Many animals use polarized light as an extra visual cue. Insects can read polarization patterns in the sky for navigation, while mantis shrimp exploit polarization information in their surroundings. For machine vision, the same type of information can reveal surface orientation, texture and material composition that conventional intensity images may miss, particularly under challenging optical conditions such as glare, haze or low illumination.Turning that biological idea into practical hardware has been difficult. Organic single crystals are attractive because their ordered molecular structures naturally respond differently to different polarization directions. Yet these devices are commonly made over small areas, their polarization contrast is usually constrained by the material's intrinsic absorption anisotropy, and sensing, memory and signal processing are often implemented in separate components.
Researchers from Tianjin University of China have developed an electret-mediated dual-functional strategy that brings these functions together in one organic device platform. The study was published online in National Science Review. The key is polyvinyl alcohol, or PVA, which performs two jobs during a biphasic dip-coating process. First, PVA regulates interfacial nucleation and stabilizes the transport of organic semiconductor molecules, helping continuous, highly oriented few-layer single crystals grow across centimeter-scale areas. Second, the hydroxyl-rich PVA layer remains directly beneath the crystal and acts as an electret - a charge-trapping layer that can retain photoexcited electrons. This combination changes how the device responds to polarized light. The organic crystal still provides the initial directional sensitivity, but the amount of charge trapped in PVA also depends on polarization. Trapped electrons create a photogating effect that nonlinearly amplifies the electrical response. In this way, a modest intrinsic optical anisotropy can be converted into a much larger electrical contrast.
The researchers quantify this contrast with the dichroic ratio, the ratio between device responses to two orthogonal polarization directions. The devices reached a dichroic ratio of 2,920, more than three orders of magnitude higher than the intrinsic dichroic ratio of the pristine crystal. Across a 7 x 7 array, the average dichroic ratio was 2,400 with a standard deviation of about 200 and a coefficient of variation of 7.3%, and all 49 transistors were operational. These results indicate that the polarization enhancement can be reproduced across an array rather than only in an isolated device. The same charge-trapping process also gives the phototransistors synaptic-like behavior. When two light pulses arrive in close succession, the second response can be much stronger than the first, mimicking paired-pulse facilitation in biological synapses. The maximum paired-pulse facilitation index reached 6,594%, while the photocurrent could be tuned over more than five orders of magnitude by changing the optical stimulus. The estimated electrical energy used to read a synaptic event was about 1 aJ.
The team then moved from device measurements to visual demonstrations. A 5 x 5 sensor array clearly resolved polarization-dependent optical patterns and retained part of the information after the light was removed, reflecting the device's memory behavior. In a separate semantic-recognition demonstration, polarization-resolved images of apples, oranges, watermelons, peaches and pineapples were acquired at 0, 45, 90 and 135 degrees. Device-derived convolution kernels were used to extract polarization-sensitive features before classification. Fusing information from all four polarization angles produced an accuracy of 95.25%, higher than any of the four single-angle inputs.
By using one polymer layer to guide crystal growth and provide charge trapping, the approach links scalable fabrication with polarization sensing, memory and hardware-level preprocessing. The researchers say this electret-mediated framework offers a route toward integrated neuromorphic polarization vision systems in which perception, storage and computation are more tightly coupled.
This work was supported by the National Key Research and Development Program of China (Grant No. 2024YFB3614500), the National Natural Science Foundation of China (Grant Nos. 52373194, 52403301, and U24A20293), the Major Project of State Key Laboratory of Advanced Materials for Intelligent Sensing (25ZXZSSS00430), and the Fundamental and Interdisciplinary Disciplines Breakthrough Plan of the Ministry of Education of China (JYB2025XDXM410).
National Science Review
Experimental study