Hundreds of new, stable, one-dimensional materials from a generative machine learning model

TL;DR

This paper demonstrates a generative neural network that creates thousands of new one-dimensional materials, many of which are stable and potentially useful, expanding the known materials space.

Contribution

The study introduces a neural network model that generates novel 1D materials, including entirely new classes, validated by DFT calculations and added to the C1DB database.

Findings

Over 500 new materials are dynamically stable.

New materials have heats of formation within 0.2 eV of known materials.

Generated materials include both element-substituted and entirely new classes.

Abstract

We use a generative neural network model to create thousands of new, one-dimensional materials. The model is trained using 508 stable one-dimensional materials from the Computational 1D Materials Database (C1DB) database. More than 500 of the new materials are shown with density functional theory calculations to be dynamically stable and with heats of formation within 0.2 eV of the convex hull of known materials. Some of the new materials could also have been obtained by chemical element substitution in the training materials, but completely new classes of materials are also produced. The band structures, electronic densities of states, work functions, effective masses, and phonon spectra of the new materials are calculated, and the data are added to C1DB.