We introduce a voltage-gated conductance model for an artificial neuron that exhibits tonic, fast, andtwo types of intrinsic burst spiking [1]. The spike generation is achieved with a single voltage-gatedchannel that exploits the conductance commutation properties of a two-terminal memristive device.Our circuit implementation is of unprecedented simplicity, using just four electronic components, allconventional, cheap and out-of-the-shelf [2]. Our bursting neuron is a two-compartment model, similarto the Pinsky-Rinzel model. We obtain the full phase diagram and discuss the origin of the differentregions. We find that the spike traces of the model bare striking similarity to experimental biologicalneuronal recordings.This work may open a new way to investigate neural pathologies, such as epilepsy and Parkinson’sdisease, from the study of the phase diagram and the transitions between spiking states of physicalneuron models. We are also currently extending our work to implement equally simple synaptichardware models that allow us to start building spiking neural networks that implement basicprimitives, such as dynamical attractors and CPGs. This may have potential applications in robotics,neuroprosthetics and AI [3].

[1] J. Wu et al. Neuromorph. Comput. Eng. 3 044008 (2023)[2] M. Rozenberg et al. Sci Rep Scientific Reports, 9, 11123(2019). Featured in Nature's Device and Material Engineering blog[3] J. Wu et al. submitted (2024)