bridging space-time scales in the field of diffusion …workshop « bridging space-time scale… »,...
TRANSCRIPT
Workshop « bridging space-time scale… », DRESDEN, September 2016
Maylise NastarCEA Saclay, France
Bridging space-time scales in the field of diffusion-controlled
phase transformations
Part of this work was supported by
o the joint program "CPR ODISSEE" funded by AREVA, CEA, CNRS, EDF and Mécachromeunder contract n°070551.
o NEEDS Materiauxo JPNM/Matisse/Mefisto
Atomic diffusion model
Macroscopic diffusion and flux coupling: diffusion equations
d ≈ aΔt ≈ a2 max(ωVA)
d ≈ naΔt ≈ d 2 DB
nano2
(10 )100 V VA
d n at n C»
D » w
Nucleation in dilute alloysCluster Dynamics, OKMC
A multiscale approach for diffusion and phase transformations
d ≈ Vat3 /CBΔt ≈ d 2 max(Mcluster )
Nanoscale diffusion & spinodal decompositionin concentrated alloys: phase field methods
𝑤"#=Γ"𝑒& '()
Growth and coarsening
d ≈ Vat3 /CBΔt ≈ d 2 max(Mcluster )
JB = − Lijj∑ ∇µ j
Atomic diffusion model
Macroscopic diffusion and flux coupling: diffusion equations
d ≈ aΔt ≈ a2 max(ωVA)
d ≈ naΔt ≈ d 2 DB
nano2
(10 )100 V VA
d n at n C»
D » w
Nucleation in dilute alloysCluster Dynamics, OKMC
A multiscale approach for diffusion and phase transformations
d ≈ Vat3 /CBΔt ≈ d 2 max(Mcluster )
Nanoscale diffusion & spinodal decompositionin concentrated alloys: phase field methods
𝑤"#=Γ"𝑒& '()
Growth and coarsening
d ≈ Vat3 /CBΔt ≈ d 2 max(Mcluster )
Interactions between cells?nanoscale fluctuations and nucleation?
Flux couplings?
JB = − Lijj∑ ∇µ j
Irradiation of Fe-15 % Cr by e- 1MeV (THT)T=773 K, F=3.9x10-5 dpa/s, dose=0.2 dpaO. Tissot, E. Meslin et al. (CEA-SRMP & Univ. de Rouen)
Atom Probe Tomography(GPM,Rouen)TEM – Bright field (CEA Saclay)
Microstructure of point defectsInterstitial loops a<100>
Radiation Induced Segregation (RIS) of Cr at loops due to flux couplings
A typical phenomenon involving bothPoint defect-solute cluster reactions and flux coupling
what is the effect of point-defect microstructure on RIS?what is the effect of RIS on the point defect microstructure?
200 nm
Continuous diffusion equations
Negative flux coupling
Spatial cluster dynamics (Crescendo)Object kinetic Monte Carlo
A typical phenomenon involving bothcluster reactions and flux coupling
DissociationCluster mobility(positive flux coupling)
VJVBJ
VAJ
sink
JAV = −
CVirr
CVeq (LAV∇µV + LAA
V ∇µA + LABV ∇µB)
VC
Onsager: positive & negative flux couplingClouet, E. ASM Handbook 2009, 22A, 203–219 .Jourdan, T. et al. Acta Mater. 2010, 58, 3295–3302.M. Nastar & F. Soisson, Comprehensive…
Continuous diffusion equations
Positive flux coupling
Spatial cluster dynamics (Crescendo)Object kinetic Monte Carlo
A typical phenomenon involving bothcluster reactions and flux coupling
DissociationCluster mobility(positive flux coupling)
VJVBJ
VAJ
sink
JAV = −
CVirr
CVeq (LAV∇µV + LAA
V ∇µA + LABV ∇µB)
VC
Onsager: positive & negative flux couplingClouet, E. ASM Handbook 2009, 22A, 203–219 .Jourdan, T. et al. Acta Mater. 2010, 58, 3295–3302.M. Nastar & F. Soisson, Comprehensive…
Outline
Ø Principles oftheself-consistentmean field theory (SCMF)
Ø Kinetic characterization ofclustersfrom theatomic scale
Ø SCMFbased - ClusterDynamicsextended toheterogeneous systems
Ø SCMFbased - PhaseFieldEquations
Principles of the Self-Consistent Mean Field theory (SCMF)
Atomic diffusion model on a rigid lattice
Application of a driving force
A mean field theory:
diffusion of particle A within a mean field
Non-equilibrium pair variables AVComputation of the coresponding fluxesIdentification of the Lij coefficients
*Nastar et al. Phil. Mag. A 2000, 80, 155–184.
