In-sensor computing aims to analyse visual or other sensory data directly where it is captured, offering real-time processing with superior energy efficiency compared with conventional architectures that shuttle data between sensors and processors. The authors note that existing devices based on oxides or two-dimensional ferroelectrics suffer from severe response loss and are difficult to fabricate at scale. The new device instead relies on heterostructures of III-nitride compounds, the same materials family used in industrial-scale LEDs and compatible with complementary metal–oxide–semiconductor integration.

The key mechanism is spatial separation of photogenerated carriers driven by polarization. In III-nitride heterostructures, intrinsic polarization creates built-in electric fields at interfaces, and the study reports that these bipolar polarization interfaces are defect-free. When light generates electron–hole pairs, the polarization-induced fields separate the opposite charges at the interfaces, which the researchers say effectively avoids carrier recombination losses that normally degrade photoresponse.

In its sensor-memristor mode, the device reaches a light sensitivity at the level of nanowatts per square centimetre, a responsivity at the level of 10^6 amperes per watt and an on/off ratio at the level of 10^5. The memristive behaviour means the device can both detect light and retain a memory of prior optical inputs, a combination suited to in-sensor computing architectures where sensing and computation are merged. Applying controlled bias allows the same device to transform into a high-speed detector, so one structure can serve different operating modes depending on the electrical conditions.

The researchers describe the work as establishing a design principle that combines interfacial-electrostatics-induced photocarrier spatial separation with asymmetric-barrier-modulated carrier transport to realize in-sensor computing. Because III-nitrides resist high temperatures, the device maintained high performance at temperatures up to 300 °C, a capability that could matter for sensing in harsh environments where conventional sensor-computing hardware would fail.

The team also emphasizes manufacturability. The approach aligns with industrial-scale III-nitride manufacturing and CMOS integration, which the authors say probably advances the feasibility of in-sensor computing architectures. III-nitride polarization is already an advantage in large-scale optoelectronic devices, but it had remained largely unexplored in sensory-memristive devices, making this demonstration a bridge between an established materials platform and an emerging computing paradigm.

The reported figures come from the published abstract and figure descriptions of the Nature Materials paper, which characterize the device across different operating modes and application demonstrations. Detailed performance data, fabrication methods and the full scope of the application demonstrations appear in the article itself, which is available by subscription. As with any newly reported device concept, the results describe a laboratory demonstration of a design principle rather than a commercially deployed technology, and the path from the reported performance levels to practical in-sensor computing systems will depend on subsequent engineering work.

Subject of Research: III-nitride heterostructure devices for polarization-enabled in-sensor computing

Article Title: III-nitride-based in-sensor computing enabled by interfacial polarization effects

Article References: Jiang, K., Wang, B., Sun, X., Liu, J., Feng, G., Duan, S., Tong, P., Fang, T., Zhang, S., Liu, J., Jia, Y., Liu, M., Ben, J., Gao, P., Tian, B., Hu, W., Tian, H., & Li, D. (2026). III-nitride-based in-sensor computing enabled by interfacial polarization effects. Nature Materials. https://doi.org/10.1038/s41563-026-02780-7

Image Credits: AI Generated

DOI: 10.1038/s41563-026-02780-7

Keywords: in-sensor computing, III-nitride semiconductors, polarization, sensor-memristor, photodetector, heterostructures, carrier separation, responsivity, high-temperature electronics, CMOS integration, optoelectronics, Nature Materials