Related Experiment Video
Updated: Jun 28, 2025
07:12
A Standard and Reliable Method to Fabricate Two-Dimensional Nanoelectronics
Published on: August 28, 2018
9.6K
Machine Learning to Promote Efficient Screening of Low-Contact Electrode for 2D Semiconductor Transistor Under
Penghui Li1,2, Linpeng Dong1,2, Chong Li3
1Shaanxi Province Key Laboratory of Thin Films Technology and Optical Test, Xi'an Technological University, Xi'an, 710032, China.
Advanced Materials (Deerfield Beach, Fla.)
|April 12, 2024
Summary
Machine learning efficiently screens low-barrier electrode materials for 2D semiconductor devices using limited data. This approach accelerates the development of high-performance transistors by reducing trial-and-error.
Area of Science:
- Materials Science
- Semiconductor Physics
- Machine Learning
Background:
- High-performance 2D semiconductor devices require low-barrier, high-injection electrodes.
- Traditional electrode material screening methods are inefficient and arbitrary.
- Machine learning (ML) offers a solution but requires substantial data.
Purpose of the Study:
- To develop an ML scheme for screening low-contact electrode materials for 2D transistors with limited data.
- To overcome the data requirements hindering practical ML application in semiconductor device development.
Main Methods:
- A combined scheme of an autoencoding regularized adversarial neural network and a feature-adaptive variational active learning algorithm was proposed.
- The model was trained using only 15% of the available data points.
- Density-of-states descriptors were employed for continual improvement and physical insight.
Main Results:
- Achieved mean square errors of 0.17 eV (vertical Schottky barrier) and 0.27 eV (lateral Schottky barrier).
- Achieved 2.88% accuracy for tunneling probability prediction.
- Demonstrated optimal predictive performance across various training dataset sizes.
- Empirical evaluations and MOSFET device construction verified the transport characteristics.
Conclusions:
- The proposed ML scheme effectively screens electrode materials for 2D semiconductor devices with limited data.
- This approach significantly accelerates the development of high-performance 2D semiconductor devices.
- The findings highlight the potential of ML in materials discovery for advanced electronic devices.
