A team demonstrates computing directly within III-nitride-based sensors, exploiting interfacial polarization effects to process signals at the point of acquisition. The architecture, published in Nature, merges sensing and computation in a single device rather than sending data to a separate processor.

Interfacial polarization, both spontaneous and piezoelectric and characteristic of III-nitrides, is turned into a functional resource here: it modulates device conductance and enables logic and learning operations without additional circuitry. This approach falls under the in-sensor computing paradigm, which seeks to reduce the energy and latency bottleneck associated with moving large volumes of data from sensors.

The implications are relevant for embedded systems and machine vision, where energy efficiency per operation is critical. By removing the separation between sensor and processor, the platform promises lower power consumption and faster response times. The work opens the door to denser integrations of smart sensors based on III-nitrides.