We use cookies.
By using the site, you agree to our Privacy Policy.

Laboratory for Microwave Photonics and Magnonics named after B. A. Kalinikos

Contract number
075-15-2021-609
Time span of the project
2021-2023

As of 01.11.2022

23
Number of staff members
10
scientific publications
1
Objects of intellectual property
General information

Artificial neural networks (ANN) are mathematical models simulating the work of the brain. The governments of various countries are fully supportive of developments in this domain, as they are able to lead to significant progress in many fields of technology. Reservoir computing is a special type of ANN that has an important advantage that is the easy and fast method of network training. The most important part of this network is the «reservoir». It has been demonstrated that a non-linear dynamic system can work as a reservoir. This has lead to the introduction of a a new research area — "physical reservoir computing (PRC). The research in this field is lead by physicists. Their ultimate goal is to find a physical system that possesses the best characteristics from the viewpoint of the requirements on PRC devices and can become a standard platform for PRC similarly to what the complementary metal-oxide-semiconductor technology has become in the field of common digital computers.

Mikhail P. Kostylev has proposed a new magnon spintronic (or simply magnonic) implementation of PRCs. It is based on active spin wave microwave active ring resonator. The main part of this physical system is a thin magnetic film in which propagating spin waves (or magnons) can spread in the strongly non-linear mode. Experiments conducted at the leading scientist's laboratory at the University of Western Australia have shown that this concept has quite a significant potential.

Name of the project: Reservoir computers based on the principles of magnonics as a new direction of artificial neural networks


Goals and objectives

The objective of the project is to transfer the experience in the field of magnonic PRCs accumulated at the University of Western Australia, the Saint Petersburg Electrotechnical University («‎LETI»), as well as to continue the study of the physics of the processes behind this concept on the grounds of «‎LETI». The new obtained knowledge will determine the focus of the international research in this field over many years after the completion of the project.

The practical value of the study

Scientific results:

  • Using the apparatus of Hamiltonian formalism, we have built a theory of nonlinear spin-wave excitation of by microstrip antennas. Relying the built theory and computations based on it, we have compared the influence of two types of nonlinearity on the amplitude of the excited spin wave. For the first time we have established the main physical mechanism responsible for the nonlinearity of the characteristics describing the excitation of spin waves by microstrip antennas. This mechanism is the four-wave parametric interaction of spin waves in the part of the magnetic film that is placed immediately beneath the antenna.
  • We have developed a theory of the formation of nonlinear transmission characteristics of an active ring resonator accounting for the nonlinear dispersion and attenuation of surface spin waves.
  • The Laboratory has conducted a research of the influence of such parameters of an active ring resonator as the coefficient of amplification of the open ring, the intensity of the external magnetic field and the input dimensionless amplitude of spin waves on the formation of nonlinear transmission characteristics.
  • It has been demonstrated that increasing the amplification coefficient up to the self-generation threshold leads to the expansion of the frequency range of bistability, while if the amplitude of spin waves is sufficiently small, the influence of nonlinear attenuation can be neglected. It was shown that increasing the magnetic field intensity ensures an increase of the nonlinear coefficient, which also leads to an extension of the frequency range of bistability. It has been found that increasing the input amplitude of spin waves leads to an extension of the frequency range of bistability, however, the impact of nonlinear attenuation in this mode is much more pronounced. The obtained results were used in the research of the nonlinear dynamics of spin waves in magnonic ring resonator systems to improve their performance.
  • We have developed a numerical model for the computation of the operational characteristics of reservoir computation system built upon a spin-wave active ring oscillator in which data input is performed by changing the coefficient of amplification of the feedback loop. The model relies on the Ginzburg–Landau equation describing the nonlinear phase shift and the nonlinear attenuation of the operating spin waves.
  • On the basis of the developed numerical model, we have created a set of computer programs to determine the efficiency of reservoir systems. As the main assessment criterion we chose the capacities observed as a result of a short-term memory check and a parity check. These tests allowed to assess two main parameters of efficiently, namely, memory and nonlinearity of the reservoir.
  • We have developed a physical implementation of a reservoir computer with a time delay based on a ring resonator with a spin-wave delay line. It has been demonstrated that this system, due to the dynamics of spin waves, is able to implement properties necessary for the reservoir, such as nonlinearity and attenuating memory. Introducing a reference line into the scheme of the reservoir computer allowed to improve the nonlinear computation capacity without a sharp decrease of the volume of attenuating memory. We have performed performance tests of the reservoir computer and it turned out that the performance of the developed reservoir computer is higher then the capacity of examples known from literature.
  • Our researchers have developed a theoretical model of the operational characteristics of a reservoir computing system relying on nonlinear interferometers. In this case, one leg of the interferometer is a spin-wave delay line operating in the nonlinear mode. Another leg of the interferometer is a microwave attenuator and a phase shifter. We have researched the transmission characteristics of a ring resonator based on a nonlinear spin-wave interferometer. It has been demonstrated that changing the phase on the phase shifter by π in the second leg of the interferometer allows to completely suppress the bistable behavior. As a result of the conducted research, we found that the use of nonlinear integro-differential schemes allows to increase the parity check capacity while preserving the short-term memory capacity, which allows to improve the performance of reservoir computation systems.

