Sensory ultraviolet encoding by non-equilibrium trap dynamics in β-Ga2O3

Deep-ultraviolet (DUV) sensing underlies safety-critical systems, especially flame and combustion detection, where detectors should both respond fast and reject false alarms from sources such as welding arcs. Conventional detectors offer no usable in-sensor dynamics, since trap-free devices offer a static response, while those laden with often-deep-level traps offer dynamics viewed as delaying a steady-state response. Thus, discriminating temporal features is deferred to downstream processing, forcing a trade-off between response time and accuracy. Here, in β-Ga2O3, we exploit such traps – DUV illumination photoactivates occupied traps, the Poole–Frenkel emission of which results in a current-controlled negative differential resistance (NDR). The NDR is out-of-equilibrium, existing only during photoactivation of traps away from a steady state. Coupled to a small external capacitance, this instability sustains voltage firing that is band-pass in effective dose rate. Reading the response at five different tuning currents distinguishes six DUV sources of identical wavelength, differing only in temporal modulation, including welding and flame that are indistinguishable by average intensity. This method offers 98.3% accuracy at 100 seconds versus 87.4% for an ideal intensity-only detector, thus discriminating dynamical information at the point of sensing versus by threshold-only. Therefore, defect dynamics treated as delaying the material’s actual response, can encode useful information.