An Efficient Biological Codon Recognition Reservoir Computing System Based on Low‐Energy Epitaxial Hf 0.52 Zr 0.48 O 2 Ferroelectric Memristors

Abstract

ABSTRACT The large demand for information processing has stimulated interest in low‐power and fast‐storage hafnium‐based ferroelectric memristors because of their ability to precisely control the state of the resistor by polarization flip‐flop without the need for electroforming. However, there is still a lack of hafnium‐based ferroelectric memristor with both high stability and ultra‐low operating energy consumption, which are the basic conditions for efficient neural network computation with high recognition rates. This article introduces a high‐quality epitaxially grown Pd/Hf 0.52 Zr 0.48 O 2 (HZO) /La 0.67 Sr 0.33 MnO 3 /SrTiO 3 ferroelectric memristor. The device offers high stability, such as multi‐stage stable storage states (16‐state retention time can exceed 10 4 s), high endurance performance (10 8 cycles), and stable pulse modulation. At the same time, the device has an ultra‐low energy consumption of 121 fJ. In addition, the HZO memristor is capable of a wide range of synaptic behaviors and logic operations. Importantly, this work is the first to apply a reservoir computing network based on HZO memristors to the field of biological genetics. The network successfully achieves a biological codon recognition accuracy of over 97% via the dual‐feature strategy. This work provides concrete system and design ideas for achieving low‐cost and high‐accuracy codon recognition in the biological field.