IIT Guwahati and IIT Kharagpur Build a Light-Sensing MXene Synapse That Sees, Remembers and Processes
An IIT Guwahati–IIT Kharagpur team has built an MXene and P3HT device that senses light, remembers it and processes it, using about 500 times less energy than conventional MXene devices in lab tests.
Researchers at IIT Guwahati and IIT Kharagpur have built an ultrathin, light-sensitive electronic device that can detect visual information, hold it as a memory, and process it, all in one component. In laboratory tests it used about 500 times less energy than conventional MXene-based devices. The work, led by Professor P. K. Giri of IIT Guwahati's Department of Physics, was reported on 12 October 2026 and has been published in the journal Carbon.
The problem it tackles
In a typical camera-based AI system, an image sensor captures light, the data is shuttled to a separate memory chip, and a processor then runs the analysis. Moving data between those components costs most of the energy and time. That is a growing problem for battery-powered devices such as drones, wearables and roadside cameras, which increasingly need to run vision AI on the device.
The human eye and brain work differently. The retina and visual cortex sense, store and process signals in the same networks of neurons and synapses. "Neuromorphic" or in-sensor computing tries to imitate that by building devices that behave like synapses, where the strength of a response depends on past signals.
How the device works
The team combined two materials:
- MXene nanosheets, an ultrathin, electrically conductive two-dimensional material, and
- P3HT, an organic semiconductor that generates electrical charges when light falls on it.
When light hits the device, P3HT produces charges and the MXene layer conducts them. The key behaviour is memory: the device keeps its response for some time after the light is switched off, much like a synapse that has been stimulated. When a second light pulse arrived, the response was 180% stronger than the first, mirroring the way repeated stimulation strengthens connections in the brain.
That combination lets one device act as sensor, memory and a basic computing element at the same time.
Tested on real AI tasks
The researchers tested the device in simulated vision workloads. It achieved 90% accuracy in facial recognition and 85% accuracy in segmenting traffic scenes, a task used in driver-assistance and smart-city cameras.
The team included Saswata Goswami and Professor Chandra Sekhar Tiwary from IIT Kharagpur.
Why it matters
India is investing heavily in semiconductors, from chip packaging plants in Gujarat and Assam to state-supported design programmes. Most of that effort focuses on conventional silicon. Work like this sits at the other end of the pipeline: new materials and device concepts that could underpin energy-efficient AI hardware a decade from now.
Low-power vision hardware has obvious uses in Indian conditions: traffic monitoring, agricultural drones, security cameras and medical devices that must work far from reliable power. A device that senses and processes images with a fraction of the energy could make such systems cheaper to run.
The caveats
The 500-fold energy saving is measured against other MXene-based devices, not against all conventional image sensors and processors. The results come from laboratory devices, and as coverage of the work noted, the wider impact will depend on further validation and on whether the lab results can be turned into reliable, manufacturable devices. Scaling from single devices to large arrays, and integrating them with standard chips, are the next hurdles.
Sources
- https://eastmojo.com/news/2026/10/12/iit-researchers-develop-brain-inspired-device-for-energy-efficient-vision/
- https://nenow.in/science-technology/assam-iit-guwahati-develops-brain-inspired-device-that-uses-500-times-less-energy-for-image-processing.html
- https://www.business-standard.com/technology/artificial-intelligence/iit-researchers-develop-new-device-for-ai-smart-computing-applications-126101201204_1.html
- https://newsable.asianetnews.com/india/iits-develop-light-sensitive-device-to-revolutionize-image-processing-articleshow-k6uvpzh
- https://thenewsmill.com/2026/10/iit-guwahati-develops-low-energy-brain-inspired-visual-device/