Twinned Locally Active Memristor-Based Chaotic Neuron

Optimized design of artificial neurons is crucial for advancing neuromorphic computing. This work presents a chaotic neuronal circuit that incorporates twinned locally active memristors and a capacitor, enhancing neuron integration density. The complexity of the circuit is attributed to the local activity of the memristors, which also facilitates the generation of chaotic neuromorphic signals. The DC V–I curve reveals three locally active domains in the memristor, establishing the biasing conditions necessary for neuronal operation. Using the small-signal method, the s-domain and frequency-domain responses are analyzed, providing essential insights into the external components and parameter ranges required for the neuron implementation. The basin of attraction demonstrates coexisting behaviors occurring under unstable locally active biasing currents. Quantitative validation of chaotic behavior is achieved using Lyapunov exponents and dynamical maps. The designed neuron exhibits diverse neuromorphic behaviors, including periodic spiking, chaotic bursting, all-or-none responses, spike adaptation, and integrator-like dynamics. Finally, circuit simulations based on the proposed design confirm the feasibility of the memristive neuron, highlighting its potential for practical applications in neuromorphic systems.