Reconfigurable mmWave microchips co-integrating hBN switches on GaN

a, I-V curves collected at constant sweep rates (0.1 V s−1 and 1 V s−1) during set cycles of Au/hBN/Au switches in chip 3 (hBN sample #B) for those devices that underwent forming at room temperature. b, Same as panel a, but for a constant sweep rate of 0.1 V s−1 and cycling performed at 175 °C. c, Same conditions as in panel b, but the data is from cycling devices that underwent forming at 175 °C. d, Statistics of the set voltage extracted from the I-V curves in panels a-c. No distinguishable differences are observed for the different forming nor testing conditions. I-V curve variability in the HRS is mostly related to variability in the reset process, and it is not a concern to device performance in RF. Some reset transitions can bring the device back to an intermediate state (see current in the I-Vs around 10−5 A) that is completely reversed to HRS withing the first 2 V of the set sweep. A few cycles show that the device resets to a pristine-like conduction state (very low leakage current), requiring another forming step (re-forming, with higher voltages than the set transitions, see circled curves in panels e and f). e, f, Room temperature I-V curves collected at constant sweep rates for 4 different devices of e, chip 1 (hBN sample #A) and f, chip 3 (hBN sample #B). Circled curves are examples of high voltage re-forming sweeps after hard reset transitions. g, Statistics of the set voltage extracted from the I-V curves in panels e-f. Differences between chips 1 and 3 (with different hBN suppliers) lay well within device-to-device variability, with no clear contribution from the hBN itself to overall device operation. h, Reset voltage statistics extracted from the reset cycles corresponding to the set statistics of chip 3 (hBN sample #B), measured at RT and 175 °C. There is an observable reduction of the reset overall voltages with increasing temperature, well inline with the filament thermal disruption process previously described (Supplementary Note 1). Data points per box: 66, 61. i, LRS resistance measured from the reset cycles analysed in panel h. Unlike reset voltage, LRS resistance shows no clear variation with temperature (median increase of 16 % but mean decrease of 6%), inline with the filamentary conduction (Supplementary Note 1). Data points per box: 66, 61. All box plot boundaries in this figure are 25-75 percentiles, with median line, mean (open symbol), and whiskers representing 1 SD.