Heterojunction nanofluidic memristors based on peptide chain valves for neuromorphic applications.

Memristors exhibit significant potential for neuromorphic computing due to their unique properties. This study introduces a heterojunction nanofluidic memristor (HJNFM) and explores its applications in simulating synapses and constructing neural networks. The HJNFM consists of a SnS2 and MoS2 heterojunction nanochannel with a peptide chain valve. The opening and closing dynamics of peptide chain valve alter ionic conductance of the nanochannel and realize the memristor characteristics. The sequence of the peptide chain also affects the electrical properties of HJNFM. Additionally, by setting up multi SnS2 strips in the nanochannel, the multi-HJNFM can achieve permanent memory and emulate synaptic features including short-term and long-term memory. Notably, we construct a convolutional neural network from multi-HJNFMs, which achieves 94 % accuracy in a digit recognition task. This study presents a new approach to constructing nanofluidic memristors, which could be advantageous for developing new forms of neuromorphic computing in the future.