Abstract
Clean, safe and economical nuclear energy requires new materials capable of withstanding severe radiation damage. One strategy of imparting radiation resistance to solids is to incorporate into them a high density of solidphase interfaces capable of absorbing and annihilating radiationinduced defects. Here we show that elastic interactions between point defects and semicoherent interfaces lead to a marked enhancement in interface sink strength. Our conclusions stem from simulations that integrate first principles, object kinetic Monte Carlo and anisotropic elasticity calculations. Surprisingly, the enhancement in sink strength is not due primarily to increased thermodynamic driving forces, but rather to reduced defect migration barriers, which induce a preferential drift of defects towards interfaces. The sink strength enhancement is highly sensitive to the detailed character of interfacial stresses, suggesting that ‘supersink’ interfaces may be designed by optimizing interface stress fields. Such interfaces may be used to create materials with unprecedented resistance to radiationinduced damage.
Similar content being viewed by others
Introduction
Damage caused by highenergy particles, such as neutrons or ions, severely limits the performance of materials used in nuclear energy applications^{1,2}. It arises from the exceedingly high concentrations—many orders of magnitude greater than under thermodynamic equilibrium—of crystal defects created by impinging radiation^{3,4}. Preventing radiationinduced damage in engineering solids requires rapid removal of these defects. Materials resistant to radiation damage would markedly improve the safety, efficiency and sustainability of nuclear energy.
One way of removing radiationinduced defects is to provide a high density of sinks, such as grain boundaries or heterophase interfaces^{5,6,7} that continually absorb defects as they are created. This motivation underlies ongoing exploration of the radiation response of nanocrystalline and nanocomposite materials^{8,9,10,11}, because of the large total interface area per unit volume they contain. These investigations have demonstrated wide variations in sink behaviour of different interfaces. Some easily absorb defects, preventing damage in neighbouring material, but become damaged themselves^{12}. Others are poor sinks for isolated defects, but excellent sinks for defect clusters^{13,14}. The sink behaviour of yet others changes with radiation dose^{15,16,17}. This wide variety of radiation responses prompts us to ask: are some specific interfaces best suited to mitigate radiation damage? Is it possible to identify them without resorting to resourceintensive irradiation experiments?
To answer these questions, we propose an improved computational method for rapidly assessing the vacancy and interstitial sink strength of semicoherent interfaces. This method builds on a reduced order model for elastic fields of heterophase bicrystals^{18}. Such interfaces are of particular interest because many of them contain a high density of defect trapping sites^{19,20,21}. Moreover, semicoherent interfaces generate elastic fields that interact directly with radiationinduced defects^{18,22,23}. We show that these elastic fields have an unexpectedly large influence on interface sink strength. Unlike previous studies, which highlighted the importance of thermodynamic driving forces for interface sink behaviour^{24,25,26}, we find that the principal effect of the elastic fields is to modify defect diffusivities, causing defects to drift preferentially towards the interface through a nonrandom walk process. Our work also demonstrates that interface sink strength is highly sensitive to the exact distribution of interface elastic fields. These findings motivate a computational search for ‘supersink’ interfaces: ones that optimally attract, absorb and annihilate radiationinduced defects.
Results
Kinetic Monte Carlo simulations with elastic interactions
Modelling the removal of radiationinduced point defects at sinks is a challenging task: on one hand, the variety and complexity of defect behaviours call for the flexibility of atomistic modelling. On the other, the relatively slow, thermally activated mechanisms of defect motion require longer simulation times than may be reached using conventional atomistic techniques, such as molecular dynamics. We employ the object kinetic Monte Carlo (OKMC) method^{27,28,29,30}, which is well suited to modelling longtime, thermally activated processes yet is also able to account for nuances of defect behaviour uncovered through atomistic modelling.
Figure 1 illustrates the set up of our simulations. Our models contain two crystalline layers—A and B—separated by semicoherent interfaces. Periodic boundary conditions are applied in all directions, so each model contains two A–B interfaces. Owing to their inherent internal structure, the interfaces create characteristic stress fields in the neighbouring crystalline layers. These stress fields interact with radiationinduced point defects, modifying their diffusion.
We calculate interface stress fields using an efficient, semianalytical method developed previously^{22,31,32}. This method accounts for the elastic anisotropy of solids A and B as well as for differences in elastic constants between them. It generates elastic field solutions consistent with a prespecified interface crystallographic character (that is, misorientation and interface plane orientation^{33}) and with vanishing farfield stresses. The method has been successfully applied to a variety of homo and heterophase interfaces^{22,31,32} and has been validated through comparisons with atomistic simulations^{18}.
Our modelling approach may be used on a wide range of semicoherent interfaces. However, for illustration, we will focus on two specific interfaces in the present work: a lowangle twist grain boundary (GB) on a (001) plane in Ag and a pure misfit (zero misorientation) heterophase interface between (001) planes of Ag and Cu. In our calculations, we use the lattice parameters and elastic constants for Ag and Cu listed in Table 1 (refs 34, 35).
