Smart polymers: use of conjugated polymers in computing tasks

• Ano de defesa: 2025
• Autor(a) principal: Souza, Rafael Francisco Santiago de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Computação; Computing; Logic gate; PEDOT; Porta lógica; Reservatório; Reservoir
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/76/76133/tde-02042025-094445/

Reservoir computing (RC) has emerged as a promising neuromorphic computing paradigm, leveraging the dynamic properties of physical systems for efficient temporal data processing. Electrically conductive polymers (ECPs) offer a unique platform for implementing RC due to their tunable electrical conductivity and ability to undergo dynamic charge transport processes, both phenomenons that can be harnesses to generate non-linear behaviors, key feature of RC. These materials enable the development of energy-efficient, flexible, and scalable reservoir systems, particularly for edge computing applications. In this work, the EDOT monomer is polymerized into a PEDOT based polymer to use in building a physical reservoir .Using a PEDOT based reservoir, we demonstrate its capability in classification tasks, specifically applied to the Iris dataset, showcasing its potential for nonlinear pattern recognition, achieving a classification accuracy of 60%. Beyond their role in reservoir computing, materials exhibiting negative differential resistance (NDR) present exciting opportunities for logic-based computing architectures. NDR, characterized by a decrease in current with increasing voltage in specific regions of a materials I-V curve, enables the implementation of bistable and multi stable electronic states essential for logic gate operations. Conducting polymers, including PEDOT:PSS, can be engineered to exhibit NDR behavior through controlled doping, electrochemical modulation, or nanoscale structuring. By harnessing this property, ECPs could serve as reconfigurable logic elements, enabling novel in-memory computing paradigms and compact, low-power logic circuits. In this work, a physical device composed of eight gold (Au) electrodes was assembled with PEDOT:PSS as the conductive layer in two different designs, vertical and planar. NDR behavior was sought in those designs and , once found it, it was used to solve a XOR classification task. Only the planar design was able to solve the task, although being an unstable solution, since the polymer degrades during the processing of the task. The synergy between RC and NDR-based logic could pave the way for advanced, adaptive computing systems, bridging the gap between neuromorphic processing and unconventional logic devices.