Highly Energy-Efficient Resistive Switching in Defect- and Strain- Engineered Mott Insulators for Neuromorphic Computing Applications

Introduction

The pressing need for beyond-von-Neumann computing paradigms has triggered intensive efforts into understanding and controlling various resistive switching (RS) mechanisms. This switching, in which the two-terminal resistance of a device is controlled by current, is at the heart of emerging technologies such as resistive random access memory and neuromorphic computation. These technologies promise to revolutionize artificial neural networks and mimic the behavior of biological brains, triggering a race for optimal RS materials.

Mott Insulators and Their Properties

In Mott insulators, electrical currents can change resistance by orders of magnitude due to an insulator-metal phase transition. The volatility of switching in Mott insulators can be adjusted by several tuning parameters, enabling both memory devices and neuron-like functionalities.

Moreover, in terms of fundamental switching timescales and energy efficiency, Mott insulators may have very significant advantages over other RS mechanisms. These unique properties have made Mott insulators prominent candidate materials for RS. However, the physical mechanisms behind RS in these materials are not well understood and often uncontrollable, hampering the realization of their full potential.

Proposed Routes for Energy Efficiency

We suggest two main routes towards Mott-insulator-based RS with ultrahigh energy efficiency:

1. Switching in the Electronic Sector: The first route is by switching purely in the electronic sector while minimizing structural distortions. Thus, the low heat capacity of electrons may enable switching with a fraction of the energy required in an insulator-metal transition coupled to a structural transition.

2. Absorption of Latent Heat: The second route involves the absorption of latent heat and/or elastic energy from the surroundings of the switching volume, thus reducing the externally supplied power consumption.

Research Goals

Our aim is to use defects, doping, and strain engineering to understand and tune RS mechanisms, and develop novel functionalities with ultra-low energy consumption.