PERFORM 60 FP7 Project Time accelerated Atomic Kinetic Monte Carlo for radiation damage modelling C. Domain, C.S. Becquart, R. Ngayam-Happy EDF R&D Dpt Matériaux & Mécanique des Composants Les Renardieres, Moret sur Loing, France UMET, Université de Lille 1 Villeneuve d Ascq, France
1nm 3 m 3 ab initio 0 - ps (10-30nm 3 ns 40 years Molecular dynamics Finite elements Multi-scale modelling s - h cm 3 Micro-macro Barbu, CEA + experimental validation PERFORM 60 FP7 Project Pareige, U. Rouen KMC cohesive model & parameterisation (30-100nm 3 µm 3 h-year Dislocation dynamics Mesoscopic 2 EURATOM European EDF Project R&D PERFECT - Workshop (FI6O-CT-2003-508840 BEMOD12 - Dresden - March 2012
Ni Radiation damage Material: Fe + alloying elements: Cu, Ni, Mn, Si, + carbon, nitrogen + dislocations 0.08 dpa Neutron irradiation Si Irradiation: Electron: Frenkel pairs Ion and neutron: displacement cascades ( 10-100 kev vacancies and interstitials: isolated and in clusters vacancies neutron displacement cascades Mn Cu Elastic interaction E energy transfert PKA PKA : primary knock-on atom interstitials TAP, Pareige, U. Rouen 15x15x50 nm Microstructure evolution: point defect clusters: nanovoids, dislocation loops solute clusters (# or \# point defects 3 TEM, Barbu, CEA
Kinetic Monte Carlo simulation of irradiation Atomic KMC Object KMC Recombination Emission + Electrons PBC or surface dislocation Interstitial loop traps Emission Vacancy cluster + Frenkel pairs surface cascades Cascades Interstitial cluster Neutrons sinks PBC PBC Vacancy loop + Annihilation Migration cascade Paires Frenkel de Frenkel pairs surface Ageing (one single vacancy [JNM 335 (2004 121 145] 4
Atomic Kinetic Monte Carlo of microstructure evolution Objective: Simulation formation of solute rich complexes (observed by TAP under irradiation Mn Ni Cu Si TAP, Pareige, U. Rouen 15x15x50 nm Ab initio Solute interactions (Cu, Ni, Mn, Si (interface energies, mixing energies Experimental data and Thermodynamical data Fe-V_1nn Parameterisation cohesive model Fe-Si_2nn AKMC Solute diffusion by - vacancy mechanisms - interstitial mechanisms Emixing (1 (2 (1 (2 (1 (2 4 ( FeFe 3 ( FeFe 8 ( FeX 6 ( FeX 4 ( X X 3 ( X X E E (1 (2 (1 (2 ( V Z formation 8 ( V Z 6 ( V Z 4 ( Z Z 3 ( Z Z (1 (1 (1 (1 (1 binding( V X ( FeV ( Fe X ( FeFe ( V X Experimental validation: TAP, SANS, SAXS, PA, TEP 5
Atomistic Kinetic Monte Carlo (AKMC Treatment of multi-component systems on a rigid lattice Substitutional elements Interstitial elements Code: LAKIMOCA Diffusion by 1nn jumps Via vacancies Via interstitials Vincent et al. NIMB 255 (2007 78 Vincent et al. JNM 382 (2008 154 6 Jump Probability: Ea X X exp kt Residence Time Algorithm applied to all events Vacancy and Interstitial jumps Frenkel Pairs and Cascade flux for irradiation Average time step: t Environment dependant form of activation energy Ea j, k 1 jk Ea X = attempt frequency Ea( X i 1,2 2,2 v v 1 2 v 3 2,1 1,1 1,1 1,8 2,1 Ef Ei 2 2,7 3,2 3,2 3,1 3,7
AKMC irradiation simulation conditions For electron irradiation: Frenkel Pair (FP flux For neutron irradiation: flux of 20 kev and 100 kev cascades debris obtained by Molecular Dynamics (R. Stoller, J. Nucl. Mater. 307-311 (2002 935 Frenkel Pairs surface cascades Cascades Typical simulation box: PBC PBC 1.01 10-17 cm 3 boxes 8.65 10 6 atoms Paires Frenkel de pairs Frenkel surface 7 7
