Highly stable self-rectifying memristor arrays: Enabling reliable neuromorphic computing via multi-state regulation

In the context of the rapid development of artificial intelligence and big data, neuromorphic computing, which mimics the working mode of the human brain, has become a research hotspot to break through the limitations of traditional computing architectures. Memristors, as core devices for constructing neuromorphic systems, have always faced challenges such as poor stability and inconsistent performance during long-term operation. A latest study published in Nano Research has made significant progress in solving these problems.

The research team developed a self-rectifying memristor (SRM) array based on the Pt/TaOx/Ti structure. What is particularly noteworthy is its outstanding stability: under AC conditions, the device can maintain stable switching performance after more than 105 cycles without obvious conductance drift or performance degradation. Even after 100 DC cycles, its key performance metrics show minimal fluctuations. For example, the coefficient of variation (CV) of the rectification ratio at 3 V is only 0.11497, which fully reflects its excellent consistency and reliability—critical qualities for large-scale integrated systems that require long-term stable operation.

Another major advantage of this memristor array is its excellent multi-state regulation capability. Through continuous DC voltage sweeps with gradually reduced stopping voltages, the device can achieve 32 consecutive and linearly quantized conductance states. Each conductance state can be stably retained for more than 104 seconds at room temperature (25 °C), and the conductance can be repeatedly switched between 359 pS and 1.51 pS with a linearity of 0.98240. This precise and controllable multi-state characteristic enables it to well simulate synaptic plasticity in the human brain, providing an ideal hardware platform for neuromorphic computing tasks such as synaptic weight adjustment.

To further explore the application potential of the device, the research team integrated its neuromorphic characteristics with a simulated annealing algorithm and optimized the annealing temperature function to make it more in line with the dynamic behavior of biological neurons. Experimental results show that this integration enables efficient image restoration: compared with traditional algorithms, it can complete restoration with higher accuracy in fewer iterations, and the structural similarity (SSIM) between the restored image and the original image can reach 99.93%.

"In the field of neuromorphic computing, stability and controllability are the prerequisites for practical application of devices. Our work focuses on improving these two key indicators, and the results are encouraging," said Shaoan Yan, one of the corresponding authors. Yingfang Zhu added, "The 32×32 array we developed can theoretically be expanded to 12.9 kbit, which provides a feasible path for large-scale in-memory computing systems."

This work was supported by by the National Natural Science Foundation of China (U23A20322), the National Key Research and Development Program of China (2023YFF0719600, 2021YFA1202600, 2021YFB4000800), the CAS Project for Young Scientists in Basic Research (YSBR-113), the Ningbo Technology Project (2022A-007-C), the Hunan Provincial Natural Science Foundation (2023JJ50009, 2025JJ60351, 2023JJ30599), the Foundation of Innovation Center of Radiation Application (KFZC2023020701), and the Major Scientific and Technological Innovation Platform Project of Hunan Province (2024JC1003).

About Nano Research

Nano Research is a peer-reviewed, open access, international and interdisciplinary research journal, sponsored by Tsinghua University and the Chinese Chemical Society, published by Tsinghua University Press on the platform SciOpen. It publishes original high-quality research and significant review articles on all aspects of nanoscience and nanotechnology, ranging from basic aspects of the science of nanoscale materials to practical applications of such materials. After 18 years of development, it has become one of the most influential academic journals in the nano field. Nano Research has published more than 1,000 papers every year from 2022, with its cumulative count surpassing 7,000 articles. In 2024 InCites Journal Citation Reports, its 2024 IF is 9.0 (8.7, 5 years), and it continues to be the Q1 area among the four subject classifications. Nano Research Award, established by Nano Research together with TUP and Springer Nature in 2013, and Nano Research Young Innovators (NR45) Awards, established by Nano Research in 2018, have become international academic awards with global influence.

END