This work presents a novel mechanistic analysis workflow by which machine-learned force fields (MLFFs) are used to accelerate interpretation of experimental results using theoretical and experimental extended X-ray adsorption fine structure spectra (EXAFS). The MLFFs are built using Bayesian active-learning, leading to accelerated time to solution for MLFF creation, as opposed to traditional <i>ab inito</i> workflows. Following validation against experimental and theoretical benchmarks, the MLFFs are employed to study the dynamics of bulk and nano-structure phenomena of Pt. The nanoparticles systems are studied under reactive atmospheres, which illustrates the power of the MLFF method to accurately capture reactions and the dynamic response of nanoparticles to these environments. Both simple and complex chemical systems are considered, demonstrating the ability of these MLFFs to interpolate across diverse atomic environments and to yield direct insight into the atomistic mechanisms. Important observations are also made regarding the choice of first-principles data that is used to train the MLFFs and how the resulting dynamics differ at finite temperature. These in-depth comparisons of different DFT training-sets are completed using bulk Pt EXAFS, and molecular dynamics (MD) simulations of periodic bulk cells. Only after sufficient trust is generated in the method, models built for NP systems are then considered and compared to available EXAFS data. Ultimately, simultaneous atomistic study of complex interfacial chemistry and structural evolution of the Pt-nanostructures is possible using this workflow, while retaining quantum-mechanical accuracy. Reaction mechanisms are uncovered as well as morphological evolution of the particles. This workflow allows for drastic acceleration of the creation of MLFFs from first-principles data and the subsequent study of atomistic mechanisms at quantum-mechanical accuracy, leading to expedited interpretation of experimental data at the atomistic level.

Linking machine-learned force-fields with EXAFS to understand dynamic nanomaterials under reactive atmospheres

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.