8 AKMC simulation of radiation damage accumulation Target dose: 0.1 dpa Irradiation duration: 2 days (10 5 s up to 40 years (10 9 s Irradiation temperature: 573 K Defect accumulation: > 100 point defects in the simulation box Events: Self interstitial migration (0.3 ev : time step : 10-10 s Vacancy migration (0.65 ev : time step : 10-7 s Rapidly: annihilation or formation point defect clusters Point defect migration within point defect - solute clusters or trapping with solutes Very large number of jumps required to have significant event (ie emission or diffusion Other jumps with high migration energies (1 ev : time step : 10-4 s Computational limitation: ~10 10 steps / month Very complex situation: many events with different time scale & long simulation required
Cohesive energy model Ea Ea( X i Ef 2 Ei Vacancy: Fe-V_1nn E j ( i ( i ( i ( i ( i ( i ( FeFe ( V V ( FeV ( Fe X ( V X ( X Y k l m n p Fe-Si_2nn RPV: 1nn and 2nn pair interactions FeCr: 2BM potential solute - dumbbell E mixed l ( X j X k E 1nnComp l + ( dumb i X j + 1nnTens E l ( X j SIA: dumbbell - dumbbell E b (dumb - dumb 1nn & 2nn Solute atoms E dumb Fe atom 1nnComp 1nnTens mixte E f El ( dumbi X j El ( X j E ( X j X k El ( dumb dumb i l j j i, j FIA (C: + + + FIA vacancy solute SIA ~ 100 ab initio data considered in the model 9 9
Cohesive model: X-Y and V-X determination Binary alloys E E E (1 (2 cohésion ( Z 4 ( Z Z 3 ( Z Z ( Z (1 (2 (1 (2 formation lac 8 ( lacz 6 ( lacz 4 ( Z Z 3 ( Z Z E E Emélange (1 (2 (1 (2 (1 (2 int erface(100 2 ( FeFe ( FeFe 4 ( FeX 2 ( FeX 2 ( X X ( X X ( i liaison ( laclac 2 4 (1 (2 (1 (2 (1 (2 ( FeFe 3 ( FeFe 8 ( FeX 6 ( FeX 4 ( X X 3 ( X X ( i ( Felac ( i ( FeFe ( i ( laclac (1 (1 (1 (1 (1 liaison( lacx ( Felac ( FeX ( FeFe ( lacx ε Fe-Cu_1nn i = 1 or 2 X, Y = solute atoms Z = Fe or solute atom Ternary alloys E ( i liaison ( X Y ( i ( Fe X ( i ( FeY ( i ( FeFe ( i ( X Y ε Si-Si_2nn Ab initio data Parameters Adjustment on thermal annealing experiment 10 10
Neutron irradiation of FeCuNiMnSi alloys Medium term evolution by atomic Kinetic Monte Carlo Fe-0.2Cu-0.53Ni-1.26Mn-0.63Si (at.% at 300 C Flux: 6.5 10-5 dpa.s -1 Dose: 1.3 10-3 dpa V-solute complex SIA-solute complexes Cu Si Ni V Small solute clusters Point defect clusters = germs for precipitation Mn 11 SIA [> 1 month on 1 CPU]
Fe CuMnNiSiP (at.% alloys Distribution and composition of clusters: Number of species in clusters 40 35 30 25 20 15 10 5 0 12 0.18Cu 1.38Mn 0.69Ni 0.43Si 0.01P (A533B plate 5.79x10-4 dpa/s - 300 C - 9.03 mdpa Cu Si Mn Ni P Vac SIA 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 Cluster Number High Nd of PD clusters Vacancy clusters are bigger and less numerous than SIA clusters Solute clusters form on PD clusters (induced segregation Clusters associated with SIA clusters are enriched in Mn Clusters associated with vacancy clusters are enriched in Si/Cu/Mn and Ni/P Cu enriched clusters observed (enhanced precipitation Mean composition of clusters 40 30 20 10 0 25 20 15 10 5 0 Vacancy SIA Pure Solute Nd of Clusters Vacancy SIA Pure Solute Mean size of Clusters Solute-Vacancy Solute-SIA Pure solute SIA Vac P Ni Mn Si Cu 12
Fe CuMnNiSiP (at.% alloys V-Solute SIA-Solute Pure solute Répartition des espèces 25 20 15 10 5 0.18Cu 1.38Mn 0.69Ni 0.43Si 0.01P 5.79x10-5 dpa/s - 300 C - 18.05 mdpa SIA Vac P Ni Mn Si Cu 0 1 7 13 19 25 31 37 43 49 55 61 67 73 79 Numéro Cluster de ID l'amas The biggest solute clusters are associated with PD clusters In agreement with induced segregation mechanism to account for solute clusters formation Clusters associated with interstitial clusters are enriched in Mn, and P/Ni Clusters associated with vacancy clusters are enriched in Si/Cu/Mn (mostly and Ni I-Solute complexes V-Solute complexes 13
