Molecular-dynamics-based modal analysis of heat transport in quasi-one-dimensional systems from a symmetry-adapted perspective

Abstract

Detailed analysis of thermal conductivity results obtained from molecular dynamics (MD) trajectories conventionally relies on knowledge of the harmonic vibrational modes of the system. Such is the case in methods like Green--Kubo modal analysis (GKMA) and homogeneous nonequilibrium modal analysis (HNEMA). This approach faces several shortcomings that become especially important when dealing with nanostructures. Chief among them is scalability, followed by the ambiguity associated to the degeneracy of the large vibrational subspaces defined by a common translational wave number k. We propose an alternative for quasi-one-dimensional (quasi-1D) systems: we construct the modal basis for the decomposition of the thermal conductivity from line-group projection operators, so that every projected component carries well-defined symmetry labels (including rotational information and parities) and the decomposition is unique at the level of irreducible representations (irreps). Applying the idea to a (10,0)-(20,0) WS2-MoS2 double-walled nanotube (DWNT) described by a neuroevolution potential (NEP), we find that at 300 K both HNEMA and GKMA yield statistically consistent total conductivities and allow the identification of several prominent symmetry-adapted conduction channels. The GKMA pair matrix shows that within-block and same-channel cross-k terms account for 75.6% of the total conductivity, while cross-channel correlations contribute 24.4%.