Neuromorphic computing based on two-dimensional materials

Neuromorphic computing systems seek to emulate the efficient and adaptive information processing of the brain through hardware that mimics synaptic plasticity. Traditional approaches rely on electrically driven memristors or phase-change materials, but these often suffer from high energy consumption and limited bi-directional tunability. Optical control offers a promising alternative, enabling parallel, low-power, and high-speed modulation of synaptic weights[1,2]. Two-dimensional semiconductors such as MoS₂ exhibit strong excitonic effects, layer-dependent bandgaps, and high carrier mobility, making them ideal platforms for optoelectronic synapses[3]. Under illumination, photogenerated carriers can modulate channel conductivity, with the sign of photoresponse often dependent on gate voltage due to competing mechanisms such as photogating, trap-assisted recombination, and band alignment shifts. While several studies have reported optical potentiation in MoS₂ devices, reliable optical depression remains elusive, limiting the implementation of fully optical neural networks capable of bidirectional learning[4,5]. This research proposes a systematic investigation of gate-dependent photoresponse inversion in MoS₂ FETs, with extended exploration of hBN encapsulation and heterostack engineering[6]. By controlling device architecture and environmental conditions, we aim to uncover the physical mechanisms underlying photoconductivity sign reversal and demonstrate functional neuromorphic behaviors with optical inputs[7].