Figure 2a shows a plan view of the Ag twist GB, where the adjacent GB planes have been rotated by ±θ/2 (θ: twist angle). The boundary plane contains two sets of parallel, pure screw dislocations: one aligned with the x=[110] direction and the other with the direction. For a relative twist angle of θ=7.5°, the spacing between dislocations within each set is ∼2.2 nm. Figure 2b shows the interface plane of the Ag–Cu pure misfit interface. Similar to the twist boundary in Fig. 2a, this interface also contains two sets of parallel dislocations aligned with the x=[110] and directions. Furthermore, the spacing between dislocations in the Ag–Cu interface is the same as in the twist boundary of Fig. 2a: ∼2.2 nm. However, unlike in the twist boundary, both sets of dislocations in the misfit interface are of pure edge type.
The two interfaces in Fig. 2 have identical dislocation arrangements, but different dislocation characters. Thus, they contain identical dislocation densities, but have differing stress fields. For instance, all normal stress components for the twist GB are zero throughout the entire bicrystal. This stress field is therefore purely deviatoric. By contrast, owing to symmetry, the shear stress σ_{12} is everywhere zero for the Ag–Cu interface, but all of its other stress components are in general nonzero. In particular, this interface generates significant hydrostatic stresses. These differences have important implications for interfacedefect interactions and defect migration pathways.
We use the force dipole moment approximation to compute elastic interaction energies between point defects and interfaces, E^{PD/int} (refs 36, 37, 38):
Here, are the components of the previously calculated interface strain field. P_{ij} are the components of the elastic dipole tensor (the ‘Ptensor’), which describes the elastic fields generated by a point defect. E^{PD/int} values are used to compute stressdependent energy barriers for defect migration at each location in the simulation cell (see Methods for details). A similar approach has been adopted in previous OKMC studies to describe point defect interactions with dislocations^{39,40}.
We use density functional theory (DFT) to calculate Ptensors for two types of point defects in Ag and Cu (see Methods for details): vacancies and selfinterstitials of lowest formation energy, namely 〈100〉split dumbbells^{41}. We obtain Ptensor values for these defects in their ground states as well as at their saddle point configurations during migration (found using the climbing image nudged elastic band method^{42}). Starting from a simulation cell containing a perfect, stressfree crystal, we insert the point defect of interest in the desired location and relax the atom positions while keeping the simulation cell shape fixed. The point defect induces stresses, σ_{ij}, in the simulation cell. They are related to the defect Ptensor through
where V is the simulation cell volume. and p^{h} are the deviatoric and hydrostatic (isotropic) Ptensor components, respectively. The former is associated with a pure shear (no volume change), whereas the latter is related to isotropic tension (interstitials) or compression (vacancies), which leads to a volume change.
Table 2 lists the Ptensors used in the present study. All of them are expressed in the Nye frame, where the X, Y and Z axes are aligned with the [100], [010] and [001] Miller index directions, respectively. The form of the Ptensor reflects the symmetry of the corresponding defect. Thus, the Ptensor for a vacancy in its ground state is isotropic, whereas that of an interstitial is tetragonal. Ptensors for defect orientations other than those given in Table 2 may be calculated using coordinate system rotations. Our Ptensors for 〈100〉split dumbbell selfinterstitials and vacancies in Cu agree with experimental data^{41,42,43,44}. Furthermore, the present calculations of relaxation volumes of a vacancy in Ag and Cu are in very good agreement with recent DFT predictions^{45}.
Figure 3 shows the distribution of groundstate interstitial and vacancy interaction energies with the Ag twist GB and the Ag–Cu misfit interface. A 〈100〉split dumbbell interstitial may take on three different orientations. Figure 3 uses the orientation with lowest E^{PD/int}. For the Ag twist GB, interstitial interaction energies are negative at all locations, as shown in Fig. 3a. Thus, all interstitials in the vicinity of this GB experience a thermodynamic driving force to migrate towards the boundary. The interstitials, however, have nearly isotropic Ptensors (see Table 2), so their interaction energies with the Ag twist GB are very small. The interaction energy of vacancies with the Ag twist GB is everywhere zero because of the absence of hydrostatic stresses near this interface. However, the anisotropy of the vacancy saddle point configuration leads to nonzero interaction energies of migrating vacancies with the GB.
Interstitial interaction energies near the Ag–Cu misfit interface, shown in Fig. 3b, may be attractive or repulsive, depending on the location of the defect. Thus, interstitials in Ag are expected to migrate towards the centre of the dislocation pattern, whereas those in Cu are expected to migrate to dislocation cores. Figure 3c shows the interaction energy between vacancies and the Ag–Cu misfit interface. The spatial variation of this interaction energy is similar to that of the interstitials, but with opposite sign.
Our OKMC simulations assume a constant, uniform defect creation rate, G. Defects diffuse until they are absorbed by an interface (see Methods for details). Only individual interstitials or vacancies are tracked in our simulations: defect reactions, such as clustering or recombination, are not considered. After a certain simulation time, defect distributions reach a steady state, whereupon the defect concentration is computed as a function of position along the zdirection (normal to the interface plane) based on the time spent by each defect on a given atomic site.
Effect of elastic interactions on interface sink strength
Figure 4 shows steadystate vacancy and interstitial concentrations for the two types of interfaces described above for models with 10nmthick Ag and Cu layers. In the absence of elastic interactions between defects and interfaces, steadystate defect concentrations may be computed analytically (see Methods for details). We compare our simulation results with these analytical solutions.
