December 1 - 6, 2024
Boston, Massachusetts
Symposium Supporters
2024 MRS Fall Meeting & Exhibit
EL05.02.04

An Integrated Organic Spiking Neuron/Synapse Architecture with an Unconventional Form-Factor for “Physiological Neuromorphic Systems”

When and Where

Dec 2, 2024
2:30pm - 2:45pm
Sheraton, Second Floor, Independence West

Presenter(s)

Co-Author(s)

Giovanni Maria Matrone1,Zachary Laswick1,Ruiheng Wu1,Abhijith Surendran1,Yoeri van de Burgt2,Jonathan Rivnay1

Northwestern University1,Technische Universiteit Eindhoven2

Abstract

Giovanni Maria Matrone1,Zachary Laswick1,Ruiheng Wu1,Abhijith Surendran1,Yoeri van de Burgt2,Jonathan Rivnay1

Northwestern University1,Technische Universiteit Eindhoven2
Signal communication mechanisms within the human body rely on the transmission and modulation of action potentials<sup>1</sup>. Neural networks exploit the interaction of spiking elements (afferent neurons and interneurons) and neuromodulator junctions (chemical and electrical synapses), to process the information acquired through biological receptors. Replicating these interdependent functions with organic/hybrid artificial neurons and biohybrid synapses is an essential first step towards merging neuromorphic circuits and biological systems<sup>2,3</sup>.<br/>Organic materials have been employed to build electronic circuits that mimic both the spiking behaviour of neurons<sup>2,4</sup> and their synaptic transmission<sup>5</sup> thus replicating afferent neurons and interneurons functions. Indeed, a combination of bioelectronic devices may recreate a “neuronal pathway” that in nature relies on the cooperation of neurons, receptors and chemical synapses<sup>2</sup>. Although the most recent organic neuromorphic circuits emulate biological functions which are essential for basic signal processing<sup>2,4</sup>, communicating seamlessly with the central nervous system (CNS) requires an even higher level of bio-realism with devices that are able to capture neurotransmitter signaling at physiological levels (synapses with tens of nM limit of detection) but also to closely match the frequency, amplitude and activation threshold of biological spikes (neurons), including attaining a bio-plausible power consumption ( at least nJ range). Moreover, synapses and neurons in the CNS operate with high energy efficiency also due to their intimate “hardware” integration, which allows to locally modulate and exchange signals without losses.<br/>Here it is presented a neuromorphic platform based on the Hodgkin-Huxley (HH) circuit model which integrates in a single architecture an organic synapse and an antimbipolar OECT to replicate both afferent neurons “sensory coding” and the interneurons neuromodulatory functions, while operating in “physiological conditions”.<br/>Exploiting a vertical transistor architecture, a bio-hybrid synapse realised through the electrodeposition of PEDOT is integrated on top of a BBL-based anti-ambipolar OECT.<br/>This architecture allows to co-allocate the fundamental neuronal mechanisms of synaptic weight modulation and spike generation. Indeed, the modulation of the conductance state of the PEDOT device (bio-hybrid synapse) through the neurotransmitters dopamine allows to selectively tune the anti-ambipolar characteristics of the whole device (BBL in series with PEDOT OECT), replicating both short-term and long-term potentiation mechanisms.<br/>Additionally, the use of different supporting polyelectrolytes for PEDOT electrodeposition is explored to tune the mixed ionic electronic conduction properties of the bio-hybrid synapse to increase its dynamic range and responsivity to electro-active neurotransmitters allowing to reach a “physiological” limit of detection of 10 nM. This in-series OECT is part of a HH circuit, representing its Na<sup>+ </sup>channel, allowing to translate physiological changes in the concentrations of dopamine and serotonin in the electrolyte environment, directly into spikes encoded signals. The total neuromorphic system’s energy consumption, as predicted by electronic circuit simulations, is reduced to 1 nJ, while the platform footprint is reduced to 7 mm x 7 mm. As such, due to its unconventional architecture, this system constitutes a building block for programmable neural pathways and it is compatible with in-sensor applications footprint and sensitivity requirements for locally executing bio-inspired pre-processing functions, thus allowing for a seamless and dynamic communication with the nervous system.<br/><b>1</b> Kandel, E. R. <i>et al.</i> 4, (McGraw-hill New York, 2000)<br/><b>2</b> Matrone, G. M. <i>et al.</i> <i>Nat Commun</i> 15, 2868 (2024)<br/><b>3</b> Gkoupidenis, P. <i>et al.</i> <i>Nat Rev Mater</i> 9, 134–149 (2023)<br/><b>4</b> Harikesh, P. C. <i>et al.</i> <i>Nat. Mater.</i> 22, 242–248 (2023)<br/><b>5</b> Matrone, G. M. <i>et al.</i> <i>Adv Materials Technologies</i> 2201911 (2023)

Keywords

electrical properties | morphology

Symposium Organizers

Paschalis Gkoupidenis, Max Planck Institute
Francesca Santoro, Forschungszentrum Jülich/RWTH Aachen University
Ioulia Tzouvadaki, Ghent University
Yoeri van de Burgt, Technische Universiteit Eindhoven

Session Chairs

Xenofon Strakosas
Yoeri van de Burgt

In this Session