OrbitAll: A Unified Quantum Mechanical Representation Deep Learning Framework for All Molecular Systems

TL;DR

OrbitAll is a physics-informed deep learning framework that accurately models all types of molecular systems, including charged and open-shell molecules, with significantly less data and much faster than traditional quantum chemistry methods.

Contribution

It introduces a unified, geometry- and physics-informed deep learning approach capable of representing any molecular system with electronic structure information.

Findings

Achieves chemical accuracy with 10x less training data

Outperforms existing models on charged and open-shell molecules

Provides 1000-10,000x speedup over density functional theory

Abstract

Despite the success of deep learning methods in quantum chemistry, their representational capacity is most often confined to neutral, closed-shell molecules. However, real-world chemical systems often exhibit complex characteristics, including varying charges, spins, and environments. We introduce OrbitAll, a geometry- and physics-informed deep learning framework that can represent all molecular systems with electronic structure information. OrbitAll utilizes spin-polarized orbital features from the underlying quantum mechanical method, and combines it with graph neural networks satisfying SE(3)-equivariance. The resulting framework can represent and process any molecular system with arbitrary charges, spins, and environmental effects. OrbitAll demonstrates superior performance and generalization on predicting charged, open-shell, and solvated molecules, while also robustly extrapolating to molecules significantly larger than the training data by leveraging a physics-informed architecture. OrbitAll achieves chemical accuracy using 10 times fewer training data than competing AI models, with a speedup of approximately $10^3$ - $10^4$ compared to density functional theory.