Isochronal annealing: pure Fe Isochronal annealing in pure Fe DT/Dt = 2K/300s (T<30K DT/TDt = 0.03K/300s (T>30K Tirrad = 4.5 K d n dt n 0 (%/K I E peak shift II peak shift (Resistivity n (number of defects Differential Fractional Recovery d/dt (D /D 0 d/dt (n/n i 2.89 x 10-4 dpa 2.31 x 10-4 dpa 4 1.74 x 10-4 dpa [C. C. Fu et al., 2004] 2 1.16 x 10-4 dpa 5.79 x 10-5 dpa Triangle: Exp. Results: S. Takaki, et al., Radiat. Eff. 79, 87-122, 1983 14 EDF R&D - Workshop BEMOD12 - Dresden - March 80 2012 100 120 140 160 180 200 220 T (K
79 K d n dt n 0 (%/K Isochronal annealing FeP I E peak shift (SIA trapping by P atoms 3K / 180s - 2.4 10-5 dpa/s Fe 0.1% P 124 K 110 Fe-P Eb = 1.02 ev Fe 0.07% P Fe 0.009% P 4 142 K 2 Fe 0.0044% P P subs 110 Fe-P Eb = 2.0 ev No modification observed on I D peak Pure Fe 15 Experimental results: H. Abe, E. Kuramoto, J. Nucl. Mater. 271 & 272 (1999 209. 80 100 120 140 T (K Solid line: sim / dash line: exp
Isochronal annealing FeP 160 K 190 K d n dt n II (%/K Fe 0.1% P 232 K 110 Fe-P Eb = 1.02 ev Fe 0.07% P Fe 0.009% P 4 250 K Fe 0.0044% P 2 P subs 110 Fe-P Eb = 2.0 ev Pure Fe 16 120 140 160 180 200 220 T (K Solid line: sim / dash line: exp
Time Accelerated Dynamics in our AKMC With solutes, vacancy clusters & interstitial clusters (SIA and carbon, many trapping situations. Associated time steps 10-10 s (other ones 10-5 s - 1 s Large number of SIAs and vacancies in the simulation box (>100 Do not know a priori which object (point defect solute cluster will be the usual suspect During irradiation simulation, several different kind objects can be trapping situations Adapted version of the TAD algorithm of Voter et al. In our AKMC. Different from pulsing algorithm of Wirth & Odette (1998 & 2007. Based on temperature increase. 17
Time Accelerated Dynamics in our AKMC 10 T ref Jump probability (s -1 Attempt frequency are all similar T TAD 10 6 1/T t ref = t TAD exp[ E m / (1 / kt ref 1 / kt TAD ] 18
Time Accelerated Dynamics algorithm in our AKMC Every N steps, determination each defect mean free path (MFP If all MFPs are very low Search of the 5 largest jump probabilities (P 1 > P 2 > P 3 > P 4 > P 5 If (P 5 / P 1 > 10 Choose T TAD in order to have P5/P1 = 10 Perform 3 AKMC steps (with adjusted time step 10 T ref t ref = t TAD exp[ E m / (1 / kt ref 1 / kt TAD ] T TAD 10 6 back to T ref 1/T 19
TAD - AKMC: performance improvements Speed-up = Nb steps AKMC / Nb steps TAD-AKMC 1000 900 800 700 600 500 400 300 200 100 0 Fe Fe 0.7Ni Fe 1.4Mn 0 50 100 150 200 250 300 350 400 T (K - speed-up : 2 order of magnitude - Simulation with solutes (Ni, Mn very long due to SIA-solute interactions 20
TAD - AKMC: preliminary results 100 90 80 70 60 50 40 30 20 10 0 T (K Fe Ni Mn Standard AKMC 0 50 100 150 200 250 300 350 400 Isochronal annealing: Fe, Fe 0.7%Ni, Fe 1.4%Mn Good agreement with experimental results standard-akmc: 1 week CPU TAD-AKMC: 12 hours CPU 100 90 80 70 60 50 40 30 20 10 0 TAD - AKMC T (K Fe Fe 0.7Ni Fe 1.4Mn 0 50 100 150 200 250 300 350 400 21 Similar physical results (some statistics are required
Conclusions & perspectives AKMC of complex alloys (with simple cohesive models under irradiation feasible for low doses and high fluxes Massive Parallelisation is a difficult issue TAD method allows to improve the performance of the AKMC tools. TAD method validated on isochronal annealing simulation. Perspectives: To use TAD for radiation damage simulation under flux. To improve cohesive models 22