Noise Performance of Diffusion Cooled Hot-Electron Bolometers: Theory vs. Experiment
Peter John Burke, Robert J. Schoelkopf, Irfan A. Siddiqi, D. E. Prober, Anders J. Skalare, B. S. Karasik, Michael C. Gaidis, William R. McGrath, Bruce Bumble, Henry George LeDuc
We have measured the spectrum of the output noise and the conversion gain from 0.1-7.5 GHz under identical conditions for both diffusion and phonon-cooled Nb bolometers, using a 20 GHz LO on a variety of devices varying in length from 0.08 izm to 3 Ara, where the gain-bandwidth varies between 100 MHz and > 6 GHz. In this paper, we will present systematic comparison between theory and experiment for the devices measured. We find the frequency dependence of the device conversion efficiency and noise is well described by a simple thermal model. We have used two methods of inferring dR/dT, and describe the predictions for the magnitude of the efficiency and noise based on these. Neither method provides consistent quantitative predictions of the magnitude of device performance for a variety of operating conditions. Thus the device performance, while excellent, must continue to be investigated experimentally. We have therefore begun a series of experiments on lower Tc devices made of Al, which may have improved performance.
We have measured the spectrum of the output noise and the conversion gain from 0.1-7.5 GHz under identical conditions for both diffusion and phonon-cooled Nb bolometers, using a 20 GHz LO on a variety of devices varying in length from 0.08 izm to 3 Ara, where the gain-bandwidth varies between 100 MHz and > 6 GHz. In this paper, we will present systematic comparison between theory and experiment for the devices measured. We find the frequency dependence of the device conversion efficiency and noise is well described by a simple thermal model. We have used two methods of inferring dR/dT, and describe the predictions for the magnitude of the efficiency and noise based on these. Neither method provides consistent quantitative predictions of the magnitude of device performance for a variety of operating conditions. Thus the device performance, while excellent, must continue to be investigated experimentally. We have therefore begun a series of experiments on lower Tc devices made of Al, which may have improved performance.
A systematic study of the intermediate frequency noise bandwidth of Nb thin-film superconducting hot-electron bolometers is presented. We have measured the spectrum of the output noise as well as the conversion efficiency over a very broad intermediate frequency range (from 0.1 to 7.5 GHz) for devices varying in length from 0.08 μm to 3 μm. Local oscillator and rf signals from 8 to 40 GHz were used. For a device of a given length, the spectrum of the output noise and the conversion efficiency behave similarly for intermediate frequencies less than the gain bandwidth, in accordance with a simple thermal model for both the mixing and thermal fluctuation noise. For higher intermediate frequencies the conversion efficiency decreases; in contrast, the noise decreases but has a second contribution which dominates at higher frequency. The noise bandwidth is larger than the gain bandwidth, and the mixer noise is low, between 120 and 530 K (double side band).
We report a systematic, comprehensive set of measurements on the dynamics and noise processes in diffusion and phonon-cooled superconducting hot-electron bolometer mixers which will serve as ultralow noise detectors in THz heterodyne receivers. The conversion efficiency and output noise of devices of varying lengths were measured with radio frequency between 8 and 40 GHz. The devices studied consist of 100-Å-thin film Nb bridges connected to thick (1000 Å), high conductivity normal metal (Au) leads. The lengths of the devices studied range from 0.08 to 3 μm. For devices longer than the electron–phonon interaction length Le–ph≡Dτe–ph, with D the diffusion constant and τe–ph−1 the electron–phonon interaction rate, the hot electrons are cooled dominantly by the electron–phonon interaction, which in Nb is too slow for practical applications. If the device length is less than πLe–ph(≈1 μm at 4.2 K), then out diffusion of heat into the high conductivity leads dominates the cooling process. In this limit, the intermediate frequency (IF) bandwidth is found to vary as L−2, with L the bridge length, as expected for diffusion cooling. The shortest device has an IF bandwidth greater than 6 GHz, the largest reported for a low-Tc superconducting bolometric mixer. The dominant component of the output noise decreases with frequency in the same manner as the conversion efficiency, consistent with a model based on thermal fluctuations. The noise bandwidth is larger than the gain bandwidth, and the mixer noise is low, ranging from 100 to 530 K (double sideband). The crossover from phonon dominated to diffusion dominated behavior is also demonstrated using noise thermometry measurements in the normal state. Scalar measurements of the device differential impedance in the intermediate state agree with a theoretical model which takes into account the thermal and electrical dynamics. We also present detailed comparisons with theoretical predictions of the output noise and conversion efficiency.
