Plasma-engineered 2D semiconductor heterostructure enabling photonic neuromorphic and 5-bit reservoir computing

Two-dimensional (2D) layered materials have gained significant attention as candidates for field-effect transistors (FETs) in next-generation electronics and ultrafast memory applications for artificial intelligence (AI) systems. In this study, we present a reliable and efficient approach to fabricate a photonic synaptic transistor for neuromorphic and reservoir computing, achieved by utilizing oxygen plasma treatment on the bottom side of hafnium disulfide (HfS 2 ). The fabricated device exhibits n-type semiconducting characteristics and a pronounced clockwise hysteresis in its transfer curves, which is attributed to intentionally engineered defect-induced traps within the HfO x layer. Notably, the device exhibits a substantial memory window of approximately 114 V during a gate voltage sweep of ± 100 V, and the memory window can be dynamically tuned from 4 V to 114 V by adjusting the gate voltage sweep range between ± 10 V and ± 100 V. Furthermore, the device exhibits low energy consumption of ~ 2.25 nJ per optical spike, demonstrating its suitability for low-power neuromorphic systems. Additionally, the device demonstrates core synaptic behaviors including paired-pulse facilitation (PPF), excitatory postsynaptic current (EPSC), optical spike number dependent plasticity (SNDP), and spike time dependent plasticity (STDP) under ultraviolet illumination (365 nm). The device further achieves 32 distinct output states, highlighting its suitability for 5-bit reservoir computing systems. These results provide an innovative approach for advancing AI-based hardware and efficient computing platforms.