Temperature-dependent trapping dynamics in AlGaN/GaN high-electron-mobility transistors are investigated across 4–600 K using current-mode deep-level transient spectroscopy (I-DLTS) method and long-duration electrical stress. A striking transition is observed from strongly persistent cryogenic trapping to near-complete high-temperature recovery: only approximately 11% of the stress-induced current deficit recovers within 10 μs at 4 K, compared with approximately 93% at 600 K. Thermal cycling confirms that the cryogenic persistent state is reversible rather than permanent. The recovery transients further resolve distinct fast-and slow-recovery contributions, with longer filling preferentially enhancing the slow component at 300–500 K, while this buildup becomes limited at 600 K. Most importantly, the 600-K response combines relatively rapid apparent population buildup with reduced saturation collapse and low persistent collapse. This behavior is consistent with thermally enhanced emission during filling and/or rapid post-fill detrapping within the unresolved initial 0.25 μs. The results reveal that temperature governs not only how trapped charge recovers after stress, but also the apparent trapped-charge population observed after filling, establishing a unified filling–recovery picture for the transition from cryogenic persistence to high-temperature recovery.
Abstract- This Letter demonstrates a unified theoretical framework of similarity laws in electron-beam-driven vacuum diodes. Two independent similarity scalings are derived based on the Boltzmann equation coupled with Poisson's equation, and validated via first-principles particle-in-cell simulations. The similarity laws for electron transport are extended from the non-relativistic to relativistic regimes with the consideration of the scale-dependent Lorentz factor (varying with geometric scales). This work uncovers the underlying similarity and symmetry nature in the physics of electron beam diodes, thereby offering novel strategies for device design and cost-effective cross-comparison among upscaled or downscaled device experiments.
This work demonstrates a single ferroelectric tunnel field-effect transistor (FeTFET) compute-in-memory (CiM) cell for Hamiltonian (H) computation in interaction-centric Ising machines. By exploiting the ambipolar transport and local ferroelectric polarization switching of the FeTFET, signed spin (σ)–interaction (J) multiplication and external field (h) term for H computation are realized within a single transistor, replacing the complementary FeFET pair required in conventional implementations. Four programmable states corresponding to J = ±1 and h = ±1 are experimentally demonstrated using a sequential write bias scheme, with the h = +1 state additionally serving as a mask state for disconnected interactions. Circuit simulations using a physics-based Verilog-A model calibrated to the measured transfer characteristics verify row-wise multiply-accumulate (MAC) operations in a 4 × 4 FeTFET CiM array. Simulations further demonstrate the solution of a 4-vertex Max-Cut problem through iterative minimization of total H, with an estimated array read power of ~0.73 nW/spin. These results highlight the potential of the single-FeTFET CiM cell for compact and energy-efficient interaction-centric Ising machines.
High-sensitivity mid-wavelength infrared (MWIR) photodetectors operating at room temperature are highly desirable for compact hydrocarbon gas sensing, because low gas concentrations produce weak absorption-induced intensity changes. In this letter, we demonstrate a nearly lattice-matched InAs/AlAsSb unipolar-barrier photodetector optimized for the 3.3-μm C–H absorption band. The unipolar-barrier design effectively suppresses generation–recombination current. At 300 K, the detector exhibits a dark current density below the empirical Rule 07 dark-current benchmark and an external quantum efficiency of 79.3%, obtained using Fourier-transform infrared spectroscopy and a 900 K blackbody calibration. Under a bias of −0.4 V, its calculated specific detectivity (D*) reaches 5.1 × 1010 cm·Hz1/2·W−1 at 3.26 μm, and the response time is below 45 ns. At 220 K, D* increases to 1.15 × 1012cm·Hz1/2·W−1 and reaches the calculated background-limited infrared photodetection level for a 300 K background under a 2π field of view. Integrated into a nondispersive infrared system, the detector achieves room-temperature detection limits of 2.45 ppm for methane and 1.09 ppm for ethane. These results demonstrate the potential of InAs detectors for compact, high-sensitivity MWIR hydrocarbon sensing.
Nanosheet field-effect transistors (FETs) have been aggressively scaled, and highly susceptible to microscopic fluctuations. Analyzing their impact on circuit performance is essential for future applications. This work proposes a lightweight Gated Fusion-temporal convolutional network (GF-TCN) model for rapid circuit performance prediction of nanosheet field-effect transistors. The model infers key metrics (delay, power, etc.) of inverter, NAND gate, and ring oscillator directly from transistor ID-VG curves, avoiding complex and time-consuming compact model parameter extraction and circuit simulation. This work shortens the design-technology co-optimization (DTCO) cycle and accelerates the development of advanced semiconductor device technologies.
