Wafer-Scale Monolayer MoS2 with Tunable Grain Size via Grain Boundary Engineering for Neuromorphic Computing

Wafer-Scale Monolayer MoS2 with Tunable Grain Size via Grain Boundary Engineering for Neuromorphic Computing

ACS Nano

10.1021/acsnano.5c17643

https://doi.org/10.1021/acsnano.5c17643

Mingxi Chen, Xufan Li, Yongli He, Haofei Zheng, Zheren Cai, Xin Ju, Musi Zeng, Shi Wun Tong, Jing Wu, Lingfei Wang, Kah-Wee Ang, Dongzhi Chi

Electrical and Electronic Engineering, General Materials Science

24 November 2025

Chen, M., Li, X., He, Y., Zheng, H., Cai, Z., Ju, X., Zeng, M., Tong, S. W., Wu, J., Wang, L., Ang, K.-W., & Chi, D. (2025). Wafer-Scale Monolayer MoS2 with Tunable Grain Size via Grain Boundary Engineering for Neuromorphic Computing. ACS Nano, 19(48), 41407–41417. https://doi.org/10.1021/acsnano.5c17643

Grain boundaries (GBs) in two-dimensional (2D) materials, once regarded as detrimental defects, are now increasingly recognized as functional features for tailoring material properties. Here, we report a grain-boundary engineering strategy based on a diameter-tunable chemical vapor deposition (DT-CVD) technique, enabling the wafer-scale growth of monolayer MoS2 with controlled grain sizes. Systematic tuning of GB density is achieved by adjusting the inner quartz tube diameter while maintaining uniform monolayer morphology and crystallinity. Structural and spectroscopic characterizations confirm that increasing GB density introduces shallow energy barriers and band bending effects, which are exploited in memtransistor architectures to induce tunable analog switching. Simulations and electrical measurements show that grain-boundary-induced charge trapping and ionic migration synergistically govern the observed resistive switching behavior. Among all devices, the grain size-tunable MoS2 memtransistor with an equivalent grain size of ∼114 ± 48 nm achieves an optimal trade-off between analog precision and cycling endurance. It supports 64-level conductance modulation along with robust long-term synaptic plasticity. When implemented in neural networks, it delivers high training and inference accuracy across diverse data sets, underscoring its promise for next-generation energy-efficient neuromorphic systems.

Publisher Copyright

This research / project is supported by the National Research Foundation - Competitive Research Programme - In-memory Computing based on Multi-terminal Memtransistors for Cognitive Internet-of-Things (C-IoT)
Grant Reference no. : NRF-CRP24-2020-0002

This document is the Accepted Manuscript version of a Published Work that appeared in final form in ACS Nano, copyright © American Chemical Society after peer review and technical editing by the publisher. To access the final edited and published work see https://doi.org/10.1021/acsnano.5c17643

https://oar.a-star.edu.sg/communities-collections/articles/23073

1936-0851
1936-086X

Institute of Materials Research and Engineering