ 1)! -1%9*#$ &'&&-*# 1%*
(2$*) 2(')! ") <()$1<$ D"$0 #(-"! ")<-'#"()#
=@M8@@8@G8@I8A58A?B6
• Developed a suite of test problems at low to higher burnup to begin testing/evaluating
7)$%19%1)'-1% &'&&-*# &*<(,* /"#"&-* ") ,"<%(9%1;0#
1# 1the continuum
!1%Z approaches
%")9 1$ 1 ;1%$"<'-1% & connecting
$*,;*%1$'%*8 D0"<0 !*:detailed bubble populations to reduced order
models
;*)!# () $0* &'%):';6 V$ $0* #1,* $",*8 $0* ")$*%9%1):
Test problem:
'-1% &'&&-*# #$1%$ ")$*%-")Z")98 1)! $0*%,1- _ab
• No spatial
<(,,*)<*# =@8G8A>8A@B6 S0* dependence,
#"\* (2 $0* ")$%19%1)'-1%
• Xe2(%
&'&&-*# %*#;()#"&-* introduced at "#2E18
$0* !1%Z %")9 m-3s!,
$ Q!M!Q!5 -1 (Fission rate ~ 8E18 m-3s-1)
=AB6
c0*) $0* 91# "# /*)$*!
• Model ~ 22%(,
years $0*(7E7
9%1")seconds),
&(')!1%"*# corresponding to ~0.83% burnup - 7.9 GWd/ton
$0%('90 $0* $'))*- )*$D(%Z 2(%,*!8 $0* ")$%19%1)'-1%
Temperatures of 1000, 1100, 1200, 1300, 1400 and 1560°C
-- Monitor bubble size, mean size as function of Temperature

*Baker, JNM 66 (1977) 283-291.

_"96 @6 P*)#"$. 1)! 1/*%19* #"\* (2 ")$%19%1)'-1% &'&&-* ;(;:
'-1$"() 1# 1 2')<$"() (2 $*,;*%1$'%* ") "%%1!"1$*! 345 8 1#
 Initial Xolotl predictions
• Initial Cluster Dynamics
modeling results (without re-
solution) are representative in
terms of intragranular bubble
density and size

• Provide a platform for

uncertainty quantification/
sensitivity analysis on Xe
diffusion mechanisms/
diffusivity

• Future efforts will expand

spatially-dependent modeling
to predict both intra-granular
and inter-granular bubble
formation & inform reduced
parameter models (c.f.,
Pastore presentation)
 Summary
• Fission gas behavior in nuclear fuels is a long-standing challenge, but
important contributor, to nuclear fuel behavior (& failure) under normal
and transient conditions
• Multiscale approach involving both an atomistic, ‘bottom-up’
perspective along with simultaneous ‘top-down’ modeling to improve
understanding of fission gas diffusion and precipitation/bubble evolution
mechanisms leading to fission gas release

Initial conclusions:
- Fission gas diffusion under irradiation remains uncertain, but evidence
points to vacancy-cluster mediated diffusion at higher temperatures and
radiation-enhanced diffusion at lower irradiation temperatures
- Re-solution of gas atoms appears dominated by heterogeneous
resolution, with a threshold dE/dx that is sensitive to bubble size &
pressure
- Developing spatially-dependent cluster dynamics for Xe fission gas
bubble populations – initial results promising, but need to demonstrate
influence of re-solution and spatial dependence to provide
“computational database” for reduced parameter models