Semi-supervised generative approach to chemical disorder: application to point-defect formation in uranium-plutonium mixed oxides.

Chemical disorder has a major impact on the characterization of the atomic-scale properties of highly complex chemical compounds, such as the properties of point defects. Due to the vast amount of possible atomic configurations, the study of such properties becomes intractable if treated with direct sampling. In this work, we propose an alternative approach, in which samples are selected based on the local atomic composition around the defect, and the defect formation energy is obtained as a function of this local composition with a reduced computational cost. We apply this approach to (U, Pu)O2 nuclear fuels. The formation-energy distribution is computed using machine-learning generative methods, and used to investigate the impact of chemical disorder and the range of influence of local composition on the defect properties. The predicted distributions are then used to calculate the concentration of thermal defects. This approach allows for the first time for the computation of the latter property with a physically meaningful exploration of the configuration space, and opens the way to a more efficient determination of physico-chemical properties in other chemically-disordered compounds such as high-entropy alloys.

Capabilities and limits of autoencoders for extracting collective variables in atomistic materials science.

Free energy calculations in materials science are routinely hindered by the need to provide reaction coordinates that can meaningfully partition atomic configuration space, a prerequisite for most enhanced sampling approaches. Recent studies on molecular systems have highlighted the possibility of constructing appropriate collective variables directly from atomic motions through deep learning techniques. Here we extend this class of approaches to condensed matter problems, for which we encode the finite temperature collective variable by an iterative procedure starting from 0 K features of the energy landscape i.e. activation events or migration mechanisms given by a minimum - saddle point - minimum sequence. We employ the autoencoder neural networks in order to build a scalar collective variable for use with the adaptive biasing force method. Particular attention is given to design choices required for application to crystalline systems with defects, including the filtering of thermal motions which otherwise dominate the autoencoder input. The machine-learning workflow is tested on body-centered cubic iron and its common defects, such as small vacancy or self-interstitial clusters and screw dislocations. For localized defects, excellent collective variables as well as derivatives, necessary for free energy sampling, are systematically obtained. However, the approach has a limited accuracy when dealing with reaction coordinates that include atomic displacements of a magnitude comparable to thermal motions, e.g. the ones produced by the long-range elastic field of dislocations. We then combine the extraction of collective variables by autoencoders with an adaptive biasing force free energy method based on Bayesian inference. Using a vacancy migration as an example, we demonstrate the performance of coupling these two approaches for simultaneous discovery of reaction coordinates and free energy sampling in systems with localized defects.

Role of Sink Density in Nonequilibrium Chemical Redistribution in Alloys.

Nonequilibrium chemical redistribution in open systems submitted to external forces, such as particle irradiation, leads to changes in the structural properties of the material, potentially driving the system to failure. Such redistribution is controlled by the complex interplay between the production of point defects, atomic transport rates, and the sink character of the microstructure. In this work, we analyze this interplay by means of a kinetic Monte Carlo (KMC) framework with an underlying atomistic model for the Fe-Cr model alloy to study the effect of ideal defect sinks on Cr concentration profiles, with a particular focus on the role of interface density. We observe that the amount of segregation decreases linearly with decreasing interface spacing. Within the framework of the thermodynamics of irreversible processes, a general analytical model is derived and assessed against the KMC simulations to elucidate the structure-property relationship of this system. Interestingly, in the kinetic regime where elimination of point defects at sinks is dominant over bulk recombination, the solute segregation does not directly depend on the dose rate but only on the density of sinks. This model provides new insight into the design of microstructures that mitigate chemical redistribution and improve radiation tolerance.

Equilibrium Vacancy Concentration Driven by Undetectable Impurities.

By means of ab initio calculations combined to statistical mechanics, we provide new evidence that an experimentally undetectable tiny amount of impurities can be responsible for drastic changes in vacancy concentrations ([V]), inducing large deviations from an Arrhenius law even at low temperature. It is the case of O and N in α-Fe. The present finding is fully compatible with existing experiments, and changes the previous common vision that C has the dominant effect. This study provides a route for bridging the longstanding theoretical-experimental gap on the prediction of [V] in metals.

A self-consistent mean field theory for diffusion in alloys.

Starting from a microscopic model of the atomic transport via vacancies and interstitials in alloys, a self-consistent mean field (SCMF) kinetic theory yields the phenomenological coefficients Lij. In this theory, kinetic correlations are accounted for through a set of effective interactions within a non-equilibrium distribution function of the system. The introduction of a master equation describing the evolution with time of the distribution function and its moments leads to general self-consistent kinetic equations. The Lij of a face centered cubic alloy are calculated using the kinetic equations of Nastar (M. Nastar, Philos. Mag., 2005, 85, 3767, ref. 1) derived from a microscopic broken bond model of the vacancy jump frequency. A first approximation leads to an analytical expression of the Lij and a second approximation to a better agreement with the Monte Carlo simulations. A change of sign of the Lij is studied as a function of the microscopic parameters of the jump frequency. The Lij of a cubic centered alloy obtained for the complex diffusion mechanism of the dumbbell configuration of the interstitial (V. Barbe and M. Nastar, Philos. Mag., 2006, in press, ref. 2) are used to study the effect of an on-site rotation of the dumbbell on the transport.

Complex precipitation pathways in multicomponent alloys.

One usual way to strengthen a metal is to add alloying elements and to control the size and the density of the precipitates obtained. However, precipitation in multicomponent alloys can take complex pathways depending on the relative diffusivity of solute atoms and on the relative driving forces involved. In Al-Zr-Sc alloys, atomic simulations based on first-principle calculations combined with various complementary experimental approaches working at different scales reveal a strongly inhomogeneous structure of the precipitates: owing to the much faster diffusivity of Sc compared with Zr in the solid solution, and to the absence of Zr and Sc diffusion inside the precipitates, the precipitate core is mostly Sc-rich, whereas the external shell is Zr-rich. This explains previous observations of an enhanced nucleation rate in Al-Zr-Sc alloys compared with binary Al-Sc alloys, along with much higher resistance to Ostwald ripening, two features of the utmost importance in the field of light high-strength materials.