Unveiling Mechanical Properties of Single-Walled Carbon Nanotubes Using Machine-Learned Interatomic Potentials

Abstract

The prime objective of computational modeling of nanomaterials is to precisely calculate material characteristics with minimal empirical data. Density functional theory (DFT) simulations are reliable methods for investigating the mechanical/electronic characteristics of nanostructures in their ground state. However, these calculations become overly expensive when used in large structures at finite temperatures. Classical molecular dynamics (MD) simulations enable the opportunity to investigate larger structures at high temperatures but highly rely on the accuracy of interatomic potentials. In the present study, we constructed machine-learned interatomic potentials (MLIPs) based on ab-initio MD trajectories and actively trained using moment tensor potential (MTP) descriptors. The performance of the developed MLIPs is investigated in this paper using MD simulations of the mechanical/failure responses of single-walled carbon nanotubes (SWCNTs) under tensile loading. Furthermore, this work includes an extensive MD simulation study on the effect of size, chirality, and temperature on the failure properties and Young’s modulus of SWCNTs with MLIP inclusion. The findings demonstrate the computational reliability and consistency of the MD simulations, which incorporate actively trained MLIPs to predict the mechanical properties of SWCNTs at temperatures between 1 and 2000 K. The observed stiffnesses correspond to Young’s modulus ranging from 1.61–0.61 TPa for different SWCNTs with diameters ranging from 1.1–2.89 nm and exhibiting chirality dependence at 1 K. Therefore, the use of MLIPs in material modeling opened up novel avenues for the development of cutting-edge nanomaterials for energy harvesting and storage applications. © 2025 Elsevier B.V., All rights reserved.