Neuromorphic Devices

Neuromorphic Devices

The slowdown of Moore’s law and the end of Dennard scaling have prompted researchers to look for alternative computer architectures that could surpass these paradigms. In recent years, this search has become more urgent thanks to the increasing amount of data generated by the popularization of Artificial Intelligence (AI) tools and Internet of Things (IoT), demanding more efficient ways to store, sort and classify data. Inspired by the human brain, neuromorphic computing emerges as a solution for these tasks, as it seeks to process high amounts of data with low power consumption while also providing parallel computing capabilities, learning capacity and absence of von Neumann bottleneck. 

Among the candidates for base units of neuromorphic computing, electrochemical devices, such as the Organic Electrochemical Transistor (OECT), stand out due to their low power consumption, small footprint, access to multiple memory levels, and superior spatio-temporal dynamics. Despite the interesting properties, however, a full control over the memory level access and stability is still elusive. Therefore, here at IMSAS we investigate the memory level modulation of OECTs by using simulation, novel material deposition strategies and optical and electrical techniques. With this, we expect to improve the control of both volatile and non-volatile memory features, thus allowing applications in neuromorphic hardware to perform not just data classification but also image and speech recognition (see https://doi.org/10.1002/aelm.202400481 and https://doi.org/10.1002/aelm.202500507).