Elastic interactions have a dramatic effect on defect concentration profiles. In all cases shown in Fig. 4 except vacancies near Ag–Cu interfaces, there are nearly no defects within ∼2nmwide zones adjacent to the interfaces. By contrast, without elastic interactions, defect concentrations are zero only at the interfaces themselves. Moreover, even though defectinterface elastic interaction energies are negligible beyond ∼2 nm, the zones depleted of defects near the interfaces have a pronounced effect on defect concentrations throughout the entire layer, markedly reducing the average defect concentration. For the simulations in Fig. 4, elastic interactions reduce defect concentrations by about a factor of two even in the middle of the layers. This effect is even more pronounced for thinner layers. For vacancies in Ag–Cu, local traps are responsible for the sharp increase in concentration near the interface.
Our simulations account for numerous aspects of defectinterface elastic interactions, such as defect anisotropy or differences in defect groundstate and saddle point properties. To discover which ones are primarily responsible for the defect concentrations shown in Fig. 4, we artificially ‘switch off’ some of these characteristics and repeat our OKMC simulations to see whether doing so changes the steadystate defect concentrations. These calculations lead us to conclude that the anisotropy of the Ptensor in the saddle point configurations is primarily responsible for the reduced defect concentrations in Fig. 4a,b.
We ‘switch off’ saddle point anisotropy by replacing the saddle point Ptensor with , where δ is the Kronecker delta and is onethird of the trace of the true saddle point Ptensor. This assumption is tantamount to modelling defects at saddle points as misfitting spherical inclusions in isotropic media. Concentration profiles obtained with this approximation are markedly different from the anisotropic case, as shown in Fig. 4. In the case of the Ag twist GB (Fig. 4c,d), isotropic saddle points yield the same defect concentrations as when there are no defectinterface interactions at all. Indeed, as the twist interface generates no hydrostatic strain field, only the deviatoric components of defect Ptensors may interact with these interfaces. Groundstate vacancies have zero deviatoric Ptensor components, so the interaction energy with the Ag twist GB vanishes, similar to groundstate interstitials with nearly isotropic Ptensors (Table 2). The same conclusions hold at saddle positions if saddle point anisotropy is ‘switched off’, as described above. Elastic interactions then do not affect migration energies, explaining why defect concentrations are identical to the case without elastic interactions.
For the Ag–Cu interface, concentration profiles computed without saddle point anisotropy lie between the noninteracting and fully anisotropic cases, as shown in Fig. 4a,b. Vacancy concentrations are only marginally lower than the noninteracting case (Fig. 4a), demonstrating the overriding importance of saddle point anisotropy in their behaviour. Interstitial concentrations obtained without saddle anisotropy lie approximately midway between the fully anisotropic and noninteracting cases (Fig. 4b), demonstrating that saddle point anisotropy is at least as important to their behaviour as are pΔV interactions, which are more commonly investigated.
To investigate the effect of saddle point anisotropy on saddle point energy, Fig. 5 shows mean stable and saddle energy landscapes for point defects as a function of distance to the interfaces. Although the scatter in saddle point energies can be high at a given distance to interfaces (specially for interstitials near the Ag–Cu misfit interface), the mean saddle point energies (blue lines) are always lower than the mean saddle energies computed with the isotropic approximation (orange lines). For the Ag twist GB, the mean of the energies of saddle positions explored by vacancies and interstitials using the full elastic model decreases dramatically with decreasing distance to the GB. The case of vacancies near the Ag–Cu misfit interface illustrates the same trend, showing that the decrease in migration energy is due entirely to the saddle point configuration. In the case of interstitials near the Ag–Cu misfit interface, both migration energies and thermodynamic driving forces contribute to the enhancement of sink strength. In particular, a reduction of barriers is obtained when z<1.5 nm, thus giving rise to a nonrandom walk drift towards the Ag–Cu interface. The migration energy effect is especially pronounced in the case of vacancies and for interfaces without hydrostatic stresses.
This analysis therefore shows that, in addition to the classical thermodynamic driving force, reduced migration barriers contribute significantly to sink efficiency and that these reduced barriers are unequivocally due to saddle point anisotropy. The effect of interface elastic fields on defect migration energies is similar to that found in the vicinity of isolated dislocations^{46,47,48,49}. However, elastodiffusion near interfaces—including semicoherent ones—is not reducible to elastodiffusion near dislocations because the stress fields of the former are in general composed of two contributions: one arising from infinite arrays of discrete dislocations and the other from coherency stresses. These contributions cancel perfectly in the far field, whereas in the near field the cancellation is imperfect^{22,31,32}. It is only this imperfect cancellation in the near field that affects defect diffusion near interfaces.
Figure 6 gives a more detailed view of defect concentrations at different locations in the Ag layer of the Ag–Cu interface and in the Ag twist GB. Close to these interfaces, concentrations vary as a function of location parallel to the interface plane, following the strain field pattern created by the interfaces. Indeed, the strain field creates preferential paths for defect migration, as shown by the grey trajectories in Fig. 6. These paths are in general different for interstitials and vacancies. For both the Ag–Cu interface and Ag twist GB, vacancies preferentially migrate to the dislocation lines, whereas interstitials are mostly absorbed between dislocations. This preferential, nonrandom walk drift of point defects to specific locations is responsible for the enhanced interface sink strengths.
