Single transistor used to implement neuromorphic behaviour

Technology News | March 31, 2025

By Peter Clarke

Neuromorphic computing has been boosted by the development of single-transistor computing cell that can mimic the behaviour of both electronic neurons and synapses.

Up until now electronic neurons and synapses implemented with traditional silicon transistors requires interconnecting multiple devices – specifically, at least 18 transistors per neuron and 6 per synapse, according to researchers at the National University of Singapore.

Now a team led by associate professor Mario Lanza of the Department of Materials Science and Engineering has shown that neuromorphic behaviour can be achieved in a single transistor. The team has written a paper titled: Synaptic and neural behaviours in a standard silicon transistor that has been published in the scientific journal Nature on March 26.

The method used by Professor Lanza’s team is related to the setting the bulk silicon resistance to a specific value that excites a physical phenomenon called impact ionization, which generates a current spike. The phenomenon is therefore analogous to the current spike characteristic of neurons and synapses.

Setting the bulk resistance to other specific values allows the transistor to store charge in the gate oxide for a length of time mimicking the behaviour a biological synapse.

Therefore making a standard silicon transistor operate as a neuron or synapse depends on selecting the appropriate resistance of the bulk terminal.

Impact ionisation had been considered a failure mechanism in silicon transistors, but Professor Lanza’s team has turned into a potentially valuable application for AI.

The team has designed a cell composed of two transistors – called the neurosynaptic random access memory (NSRAM) – that allows switching between neuron or synapse operating modes. This offers versatility in manufacturing.

Professor Lanza’s team has implemented the single- and two-transistor neuromorphs on 180nm CMOS. The first-named author of the paper is Dr Sebastián Pazos from King Abdullah University of Science and Technology.