An artificial ionic neuron for tomorrow's electronic memories

With an energy consumption equivalent to two bananas per day, the human brain can perform many complex tasks. Its high energy efficiency depends in particular on its base unit, the neuron, which has a membrane with nanometric pores called ion channels, which open and close according to the stimuli received. The resulting ion flows create an electric current responsible for the emission of action potentials, signals that allow neurons to communicate with each other.

Artificial intelligence can do all of these tasks but only at the cost of energy consumption tens of thousands of times that of the human brain. So the entire research challenge today is to design electronic systems that are as energy efficient as the human brain, for example, by using ions, not electrons, to carry the information. For this, nanofluidics, the study of how fluids behave in channels less than 100 nanometers wide, offer many perspectives. In a new study, a team from the ENS Laboratoire de Physique (CNRS/ENS-PSL/Sorbonne Université/Université de Paris) shows how to construct a prototype of an artificial neuron formed of extremely thin graphene slits containing a single layer of water molecules1 . The scientists have shown that, under the effect of an electric field, the ions from this layer of water assemble into elongated clusters and develop a property known as the memristor effect: these clusters retain some of the stimuli that have been received in the past. To repeat the comparison with the brain, the graphene slits reproduce the ion channels, clusters and ion flows. And, using theoretical and digital tools, scientists have shown how to assemble these clusters to reproduce the physical mechanism of emission of action potentials, and thus the transmission of information.

This theoretical work continues experimentally within the French team, in collaboration with scientists from the University of Manchester (UK). The goal now is to prove experimentally that such systems can implement simple learning algorithms that can serve as the basis for tomorrow’s electronic memories.

• 1Recently invented in Manchester by the group of André Geim (Nobel Prize in Physics 2010)