In recent years, two-dimensional (2D) materials-based RRAMs have gained high importance because of their thermal and mechanical stability, and better potentiation-depression controllability. 2D materials based conductive bridge random access memory (CBRAM) has been considered as promising approach for neuromorphic and image processing technology ^[1]. Despite much progress in CMOS technology, the growth and deposition technology of 2D materials for semiconductor integrated circuit are much complex and is generally available at wafer scale ^[2]. In addition, high growth temperature for high quality of 2D materials complicates direct wafer growth and makes transfer process desirable. At the device level, challenges are linked to controlled and uniform growth of 2D material for high density electronic structure.

Recently, discreet 2D based memristor have been used in crossbar structure as synapse for neuromorphic computing. However, the plasma-assisted chemical vapor reaction (PACVR) based memristor for neuromorphic application are rarely demonstrated. Here, we report the co-integration of plasma-assisted chemical vapor reaction (PACVR) with silicon CMOS technology to provide brain-inspired computing device. PACVR offers compatibility with temperature limited 3D integration process and also provides much better thickness control over a large area. Furthermore, it an easy platform for direct and controlled synthesis of TMDs compared to conventional CVD approach. The PACVR grown WSe₂ layer (~2 nm) on silicon substrate is realized, which exhibits both threshold and bipolar switching. The threshold and bipolar switching emulate integrate-fire neuron function and is obtained by modulating the compliance current in the device. The dynamics of the switching is closely related to the diffusive dynamics of the active metal (Ag or Cu) which can be controlled by device current. As a result, the WSe₂/Si memristor shows synaptic behavior for neuromorphic system with learning accuracy of 96%.

References:

1. Wang, C.-Y. et al. 2D layered materials for memristive and neuromorphic applications. Electron. Mater. 6, 1901107 (2020)
2. Zhang, X. et al. Two-dimensional MoS₂-enabled flexible rectenna for Wi-Fi-band wireless energy harvesting. Nature566, 368–372 (2019).