Neuromuscular Junction Inspired Mechanosensitive hBN-based Memristors for Flexible Neuromorphics

Mechanosensitive neurons in neuromuscular junctions transduce bending into graded changes in synaptic strength, a biological design principle that artificial neuromorphic hardware has yet to exploit. Here, we demonstrate two-terminal mechanosensitive memristor with a vertical Ag/hBN/ITO architecture on flexible PET substrate, where its bending directly and reversibly programs synaptic plasticity through a non-destructive mechanism. Mechanical deformation drives interlayer sliding of the hBN flakes, perturbing the stacking and the inter-flake tunnelling distance without straining lattice. This programs synaptic plasticity, quantitatively mirroring the amplitude and dynamics co-modulation like in biological mechanosensitive synapses. Conductance states under varying bending conditions also serve as analog synaptic weights in a fully connected neural network trained on the MNIST digit-recognition benchmark, achieving more than 91% classification accuracy across all bending conditions. These results establish interlayer-sliding mediated tunnelling as a new mechanoelectronic design principle for flexible neuromorphic devices whose synaptic state adapts physical interactions with the environment.