A carregar...
Publicação
Improved reliability and stability of IGZO memristors performance for artificial intelligent (AI) applications
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.2 MB | Adobe PDF |
Orientador(es)
Resumo(s)
A rápida expansão das aplicações de internet-das-coisas (IoT) e inteligência artificial exige um processamento eficiente de dados, impulsionando uma transição de arquiteturas de computação convencionais para superar limitações de tempo e energia. À medida que a miniaturização enfrenta limitações físicas, paradigmas alternativos, como a computação in-memory, que imita o comportamento sinapse-neurónio semelhante ao cérebro, têm emergido. Este estudo explora o aprimoramento do desempenho elétrico de dispositivos de comutação resistiva baseados em óxido de índio-gálio-zinco (IGZO), produzidos por métodos de deposição física de vapor a temperatura ambiente, com foco na retenção, atribuindo também grande relevância à sua razão de condutância, confiabilidade e resposta a esquemas de pulsos. Vários parâmetros foram investigados, incluindo proporções de composição da camada ativa, tratamentos de oxigénio no contacto inferior e diferentes materiais de elétrodo, molibdénio e alumínio.
As descobertas-chave indicam que o tratamento com oxigénio e a composição alvo de IGZO influenciam significativamente as características do memristor, nomeadamente nas razões de condutância e retificação, as quais foram melhoradas com o uso de um tratamento alternativo de plasma de oxigénio a uma pressão mais elevada. O recozimento como tratamento com oxigénio resultou num contacto quase óhmico sem comutação resistiva. As tentativas de resolver a baixa retenção usando alumínio como um contacto de baixo alternativo foram frutíferas; no entanto, levaram à perda do comportamento analógico confiável verificado ao usar molibdénio. Os resultados alcançados estão em linha com o estado-da-arte, uma vez que não foram realizadas melhorias inovadoras. No entanto, é provável que este tratamento alternativo de plasma seja implementado daqui em diante.
The rapid expansion of internet-of-things (IoT) and artificial intelligence (AI) applications demands efficient data processing, prompting a shift from conventional computing architectures to overcome time and energy constraints. As miniaturization faces physical limitations, alternative paradigms like in-memory computing, mimicking brain-like synapse-neuron behavior, have emerged. This study explores enhancing the electrical performance of resistive switching (RS) devices based on indium-gallium- zinc-oxide (IGZO), produced by physical vapor deposition methods at room temperature, with a focus on retention, while also giving great relevance to their conductance ratio, reliability, and response to pulse schemes. Various parameters were investigated, including active layer composition ratios, oxygen treatments on the bottom contact, and different electrode materials, molybdenum and aluminum. Key findings indicate that oxygen treatment and the IGZO target composition significantly influence memristor characteristics, particularly conductance and rectification ratios, which were improved with the use of an alternative oxygen plasma treatment at higher pressure. Annealing as a oxygen treatment resulted in an ohmic-like contact without resistive switching. Attempts to address poor retention using aluminum as an alternative bottom contact (BC) were fruitful; however, they led to the loss of reliable analog behavior verified when using molybdenum. The results achieved are in line with the state-of-the-art, as no groundbreaking improvements were accomplished. Nonetheless, this alternative plasma treatment will most likely be implemented onwards.
The rapid expansion of internet-of-things (IoT) and artificial intelligence (AI) applications demands efficient data processing, prompting a shift from conventional computing architectures to overcome time and energy constraints. As miniaturization faces physical limitations, alternative paradigms like in-memory computing, mimicking brain-like synapse-neuron behavior, have emerged. This study explores enhancing the electrical performance of resistive switching (RS) devices based on indium-gallium- zinc-oxide (IGZO), produced by physical vapor deposition methods at room temperature, with a focus on retention, while also giving great relevance to their conductance ratio, reliability, and response to pulse schemes. Various parameters were investigated, including active layer composition ratios, oxygen treatments on the bottom contact, and different electrode materials, molybdenum and aluminum. Key findings indicate that oxygen treatment and the IGZO target composition significantly influence memristor characteristics, particularly conductance and rectification ratios, which were improved with the use of an alternative oxygen plasma treatment at higher pressure. Annealing as a oxygen treatment resulted in an ohmic-like contact without resistive switching. Attempts to address poor retention using aluminum as an alternative bottom contact (BC) were fruitful; however, they led to the loss of reliable analog behavior verified when using molybdenum. The results achieved are in line with the state-of-the-art, as no groundbreaking improvements were accomplished. Nonetheless, this alternative plasma treatment will most likely be implemented onwards.
Descrição
Palavras-chave
amorphous oxide semiconductor (AOS) IGZO neuromophic computing parylene-C resistive switching devices