Si-based Resonant Interband Tunnel Diode with Cutoff Frequency over 20 GHz and Estimated Peak Current Density of 218 kA/cm/sup 2'/

Abstract

++ 5×10 19 cm -3 , while both the B and P δ-doping sheet concentrations were maintained at 1×10 14 cm -2 . Due to the stability criteria for the RF measurements stemming from the ultra-high PCD, the device size and its contact resistance need to be very small to avoid self-oscillations at bias points in the negative differential resistance (NDR) region [7]. Fig. 3 shows the measured I-V characteristics obtained for two different device sizes, estimated to be 3.13 µm 2 and 0.36 µm 2 , respectively. The sizes of the active area were estimated from the PCD of the DC test device (5×5 µm 2 ) assuming the PCD is independent of the device size. For the PCD calculation, the lateral extent (~0.5 µm) of the etch undercut was taken into account, which was measured by scanning electron microscopy. Weak self-oscillation-induced steps in the NDR region were observed for the device size of 3.13 µm 2 , but not for the smaller sized device of 0.36 µm 2 . For the fabrication of small devices, a two-mask process [1] was adopted. For the small contact resistance, a P δ-doped layer was inserted 5 nm below the cap layer during the LT-MBE growth, followed by Ni silicidation through this P δ-doped quantum well by rapid thermal sintering at 430 o C for 30 seconds, resulting in a specific contact resistivity of 5.3×10 -7 Ω-cm 2 extracted from RF measurements assuming that the contact resistance dominates the series resistance. The microwave performance of the RITDs was assessed using on-wafer bias-dependent sparameter measurements from 1 to 25 GHz in the bias voltage range from 0 to 0.62 V in 10 to 50 mV steps. The 50 Ω microwave probe was calibrated by the short-open-load-through method before RF measurements. The signal delivered to the diodes was ~10 mV peak-to-peak. Shown in the inset of Fig. 4 is an equivalent small signal circuit model of the RITD and the probe pads. The probe pad parasitics were determined by fitting the pad circuit model to the measured S11 of an on-wafer open pad test structure using ADS software. Then, the de-embeded s-parameters of the RITDs were used to obtain the intrinsic device parameters including rs, Cj, and gj by ADS optimization over the frequency range of 1 to 25 GHz at each bias point. Figure 4 shows the extracted Cj and gj as a function of the junction voltage (external voltage – current × rs) with the I-V characteristic of the intrinsic tunnel diode and the gj calculated from the DC I-V characteristics by first order differentiation. Very good agreement between the gj obtained from RF measurements and DC measurements was observed, which suggests that the RITDs are very good in terms of their low-frequency noise mainly stemming from trapping. Figure 5 shows the measured and the modeled reflection gain (20log10Г) for several representative external bias points where the model for the intrinsic RITD in the inset of Fig. 4 is well matched with the measured s-parameters. The inset in Fig. 5 displays cutoff vs. external bias. The cutoff of 20.2 GHz was observed at a bias of 350 mV.