Accelerating explicit solvent models of heterogeneous catalysts with machine learning interatomic potentials

Abstract

Realistically modelling how solvents affect catalytic reactions is a longstanding challenge due to the prohibitive computational cost. Typically, an explicit atomistic treatment of the solvent molecules is needed together with...

Redefining the Stability of Water Oxidation Electrocatalysts: Insights from Materials Databases and Machine Learning

“…65,66 MLIPs have been used to study complex catalytic scenarios, such as the nucleation and crystallization of Pt nanoparticles 67 and the solvation of adsorbates at metal-water interfaces. 68 We believe that MLIPs could be extremely valuable understanding the dynamics of TM X-ide catalysts and encourage further research in this area. However, it is important to note that the accuracy and robustness of MLIPs depend on the quality of the training data, so researchers must carefully select materials and reaction environments.…”

“…Recent theoretical advances in atomic simulations assisted by ML methods offer promising avenues for understanding complex catalytic systems. , For instance, machine learning interatomic potentials (MLIPs) can parametrize the potential energy surface of atomic systems based on local environment descriptors, achieving accuracy comparable to ab initio methods but at speeds orders of magnitude faster. , Electronic-structure data from DFT data sets can train MLIPs to predict energies and forces for larger ensembles of atoms, thereby making more complex structural problems computationally manageable. , MLIPs have been used to study complex catalytic scenarios, such as the nucleation and crystallization of Pt nanoparticles and the solvation of adsorbates at metal-water interfaces . We believe that MLIPs could be extremely valuable for understanding the dynamics of TM X-ide catalysts and encourage further research in this area.…”

Solvate Suite: A Command-Line Interface for Molecular Simulations and Multiscale Microsolvation Modeling

“…From theoretical and computational standpoints, a crucial aspect of the microsolvation treatment lies in the arrangement of solvent molecules around the solute. This process can be carried out manually − or through automated methods, − which are often employed specifically in cases where water serves as the solvent molecule.…”

Enhancing the Quality and Reliability of Machine Learning Interatomic Potentials through Better Reporting Practices

“…Recently, machine learning interatomic potentials (MLIPs) that are trained on DFT data have emerged as a promising method to expand the time and length scale of atomistic simulations while potentially maintaining DFT-like accuracy. , Many ready-to-use software packages − are now available for prospective practitioners to download, install, and use within hours. The relative ease of using such packages has generated considerable interest, resulting in a rapid increase in scientific applications such as accelerating molecular dynamics, ,, probing chemical reactivity, − and investigating long-time-scale and length-scale phase transitions. − Some MLIPs may now be extended to provide estimates on other properties such as dipole moments, charge distribution, the density of states, , magnetic moments of atoms, , and the local electron density around an atom …”