IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 17, NO. 2, JUNE 2007
399
Optimized Sensitivity of NbN Hot Electron Bolometer Mixers by Annealing M. Hajenius, Z. Q. Yang, J. R. Gao, J. J. A. Baselmans, T. M. Klapwijk, B. Voronov, and G. Gol’tsman
Abstract—We report that the heterodyne sensitivity of superconducting hot-electron bolometers (HEBs) increases by 25–30% after annealing at 85 C in high vacuum. The devices studied are twin-slot antenna coupled mixers with a small area NbN bridge of 1 m 0 15 m, above which there is a SiO2 passivation layer. The mixer noise temperature, gain, and resistance versus temperature curve of a HEB before and after annealing are compared and analysed. We show that the annealing reduces the intrinsic noise of the mixer by 37% and makes the superconducting transition of the bridge and the contacts sharper. We argue that the reduction of the noise is mainly due to the improvement of the transparency of the contact/film interface. The lowest receiver noise temperature of 700 K is measured at a local oscillator frequency of 1.63 THz and at a bath temperature of 4.2 K. Index Terms—HEB, heterodyne mixing, terahertz.
I. INTRODUCTION
S
IGNIFICANT progress in their development has made NbN hot electron bolometer (HEB) mixers suitable for diverse ground and space applications [1]–[4]. HEBs based on a thin superconducting NbN film with a fast electronphonon cooling are particularly attractive due to their high sensitivity and large intermediate frequency (IF) bandwidth. Theoretically it has been predicted that such mixers could have an extremely low mixer noise temperature, determined by the sharpness of the transition of the superconductor and the critical temperature [5]. The best ) double sideband (DSB) receiver noise temperature ( , reported so far is 950 K at 2.5 THz, corresponding to 8 where is Planck’s constant, the frequency and is Boltzmann’s constant. This result was achieved in a spiral-antenna m) by coupled to a relatively large NbN HEB ( m cleaning the surface of the NbN film, followed by the deposition of a superconducting NbTiN inter-layer before a standard Au layer is deposited to form the contacts [4], [6]. Using the of 900 K has been obtained for a same technology, a m) twin-slot antenna coupled small NbN HEB ( m mixer at 1.6 THz [7]. Despite of the fact that similar results have Manuscript received August 29, 2006. This work was supported by Radionet (EU) and INTAS. M. Hajenius and J. R. Gao are with the Kavli Institute of Nanoscience, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, the Netherlands, and also with the Space Research Organisation of the Netherlands (SRON), Sorbonnelaan 2, 3584 CA Utrecht, the Netherlands (e-mail:
[email protected]). Z. Q. Yang and J. J. A. Baselmans are with SRON, Sorbonnelaan 2, 3584 CA Utrecht, the Netherlands. T. M. Klapwijk is with the Kavli Institute of Nanoscience, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, the Netherlands. B. Voronov and G. Gol’tsman are with the Department of Physics, Moscow State Pedagogical University, M. Pirogovskaya st. 1, Moscow 119435, Russia. Digital Object Identifier 10.1109/TASC.2007.898064
Fig. 1. Current-voltage characteristics of a NbN hot electron bolometer mixer without and with local oscillator power in nW, before annealing. Insets: (a) cross-sectional view of the HEB mixer, which shows the NbN bridge, the NbN/NbTiN/Au contacts, and the NbN/Au layer for RF CPW transmission line; (b) top view of the HEB mixer before the deposition of the SiO layer, where the bridge, antenna, and RF filter structure are indicated. The transmission line (not indicated) connects the HEB to the antenna.
