A Multimodal Humidity Adaptive Optical Neuron Based on a MoWS2/VOx Heterojunction for Vision and Respiratory Functions

Advancements in computing have progressed from near-sensor to in-sensor computing, culminating in the development of multimodal in-memory computing, which enables faster, energy-efficient data processing by performing computations directly within the memory devices. A bio-inspired multimodal in-memory computing system capable of performing real-time low power processing of multisensory signals, lowering data conversion and transmission across several modules in conventional chips is introduced. A novel Cu/MoWS₂/VO_x/Pt based multimodal memristor is characterized by an ON/OFF ratio as high as 10⁸ with consistent and ultralow operating voltages of ±0.2 surpassing conventional single-mode memory functions. Apart from observing electrical synaptic behavior, photonic depression and humidity mediated optical synaptic learning is also demonstrated. The heterojunction with MoWS₂ also enables reconfigurable modulation in both memory and optical synaptic functionalities with changing humidity. This behavior provides tunable conductance modulation capabilities emulating synaptic transmission in biological neurons while showing potential in respiratory detection module for healthcare application. The humidity sensing capability is implemented to demonstrate vision clarity using a convolutional neural network (CNN), with different humidity levels applied as a data augmentation preprocessing method. This proposed multimodal functionality represents a novel platform for developing artificial sensory neurons, with significant implications for non-contact human–computer interaction in intelligent systems.