Publications

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Al0.26Ga0.74N/GaN on SiC lateral Schottky diodes were fabricated with variable anode-to-cathode spacing and were analyzed for blocking and on-state device performance. On-chip normally-on High Electron Mobility Transistor (HEMT) structures were also fabricated for a comparison of blocking characteristics. The Schottky diode displayed an ideality factor of 1.59 with a Ni/AlGaN zero bias barrier height of 1.18 eV and a flat band barrier height of 1.59 eV. For anode-to-cathode spacings between 10 and 100 μm, an increase in median breakdown voltages from 529 V to 8519 V and median specific on-resistance (Ron-sp) from 1.5 to 60.7 mΩ cm2 was observed with an increase in spacing. The highest performing diode had a lateral figure of merit of 1.37 GW/cm2 corresponding to a breakdown voltage upwards of 9 kV and a Ron-sp of 59 mΩ cm2. This corresponds to the highest Schottky diode breakdown voltage reported thus far with an Al0.26Ga0.74N/GaN lateral structure.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Resistive memory (ReRAM) shows promise for use as an analog synapse element in energy-efficient neural network algorithm accelerators. A particularly important application is the training of neural networks, as this is the most computationally-intensive procedure in using a neural algorithm. However, training a network with analog ReRAM synapses can significantly reduce the accuracy at the algorithm level. In order to assess this degradation, analog properties of ReRAM devices were measured and hand-written digit recognition accuracy was modeled for the training using backpropagation. Bipolar filamentary devices utilizing three material systems were measured and compared: one oxygen vacancy system, Ta-TaO_x, and two conducting metallization systems, Cu-SiO₂, and Ag/chalcogenide. Analog properties and conductance ranges of the devices are optimized by measuring the response to varying voltage pulse characteristics. Key analog device properties which degrade the accuracy are update linearity and write noise. Write noise may improve as a function of device manufacturing maturity, but write nonlinearity appears relatively consistent among the different device material systems and is found to be the most significant factor affecting accuracy. As a result, this suggests that new materials and/or fundamentally different resistive switching mechanisms may be required to improve device linearity and achieve higher algorithm training accuracy.

Resistive memory (ReRAM) shows promise for use as an analog synapse element in energy-efficient neural network algorithm accelerators. A particularly important application is the training of neural networks, as this is the most computationally-intensive procedure in using a neural algorithm. However, training a network with analog ReRAM synapses can significantly reduce the accuracy at the algorithm level. In order to assess this degradation, analog properties of ReRAM devices were measured and hand-written digit recognition accuracy was modeled for the training using backpropagation. Bipolar filamentary devices utilizing three material systems were measured and compared: one oxygen vacancy system, Ta-TaOx, and two conducting metallization systems, Cu-SiO2, and Ag/chalcogenide. Analog properties and conductance ranges of the devices are optimized by measuring the response to varying voltage pulse characteristics. Key analog device properties which degrade the accuracy are update linearity and write noise. Write noise may improve as a function of device manufacturing maturity, but write nonlinearity appears relatively consistent among the different device material systems and is found to be the most significant factor affecting accuracy. This suggests that new materials and/or fundamentally different resistive switching mechanisms may be required to improve device linearity and achieve higher algorithm training accuracy.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

The switching characteristics of vertical Gallium Nitride (v-GaN) diodes grown on GaN substrates are reported. v-GaN diodes were tested in a Double-Pulse Test Circuit (DPTC) and compared to test results for SiC Schottky Barrier Diodes (SBDs) and Si PiN diodes. The reported switching characteristics show that GaN diodes, like SiC SBDs, exhibit nearly negligible reverse recovery current compared to traditional Si PiN diodes. The reverse recovery for the v-GaN PiN diodes is limited by parasitics in the DPTC, precluding extraction of a meaningful recovery time. These results are very encouraging for power electronics based on v-GaN and demonstrate the potential for very fast, low-loss switching for these devices.

Abstract not provided.

Abstract not provided.

Abstract not provided.

The realisation of a GaN high voltage vertical p-n diode operating at >3.9 kV breakdown with a specific on-resistance <0.9 mΩ cm2 is reported. Diodes achieved a forward current of 1 A for on-wafer, DC measurements, corresponding to a current density >1.4 kA/cm2. An effective critical electric field of 3.9 MV/cm was estimated for the devices from analysis of the forward and reverse current-voltage characteristics. This suggests that the fundamental limit to the GaN critical electric field is significantly greater than previously believed.