been obtained in devices fabricated in different runs, fluctuations in the sensitivity occur. Insufficient understanding of the performance related device physics makes a further improvement of the mixer’s properties difficult. The mixer’s sensitivity must relate to the NbN bridge, the contacts, and the antenna plus RF coupling circuit. But it is hard in practice to distinguish among the influence of any of these three factors in case different devices show different performance. We therefore propose a way to modify mixer parameters of one and the same device. We apply soft annealing at 85 C under high vacuum to completed NbN HEB mixers (twin-slot antenna coupled), allowing comparison of the performance related exclusively to the bridge and its contacts. We show that the sensitivity of “poor” NbN HEB mixers can be improved by annealing, which changes their properties as reflected in sharper resistive transitions of the complete device. We also observe that annealing can reduce the receiver noise temperature at 1.6 THz by 25–30% [8]. II. EXPERIMENTS The NbN HEB mixers studied in the present work are fabricated using a standard thin NbN film with a thickness of 5.5 nm (determined from a transmission electron microscopy study [9], although 3.5 nm is expected from sputter-rate calibration) on a Si substrate. The mixer’s NbN bridge has a superconducting transition temperature of 9.5 K, and is coupled to a twin-slot antenna designed for 1.6 THz (see the inset in Fig. 1). The bridge is 1 m wide and 0.15 m long and the contacts are made of
1051-8223/$25.00 © 2007 IEEE
400
IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 17, NO. 2, JUNE 2007
NbTiN/Au. Before deposition of the contact layers, a short (6 s) O plasma etch is applied to remove resist remnants followed by a short (15 s) in-situ Ar sputter etch to clean the film surface. These are critical technological steps, which can result in fluctuations of the interface quality from batch to batch. The fabrication process is very similar to our early work [6], except for the definition of the antenna and the passivation layer. The passivation layer consists of 500 nm SiO and is sputtered on top of the active region of the HEB mixer, to prevent aging of the NbN bridge under normal laboratory conditions. Note that the e-beam resist used to define the bridge width [6] still remains. Although several devices were measured, we focus here on one ) of 153 at 16 K and a having a normal state resistance ( of the NbN film of around 9 K. Annealing slightly reduced has been performed in an oven connected to a turbomolecular pump. Before switching on and fixing it to 85 C, the oven has mbar. been pumped for 24 hours reaching a pressure of Under these conditions the devices were annealed for 72 hours. This is a standard procedure for space qualification testing. We measure the receiver noise temperature, the gain, and the curve before and after DC resistance versus temperature annealing. We use a RF (radio frequency) test setup for a typical -factor measurement. A standard quasi-optical technique is applied to couple RF signal from free space to the HEB. The HEB chip is glued on the backside of an elliptical silicon lens. The lens is placed in a metal mixer block, thermally anchored to the 4.2 K cold plate of a vacuum cryostat. Blackbody signal sources (Eccosorb) are used to define a hot thermal load at 295 K and a cold thermal load at 77 K. The signal is combined with the local oscillator (LO) signal via a 3.5 m thick Mylar beam splitter. Both of the signals pass further into the lens through a 1.1 mm thick high density polyethylene (HDPE) window (RF loss of 1.1 dB) and a Zytex G104 ( 0.45 dB) heat filter at 77 K. The LO source is an optically pumped gas laser at 1.63 THz and is attenuated with a rotable grid. The IF signal is amplified by a cryogenic low noise amplifier with 40 dB gain and a noise temperature of 5 K. The signal is further amplified by a room temperature amplifier with 41 dB gain and is filtered in a 80 MHz bandwidth at 1.4 GHz before detection with a power meter. The -factor used for calculating the receiver noise temperature, is the ratio of the measured IF output powers responding to the hot and cold load [10]. The DC resistance is measured using a standard lock-in technique. III. RESULTS AND DISCUSSION A. Current-Voltage Characteristics In Fig. 1 we plot the current-voltage ( ) curves of the device, before annealing, without and with applying LO power. We also measured these curves after annealing and find a small reduction of the critical current from 68 A to 65 A and a small (4 ) increase of . The change of is so small that no effects on the impedance matching between antenna and curves are similar. In bolometer are expected. The pumped curve, where the lowest noise temparticular, the optimal perature is obtained, remains unchanged (although it depends curves at different on the LO power level). Based on the
Fig. 2. Double sideband receiver noise temperature of the hot electron bolometer mixer at 1.63 THz as a function of bias voltage obtained before and after annealing. Symbol notation: squares and triangles—before and after annealing, measured using an uncoated lens; filled triangles—after annealing measured using an anti-reflection coated lens and a cold bandpass filter (BF) to reduce the direct detection effect.
