Laboratory of Physics for Neuromorphic Computation Systems

The Laboratory's research program is based on the notion that the human brain with its complex architecture that is capable of both data processing and storage only consumes about 10 W of power while possessing, at least for some tasks, the same capacity as a supercomputer consuming 10 MW. The urgency and necessity to transit to neuromorphic devices have already lead to the commercial launch of neuromorphic chips such as Loihi (Intel) and TrueNorth (IBM). However, these devices are based on the CMOS silicon technology and, despite some advantages from the manufacturing point of view, runs into the same limitations in terms of energy consumption that standard CMOS devices also face. Moreover, they still use digital technologies, while the brain in its essence is an analog device. To better imitate the brain and to use the advantages of its architecture, analog neuromorphic data processing should be emulated directly in the material of the device. Therefore, a program is proposed that will research neuromorphic concepts based on switching between multiple states in multiferroic and 2D materials controlled by an optical pulse, magnetic field, electric field and deformation. The objective of the research is to create ICT devices with low power consumption.

Name of the project:

Multiferroics and 2D materials for neuromorphic computations

Research directions: Electrical engineering and electronics

Goals of project:

To lay the scientific foundations for drastically improving the energy efficiency and the rate of operation of information and communication technologies (ICT).

Project objective:

1. The development of the notion of controlled multitude and plasticity in ferroic materials based on: (а) the mobility of ferroic domain walls and (b) the response of magnetic and spin structures to the impact of an electric field via magnetoelectric interaction, (c) the response of magnetic and ferroelectric structures to mechanical strains (straintronics), (d) the interface effects at the ferroic-2D materials interface;
2. The development of concepts of improving the performance and energy efficiency of ferroics by controlling the magnetic spin structures/degrees of freedom using ultrashort laser pulses;
3. The development of methods of the visualization of the above-mentioned effects, including the (nonlinear) dynamics of spins, mechanical strains and domain walls;
4. The development of elements relying on the principles of the organization of the brain, such as synapses and neurons based on (2D) ferroic materials, spin valves, straintronics elements;
5. The development of concepts of energy efficient ICT architectures based on the developed materials and effects.