Abstract: Hot-electron bolometer mixers (HEB’s) are the most promising devices as mixing element for terahertz spectroscopy and astronomy at frequencies beyond 1.4 THz. They have a low noise temperature and low demands on local oscillator (LO) power. 1,2 An important limitation is the IF bandwidth, of the order of a few GHz, and which in principle depends on energy relaxation due to electron- phonon processes and on diffusion-cooling. It has been proposed by Prober that a reduction in length of the HEB would lead to an increased bandwidth. 3 This appeared to be achieved by Tretyakov et al by measuring the gain bandwidth close to the critical temperature of the NbN. 2 Unfortunately, the noise bandwidth of similar devices operated at temperatures around 4.2 K appear not depend on the length. The fundamental problem to be addressed is the position-dependent superconducting state of the HEB- devices under operating conditions, which determines the conditions for the cooling of the hot quasiparticles. Some progress has been made by Barends et al in a semi-empirical model to describe the I,V curves under operating conditions at a bath temperature around 4.2 K. 4 In more recent work Vercruyssen et al have analyzed the I,V curve, without any LO-equivalent bias, of a model NSN system. 5 This work suggests that the most appropriate model for an HEB under operating conditions is that of a potential-well in the superconducting gap in the center of the NbN, analogous the bimodal superconducting state described by Vercruyssen et al. Hot quasiparticles in the well can not diffuse out and can only cool by electron-phonon processes, those with higher energies than the heights of the walls of the well can diffuse out. Using this working hypothesis we have carried out experiments on a sub-micrometer NbN bridge connected to a gold (Au) planar spiral antenna. An in situ process is used to deposit Au on NbN. The Au is removed in the center to define the uncovered NbN, which will act as the superconducting mixer itself. The antenna is deposited on the remaining Au layer on the NbN. The Au contacts suppress the energy gap of the NbN film located underneath the gold layer 7,8 . The measured resistive transition is shown in Fig.1. It clearly shows a T c of the bilayer at 6.2 K and the resistive transition of the NbN itself around 9 K. In addition we show the measured noise bandwidth (red squares) for different bath temperatures. Clearly the noise bandwidth increases strongly by increasing the bath temperature from 5 K to 8 K, up to 13 GHz. We interpret this pattern as evidence for improved out-diffusion of hot electrons due to normal banks and a shallow superconducting potential well compared to k B T. As expected the noise temperature in this regime is much bigger than when biased at 4.2 K. R EFERENCES 1 W. Zhang, P. Khosropanah, J. R. Gao, E. L. Kollberg, K. S. Yngvesson, T. Bansal, R. Barends, and T. M. Klapwijk Appl. Phys. Lett. 96, 111113, (2010). 2 Ivan Tretyakov, Sergey Ryabchun, Matvey Finkel, Anna Maslennikova, Natalia Kaurova, Anastasia Lobastova, Boris Voronov, and Gregory Gol’tsman Appl. Phys. Lett. 98, 033507 (2011). 3 D. E. Prober, Appl. Phys. Lett. 62, 2119 (1992). 4 R. Barends, M. Hajenius, J. R. Gao, and T. M. Klapwijk, Appl. Phys. Lett. 87, 263506 (2005). 5 N. Vercruyssen, T. G. A. Verhagen, M. G. Flokstra, J. P. Pekola, and T. M. Klapwijk Physical Review B 85, 224503 (2012).

Abstract: We present a detailed experimental study of the direct detection effect in a small volume (0.15pm x lpm) NbN hot electron bolometer mixer. It is a quasioptical mixer with a twin slot antenna designed for 700 GHz and the measurement was done at a LO frequency of 670 GHz. The direct detection effect is characterized by a change in the mixer bias current when switching broadband radiation from a 300 K hot load to a 77 K cold load in a standard Y factor measurement. The result is, depending on the receiver under study, an increase or decrease in the receiver noise temperature. We find that the small signal noise temperature, which is the noise temperature that would be observed without the presence of the direct detection effect, and thus the one that is relevant for an astronomical observation, is 20% lower than the noise temperature obtained using 300 K and 77 K calibration loads. Thus, in our case the direct detection effect reduces the mixer sensitivity. These results are in good agreement with previous measurement at THz frequencies [1]. Other experiments report an increase in mixer sensitivity [2]. To analyze this discrepancy we have designed a separate set of experiments to find out the physical origin of the direct detection effect. Possible candidates are the bias current dependence of the mixer gain and the bias current dependence of the IF match. We measured directly the change in mixer IF match and receiver gain due to the direct detection effect. From these measurements we conclude that the direct detection effect is caused by a combination of bias current reduction when switching form the 77 K to the 300 K load in combination with the bias current dependence of the receiver gain. The bias current dependence of the receiver gain is shown to be mainly caused by the current dependence of the mixer gain. We also find that an increase in receiver sensitivity due to the direct detection effect is only possible if the noise temperature change due to the direct detection is dominated by the mixer-amplifier IF match. [1] J.J.A. Baselmans, A. Baryshev, S.F. Reker, M. Hajenius, J.R. Gao, T.M. Klapwijk, Yu.Vachtomin, S. Maslennikov, S. Antipov, B. Voronov, and G. Gol'tsman., Appl. Phys. Lett. 86, 163503 (2005). [2] S. Svechnokov, A. Verevkin, B. Voronov, E. Menschikov. E. Gershenzon, G. Gol'tsman, 9th Int. Symp. On Space THz. Techn., 45, (1999).

Abstract: We present a detailed experimental study of the direct detection effect in a small volume (0.15μm×1μm×3.5nm) quasioptical NbN phonon cooled hot electron bolometer mixer at 673GHz. We find that the small signal noise temperature, relevant for an astronomical observation, is 20% lower than the noise temperature obtained using 300 and 77K calibration loads. In a separate set of experiments we show that the direct detection effect is caused by a combination of bias current reduction when switching from the 77 to the 300K
load in combination with the bias current dependence of the receiver gain. The bias current dependence of the receiver gain is shown to be mainly caused by the current dependence of the mixer gain.