Bottom-Up Prediction of Amorphous Poly(1,3-dioxolane) using Ab Initio Reactive Machine-Learning Force Fields

Abstract

Poly(1,3-dioxolane) (pDXL) is a chemically recyclable polyether that can be synthesized by living cationic ring-opening polymerization with precise control of chain length into the ultra-high-molecular-weight (UHMW) regime, where it acquires enhanced mechanical properties. Like most polymers, however, it has amorphous condensed-phase structure, and diffraction measurements alone cannot resolve the atomistic microstructure that underlies them. In this work, we predict the ensemble of the pDXL microstructures by applying a recently developed AI-accelerated ab initio bottom-up polymer structure prediction (AI$^{2}$-BPSP) framework that simulates the living polymerization of pDXL under experimental synthetic conditions with machine-learning force fields trained within van der Waals-corrected hybrid density functional theory. Partitioning the predicted X-ray and neutron diffraction signal between the growing chain and the surrounding monomer resolves its dominant feature into two counteracting modes: a monomer contribution near $q \approx 1.5$ Å$^{-1}$ that decays as monomer is consumed, and a polymer contribution near $q \approx 1.6$ Å$^{-1}$ that is absent in oligomers and grows with chain length. This work paves the way to direct accuracy validation of these statistically significant, first-principles, and chain-length resolved diffraction predictions against experiments.