文献类型：期刊文献

英文题名：Physics-informed machine learning for predicting p-type schottky barriers

作者：Wang, Xinjie[1];Sun, Chenghua[2];Chen, Qian[1];Wang, Lixuan[1];Li, Hongbo[1];Hao, Guoqiang[1];Hu, Jisong[3,5];Zhang, Rui[4]

机构：[1]East China Univ Sci & Technol, Sch Mat Sci & Engn, Shanghai 200237, Peoples R China;[2]Swinburne Univ Technol, Dept Chem & Biotechnol, Hawthorn, Vic 3122, Australia;[3]Hanshan Normal Univ, Sch Mat Sci & Engn, Sch Chem & Environm Engn, Adv Mat & Devices Lab, Chaozhou 521041, Peoples R China;[4]Shanghai Inst Technol, Fac Chem Engn & Energy Technol, Shanghai 201418, Peoples R China;[5]Shenzhen Univ, Coll Phys & Optoelect Engn, Coll Mat Sci & Engn, Shenzhen 518060, Peoples R China

年份：2026

卷号：80

外文期刊名：SURFACES AND INTERFACES

收录：;WOS:【SCI-EXPANDED(收录号:WOS:001641838200001)】；

基金：This work was supported by the National Natural Science Foundation of China (22303031) .

语种：英文

外文关键词：Schottky barrier height; Heterostructures; Machine learning; Transition Metal Dichalcogenides

摘要：P-type Schottky barrier height (p-SBH) is critical for high-performance 2D electronics, but its modulation is particularly challenging due to complicated interfacial contact associated with lattice mismatch and electronic interactions. In this report, we reported a robust model, namely XGBoost, to predict p-SBHs after an investigation of 60 graphene/transition metal dichalcogenide MX2 (TMDC, M=Mo, W; X = S, Se, Te) heterointerfaces with a combined density functional theory (DFT) and machine learning. Specifically, XGBoost achieves an accuracy R-2 >0.95 and MAE <0.06 eV primarily due to the use of key descriptors (interlayer spacing, tunneling probability and chalcogen properties). Such physics-informed approach advances the knowledge of heterojunction structures and enhances the predictive capacity, providing insights for rational design of functional nanoelectronic interfaces.