First-principles and deep learning frameworks to predict the electronic and magnetic properties of V-doped SiC nanotube

Abstract

Abstract In this study based on Density Functional Theory (DFT) and Local Spin Density Approximation (LDA) methods within Hubbard U corrections have been theoretically studied electronic and magnetic properties of single wall silicon carbide nanotube doped by vanadium. These properties were simulated for cases that single or double silicon atoms of the SiC nanotube replaced with V atoms. Using Deep Learning (DL) Algorithms are the boon to provide prediction of quantum-confined electronic structure calculations, however first-principles simulation methods more accurate. ML based regression model shows the accuracy and prediction model for the quantum-confined nanotube. Among the various neural network algorithms, tri-layered and medium neural netwok algorithms provide more accuracy and less error rate for this molecular nanotube. The comparison between ML based approach and DFT based procedure reveals the similarity and accuracy of the proposed algorithm. The first-principles calculated energy spin-up and spin-down band gap values for single wall chiral (6,0) SiC:V nanotube systems are about of 0.6 and 1.4 eV, respectively. Although the undoped SiC system is a nonmagnetic, the V-doped SiC nanotube induces magnetism and total magnetic moment of this magnetic material equal to ~ 1.001 µB. Density of states calculations indicated that the magnetization of SiC:V single wall nanotube mainly come from the 2p orbitals of carbon atoms and 3d orbitals of V dopant. From the total energy calculations for ferromagnetic and antiferromagnetic phases for V-doped SiCNT systems obtained that the ferromagnetic phase more stable.