Knowing the steadystate defect concentrations obtained by OKMC, we derive sink strengths for the two interfaces considered above. In the mean field rate theory formalism^{50}, ‘sink strengths’ quantify the ability of sinks, such as interfaces, to absorb defects. Within this formalism, the evolution equation for the average defect concentration, , follows
where G is the defect creation rate and D is bulk defect diffusivity. The second term on the right hand side is related to the loss of defects at sinks with associated sink strength, k^{2}. At steady state, the sink strength may be computed from the average concentration:
Using the average of the concentration profile computed for defect removal at interfaces in the absence of elastic interactions (see Methods for details), the interface sink strength is analytically found to be k^{2}=12/d^{2} (ref. 51). When interactions between interfaces and defects are present, we numerically determine the sink strength through equation (4), by using the average steadystate concentration obtained by OKMC simulations and the diffusion coefficient without elastic interactions. The resulting vacancy and interstitial sink strengths for both interfaces are shown in Fig. 7a–f as a function of layer thickness.
In all cases, the sink strength increases significantly when elastic interactions are taken into account. This effect is especially pronounced for thinner layers, as defects undergo elastic interactions with interfaces over a larger fraction of the layer. It is particularly strong for interstitials, whatever the interface type, and for vacancies for the twist interface. These results also confirm the importance of saddle point anisotropy: by comparing with OKMC simulations that use isotropic saddlepoint Ptensors, we see that it yields orderofmagnitude increases in sink strength, in some cases.
Another quantity of interest for radiation response is the bias factor, B, which expresses the propensity of a given sink to absorb more interstitials than vacancies. It is defined as
where and are the sink strengths for vacancies and interstitials, respectively. For example, small interstitial clusters and dislocations exhibit positive bias factors (typically between 0.01 and 0.3 (refs 52, 53)) and thus absorb more interstitials than vacancies. The preferential absorption of interstitials by biased sinks leads to an excess of remaining vacancies, which cluster and eventually aggregate into voids^{52,54}.
Bias factors for the semicoherent interfaces we studied are shown in Fig. 7g–i. Values larger than 0.2 are obtained for the fully anisotropic interaction model in the case of the Ag–Cu interface. Such interfaces would compete for interstitials with dislocations. The presence of two sinks of differing bias magnitude has been given as a possible cause for void swelling suppression in ferritic steels^{55}. Interestingly, for the Ag twist GB, the bias factor is negative, meaning that these interfaces tend to absorb more vacancies than interstitials. Similar observations have been made in ref. 56, where the bias factor for single screw dislocations is negative when using anisotropic elasticity theory and zero in the isotropic approximation. Such GBs may therefore deplete excess vacancy concentrations sufficiently to inhibit void nucleation.
Discussion
Our work demonstrates that elastic interactions between radiationinduced point defects and semicoherent interfaces lead to significant increases in interface sink strength, compared with the case with no defectinterface interactions. These conclusions are consistent with other recent simulations, which show that elastic interactions also have a significant influence on defectdislocation interactions^{39,40,56} and on mutual interactions between radiationinduced dislocation loops^{57,58,59}.
Our simulations identify nonhydrostatic interactions of point defects in saddle point configurations as the main cause of enhanced interface sink strengths. How might one incorporate such interactions into continuumlevel diffusion models? Compared with diffusion in the absence of stress fields, the effect of local variations in defect groundstate and saddle point energies is twofold^{44,60}: the diffusivity gains a spatial dependence and the continuum diffusion equation gains a drift term. These modifications are sufficient to model defects that undergo purely hydrostatic, pΔV interactions with elastic fields. When nonhydrostatic interactions are considered, local diffusivities are, in general, no longer isotropic and the scalar diffusivity must be replaced by a diffusivity tensor. The effect of saddle point anisotropy on vacancy diffusion may be modelled this way.
Modelling the effect of saddle point anisotropy on the diffusion of 〈100〉split dumbbell selfinterstitials furthermore requires tracking the orientation of the interstitials. Therefore, three separate interstitials concentrations (corresponding to the three different dumbbell orientations) must be defined^{44}. As these interstitials reorient during each migration step (see Methods for details), the evolution of their concentrations is described by three coupled diffusion equations, each with a different locationdependent diffusivity tensor. It may also be necessary to account for thermally activated interstitial reorientation without migration^{61}. Posing and solving a continuum model for such a complex process is likely to remain beyond the scope of most continuum modelling studies in the near future. Thus, OKMC simulations are likely to remain the workhorse method for investigating the effect of elastic interactions on point defect diffusion, especially of selfinterstitials.
Our simulations show that the Ag twist GB and the Ag–Cu interface have markedly different sink strengths, even though both have identical dislocation densities. These differences arise from unlike detailed strain distributions, which are due to the different dislocation characters in these interfaces and the unlike coherency strains in the reference states of each interface. The orientations of the adjacent crystals in the twist GB are related through a pure rotation, so the GB has no coherency strains. By contrast, in the Ag–Cu pure misfit interface, unequal stresses of opposite sign are needed to impose coherency^{22,31,32}.
The sensitivity of interface sink strength to interface elastic fields opens up new opportunities for materials design. For instance, ‘supersink’ interfaces with maximal sink strengths may be created by optimizing interface elastic fields. Interface elastic fields may also be designed to yield a desired bias factor. The dependence of interface elastic fields on interface crystallography and the physical properties of the adjoining crystals are wellestablished^{18,22}. Strategies for scanning the interface design space and locating optimal interface crystallographies and compositions have also been developed^{18,62}.
