High-performance ternary logic circuits and neural networks based on carbon nanotube source-gating transistors

Xuehao Zhu; Meiqi Xi; Jianyu Wang; Panpan Zhang; Yi Li; Xiao Luo; Lan Bai; Xingxing Chen; Lian-Mao Peng; Yu Cao; Qiliang Li; Xuelei Liang

doi:10.1126/sciadv.adt1909

High-performance ternary logic circuits and neural networks based on carbon nanotube source-gating transistors

Sci Adv. 2025 Jan 10;11(2):eadt1909. doi: 10.1126/sciadv.adt1909. Epub 2025 Jan 10.

Authors

Xuehao Zhu¹, Meiqi Xi^{1

2}, Jianyu Wang³, Panpan Zhang⁴, Yi Li^{1

5}, Xiao Luo¹, Lan Bai^{1

2}, Xingxing Chen^{1

2}, Lian-Mao Peng^{1

2

6}, Yu Cao^{1

2

6}, Qiliang Li^{2

3}, Xuelei Liang^{1

2

6}

Affiliations

¹ Key Laboratory for the Physics and Chemistry of Nanodevices and Center for Carbon-Based Electronics, School of Electronics, Peking University, Beijing 100871, China.
² Institute for Carbon-Based Thin Film Electronics, Peking University, Shanxi (ICTFE-PKU), Taiyuan 030012, China.
³ Department of Advanced Manufacturing and Robotics, College of Engineering, Peking University, Beijing 100871, China.
⁴ State Key Laboratory of Information Photonics and Optical Communications, Beijing University of Post and Telecommunications, Beijing 100876, China.
⁵ Academy for Advanced Interdisciplinary Studies, Peking University, Beijing 100871, China.
⁶ Institute of Advanced Functional Materials and Devices, Shanxi University, Taiyuan 030006, China.

PMID: 39792666
DOI: 10.1126/sciadv.adt1909

Abstract

Multi-valued logics (MVLs) offer higher information density, reduced circuit and interconnect complexity, lower power dissipation, and faster speed over conventional binary logic system. Recent advancement in MVL research, particularly with emerging low-dimensional materials, suggests that breakthroughs may be imminent if multistates transistors can be fabricated controllably for large-scale integration. Here, a concept of source-gating transistors (SGTs) is developed and realized using carbon nanotubes (CNTs). By extending the source electrode into the channel of conventional CNT transistors, a controllable p-n homojunction is formed, allowing CNT-SGTs to reliably switch between three distinct states. Capitalizing on the straightforward fabrication process of CNT-SGTs, ternary inverters, NMIN and NMAX logic gates, ternary SRAM cells, and a ternary neural network achieving 100% image classification accuracy have been successfully implemented. This study represents the most advanced and highest-performing ternary circuits realized with low-dimensional materials to date. This progress highlights the potential of CNT-SGTs in driving the future of MVL architectures.