Device-level modelling for predicting the total density of states of single-walled CNTs with increasing chirality: a fusion of ab initio modeling and a machine learning framework

Abstract

The electronic properties of single-walled carbon nanotubes (SWCNTs) are highly sensitive to their chirality, influencing their potential applications in nanoelectronics and energy storage. This study presents a novel approach for predicting the distribution of the total density of states (TDOS) in SWCNTs as a function of chirality, integrating ab initio modeling with machine learning techniques. First-principles calculations based on density functional theory (DFT) are employed to establish a comprehensive dataset of TDOS values across various chiral indices. Machine learning models, trained on this dataset, are then utilized to generalize and predict trends in the electronic behavior of previously computed chirality configurations. The integration of computational physics with artificial intelligence enables a more efficient exploration of the electronic structure of SWCNTs, significantly reducing computational costs while maintaining high predictive accuracy. The proposed framework enhances the understanding of chirality-dependent electronic properties and paves the way for the tailored design of carbon-based nanomaterials for advanced technological applications. In this study, we simulated the electronic properties of carbon nanotubes (CNTs) with chirality of (n,m) (here, n = 4, 5, 6, …, 10; m = 0). Our first-principles simulations predicted that SWCNT systems with (n = 4, 5, 6; m = 0) chirality have a metallic character. The metallicities of the (4,0), (5,0), and (6,0) systems are due to the strong σ*- and π*-mixing caused by the large curvature of the tube. In contrast, in the SWCNT (n = 4, 5, 6; m = 0) systems, the SWCNT compounds with (n = 7, 8, 9, 10; m = 0) chirality demonstrate semiconducting characteristics with narrow band gaps of 0.10–0.82 eV, and we conclude that these systems are direct band gap materials.