Machine-learning-assisted density functional theory for predicting transparent conducting materials
Phys. Rev. Applied 25, 014071 – Published 29 January, 2026
DOI: https://doi.org/10.1103/br17-vmg2
Abstract
Despite the success of n-type transparent conducting oxides (TCOs), there is a notable scarcity of new p-type TCOs and transparent conducting materials (TCMs) in optoelectronics. We utilized 10 333 Heyd-Scuseria-Ernzerh (HSE) band gap data points for transfer learning to enhance the Perdew-Burke-Ernzerhof (PBE) band gap model, resulting in a reliable HSE model that aligns closely with 640 calculated values. We screened 518 materials with band gaps of 3.0 eV to 4.3 eV from 5276 compounds in the Materials Project, identifying two potential p-type (nine potential n-type) transparent conductors via high-throughput density functional theory. For n-type TCMs, cations of , , , , and possess a closed-shell electronic configuration, with a forbidden dipole transition between the valence band maximum (VBM) and the conduction band minimum (CBM). Furthermore, p-type TCMs and feature anions with delocalized p orbitals that hybridize with cations’ configurations, leading to a low hole effective mass. Moreover, was identified as a potential p-type material based on intrinsic point defect properties analyzed using the HSE06 hybrid functional. This work expands the database of inorganic transparent conductors, necessitating further validation.