Predicting the Thermal Conductivity Collapse in SWCNT Bundles: The Interplay of Symmetry Breaking and Scattering Revealed by Machine-Learning-Driven Quantum Transport

TL;DR

This study uses machine learning-enhanced quantum transport methods to understand how symmetry breaking and new scattering channels drastically reduce thermal conductivity in SWCNT bundles, aligning theory with experiments.

Contribution

It introduces a combined ML and ALD-BTE framework to accurately model and predict thermal transport in SWCNTs and their bundles, highlighting symmetry effects.

Findings

Symmetry breaking enhances phonon scattering in isolated SWCNTs.

Emergence of inter-tube phonon modes increases scattering channels.

Quantum Bose-Einstein statistics are crucial for accurate modeling.

Abstract

We combine machine learning (ML)-based neuroevolution potentials (NEP) with anharmonic lattice dynamics and the Boltzmann transport equation (ALD-BTE) to achieve a quantitative and mode-resolved description of thermal transport in individual (10, 0) zigzag single-walled carbon nanotubes (SWCNTs) and their bundles. Our analysis reveals a dual mechanism behind the drastic suppression of thermal conductivity in bundles: first, the breaking of rotational symmetry in isolated SWCNTs dramatically enhances the scattering rates of symmetry-sensitive phonon modes, such as the twist (TW) mode. Second, the emergence of new inter-tube phonon modes introduces abundant additional scattering channels across the entire frequency spectrum. Crucially, the incorporation of quantum Bose-Einstein (BE) statistics is essential to accurately capture these phenomena, enabling our approach to quantitatively reproduce experimental observations. This work establishes the combination of ML-driven interatomic potentials and ALD-BTE as a predictive framework for nanoscale thermal transport, effectively bridging the gap between theoretical models and experimental measurements.