We present studies of the input and output noise of diffusion cooled hot-electron bolometer mixers. By simultaneously measuring the gain and noise (with a 14 GHz LO) as a function of intermediate frequency for a 0.16 /spl mu/m diffusion cooled Nb device, we show that the noise bandwidth (4 GHz) is larger than the gain bandwidth (2.4 GHz). The output noise is 55 K, and the mixer noise is very low, 340 K DSB. This shows that diffusion cooled devices have low noise over a broad enough intermediate frequency band for practical applications in THz receivers.
We report on microwave measurements of superconducting aluminum hot-electron bolometers (Al HEBs). Diffusion-cooled Al HEB mixers are good candidates for space-borne applications in the Terahertz frequency range since they are predicted to have small local oscillator (LO) power requirements, intermediate frequency (IF) bandwidths /spl gsim/10 GHz, and a noise temperature lower than that of Nb and NbN HEB mixers. Mixer measurements were made at an LO frequency /spl sim/30 GHz, with an IF in the range 0.1 to 7.3 GHz. For T<0.8 K, a magnetic field H=0.1-0.3 T was applied to suppress the superconductivity in the contact pads, and partly in the bridge. For a 0.6 /spl mu/m long device, we measure an IF bandwidth of 4 GHz, a conversion efficiency /spl eta/=-8 dB, and a mixer noise temperature T/sub M//spl gsim/4 K, DSB (T/sub mixer/=T/sub output noise//2/spl eta/). These results are shown to be in quantitative agreement with simple theoretical predictions.
Measurements of the spectrum of the output noise of Nb hot-electron mixers are presented. The device lengths vary from 0.08 pm to 3 pm, allowing the investigation of the crossover from phonon to diffusion cooling. The data span the IF range from 75 MHz to 7.5 GHz. The dominant contribution to the noise is thermal fluctuation noise, which is expected to display a one pole rolloff. We present data with 20 GHz LO applied in two cases: with LO power optimized for gain, and LO power optimized for mixer noise. We find that in the optimum gain case, the spectrum is in agreement with simple theoretical predictions, while in the minimum mixer noise case, the output noise is too low (< 15 K) to allow accurate measurements of the spectrum.
A reduction in the mixer noise is expected when using superconductors with a lower transition temperature (Tc) since the thermal noise components of the mixer noise should scale with Tc. Also, the local oscillator (LO) power required for a diffusion-cooled device should decrease as Tc when Tbath<<Tc. We previously studied mixing in aluminum based hot-electron bolometers (HEBs) at microwave frequencies (~30 GHz), and observed a significant improvement in noise performance, and a reduction in LO power as predicted. However, the bias voltage range over which good mixer performance was observed was ~ 5-10 μV. These devices are thus susceptible to saturation effects, in particular output saturation. In the present work, we have investigated Nb HEBs whose Tc is lowered by applying a magnetic field. The goal is to study a case intermediate between Nb and Al, and hopefully to find properties that will allow use in practical receivers. A 15 kOe perpendicular magnetic field was applied to a Nb HEB (L=0.16 μm, W=0.08 μm, RN=90Ω) to reduce Tc from 5.2 K to 2.4 K. The mixer noise, as inferred from the output noise and the conversion efficiency, decreased from 390 K, DSB to 171 K, DSB. The LO power required for near optimum mixer conversion efficiency (ηmixer = 9 dB in this device) was 8nW in zero field, and ~ 2 nW when Tc was reduced to 2.4K. Tbath=0.22K. The conversion bandwidth was previously measured to be 2.4 GHz and the same bandwidth was observed in the presence of a magnetic field. By lowering Tc, the voltage range over which good mixing was observed also decreased. However, even with Tc reduced to 2.4 K, the conversion efficiency dropped by 3dB from its maximum value only when the bias voltage was changed by ~ 90 μV. Saturation effects should thus be much less of a concern in these devices than in Al HEBs. In situations where the application of a large magnetic field is not feasible, we suggest using Ta based HEBs. Ta HEBs should have Tc ~ 3-3.5K and material properties very similar to Nb.