Accueil > English > News

Out-of-equilibrium Mott insulators : a step towards artificial neurons

How do Mott insulators, a type of insulating quantum materials, become conductive under strong voltage ? A researcher from the LPS and his collaborators explained how, and how it may open the way for implementing artificial neurons for the electronics of tomorrow.

For over 50 years, conventional microelectronics have followed the so called Moore’s Law that predicts a doubling of the number of transistors per unit area every 18 months. But today there is a consensus that the already achieved extreme miniaturisation implies the end of this law. The microelectronics industry is searching for "more than Moore" which consists of developing new functionalities based on the physical properties of new materials, such as Mott insulators, which are still not well-understood. Researchers from the LPS, the Institut des Matériaux Jean Rouxel and the Institut des Sciences Chimiques de Rennes have proposed a model to explain the behaviour of these insulators when strong voltages are applied on them.

Studied by physicists for several decades for their exceptional physical properties, Mott insulators can become conductive under high pressure or by chemical doping, two parameters that are difficult to reconcile with the technological constraints of microelectronics. However, the application of a voltage can destabilise them and also make them electrically conductive. Previous work of the team has shown that this destabilisation can be used to make components for the microelectronics of tomorrow, such as artificial neurons for bioinspired computing.

However, the mechanism behind the driving out of equilibrium of Mott insulators by electric field remained controversial. Researchers have now shown that it is the massive creation of hot electrons in an electric field and their strong interactions that is at the base of the effect [1]. Beyond a certain voltage threshold, an electronic avalanche phenomenon occurs, and due to the strong electron-electron interaction that characterises Mott materials, the electrons further excite their neighbours and so on, leading to the full destabilisation of the insulator state.

This new insight enables a better control of the associated functionalities of these materials. Indeed, in a second recent work, the research teams have shown how to model the relaxation of an artificial neuron based on this material [2]. These works are an important step towards neuromorphic computers, the artificial intelligence machines that emulate the brain.

Figure 1. Left : Resistive collapse of a Mott insulator (photo) under a strong electric field. The bottom panel shows the dramatic drop in voltage on the sample, and the top panel the concomitant surge in current. After the applied voltage is terminated, the system recovers to the initial state. Right : Non-linear conductivity as a function of applied field and temperature. The line and the dot indicate the theoretical model prediction.

Figure 2. Model simulations of the metallic filamentary formation in a model of the resistive collapse of a Mott insulator material. As a function of the parameters, the model shows either multi-filamentary formation, or the thickening of a single filament. This behaviour permits to understand the long relaxation time of the collapsed-resistance states.


[1] How a dc Electric Field Drives Mott Insulators Out of Equilibrium
P. Diener, E. Janod, B. Corraze, M. Querré, C. Adda, M. Guilloux-Viry, S. Cordier, A. Camjayi, M. Rozenberg, M. P. Besland et L. Cario
Physical Review Letters 121, 016601 (2018)

[2] Relaxation of a Spiking Mott Artificial Neuron
F. Tesler, C. Adda, J. Tranchant, B. Corraze, E. Janod, L. Cario, P. Stoliar, and M. Rozenberg
Physical Review Applied 10, 054001 (2018)


Marcelo Rozenberg