Search for: All records

The demand for high power and high-frequency radio frequency (RF) power amplifiers makes AlGaN/GaN high electron mobility transistors (HEMTs) an attractive option due to their large critical field, high saturation velocity, and reduced device footprint as compared to Si-based counterparts. However, due to the high operating power densities, intense device self-heating occurs, which degrades the electrical performance and compromises the device’s reliability. The self-heating behavior of AlGaN/GaN HEMTs is known to be not solely a function of the dissipated power but is highly bias-dependent. As the operation of RF power amplifiers involves alteration of the device operation from fully-open to pinched-off channel conditions, it is critical to experimentally map the full channel temperature profile as a function of bias conditions. However, such measurement is difficult using optical thermography techniques due to the lack of optical access underneath the gate electrode, where the peak temperature is expected to occur.

To address this challenge, an AlGaN/GaN HEMT employing a transparent gate made of indium tin oxide (ITO) was fabricated, which enables full channel temperature mapping using Raman spectroscopy. It was found that the maximum channel temperature rise under a partially pinched-off condition is more than ∼93% higher than that for an openmore »channel condition, although both conditions would lead to an identical power dissipation level. The channel peak temperature probed in an ITO-gated device (underneath the gate) is ∼33% higher than the highest channel temperature that can be measured for a standard metal-gated AlGaN/GaN HEMT (i.e., next to the metal gate structure) operating under an identical bias condition. This indicates that one may significantly underestimate the device’s thermal resistance when solely relying on performing thermal characterization on the optically accessible region of a standard AlGaN/GaN HEMT. The outcomes of this study are important in terms of conducting a more accurate lifetime prediction of the device lifetime and designing thermal management solutions.

« less

AlN thin films are enabling significant progress in modern optoelectronics, power electronics, and microelectromechanical systems. The various AlN growth methods and conditions lead to different film microstructures. In this report, phonon scattering mechanisms that impact the cross-plane (κ z ; along the c-axis) and in-plane (κ r ; parallel to the c-plane) thermal conductivities of AlN thin films prepared by various synthesis techniques are investigated. In contrast to bulk single crystal AlN with an isotropic thermal conductivity of ∼330 W/m K, a strong anisotropy in the thermal conductivity is observed in the thin films. The κ z shows a strong film thickness dependence due to phonon-boundary scattering. Electron microscopy reveals the presence of grain boundaries and dislocations that limit the κ r . For instance, oriented films prepared by reactive sputtering possess lateral crystalline grain sizes ranging from 20 to 40 nm that significantly lower the κ r to ∼30 W/m K. Simulation results suggest that the self-heating in AlN film bulk acoustic resonators can significantly impact the power handling capability of RF filters. A device employing an oriented film as the active piezoelectric layer shows an ∼2.5× higher device peak temperature as compared to a device based on an epitaxial film.

The label noise transition matrix, characterizing the probabilities of a training instance being wrongly annotated, is crucial to designing popular solutions to learning with noisy labels. Existing works heavily rely on finding “anchor points” or their approximates, defined as instances belonging to a particular class almost surely. Nonetheless, finding anchor points remains a non-trivial task, and the estimation accuracy is also often throttled by the number of available anchor points. In this paper, we propose an alternative option to the above task. Our main contribution is the discovery of an efficient estimation procedure based on a clusterability condition. We prove that with clusterable representations of features, using up to third-order consensuses of noisy labels among neighbor representations is sufficient to estimate a unique transition matrix. Compared with methods using anchor points, our approach uses substantially more instances and benefits from a much better sample complexity. We demonstrate the estimation accuracy and advantages of our estimates using both synthetic noisy labels (on CIFAR-10/100) and real human-level noisy labels (on Clothing1M and our self-collected human-annotated CIFAR-10). Our code and human-level noisy CIFAR-10 labels are available at https://github.com/UCSC-REAL/HOC.

Abstract Gallium nitride (GaN) has emerged as one of the most attractive base materials for next-generation high-power and high-frequency electronic devices. Recent efforts have focused on realizing vertical power device structures such as in situ oxide, GaN interlayer based vertical trench metal–oxide–semiconductor field-effect transistors (OG-FETs). Unfortunately, the higher-power density of GaN electronics inevitably leads to considerable device self-heating which impacts device performance and reliability. Halide vapor-phase epitaxy (HVPE) is currently the most common approach for manufacturing commercial GaN substrates used to build vertical GaN transistors. Vertical device structures consist of GaN layers of diverse doping levels. Hence, it is of crucial importance to measure and understand how the dopant type (Si, Fe, and Mg), doping level, and crystal quality alter the thermal conductivity of HVPE-grown bulk GaN. In this work, a steady-state thermoreflectance (SSTR) technique was used to measure the thermal conductivity of HVPE-grown GaN substrates employing different doping schemes and levels. Structural and electrical characterization methods including X-ray diffraction (XRD), secondary-ion mass spectrometry (SIMS), Raman spectroscopy, and Hall-effect measurements were used to determine and compare the GaN crystal quality, dislocation density, doping level, and carrier concentration. Using this comprehensive suite of characterization methods, the interrelation among structural/electrical parameters andmore »the thermal conductivity of bulk GaN substrates was investigated. While doping is evidenced to reduce the GaN thermal conductivity, the highest thermal conductivity (201 W/mK) is observed in a heavily Si-doped (1–5.00 × 1018 cm−3) substrate with the highest crystalline quality. This suggests that phonon-dislocation scattering dominates over phonon-impurity scattering in the tested HVPE-grown bulk GaN substrates. The results provide useful information for designing thermal management solutions for vertical GaN power electronic devices.« less

Abstract Researchers have been extensively studying wide-bandgap (WBG) semiconductor materials such as gallium nitride (GaN) with an aim to accomplish an improvement in size, weight, and power of power electronics beyond current devices based on silicon (Si). However, the increased operating power densities and reduced areal footprints of WBG device technologies result in significant levels of self-heating that can ultimately restrict device operation through performance degradation, reliability issues, and failure. Typically, self-heating in WBG devices is studied using a single measurement technique while operating the device under steady-state direct current measurement conditions. However, for switching applications, this steady-state thermal characterization may lose significance since the high power dissipation occurs during fast transient switching events. Therefore, it can be useful to probe the WBG devices under transient measurement conditions in order to better understand the thermal dynamics of these systems in practical applications. In this work, the transient thermal dynamics of an AlGaN/GaN high electron mobility transistor (HEMT) were studied using thermoreflectance thermal imaging and Raman thermometry. Also, the proper use of iterative pulsed measurement schemes such as thermoreflectance thermal imaging to determine the steady-state operating temperature of devices is discussed. These studies are followed with subsequent transient thermal characterization tomore »accurately probe the self-heating from steady-state down to submicrosecond pulse conditions using both thermoreflectance thermal imaging and Raman thermometry with temporal resolutions down to 15 ns.« less