Tellurium‐Based Artificial Neuron: Capturing Biological Complexity While Keeping It Simple

Abstract

Neuronal oscillation is critically involved in almost every cognitive task, including information coding, memory formation, and perception. This phenomenon is rooted in the spiking activity of neuron that is mediated by two principal ion‐channel mechanisms, i.e., fast Na+ mediated depolarization and delayed K+ mediated repolarization. In neuromorphic circuits, however, subcellular level of biocomplexity has routinely been traded off for smaller hardware overhead. Even so, the simplest artificial oscillatory neuron by far is still made of no less than two devices, a volatile memristor and a parallelly (serially) connected capacitor (resistor) that typically occupies relatively large circuit area. Herein, a simplest imaginable implementation is reported of the neuronal oscillation function in a single threshold switching device based on tellurium (Te). Thanks to the unique electrochemical‐thermal properties combined in Te, this device can exhibit self‐sustained oscillatory behavior when stimulated by a steady DC voltage as a result of the repetitive electrochemical growth and the ensuing Joule heat‐induced thermal melting of the Te filament, which phenomenologically capture the Na+ and K+ ion‐channel functions, respectively. This work broadens the neuromorphic applications of Te resistive switching devices from synaptic array devices to artificial neurons.

Artificial Neuronal Devices Based on Emerging Materials: Neuronal Dynamics and Applications

“…The output voltage would go back and forth between the two threshold voltages and forms a continuous oscillation waveform. Volatile memristors, for example, TaO x , [9b] HfO 2 , [31] PCMO, [32] Te/TiTe 2 , [33] and GeSe [34] memristors, share a similar hysteretic switching characteristic as IMT materials. Despite the switching physics is different, same design of oscillatory circuits also applies to these devices (Figure 3b).…”