Fortnet, a software package for training Behler-Parrinello neural networks

TL;DR

Fortnet is an open-source software package that facilitates the training and evaluation of Behler-Parrinello neural networks, enabling improved correction functions in density functional tight binding methods with enhanced transferability.

Contribution

This work introduces Fortnet, a comprehensive software tool for neural network training in atomic simulations, and demonstrates its application to improve DFTB correction functions.

Findings

Fortnet successfully trains neural networks for atomic potential correction functions.

The DFTB+ANN approach enhances transferability across diverse chemical environments.

Neural network corrections improve the accuracy of DFTB simulations.

Abstract

A new, open source, parallel, stand-alone software package (Fortnet) has been developed, which implements Behler-Parrinello neural networks. It covers the entire workflow from feature generation to the evaluation of generated potentials, coupled with higher-level analysis such as the analytic calculation of atomic forces. The functionality of the software package is demonstrated by driving the training for the fitted correction functions of the density functional tight binding (DFTB) method, which are commonly used to compensate the inaccuracies resulting from the DFTB approximations to the Kohn-Sham Hamiltonian. The usual two-body form of those correction functions limits the transferability of the parameterizations between very different structural environments. The recently introduced DFTB+ANN approach strives to lift these limitations by combining DFTB with a near-sighted artificial neural network (ANN). After investigating various approaches, we have found the combination of DFTB with an ANN acting on-top of some baseline correction functions (delta learning) the most promising one. It allowed to introduce many-body corrections on top of two-body parametrizations, while excellent transferability to chemical environments with deviating energetics could be demonstrated.