Berlin 2024 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

HL: Fachverband Halbleiterphysik

HL 36: Poster III

HL 36.42: Poster

Mittwoch, 20. März 2024, 18:00–20:30, Poster E

Thermal Conductivities of Crystalline Polymers calculated with Machine-Learned Potentials — •Lukas Reicht, Lukas Legenstein, Sandro Wieser, and Egbert Zojer — Institute of Solid State Physics, Graz University of Technology, Graz, Austria

Disordered polymers are typically characterized by a very low thermal conductivity on the order of 0.1 W/mK. In contrast, recent experiments showed that, when polymers are highly aligned (crystalline), polyethylene (PE) can reach a thermal conductivity of ~104 W/mK, which would be interesting for applications. Newly developed machine-learned potentials (MLP) promise to be an efficient and accurate tool for calculating these thermal conductivities. Applying a new methodology, however, requires a thorough benchmarking. We performed such a benchmarking for moment tensor potentials (MTPs), which are a flavour of machine-learned potential, by calculating various phonon related properties of polyethylene (PE), polythiophene (PT), and poly(3-hexyl-thiophene) (P3HT). Based on the calculated phonon band-structures, elastic constants, thermal expansion coefficients, and thermal conductivities, we conclude that the accuracy of MTPs can be substantially increased by a deliberate choice of training data adapted to the intended use case. Having established the accuracy of the trained MTPs, they are used to calculate thermal conductivities of PE and PT using the Boltzmann transport equation (BTE), non-equilibrium molecular dynamics (NEMD), and the approach-to-equilibrium molecular dynamics (AEMD). This provides complementary atomistic insights into the factors determining heat transport.

Keywords: Machine-learned potential; Thermal conductivity; Heat transport; Polymer; Moment tensor potential