Henrique Frulani de Paula Barbosa, Andreas Schander, Andika Asyuda, Luka Bislich, Sarah Bornemann, Björn Lüssem

Advanced Electronic Materials (2024): 2400481

https://doi.org/10.1002/aelm.202400481

Artificial Neural Networks (ANN) require a better platform to reduce their energy consumption and achieve their full potential. Electrochemical devices like the Electrochemical Neuromorphic Organic Device (ENODe) stand out as a potential building block for ANNs, due to their lower energy demand, in addition to their biocompatibility and access to multiple and stable memory levels. However, the non-volatile effect observed in these devices is not yet fully understood. Hence, here we propose a 2D drift-diffusion model that is capable to reproduce the device behavior. The model relies on the assumption of trapping sites for cations, which are increasingly filled or emptied during subsequent pre-synaptic pulses. The model is verified by experiments on devices with varying post-synaptic dimensions. Overall, the results provide a framework to discuss ENODe operation and design strategies for ENODes with well-controlled memory states.

© 2024 The Author(s). Advanced Electronic Materials published by Wiley-VCH GmbH