MAISE: Construction of neural network interatomic models and evolutionary structure optimization

Abstract

MAISE is an open-source package for materials modeling and prediction. The code's main feature is an automated generation of neural network (NN) interatomic potentials for use in global structure searches. The systematic construction of Behler-Parrinello-type NN models approximating ab initio energy and forces relies on two approaches introduced in our recent studies. An evolutionary sampling scheme for generating reference structures improves the NNs' mapping of regions visited in unconstrained searches, while a stratified training approach enables the creation of standardized NN models for multiple elements. A more flexible NN architecture proposed here expands the applicability of the stratified scheme for an arbitrary number of elements. The full workflow in the NN development is managed with a customizable 'MAISE-NET' wrapper written in Python. The global structure optimization capability in MAISE is based on an evolutionary algorithm applicable for nanoparticles, films, and bulk crystals. A multitribe extension of the algorithm allows for an efficient simultaneous optimization of nanoparticles in a given size range. Implemented structure analysis functions include fingerprinting with radial distribution functions and finding space groups with the SPGLIB tool. This work overviews MAISE's available features, constructed models, and confirmed predictions.

Electronic, vibrational, and electron-phonon coupling properties in SnSe$_2$ and SnS$_2$ under pressure

“…4 along with an estimate of the similarity factor (defined as the RDF dot product between two structures) provide further evidence that the H2-1 and H2-2 configurations are indeed distinct despite having similar enthalpies. In this case, using a Gaussian spread of 0.008 56 , we find a similarity factor of 0.5358 and 0.6102 between the H2-1 and H2-2 structures for the SnSe 2 and SnS 2 systems, respectively. Noticeably, the main difference in the RDFs comes from the second-nearest neighbor, explaining the similar covalent network and, therefore, the very similar vibrational and low-energy electron properties of the H2 derivatives.…”

“…We displace the atomic coordinates according to the eigenvectors of A 2g (H2-1 phase) and A 2u (H2-2 phase) modes as well as a linear combination of the two eigenvectors (H2-3 phase) to better explore the adiabatic potential energy surface. As for SnSe 2 , the resulting distorted structures for the √ 3 × √ 3 × 1 supercell are fully relaxed and found to converge to three distinct configurations based on the structure analysis performed with the MAISE package 56 . The lattice parameters, space groups, and Wyckoff positions of H2 structures at 30 GPa are given in Table . S1 and the crystal structures are shown in Fig.…”