Polarization-sensitive ferroelectric photomemristors boost low-contrast image recognition for next-generation AI vision
image: Inspired by the compound eyes of dragonflies, the new device integrates anisotropic ReSe₂ and ferroelectric DIPAB thin films to enhance low-contrast image recognition. It mimics biological polarization vision to achieve accurate, real-time recognition in autonomous driving and medical imaging.
Credit: Nano Research, Tsinghua University Press
The ability to see clearly in complex environments—such as foggy roads, dimly lit medical scans, or distant astronomical observations—remains a formidable challenge for artificial intelligence (AI) vision systems. Existing processors based on the von Neumann architecture are hampered by latency and high energy costs, while conventional photomemristors often fail to distinguish low-contrast targets.
A research team led by Professors Chunxiao Cong, Laigui Hu and Ran Liu, at Fudan University has now introduced a self-driven polarization-sensitive ferroelectric photomemristor (PSFP) that overcomes these limitations. By mimicking the way dragonflies detect polarized light to track prey, the device directly integrates polarization sensitivity into hardware, enabling real-time recognition of subtle features that traditional systems often miss.
The device was engineered using a heterostructure of two materials: the anisotropic two-dimensional semiconductor rhenium diselenide (ReSe₂) and a thin single-crystalline layer of the ferroelectric material diisopropylammonium bromide (DIPAB). This unique design allows the memristor to function simultaneously as a sensor, memory, and processor. Crucially, the built-in ferroelectric polarization enables self-powered operation, eliminating the need for external bias during photo-detection.
Polarization-sensitive photomemristor offers a hardware-level solution for enhancing AI vision, especially under challenging low-contrast conditions. By extracting polarization information directly in the sensor, it can bypass the delays of preprocessing algorithms and achieve accurate recognition in real time.
Performance tests confirmed the breakthrough: in recognizing road signs under low-contrast conditions, the PSFP achieved an average accuracy of 85.9%, compared with just 47.5% for conventional photomemristors. Neural network simulations further demonstrated that the device could process Gaussian-blurred patterns with rapid convergence, even in noisy environments.
The device also exhibited key neuromorphic features, including synaptic plasticity (long-term potentiation and depression), multilevel data storage, and millisecond-level response speeds. These capabilities make it an ideal candidate for integration into neuromorphic computing systems, autonomous vehicles, intelligent medical imaging, and other advanced AI applications.
The researchers emphasize that this work represents a proof-of-concept, but its potential impact is significant. By moving beyond traditional grayscale or color-based imaging to harness polarization information, the PSFP provides a new dimension for machine vision.
This study was supported by the National Key Research and Development Program of China for International Cooperation (Grant 2023YFE0117100) and the National Natural Science Foundation of China (Nos. 62074040, 62074045).
Contributors include Chenxu Sheng, Shuwen Shen, Laigui Hu, Xiaofei Yue, Shoaib Awan, Dacheng Xia, Jiao Wang, Zhi-Jun Qiu, Chunxiao Cong, Ran Liu from School of Information Science and Technology, Fudan University.
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.