Implemented results of research: 

We are conducting negotiations with enterprises of the industry to develop an industrial sample of the reservoir computer .

Education and career development: 

  • We have developed education programs (including lecture courses, laboratory case studies, practical classes) in the area of our project: «Artificial neural networks based on principles of physical electronics», «Radiophotonics and fiber optics». These education programs have been implemented into the education process.
  • The academic team of the Laboratory organizes and stages the annual international conference «Microwave electronics and microelectronics».

Collaborations: 

«Inzarus» Ltd, «Powplay Systems» Ltd (Russia): conducting R&D in the domain of the project 

Hide Show full
watt s., kostylev m., ustinov a.b., kalinikos b.a.
Implementing a magnonic reservoir computer model based on time-delay multiplexing //Physical Review Applied. – 2021. – Т. 15. – №. 6. – С. 064060
ustinov a. b., kondrashov a. v., tatsenko i., nikitin a. a., and kostylev m. p.
Progressive development of spin wave chaos in active-ring oscillators //Physical Review B. – 2021. – Т. 104. – №. 14. – С. L140410
nikitin a.a., nikitin a.a., ustinov a.b., watt s., kostylev m. p.
Theoretical model for nonlinear spin-wave transient processes in active-ring oscillators with variable gain and its application for magnonic reservoir computing. Journal of Applied Physics, 131(11), 113903 (2022)
chumak a. v., kabos p., wu m., abert c.... kostylev m.p., ustinov a.b., nikitin a.a.
Roadmap on spin-wave computing, IEEE Transactions on Magnetics, Vol. 58, No. 6, P. 0800172 (2022)
vitalii v. vitko, andrey a. nikitin, roman v. haponchyk, andrey a. stashkevich, mikhail p.kostylev and alexey b. ustinov
Bistable behavior of active ring resonator on surface spin waves // The European Physical Journal Plus, Vol.137(9), pp.1-12. (2022)
alexandr v. kondrashov, alexey a. nikitin, andrey a. nikitin, mikhail kostylev and alexey b. ustinov
“Numerical simulation of performance of magnonic reservoir computer based on active-ring oscillator” //Journal of Magnetism and Magnetic Materials. –– С. 169968 (2022).
Other laboratories and scientists
Hosting organization
Field of studies
City
Invited researcher
Time span of the project
Laboratory for Crystal Photonics

Saint Petersburg State University - (SPbU)

Physics

St. Petersburg

Stoumpos Constantinos

Greece

2022-2024

Laboratory for Synchrotron Radiation Detectors

Tomsk State University - (TSU)

Physics

Tomsk

Shekhtman Lev Isayevich

Russia

2022-2024

Laboratory «Quantum engineering of light»

South Ural State University (national research university) - (SUSU)

Physics

Chelyabinsk

Kulik Sergey Pavlovich

Russia

2022-2024