Multi-gate organic dendritic transistors for spatiotemporal BCM learning

The Bienenstock–Cooper–Munro (BCM) learning rule effectively mitigates the issue of synaptic weight oversaturation in Hebbian learning via a sliding threshold dependent on historical neuronal activity. However, existing hardware implementations focus primarily on temporal history while often overlooking the spatial attenuation pivotal for signal integration. Here, we report a bio-inspired multi-gate organic dendritic transistor based on poly(3-hexylthiophene) with electrochemical gating that integrates spatiotemporal information to implement the BCM rule via its spatial architecture. By tuning the physical distances between different gates and the channel, the device mimics the nonlinear signal attenuation mechanism of biological dendrites. Beyond reproducing fundamental synaptic plasticity, the device establishes a sliding threshold mechanism for homeostatic regulation by using physical distances to modulate historical activity states. This work demonstrates the feasibility of defining synaptic historical states using spatial factors, providing a strategy for neuromorphic computing systems that combine homeostatic regulation with spatiotemporal processing.