An advantage of the OKMCbased simulation method presented here is that it may be systematically improved to enhance its accuracy. For example, our simulations did not take into account the interactions of defects with each other. One extension of the model would therefore be to include defect reactions, such as vacancyinterstitial recombination or clustering of like defects^{63}. Another simplifying assumption of the current simulations is that they consider only the firstorder term in the interaction energy between interfaces and defects^{44}. An improved model might account for other interactions, such as higher order moments of the multipole expansion^{37} or heterogeneity interactions^{64}. The internal structure of interfaces may change because of loading of point defects^{15}. Our computational method may be further improved by modelling this structure evolution, for example, following the approach of Uberuaga et al.^{65}. Finally, all of our simulations assumed perfect trapping of defects that arrive at an interface (see Methods for details), which corresponds to an interface sink efficiency of unity^{66}. This assumption may be relaxed to model the sink efficiency of interfaces that are not perfect defect sinks.
Methods
Interface elastic field calculation
We compute the complete elastic strain fields ɛ^{int} of interface dislocation arrays using the method described in ref. 22. The geometry of the interface dislocation pattern is found by solving the quantized Frank–Bilby equation^{33}. Interface dislocations are viewed as Volterra dislocations that have been inserted into a single crystal, coherent reference interface. The complete interface strain is the superposition of the coherency strain ɛ^{coh} in the reference interface and the strain field of the interface dislocations ɛ^{int}:
ɛ^{dis} is represented using a Fourier expansion on the right hand side of the expression above. The dislocation strain field must satisfy mechanical equilibrium in both of the adjacent crystals:
where the fourthorder anisotropic elasticity tensor. Substituting equation (6) into equation (7), the dislocation strain field is obtained by solving the sextic eigenvalue problem developed by Stroh^{67} with specific boundary conditions dedicated to interface dislocations:

No net farfield strains.

Consistency of farfield rotations with the prescribed interface misorientation.

No net tractions along the interface.

Interface displacement discontinuity matches the disregistry of the desired dislocation pattern.
Elastic dipole tensor calculation
Defect Ptensors are calculated using VASP (the Vienna Ab initio Simulation Package^{68}), a plane wavebased, first principles DFT code. A facecentred cubic supercell containing 256±1 atoms (+1 and −1 for interstitial and vacancy, respectively) is used. We also performed LAMMPS (Largescale Atomic/Molecular Massively Parallel Simulator^{69}) classical potential simulations using embedded atom method potentials for Ag (ref. 70) and Cu (ref. 71) to study the convergence of the elastic dipole tensors up to supercell sizes of 2,048 atoms. We find that the discrepancy in Ptensor components between the 256atom supercell and that of 2,048atom supercell is lower than 4%. This supercell size ensures the convergence of defect formation energies to within few meV, as detailed in the Supplementary Note and Supplementary Figs 1 and 2. We therefore view the 256atom DFT simulations as well converged with respect to model size.
A 3 × 3 × 3 shifted Monkhorst–Pack kpoint grid mesh, a Hermite–Gaussian broadening of 0.25 eV (ref. 72) and a plane wave cutoff energy of 400 eV are used. The change of the elastic dipole tensors is less than 0.5% compared with tighter settings. We use the Perdew–Burke–Ernzerhof^{73} exchangecorrelation functional within the projectoraugmentedwave approach^{74}. The structures are internally relaxed with a force convergence criterion of 10^{−3} eV Å^{−1}. The climbing imagenudged elastic band method^{42} is employed to find the saddle points for defect migration.
In the migration of a vacancy, one of the atoms directly neighbouring the vacancy travels to the vacant site, leaving behind a new vacancy. The more complex migration mechanism of a 〈100〉split dumbbell interstitial is shown in Fig. 8. Here, one atom (B) in the initial dumbbell configuration (A−B, Fig. 8a) migrates to a neighbouring lattice site (C), forming a new 〈100〉 dumbbell (B−C, Fig. 8c). Thus, three atoms (A, B and C) are involved in this migration mechanism. The initial and final dumbbells, as well as the transition path, are confined to the same lattice plane. The initial and final dumbbell orientations are orthogonal to each other.
OKMC algorithm
We model defect diffusion using an OKMC code with a residence time algorithm to advance the simulation clock^{27,28}. At time t, the time step is chosen according to Δt=−(ln r_{1})/w_{tot}, where r_{1} is a random number with r_{1}∈]0,1] and w_{tot} is the sum of frequencies of all events that may occur at t. The chosen event j is such that , where r_{2} is another random number with r_{2}∈]0,1].
Three kinds of events are considered in the simulations: the jump of a point defect from one stable point to a neighbouring one, the absorption of a defect by an interface and the creation of a new point defect through irradiation. Jump frequencies are given by w_{i}=νexp(−ΔE_{i}/(kT)), where ν is an attempt frequency and is the energy difference between the saddle position and the initial stable position of the jump considered. The stable point energy is
whereas the saddle point energy is
with E^{m} the migration energy in the absence of elastic interactions.
Here, P^{sta} and P^{sad} are the defect Ptensors in the groundstate and saddle point configurations, respectively. For simplicity, the position of the saddle point is taken midway between the two stable points explored by the jump^{40}.
The defect is considered to have been absorbed by an interface if it reaches the nearest atomic row to the interface. It is then simply removed from the simulation. This absorption condition is used to obtain a first estimate of sink strength, without taking into account the diffusion of point defects along interfaces or their possible reemission. The irradiation rate is fixed at the beginning of each simulation to keep the average number of point defects equal to 200 in the material where the measurements are performed, if no elastic interactions are considered. The actual number of point defects in the system, averaged over the simulation time when steady state is reached, constitutes the basis for our sink strength calculation.
