Intrinsic and tunable lattice thermal conductivity of single- and multi-layer goldene: A machine-learning molecular dynamics study

Abstract

The recent successful fabrication of single- and multi-layer gold, referred to as goldene, represents a significant breakthrough in low-dimensional materials research. A fundamental understanding of its phonon transport properties, both intrinsic and extrinsic, is crucial for assessing its potential in applications such as thermal management, sensing, solar cells, and photothermal therapy. Here, we perform extensive machine-learning molecular dynamics simulations using an accurate and efficient unified neuroevolution potential to investigate the intrinsic lattice thermal conductivity of goldene and its tunability through external strain, defects, and layer thickness, drawing close comparisons with bulk gold. We find that single-layer goldene exhibits a high intrinsic lattice thermal conductivity of 34.4±0.4Wm−1K−1, which is approximately 17 times as large as that of bulk gold. While both systems exhibit a κ∼T−p dependence (p=1.69 for goldene, p=1.30 for bulk), they respond oppositely to tensile strain: κ decreases in bulk gold but increases in goldene. We attribute this contrasting behavior to the fundamentally different effects of strain on their phonon modes, hardening the dominant flexural phonons in goldene while softening all the phonon modes in bulk gold. Furthermore, both vacancy and adatom defects significantly reduce κ, with adatoms having a more pronounced effect. Finally, κ decreases monotonically with increasing layer number, rapidly converging toward the bulk value. Our results elucidate the intrinsic lattice thermal transport in goldene and demonstrate versatile strategies for tuning it, thereby enriching the understanding of heat transport in two-dimensional metallic systems.