CogniGron Seminar: Merlyne de Souza (Sheffield University) - "Physical reservoir computing based on a nano-ionic solid electrolyte FET"

Reservoir computing is a versatile framework for applications in diverse domains, ranging from time-series prediction for real-time decisions at the edge, to pattern recognition, speech processing and robotics. Physical reservoir computing took off when it was demonstrated that a single volatile memristor can be effectively utilized in place of recurrent neural networks, without the need for any interconnected reservoir nodes [1].  The input signal is projected onto the reservoir (i.e. a transistor), where its time dependent characteristics project not just the present input, but also its past history.  This enriched input is fed to a linear readout layer that maps the reservoir's dynamic state into an output signal. Crucially, the learning process is simplified, as only the readout layer weights need to be trained, (usually one-shot training) making reservoir computing computationally efficient and particularly well-suited for real-time applications.

In this talk I will showcase a three terminal volatile  ZnO/Ta₂O₃ Thin Film Transistor from our lab [2,3] which has proven versatile across many tasks such as image and audio signal processing, with high accuracy [4].  The transistor has a unique negative differential resistor in its gate current characteristics which makes it an active memristor, potentially suitable to be operated as a neuron.

References:

[1]L. Appeltant, M. Soriano, G. Van Der Sande, J. Danckaert, S. Massar, J. Dambre, B. Schrauwen, C. R. Mirasso, I. Fischer 2, 468 (2011).

[2] “Nanoionics Based three terminal synaptic device using ZnO”, PB Pillai and M. M. De Souza, ACS Appl. Mater. Interfaces,  9 (2), pp 1609–1618 (2017).

[3] “Diffusion controlled Faradaic charge storage in high performance solid electrolyte gated Zinc Oxide thin film transistors”, P. B. Pillai, A. Kumar, X. Song, M. M. De Souza, ACS Materials and Interfaces, 10 (11), pp 9782–9791 (2018)

[4] Ankit Gaurav, Xiaoyao Song , Sanjeev Manhas , Aditya Gilra , Eleni Vasilaki , Partha Roy and Maria Merlyne De Souza Frontiers in Electronics, doi: 10.3389/felec.2022.869013