The concentration of defects is recorded every 10^{4} iterations, after the concentration has become stationary. At the end of the simulation, an estimate of is computed by averaging over the values C_{j} (j=1,…,n):
The final time is adjusted to obtain sufficient accuracy on and thus on k^{2}. For this purpose, the estimation of the error on the concentration is given by the standard error of the mean value, that is,
where
The final time for each system is chosen so that the relative error on and k^{2} is less than 0.5%.
Analytical solution in the absence of elastic interactions
When there are no defect–interface interactions, the steadystate concentration in a flat crystalline layer may be found analytically. We consider a layer of thickness d with interfaces at z=−d/2 and z=d/2, where zero concentration Dirichlet conditions are imposed. Defects are created by radiation at a constant rate, G (per atomic site and per second) and diffuse with a diffusion coefficient D. Solving the steadystate diffusion equation leads to
from which the average concentration per atomic site is readily deduced:
Taking into account admissible jumps for a 〈100〉split dumbbell interstitial in a facecentred cubic lattice, this interstitial diffusion coefficient is
whereas in the case of vacancies it is
Additional information
How to cite this article: Vattré, A. et al. Nonrandom walk diffusion enhances the sink strength of semicoherent interfaces. Nat. Commun. 7:10424 doi: 10.1038/ncomms10424 (2016).
References
Zinkle, S. J. & Was, G. S. Materials challenges in nuclear energy. Acta Mater. 61, 735–758 (2013).
Zinkle, S. J. & Snead, L. L. Designing radiation resistance in materials for fusion energy. Annu. Rev. Mater. Res. 44, 241–267 (2014).
Averback, R. S. & Diaz de la Rubia, T. in Solid State Physics: Advances in Research and Applications eds Ehrenreich H., Spaepen F. Academic Press Inc (1998).
Bacon, D. J. & Osetsky, Y. N. Modelling atomic scale radiation damage processes and effects in metals. Int. Mater. Rev. 47, 233–241 (2002).
Singh, B. N. Effect of grain size on void formation during highenergy electron irradiation of austenitic stainless steel. Philos. Mag. 29, 25–42 (1974).
Demkowicz, M. J., Misra, A. & Caro, A. The role of interface structure in controlling high helium concentrations. Curr. Opin. Solid State Mat. Sci. 16, 101–108 (2012).
Han, W. Z. et al. Design of radiation tolerant materials via interface engineering. Adv. Mater. 25, 6975–6979 (2013).
Odette, G. R., Alinger, M. J. & Wirth, B. D. Recent developments in irradiationresistant steels. Ann. Rev. Mater. Res. 38, 471–503 (2008).
Demkowicz, M. J., Bellon, P. & Wirth, B. D. Atomicscale design of radiationtolerant nanocomposites. MRS Bull. 35, 992–998 (2010).
Beyerlein, I. J. et al. Radiation damage tolerant nanomaterials. Mater. Today 16, 443–449 (2013).
Chen, Y. et al. Damagetolerant nanotwinned metals with nanovoids under radiation environments. Nat. Commun. 6, 7036 (2015).
Han, W. Z., Fu, E. G., Demkowicz, M. J., Wang, Y. Q. & Misra, A. Irradiation damage of single crystal, coarsegrained, and nanograined copper under helium bombardment at 450°C. J. Mater. Res. 28, 2763–2770 (2013).
Demkowicz, M. J., Anderoglu, O., Zhang, X. & Misra, A. The influence of Σ3 twin boundaries on the formation of radiationinduced defect clusters in nanotwinned Cu. J. Mater. Res. 26, 1666 (2011).
Yu, K. Y. et al. Removal of stackingfault tetrahedra by twin boundaries in nanotwinned metals. Nat. Commun. 4, 1377 (2013).
Bai, X. M., Voter, A. F., Hoagland, R. G., Nastasi, M. & Uberuaga, B. P. Efficient annealing of radiation damage near grain boundaries via interstitial emission. Science 327, 1631–1634 (2010).
Bai, X. M. et al. Role of atomic structure on grain boundarydefect interactions in Cu. Phys. Rev. B 85, 214103 (2012).
Yu, W. S. & Demkowicz, M. J. Noncoherent Cu grain boundaries driven by continuous vacancy loading. J. Mater. Sci. 50, 4047–4065 (2015).
Vattré, A. J., Abdolrahim, N., Kolluri, K. & Demkowicz, M. J. Computational design of patterned interfaces using reduced order models. Sci. Rep. 4, 6231 (2014).
Demkowicz, M. J., Wang, J. & Hoagland, R. G. in Dislocations in Solids (ed. Hirth J. P. Elsevier (2008).
Demkowicz, M. J., Hoagland, R. G. & Hirth, J. P. Interface structure and radiation damage resistance in CuNb multilayer nanocomposites. Phys. Rev. Lett. 100, 136102 (2008).
Shao, S., Wang, J., Misra, A. & Hoagland, R. G. Spiral patterns of dislocations at nodes in (111) semicoherent FCC interfaces. Sci. Rep. 3, 2448 (2013).
Vattré, A. J. & Demkowicz, M. J. Determining the Burgers vectors and elastic strain energies of interface dislocation arrays using anisotropic elasticity theory. Acta Mater. 61, 5172–5187 (2013).
