Rapid mapping of alloy surface phase diagrams via Bayesian evolutionary multitasking

Surface phase diagrams(SPDs)are essential for understanding the dependence of surface chemistry on reaction condition.For multi-component systems such as metal alloys,the derivation of such diagrams often relies on separate first-principles global optimization tasks under different reaction conditions.Here we show that this can be significantly accelerated by leveraging the fact that all tasks essentially share a unified configurational search space,and only a single expensive electronic structure calculation is required to evaluate the stabilities of a surface structure under all considered reaction conditions.As a general solution,we propose a Bayesian evolutionary multitasking(BEM)framework combining Bayesian statistics with evolutionary multitasking,which allows efficient mapping of SPDs even for very complex surface systems.As proofs of concept,we showcase the performance of our methods in deriving the alloy SPDs for two heterogeneous catalytic systems:the electrochemical oxygen reduction reaction(ORR)and the gas phase steam methane reforming(SMR)reaction.

Genetic algorithms for computational materials discovery accelerated by machine learning

Materials discovery is increasingly being impelled by machine learning methods that rely on pre-existing datasets.Where datasets are lacking,unbiased data generation can be achieved with genetic algorithms.Here a machine learning model is trained on-the-fly as a computationally inexpensive energy predictor before analyzing how to augment convergence in genetic algorithm-based approaches by using the model as a surrogate.This leads to a machine learning accelerated genetic algorithm combining robust qualities of the genetic algorithm with rapid machine learning.The approach is used to search for stable,compositionally variant,geometrically similar nanoparticle alloys to illustrate its capability for accelerated materials discovery,e.g.,nanoalloy catalysts.The machine learning accelerated approach,in this case,yields a 50-fold reduction in the number of required energy calculations compared to a traditional“brute force”genetic algorithm.This makes searching through the space of all homotops and compositions of a binary alloy particle in a given structure feasible,using density functional theory calculations.

Unfolding the structural stability of nanoalloys via symmetry-constrained genetic algorithm and neural network potential

The structural stability of nanoalloys is a challenging research subject due to the complexity of size,shape,composition,and chemical ordering.The genetic algorithm is a popular global optimization method that can efficiently search for the ground-state nanoalloy structure.However,the algorithm suffers from three significant limitations:the efficiency and accuracy of the energy evaluator and the algorithm’s efficiency.Here we describe the construction of a neural network potential intended for rapid and accurate energy predictions of Pt-Ni nanoalloys of various sizes,shapes,and compositions.We further introduce a symmetry-constrained genetic algorithm that significantly improves the efficiency and viability of the algorithm for realistic size nanoalloys.The combination of the two allows us to explore the space of homotops and compositions of Pt-Ni nanoalloys consisting of up to 4033 atoms and quantitatively report the interplay of shape,size,and composition on the dominant chemical ordering patterns.

Structural and chemical mechanisms governing stability of inorganic Janus nanotubes

One-dimensional inorganic nanotubes hold promise for technological applications due to their distinct physical/chemical properties,but so far advancements have been hampered by difficulties in producing single-wall nanotubes with a well-defined radius.In this work we investigate,based on Density Functional Theory(DFT),the formation mechanism of 135 different inorganic nanotubes formed by the intrinsic self-rolling driving force found in asymmetric 2D Janus sheets.We show that for isovalent Janus sheets,the lattice mismatch between inner and outer atomic layers is the driving force behind the nanotube formation,while in the non-isovalent case it is governed by the difference in chemical bond strength of the inner and outer layer leading to steric effects.From our pool of candidate structures we have identified more than 100 tubes with a preferred radius below 35Å,which we hypothesize can display distinctive properties compared to their parent 2D monolayers.Simple descriptors have been identified to accelerate the discovery of small-radius tubes and a Bayesian regression approach has been implemented to assess the uncertainty in our predictions on the radius.

Computational framework for a systematic investigation of anionic redox process in Li-rich compounds

Anionic redox processes play a key role in determining the accessible capacity and cycle life of Li-rich cathode materials for batteries.We present a framework for investigating the anionic redox processes based on data readily available from standard DFT calculations.Our recipe includes a method of classifying different anionic species,counting the number of species present in the structure and a preconditioning scheme to promote anionic redox.