Learning glass transition temperatures via dimensionality reduction with data from computer simulations: Polymers as the pilot case

Artem Glova,Mikko Karttunen

2024-06-29

Abstract:Machine learning (ML) methods provide advanced means for understanding inherent patterns within large and complex datasets. Here, we employ the principal component analysis (PCA) and the diffusion map (DM) techniques to evaluate the glass transition temperature ($T_\mathrm{g}$) from low-dimensional representations of all-atom molecular dynamic (MD) simulations of polylactide (PLA) and poly(3-hydroxybutyrate) (PHB). Four molecular descriptors were considered: radial distribution functions (RDFs), mean square displacements (MSDs), relative square displacements (RSDs), and dihedral angles (DAs). By applying a Gaussian Mixture Model (GMM) to analyze the PCA and DM projections, and by quantifying their log-likelihoods as a density-based metric, a distinct separation into two populations corresponding to melt and glass states was revealed. This separation enabled the $T_\mathrm{g}$ evaluation from a cooling-induced sharp increase in the overlap between log-likelihood distributions at different temperatures. $T_\mathrm{g}$ values derived from the RDF and MSD descriptors using DM closely matched the standard computer simulation-based dilatometric and dynamic $T_\mathrm{g}$ values for both PLA and PHB models. This was not the case for PCA. The DM-transformed DA and RSD data resulted in $T_\mathrm{g}$ values in agreement with experimental ones. Overall, the fusion of atomistic simulations and diffusion maps complemented with the Gaussian Mixture Models presents a promising framework for computing $T_\mathrm{g}$ and studying the glass transition in a unified way across various molecular descriptors for glass-forming materials.

Soft Condensed Matter

What problem does this paper attempt to address?

This paper mainly discusses how to use machine learning (ML) methods, especially Principal Component Analysis (PCA) and Diffusion Mapping (DM) techniques, to learn the glass transition temperature (Tg) of polymers (such as polylactic acid and poly-3-hydroxybutyrate) from computer simulation data. Four molecular descriptors are considered in the study: radial distribution functions (RDFs), mean square displacements (MSDs), relative square displacements (RSDs), and dihedral angles (DAs).

By applying Gaussian Mixture Models (GMM) to analyze PCA and DM projections, researchers found that Tg can be assessed through a sharp increase induced by cooling in the log-likelihood distribution at different temperatures. Tg values obtained from RDF and MSD descriptors in PCA and DM match well with standard thermomechanical and dynamic Tg values based on computer simulation, but PCA is inconsistent in some cases. Tg values obtained from DA and RSD data transformed by DM are consistent with experimental Tg values.

The paper points out that combining atomistic simulations, diffusion mapping, and GMM provides a promising framework for calculating Tg and studying the glass transition of various glass-forming materials in a unified manner. Although various methods have been used to determine Tg from computer simulations, the choice of appropriate molecular descriptors and dimensionality reduction techniques still affects the accuracy of the results. By comparing the projection effects of PCA and DM at different temperatures, the paper emphasizes the importance of selecting the appropriate dimensions and provides density-based metrics for determining Tg.

In the study, the authors conducted detailed molecular dynamics simulations and revealed the patterns of transition from the melt state to the glass state by comparing data changes at different temperatures. Through these analyses, they proposed a new method to estimate Tg, which contributes to a better understanding of the glass transition phenomenon and provides support for applications in materials science.