Hirth, J. P., Pond, R. C., Hoagland, R. G., Liu, X. Y. & Wang, J. Interface defects, reference spaces and the FrankBilby equation. Prog. Mater. Sci. 58, 749–823 (2013).
Balluffi, R. W. & Granato, A. V. in Dislocations in Solids ed. Nabarro F. R. N. NorthHolland Pub Co (1979).
King, A. H. & Smith, D. A. Calculations of sink strength and bias for pointdefect absorption by dislocations in arrays. Radiat. Eff. 54, 169–176 (1981).
Jiang, C., Swaminathan, N., Deng, J., Morgan, D. & Szlufarska, I. Effect of grain boundary stresses on sink strength. Mater. Res. Lett. 2, 100–106 (2013).
Bortz, A. B., Kalos, M. H. & Lebowitz, J. L. A new algorithm for Monte Carlo simulation of Ising spin systems. J. Comput. Phys. 17, 10–18 (1975).
Gillespie, D. T. A general method for numerically simulating the stochastic time evolution of coupled chemical reactions. J. Comput. Phys. 22, 403–434 (1976).
Caturla, M. J. et al. Comparative study of radiation damage accumulation in Cu and Fe. J. Nucl. Mater. 276, 13–21 (2000).
Jourdan, T., Bocquet, J. L. & Soisson, F. Modeling homogeneous precipitation with an eventbased Monte Carlo method: application to the case of FeCu. Acta Mater. 58, 3295–3302 (2010).
Vattré, A. J. & Demkowicz, M. J. Effect of interface dislocation Burgers vectors on elastic fields in anisotropic bicrystals. Comput. Mater. Sci. 88, 110–115 (2014).
Vattré, A. J. & Demkowicz, M. J. Partitioning of elastic distortions at a semicoherent heterophase interface between anisotropic crystals. Acta Mater. 82, 234–243 (2015).
Sutton, A. P. & Balluffi, R. W. Interfaces in Crystalline Materials Oxford Univ. Press (1995).
Gray, D. E. American Institute of Physics Handbook McGrawHill (1957).
Hirth, J. P. & Lothe, J. Theory of Dislocations 2nd edn Wiley (1982).
Kanzaki, H. Point defects in facecentered cubic latticeI distortion around defects. J. Phys. Chem. Solids 2, 24–36 (1957).
Siems, R. Mechanical interactions of point defects. Phys. Status Solidi 30, 645–658 (1968).
Dederichs, P. H., Lehmann, C., Schober, H. R., Scholz, A. & Zeller, R. Latticetype of pointdefects. J. Nucl. Mater. 6970, 176–199 (1978).
Sivak, A. B., Chernov, V. M., Romanov, V. A. & Sivak, P. A. Kinetic MonteCarlo simulation of selfpoint defect diffusion in dislocation elastic fields in bcc iron and vanadium. J. Nucl. Mater. 417, 1067–1070 (2011).
Subramanian, G., Perez, D., Uberuaga, B. P., Tome, C. N. & Voter, A. F. Method to account for arbitrary strains in kinetic Monte Carlo simulations. Phys. Rev. B 87, 144107 (2013).
Ehrhart P., Jung P., Schultz H., Ullmaier H. in. LandöltBornstein (ed. Ullmaier H. SpringerVerlag (1991).
Henkelman, G., Uberuaga, B. P. & Jonsson, H. A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys. 113, 9901 (2000).
Haubold, H. G. & Martinsen, D. Structure determination of self interstitials and investigation of vacancy clustering in copper by diffuse Xray scattering. J. Nucl. Mater. 6970, 644–649 (1978).
Wolfer, W. G. in Comprehensive Nuclear Materials Elsevier (2012).
Nazarov, R., Hickel, T. & Neugebauer, J. Vacancy formation energies in fcc metals: Influence of exchangecorrelation functionals and correction schemes. Phys. Rev. B 85, 144118 (2012).
Dederich, P. H. & Schroeder, K. Anisotropic diffusion in stress fields. Phys. Rev. B 17, 2524 (1978).
Koehler, J. S. Diffusion of lattice defects in stress field. Phys. Rev. 181, 1015 (1969).
Chen, I. W. Anisotropic diffusion of point defects to edge dislocations. J. Nucl. Mater. 125, 52–63 (1984).
Borodin, V. A. & Ryazanov, A. I. The effect of diffusion anisotropy on dislocation bias and irradiation creep in cubic lattice materials. J. Nucl. Mater. 210, 258–267 (1994).
Brailsford, A. D. & Bullough, R. The theory of sink strengths. Phil. Trans. R. Soc. A 302, 87–137 (1981).
Bullough, R., Hayns, M. R. & Wood, M. H. Sink strengths for thin film surfaces and grain boundaries. J. Nucl. Mater. 90, 44–59 (1979).
Bullough, R. & Perrin, R. C. The mechanism and kinetics of void growth during neutron irradiation. ASTM Special Technical Publications 484, 317–331 (1970).
Heald, P. T. & Speight, M. V. Steadystate irradiation creep. Philos. Mag. 29, 1075–1080 (1974).
Mansur, L. K. Void swelling in metals and alloys under irradiation: an assessment of the theory. Nucl. Technol. 40, 5–34 (1978).