𝜔+,"#
𝛻𝜇"
𝐽"
𝜔+,#
𝛻𝜇"
𝐽"
𝜔+,"#
𝛻𝜇"
𝐽"
i
i j
j
ji
SCMF theory: introduction of a non equilibriumdistribution function
Non equilibrium distribution function
Equilibrium distribution function
niα = 1, ni
V = 0njV = 1, nj
α = 0
𝜔+,"#
𝛻𝜇"
𝐽"i j
Pair correlators versus effective interactions
niα = ni
αP(n,t)n∑
niαnj
β = niαnj
βP(n,t)n∑
P(1)(n,t) = exp δΩ + δµiαni
α − 12
vijαβ (t)ni
αnjβ
i, jα ,β
∑i,α∑
⎛
⎝
⎜⎜
⎞
⎠
⎟⎟
kBT⎡
⎣
⎢⎢⎢
⎤
⎦
⎥⎥⎥
P(0)(n) = exp Ω + µα niα − H (n)
i∑
α∑⎛
⎝⎜⎞⎠⎟
kBT⎡
⎣⎢
⎤
⎦⎥
P(n,t) = P(0)(n)P(1)(n,t)
Kijαβ = ni
αnjβ − ni
α njβ
Kijαβ − Kij
βα = niαnj
β (0)vijαβ − vij
βα( )
SCMF Kinetic equations from the Master equation
Master equation
10
Identification with Onsager’s flux and computation of Lab
𝒗𝒊𝒋𝑽𝜶 solution of the pair kinetic equations
dP (n,t)dt
= W n '→ n( )P(n ')−W n→ n '( )P(n)⎡⎣ ⎤⎦n '∑
JA = − Lαββ∑ β∇(µβ − µV )
Atomic fluxes
Kinetic correlations = jump efficiency
d niB
dt= −∇i JB ⇒ JB = −z ω i,i+1
BV niBni+1
V β ∇i (µB − µV )+ vi,i+1VB − vi,i+1
BV⎡⎣ ⎤⎦
d niBni+1
V
dt= ω i, i+1
BV niBni+1
V (0)∇i (µB − µV )− ni
Bω i+1,i+2V A ni+1
V ni+2A (0)
∇i+1(µA − µV )
+q1(vi,i+1AV − vi,i+1
VA )+ q2 (vi,i+2AB − vi,i+2
VB )
KINETIC CHARACTERIZATION OF CLUSTERS FROM THE ATOMIC SCALE
*T. SCHULER and M. NASTAR, PRB 93 (22), 224101
Breakdown of the Onsager matrix for a dilute alloy
Equal because of detailed balance
Mobility without dissociation
Kinetic correlations = jump efficiency
From the SCMF theory:
Dilute alloy
*T. Schuler & M. Nastar, PRB 93 (22), 224101
Equilibrium clusterTransport coefficients
Lαβ = ci[ ]i∑ Lαβ
eq (ci )
Non equilibrium Onsager transport coefficient
Cluster reactions
The mobility of a cluster 13
Mobility without
dissociation
A consistent definition of the cluster radius (Rc) is not only
thermodynamic, but also depends on kinetic correlations
SCMF in the cluster volume (no association/dissociation jumps) :
Rc
*T. Schuler & M. Nastar, PRB 93 (22), 224101
Model I: Highest barrier modelModel IV: R=Rth=1.23 aModel V: R=3.2 a
SCMF mobility compared to the highest barrier model
Mobility of VC pairs in Fe*T. Schuler & M. Nastar, PRB 93 (22), 224101
Fe(C)
The association-dissociation term 15
Rc
Dissociation - Association
Equal because of detailed balance
There are different dissociation paths => the AD matrix does not reduce to a scalar
Systems Fe(X): Monte Carlo versus SCMF
XVM
*L. Messina, T. Schuler, P. Olsson, M. Nastar
τ = a2
LVV (XV ) AD
l = MXVτ
Monte Carlo simulations: in coll. with M. Chiapetto et al. (SCK, Belgium)
Physics of flux coupling within cluster dynamics
Breakdown of the Onsager matrix into cluster contributions for dilute systems
• Exchange Correlations between A/D and M jumps• Contains the physics of negative flux coupling
Association jump Mobility jump Dissociation jump
vacancy
soluteNegative flux couplingDue to exchange correlations
SCMF-BASED CLUSTER DYNAMICSEXTENDED TO HETEROGENEOUSSYSTEMS
*T. SCHULER, T. JOURDAN, M. NASTAR
Homogeneous Cluster Dynamics
Association coefficient GAWaite formulae
κ =2MCj
λγ
Lγγ (Ci ,Cm )ADm= j ,k∑
γ =α ,β∑
Ck
Cj
Ci
Dissociation coefficient GDGlobal detailed balance condition
ΓA Cj +Ck →Ci( ) = 4πR(Ck )2MCj
(R(Ck )+κ )
ΓD Ci →Cj +Ck( ) = 1Ωat
ΓA Cj +Ck →Ci( )exp F(Cj )+ F(Ck )− F(Ci )kBT
⎛⎝⎜
⎞⎠⎟
*Clouet, E. ASM Handbook 2009, 22A, 203–219 .*Jourdan, T. et al. Acta Mater. 2010, 58, 3295–3302.
