top of page
Katherine Hollingsworth

A Dual-cutoff Machine-learned Potential for Condensed Organic Systems Obtained via Uncertainty-guided Active Learning

Updated: Nov 19


As a result of a collaboration between scientists at BP and Materials Design, we are thrilled to announce that an article by authors Leonid Kahle, Benoit Minisini, Tai Bui, Jeremy T. First, Corneliu Buda, Thomas Goldman, and Erich Wimmer was

published in Physical Chemistry Chemical Physics in a few weeks! Titled "A Dual-cutoff Machine-learned Potential for Condensed Organic Systems Obtained via Uncertainty-guided Active Learning," this significant work explores the innovative application of machine-learned potentials (MLPs) to improve the accuracy and efficiency of predicting the behavior of organic compounds.


this significant work explores the innovative application of machine-learned potentials (MLPs) to improve the accuracy and efficiency of predicting the behavior of organic compounds.

We train a machine-learned potential (MLP) using a dual-cutoff architecture to capture interactions at different length scales, employing active learning to identify which new configurations should be added to the training set. After final hyperparameter optimization, the MLP is used to calculate physical properties, achieving good accuracy in densities, heat capacity, and other properties compared to experimental and electronic structure theory results.



Abstract: Machine-learned potentials (MLPs) trained on ab initio data combine the computational efficiency of classical interatomic potentials with the accuracy and generality of the first-principles method used in the creation of the respective training set. In this work, we implement and train a MLP to obtain an accurate description of the potential energy surface and property predictions for organic compounds, as both single molecules and in the condensed phase. We devise a dual descriptor, based on the atomic cluster expansion (ACE), that couples an information-rich short-range description with a coarser long-range description that captures weak intermolecular interactions. We employ uncertainty-guided active learning for the training set generation, creating a dataset that is comparatively small for the breadth of application and consists of alcohols, alkanes, and an adipate. Utilizing that MLP, we calculate densities of those systems of varying chain lengths as a function of temperature, obtaining a discrepancy of less than 4% compared with experiment. Vibrational frequencies calculated with the MLP have a root mean square error of less than 1 THz compared to DFT. The heat capacities of condensed systems are within 11% of experimental findings, which is strong evidence that the dual descriptor provides an accurate framework for the prediction of both short-range intramolecular and long-range intermolecular interactions.


For more details, you can access the article here.


This work not only highlights the potential of MLPs in materials science but also establishes a robust framework for predicting both short-range intramolecular and long-range intermolecular interactions. As we continue to explore the implications of these findings, the Materials Design team remains committed to advancing computational materials science and enhancing our understanding of organic compounds.






15 views

Commentaires


bottom of page