Little, E. A., Bullough, R. & Wood, M. H. On the swelling resistance of ferritic steel. Proc. R. Soc. A 372, 565–579 (1980).
Sivak, A. B. & Sivak, P. A. Efficiency of dislocations as sinks of radiation defects in fcc copper crystal. Crystallogr. Rep. 59, 407–414 (2014).
Mason, D. R., Yi, X., Kirk, M. A. & Dudarev, S. L. Elastic trapping of dislocation loops in cascades in ionirradiated tungsten foils. J. Phys. Condens. Matter. 26, 375701 (2014).
Derlet, P. M., Gilbert, M. R. & Dudarev, S. L. Simulating dislocation loop internal dynamics and collective diffusion using stochastic differential equations. Phys. Rev. B 84, 134109 (2011).
Dudarev, S. L. et al. Langevin model for realtime Brownian dynamics of interacting nanodefects in irradiated metals. Phys. Rev. B 81, 224107 (2010).
Schroeder, K. & Dettmann, K. Diffusion reactions in longrange potentials. Z. Phys. B 22, 343–350 (1975).
Derlet, P. M., NguyenManh, D. & Dudarev, S. L. Multiscale modeling of crowdion and vacancy defects in bodycenteredcubic transition metals. Phys. Rev. B 76, 054107 (2007).
Yuryev, D. V. & Demkowicz, M. J. Computational design of solidstate interfaces using Olattice theory: An application to mitigating heliuminduced damage. Appl. Phys. Lett. 105, 221601 (2014).
Was, G. S. Fundamentals of Radiation Materials Science: Metals and Alloys Springer (2007).
Eshelby, J. D. The continuum theory of lattice defects. Solid State Phys. Adv. Res. Appl. 3, 79–144 (1956).
Uberuaga, B. P., Vernon, L. J., Martinez, E. & Voter, A. F. The relationship between grain boundary structure, defect mobility, and grain boundary sink efficiency. Sci. Rep. 5, 9095 (2015).
Beyerlein, I. J., Demkowicz, M. J., Misra, A. & Uberuaga, B. P. Defectinterface interactions. Prog. Mater. Sci. 74, 125–210 (2015).
Stroh, A. N. Dislocations and cracks in anisotropic elasticity. Philos. Mag. 3, 625–646 (1958).
Kresse, G. & Furthmuller, J. Efficient iterative schemes for ab initio totalenergy calculations using a planewave basis set. Phys. Rev. B 54, 11169 (1996).
Plimpton, S. Fast parallel algorithms for shortrange moleculardynamics. J. Comput. Phys. 117, 1–19 (1995).
Foiles, S. M., Baskes, M. I. & Daw, M. S. Embeddedatommethod functions for the FCC metals Cu, Ag, Au, Ni, Pd, Pt, and their alloys. Phys. Rev. B 33, 7983 (1986).
Mishin, Y., Mehl, M. J., Papaconstantopoulos, D. A., Voter, A. F. & Kress, J. D. Structural stability and lattice defects in copper: Ab initio, tightbinding, and embeddedatom calculations. Phys. Rev. B 63, 224106 (2001).
Methfessel, M. & Paxton, A. T. Highprecision sampling for Brillouinzone integration in metals. Phys. Rev. B 40, 3616 (1989).
Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865 (1996).
Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmentedwave method. Phys. Rev. B 59, 1758 (1999).
Acknowledgements
H.D. acknowledges support from the DOE, Office of Nuclear Energy, Nuclear Energy Enabling Technologies, Reactor Materials program, under contract No. DENE0000533. H.D. also thanks the DOENERSC for computational support. M.C.M. acknowledges support from the GENCI(CINES/CCRT/IDRIS) computer centre under Grant No. x2015096973 and from PRACE (Partnership for Advanced Computing in Europe) for access to the resource MareNostrum III at the Barcelona supercomputing centre (TRANSOM project). M.J.D. acknowledges support from the National Science Foundation under Grant No. 1150862.
Author information
Authors and Affiliations
Contributions
A.V. carried out elasticity calculations of stress fields and interaction energies. T.J. performed OKMC simulations. H.D. and M.C.M. computed defect Ptensors using DFT simulations. M.J.D. supervised the manuscript. All the authors contributed to data interpretation and to the manuscript preparation and approved the final version of the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Supplementary information
Supplementary Information
Supplementary Figures 1 2, Supplementary Note 1 and Supplementary References. (PDF 460 kb)
Rights and permissions
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
About this article
Cite this article
Vattré, A., Jourdan, T., Ding, H. et al. Nonrandom walk diffusion enhances the sink strength of semicoherent interfaces. Nat Commun 7, 10424 (2016). https://doi.org/10.1038/ncomms10424
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/ncomms10424
This article is cited by

Direct imaging of the disconnection climb mediated point defects absorption by a grain boundary
Nature Communications (2022)

Antagonist effects of grain boundaries between the trapping process and the fast diffusion path in nickel bicrystals
Scientific Reports (2021)

Segregation energy of the hydrogen at Ni Σ3 grain boundaries: some implications of the atomic volume and the interstitial selfstress
Journal of Materials Science (2018)

The Multiple Roles of SmallAngle Tilt Grain Boundaries in Annihilating Radiation Damage in SiC
Scientific Reports (2017)

Effects of Helium Ion Irradiation on Properties of Crystalline and Amorphous Multiphase Ceramic Coatings
Journal of Materials Engineering and Performance (2017)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.