Short-range $\Delta$-Machine Learning: A cost-efficient strategy to transfer chemical accuracy to condensed phase systems

TL;DR

This paper introduces short-range Δ-ML, a cost-effective machine learning approach that enhances the accuracy of condensed phase system simulations by building on existing periodic MLPs to replicate high-level method observables.

Contribution

It presents a novel short-range Δ-ML strategy that improves the transfer of chemical accuracy to condensed phase systems efficiently.

Findings

srΔML accurately reproduces high-level method observables

Cost-efficient approach for condensed phase simulations

Builds on periodic MLPs for enhanced accuracy

Abstract

DFT-based machine-learning potentials (MLPs) are now routinely trained for condensed-phase systems, but surpassing DFT accuracy remains challenging due to the cost or unavailability of periodic reference calculations. Our previous work (PRL 2022, 129, 226001) demonstrated that high-accuracy periodic MLPs can be trained within the CCMD framework using extended yet finite reference calculations. Here, we introduce short-range $\Delta$-Machine Learning (sr$\Delta$ML), which builds on periodic MLPs while accurately reproducing the observables of the high-level method.