April 22 - 26, 2024
Seattle, Washington
May 7 - 9, 2024 (Virtual)
Symposium Supporters
2024 MRS Spring Meeting
SB03.09.05

Brain Inspired Floating Redox Wiring for Neural Networks

When and Where

Apr 25, 2024
9:30am - 9:45am
Room 436, Level 4, Summit

Presenter(s)

Co-Author(s)

Roshani Madurawala1,Maik-Ivo Terasa1,Anna Lina Wyschkon1,Soeren Kaps1,Rainer Adelung1

University of Kiel, Germany1

Abstract

Roshani Madurawala1,Maik-Ivo Terasa1,Anna Lina Wyschkon1,Soeren Kaps1,Rainer Adelung1

University of Kiel, Germany1
The human brain is a remarkably complex and adaptable organ and its capability to change and adapt, is due to a phenomenon known as "neuroplasticity." Neuroplasticity in synapses refers to the brain's outstanding ability to change the strength and structure of the connections between neurons, which is crucial for information transfer. This has inspired the design and training of artificial neural networks (ANN), allowing them to learn and improve their performance over time. In the trend of implementing ANN in hardware, one specific challenge is the implementation of plasticity in the artificial synapses. Here we show a concept and a mechanism to incorporate plasticity into a system by means of electrochemical metallization in a liquid matrix, Dimethyl sulfoxide (DMSO). A liquid matrix inspired by the brain has been employed to facilitate the mobility of the ions in addition to its cooling purposes in this ionotronic system. These bio-inspired, dynamic and reconfigurable electronic connections are robust in DMSO while still being prone to manipulation upon applied external stimuli, because they float in DMSO. These connections are wires that grow as dendrites due to the redox reaction taking place at the electrodes. The thickness and the growth conditions of these wires influence their electrical properties, which could be altered by varying the initial experimental conditions such as the applied voltage. The manipulation of these wires could be done during the growth of the wire (filament) between two nodes or more, simultaneously. Furthermore, after the growth of one filament, the filament could be manipulated in a way to induce filament breakage which resembles synaptic properties. This process of plasticity in artificial synaptic connection equivalents, serves as a fundamental step in the refinement of network connectivity. Its replication within artificial systems marks a significant milestone in the quest to emulate the brain's adaptability. Intriguingly, this filament can grow in a span of over 200 µm. What is even more remarkable is their ability to overcome the limitations of traditional 2D growth, as they readily extend their presence into the 3D realm. Growth and manipulation of single 2D and 3D filaments and in extension, a filament network, using external stimuli will be shown along with electrical characterization of these filaments and possible applications. In essence, the groundwork laid in the field of dynamic filament growth within a liquid matrix holds the potential to reshape the landscape of technological advancement. With its capacity for synaptic mimicry, flexible growth, and 3D expansion, this innovation ushers in an era of transformative possibilities, making inroads into fields as diverse as neuromorphic computing, cognitive sciences, and advanced artificial intelligence. [1]<br/>[1] M.-I. Terasa et al., Materials Today (2023), https://doi.org/10.1016/j.mattod.2023.07.019<br/><i>The authors acknowledge financial support by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 434434223 – SFB 1461</i>

Keywords

electrochemical synthesis

Symposium Organizers

Dimitra Georgiadou, University of Southampton
Paschalis Gkoupidenis, Max Planck Institute
Francesca Santoro, Forschungszentrum Jülich/RWTH Aachen University
Yoeri van de Burgt, Technische Universiteit Eindhoven

Session Chairs

Paschalis Gkoupidenis
Bjorn Lussem

In this Session