Polarization-driven multimodal synapses with homologous regulation for emotion-vision mapping

Multimodal synapses with hybrid signal processing mechanisms are promising to enable complex functions of neural computing. However, the absent homology of physical mechanisms for bionic-coupled mapping behavior hinders multimodal synaptic application. Here, based on polarization dynamics, we explore a homeotic mechanism to achieve the multimodal mapping of emotion simulation and visual bionics in BaTiO3-based photo-ferroelectric synapses. Synaptic plasticity for multi-emotional evolution arises from weight-featured resistance switching in various ferroelectric memristors through the polarization-modulated interfacial barrier. Based on the regulation homology of polarization dynamics for the bulk photovoltaic effect, the optical synaptic responses of visual bionics are coupled into a precise emotion mapping with arousal-valence dimensions. These multimodal photo-ferroelectric synapses complete precise digit-pattern recognition within an artificial neural network system and possess rectifications against crosstalk and superior high thermal stability. Circuit simulations of optoelectronic mapping response display the application potential of multimodal synapses as smart decoupling components for balanced ternary computing. Our results provide a paradigm for designing multimodal devices with synergistic and coupled multi-synaptic functions.