Deoxyribonucleic acid brick crystals-based memristor as an artificial synapse for neuromorphic computing

Abstract

Biomaterial-based memristor synapses offer new possibilities for implementing high-performance organic resistive memory devices, which are essential for the development of green neuromorphic computing. Deoxyribonucleic acid (DNA), a biomaterial that is abundantly present in nature and possesses outstanding biocompatibility and stable physicochemical features, has been utilized as resistive switching (RS) layers for the fabrication of memristors. However, the presented DNA-based memristors are limited by stranded DNA with a single layer depth of ∼2 nm and a lateral size of ∼0.3 nm for a single base. The characteristics of memristors with DNA alone as the functional layer still need to be further improved, such as the switching ratios and power consumption. Herein, we propose a strategy for improving the performance of stranded DNA-based memristors by using three-dimensional DNA brick crystals (DNAbc) of micrometer lateral size as the RS layers. With high switching ratios (∼10⁴), low set power consumption (∼2.17 μW), good cycling stability and data retention capability, the DNAbc-based device exhibits superior RS memory characteristics compared to the double stranded DNA-based memristor with a similar device structure, which can be ascribed to the compact and oriented structures in DNAbc films that contribute to the formation of more efficient ion channels for Ag⁺ transport. The ability of the DNAbc-based memristor to simulate various bio-synaptic functions is demonstrated. Fundamental logic calculations of “AND” and “OR” are realized successfully by integrating two DNAbc-based memristors. A single-layer artificial neural network built with the DNAbc-based memristor achieves an accuracy of up to 91.23% in digit image recognition. This work paves the way for the construction of high-performance DNA-based storage devices, which are critical for future bio-realistic neuromorphic computing.