Publicação
Zinc-tin oxide devices for sustainable electronic systems
| Resumo: | Sustainable, earth-abundant materials offer a promising route toward addressing the economic, environmental, and energy challenges of modern computing. This thesis explores zinc-tin oxide (ZTO), an amorphous semiconductor, as a versatile platform for energy-efficient and scalable neuromorphic computing systems. Initially, ZTO memristors were designed to show the necessary qualities for their use as in-memory computing units in neural network accelerating applications. However, with the aim of more faithfully emulating biological neuronal structures, the research of physical reservoir systems (PRC) was pursued. By exploiting the inherent non-linear mechanisms of tailor designed memristors, the resulting PRC implementations were able to achieve considerable improvements in both network size and accuracy in a practical digit recognition task. ZTO-based diodes were also developed in tandem with the previously mentioned systems. With potential use as sneak-path-preventing selective devices in active crossbars for neuromorphic applications or other high-frequency tasks, the Schottky diodes showed record breaking metrics when compared with the vertically stacked solution-based oxide structures currently available in the literature. |
|---|---|
| Autores principais: | Silva, Carlos Miguel de Castro |
| Assunto: | Zinc-tin oxide memristor neuromorphic computing reservoir computing diode high-frequency |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | tese de doutoramento |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Sustainable, earth-abundant materials offer a promising route toward addressing the economic, environmental, and energy challenges of modern computing. This thesis explores zinc-tin oxide (ZTO), an amorphous semiconductor, as a versatile platform for energy-efficient and scalable neuromorphic computing systems. Initially, ZTO memristors were designed to show the necessary qualities for their use as in-memory computing units in neural network accelerating applications. However, with the aim of more faithfully emulating biological neuronal structures, the research of physical reservoir systems (PRC) was pursued. By exploiting the inherent non-linear mechanisms of tailor designed memristors, the resulting PRC implementations were able to achieve considerable improvements in both network size and accuracy in a practical digit recognition task. ZTO-based diodes were also developed in tandem with the previously mentioned systems. With potential use as sneak-path-preventing selective devices in active crossbars for neuromorphic applications or other high-frequency tasks, the Schottky diodes showed record breaking metrics when compared with the vertically stacked solution-based oxide structures currently available in the literature. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.