Heterogeneous Cluster Dynamics*T. Schuler, T. Jourdan, M. Nastar
Vacancy diffusion mechanism, 3 species: monovacancy V, solute B, pair BV
Flux – part 1: migration of clusters
∂ V[ ]∂t
= −∇JVCD − ΓA V[ ] B[ ]+ ΓD VB[ ]
∂ B[ ]∂t
= −∇JBCD − ΓA V[ ] B[ ]+ ΓD VB[ ]
∂ BV[ ]∂t
= −∇JBVCD + ΓA V[ ] B[ ]− ΓD VB[ ]
⎧
⎨
⎪⎪⎪
⎩
⎪⎪⎪
JVCD = −MV∇ V[ ]
JBVCD = −MBV∇ BV[ ]
Heterogeneous Cluster Dynamics
Different cluster concentrations between two adjacent cells leads to asymmetric A-D reactions and fluxes between cells.
*T. Schuler, T. Jourdan, M. Nastar
JV
Flux – part 2: Association-Dissociation of clusters
AD fluxes obtained from TIPat local equilibrium, CD and TIP fluxes are put equalkBT
∇ B[ ]B[ ] = ∇µB
kBT∇ V[ ]V[ ] = ∇µV
kBT∇ BV[ ]BV[ ] = ∇µB +∇µV
⎧
⎨
⎪⎪⎪⎪
⎩
⎪⎪⎪⎪
JVCD + JBV
CD = JVTIP
JBCD + JBV
CD = JBTIP
⎧⎨⎪
⎩⎪
Cluster Dynamics versus Monte Carlo
0 50 100 150 200 250 300 350 400 450 5000,0
5,0x10-5
1,0x10-4
1,5x10-4
2,0x10-4
2,5x10-4
3,0x10-4
3,5x10-4
CB
plane number
Atomic Monte Carlo Cluster Dynamics Crescendo
0 50 100 150 200 250 300 350 400 450 5000,0
1,0x10-4
2,0x10-4
3,0x10-4
4,0x10-4
5,0x10-4
6,0x10-4
7,0x10-4
CB
plane number
Atomistic Monte Carlo Cluster Dynamics Crescendo
GB=GA
EBV=0.1 eV
Monte Carlo simulations N=512*256*256/4
Monte Carlo eventsRandom replacement of an atom by a vacancyatrateF*NExchange of a vacancy with a nn atom (A or B)
Boundary conditionsVacancy at sinks is removed and randomly replaced by atom A or B with the constraint of constant solute concentration.
ClusterDynamicsBoundary conditionsCB=CBV=0 at sinks
Simulation of vacancy irradiation segregationT=500 K, CB=2.10-4, F=5. 10-10 dpa.s-1
𝑤𝑖𝑡ℎ𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 ∶ 𝑤"#=Γ"𝑒&'(GHI)() andWBV
GB=10 GA
Steady state profiles of solute
perfect sink
perfect sink
SCMF-BASED PHASE FIELDEQUATIONS
*M. NASTAR, PRB 90 (14), 144101
Nanoscale interdiffusion experiments*O Senninger, F Soisson, E Martínez, M Nastar, CC Fu, Y Bréchet,Acta Materialia 103, 1-11
time
Slope= -Log(wave-amplitude)
Phenomenological Cahn-Hilliard theory
slope=2Worder (=stiffness k)
annealing
D
Multilayers of isotopic atomsWorder = 0 ⇒ D (Λ) = D (0) ?
Something is missingin the Cahn-Hilliard theory
A kinetic stiffness parameter challenging the Cahn-Hilliard phenomenological equation
*M. Nastar, PRB 90 (2014), 144101
time
Slope=Log(wave-amplitude)
annealing
Composition gradient-energy Composition gradient-correlation
Interdiffusion coefficient as a function of k
κ E :κC :
Fe-Cr, T=700 K
2Dk
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
Kapp
a (e
V)
cCr
AKMC-Dumbbell AKMC-Vacancy AKMC-Direct exchange
D = M f "+ (κ th +κ kin )k2⎡⎣ ⎤⎦
Interdiffusion in binary alloys with non uniform driving forces
Beyond the Onsager coefficients
JB = −LBB∇µBV − LAB∇µAV −QBB∇3µBV −QAB∇
3µAV
A purely correlation term
MRS 2012
Uniform driving force Non uniform driving force
Conclusion and perspectives
Atomic diffusion theory SCMF could be used to properlybridge space-time scales in diffusion controlled phenomena
-Heterogeneous object oriented simulation methods-A rigourous definition of cluster kinetic properties-Modeling of flux coupling phenomena
-New diffusion equation for the phase field method-Nanoscale diffusion experiments: a way to get detailedinformation on the alloy diffusion properties
-Perspectives: Off-lattice modeling: a network specific to each cluster