Biological leaky integrate-and-fire (LIF) neurons are dynamic, living computational systems that modulate signal strength via synaptic plasticity and generate action potentials through somatic firing. To emulate LIF functionality, a memristor capable of reversible analog (synaptic function) and threshold switching (somatic function) states is essential. Here, we present a reconfigurable hydrogel-based memristor that can be configured to emulate either synaptic or somatic functions by tuning its composition. Within the flexible hydrogel matrix, polyvinyl alcohol (PVA) content governs the dispersion state of silver nanoflakes (Ag NF), enabling distinct Ag conductive pathways. In low PVA content, the memristor exhibits analog characteristics; in high PVA content, the memristor shows threshold switching characteristics with a low activation voltage of 0.56 V. Furthermore, by integrating the analog memristor, threshold switching memristor, capacitor, and resistor, we construct an artificial LIF neuron that dynamically adjusts firing probability based on historical stimuli. This system achieves 95.46% accuracy in image classification, closely mimicking biological neuron behavior and advancing hardware for artificial neural networks.
Lan et al. (2026) studied this question.