Deep-neural-network molecular dynamics investigation of phonon thermal transport in polyether ether ketone

Abstract

Polyether ether ketone (PEEK) is an important high-performance engineering thermoplastic, yet the thermal transport properties of its crystalline and single-chain forms remain elusive. In this work, a deep neural network interatomic potential is trained using ab initio molecular dynamics to accurately model thermal transport in bulk crystalline, bulk amorphous, and single-chain PEEK. Additionally, phonon thermal transport across chains, which are grouped together through van der Waals (vdW) interactions, exhibits a weak dependence of thermal conductivity ($\kappa$) on the number of chains, i.e., weakly ballistic transport in cross-chain directions. This behavior contrasts with many layered materials bonded by vdW interactions, which often show a strong dependence of cross-plane $\kappa$ on the number of layers. This work facilitates the understanding of thermal transport properties of PEEK and phonon transport in vdW-bonded materials in general, offering a theoretical guideline for predicting optimal conditions for PEEK processing and beyond.