Radiation-Induced Charge Trapping Effect in BST/AlGaN/GaN Transistor: Gamma-Ray Dosimeter

Abstract

Reliable gamma-ray dosimetry across nuclear, medical, space, and defense environments demands sensing platforms that combine radiation hardness with sensitive electrical transduction and real-time readout capability. This work reports a solid-state gamma-ray dosimeter based on a two-terminal AlGaN/GaN high-electron-mobility transistor (HEMT) functionalized with a sputtered barium strontium titanate (Ba0.5Sr0.5TiO3, BST) dielectric layer. The AlGaN/GaN HEMT was evaluated for radiation hardness up to 10 Gy, with no measurable change in saturation current or threshold voltage, thereby establishing it as a stable electrostatic transducer. The BST layer, deposited over the gate region, served as the radiation-sensitive medium: gamma irradiation generates electron-hole pairs and donor-type oxygen-vacancy-related defects in the BST, and preferential electron capture at these defects and interface states leaves a net negative trapped charge. This net negative charge modulates the surface potential at the BST/AlGaN interface, progressively depleting the 2DEG and reducing the drain current in proportion to the absorbed dose. Dose-dependent transfer characteristics of the device confirm the mechanism: the drain current decreases and saturates with dose while the transconductance is progressively suppressed, indicating a net negative trapped charge that both depletes and scatters the 2DEG channel. High-resolution XPS shows that the near-surface BST is chemically and structurally intact under irradiation-the A-site (Ba, Sr) core levels and the Ti oxidation state are invariant. Grazing-incidence XRD confirms that the perovskite lattice is preserved over the full dose range. BST thicknesses of 50 to 200 nm were evaluated; the 100 nm configuration yielded the optimal response, reflecting a balance between the total trap population and the charge-to-interface electrostatic coupling efficiency. Six independently fabricated devices exhibit a consistent, monotonic drain-current response over 0–10 Gy, with ≤±6% inter-device variation, and a quasi-linear sensitivity of ∼25%/Gy below 2 Gy, extending to ∼7.5 Gy. A machine-learning dose-prediction pipeline, validated via leave-one-device-out cross-validation, complements a simple calibration baseline. System-level integration with a hardware-differential sensing architecture and an edge-computing IoT module enables cumulative dosimetry with continuous real-time wireless readout at ∼200 ms latency and ∼300 mW power consumption, bridging lab-scale radiation sensing and field-deployable dosimetry.