Orbital-interaction-aware deep learning model for efficient surface chemistry simulations

TL;DR

This paper introduces DOTA, a deep learning model that accurately predicts adsorption energies in surface chemistry by leveraging orbital interaction patterns, thus reducing dependence on scarce high-precision data and enabling efficient materials screening.

Contribution

The paper presents a novel DOS Transformer model that aligns experimental and quantum data for surface chemistry, achieving high accuracy with limited high-precision training data.

Findings

Achieves chemical accuracy in adsorption energy predictions

Bridges the gap between experimental and quantum data

Enables efficient high-throughput materials screening

Abstract

Deep learning has advanced efficient chemical process simulations on the surfaces, accelerating high-throughput materials screening and rational design in heterogeneous catalysis, energy storage and conversion, and gas separation. However, the accuracy of the deep learning model generally depends on the quality of the training data. Unfortunately, precise experimental data in surface chemistry, such as adsorption energies, are scarce, while accurate quantum chemistry simulations remain computationally prohibitive for large-scale studies. Herein, we present a deep learning model of DOS Transformer for Adsorption (DOTA) for efficient surface chemistry simulations with chemical accuracy. It enables the alignment of experimental data and multi-fidelity quantum chemistry calculation data by capturing latent orbital interaction patterns based on the map between local density of states (LDOS) and adsorption energy. This minimizes the reliance on scarce high-precision training data in surface chemistry to accomplish efficient prediction of adsorption energies rivaling the high-precision experimental data, resolving the long-standing challenge of "CO puzzle". It provides a robust framework for efficient materials screening, effectively bridging the gap between computational and experimental data.