pumping levels, the LO power absorbed by the HEB is estimated using the isothermal technique [11]. The optimal LO pumping level is around 50 nW for both cases. B. Double Sideband Receiver Noise Temperature The is calculated according to the Callen and Welton definition [10] and is plotted in Fig. 2 as a function of bias voltage at the optimal LO power before and after annealing. Note that in both cases an uncoated Si lens is used and the total RF loss in the signal path is 4.35 dB. As indicated after annealing decreases signifiin the figure, the cantly over the whole voltage bias range. At the optimal bias decreases by 27% and becomes point (0.6 mV), the 1050 K, which is slightly lower than the previous best result for the case of the uncoated lens [7]. Similar improvement was observed in four other devices, showing a decrease of 25–30% . in C. Single Sideband Mixer Noise Temperature and Conversion Gain The reflects the effective noise temperature of the cascade of the optics (all of the optical components and the transmission efficiency of the atmosphere), mixer, and IF amwas plifier. The single sideband (SSB) mixer gain obtained by measuring the difference in total IF output power between a hot load (295 K) and cold load (77 K) and using the effective temperatures of the hot/cold load at the mixer input. After subtracting the contributions from the optics and IF amplifier, we can derive the SSB mixer noise temperature . Both are plotted as a function of bias voltage in Fig. 3. We observe that after annealing, the mixer gain hardly changes. Around the optimal bias point, it increases by less than decreases by 13%. The mixer noise temperature is 590 K. Since the output noise 44%. The lowest of a HEB equals , the annealing essentially reduces the intrinsic noise by 37%.
HAJENIUS et al.: OPTIMIZED SENSITIVITY OF Nbn HOT ELECTRON BOLOMETER MIXERS BY ANNEALING
401
Fig. 5. The resistance of an initially “poor” HEB mixer as a function of temperature before (solid line) and after (dash-dotted line) annealing. T , T , and T correspond to the transition temperature of the bridge, the contact pads and the transmission line/RF filter structure respectively (see Fig. 1). The inset shows the R T of an initially already good device which does not show the sensitivity improvement after annealing.
( )
Fig. 3. (a) Conversion gain (SSB) of the HEB mixer as a function of bias voltage; (b) mixer noise temperature (SSB). The open squares denote the value before annealing; the open triangles—after annealing.
Fig. 4. Measured direct response of the twin slot antenna coupled mixer before and after annealing using an evacuated Fourier transform spectrometer. No corrections to the data are performed.
D. Discussions In analyzing these results, we assume that the coupling of the RF signal via the antenna and transmission line into the HEB remains unchanged by the annealing. This is supported by the gain data and by the measured direct response of the HEB using a Fourier transform spectrometer (see Fig. 4), which shows essentially the same spectrum after annealing. Hence, we assume that the dominant cause of improvement is due to changes of the device properties. For this reason, we compare the measured characteristics before and after annealing (the main plot in Fig. 5). We observe three superconducting transitions in the curve. According to our earlier study [12] we can define (the highest) for the bridge, (the middle) for the contact-pads, and (the lowest) for the transmission line plus RF filter structure. Obviously, annealing affects all three aspects of and a the HEB devices. The HEB’s NbN film has a reduced . The annealing also makes the supernarrower transition conducting transition in the curve at (the contacts)
TABLE I THE CRITICAL TEMPERATURE AND THE TRANSITION WIDTH OF THE NBN BRIDGE (T , T ), THE CONTACTS (T , T ), AND THE “RF FILTER STRUCTURE” (T , T ), BEFORE AND AFTER ANNEALING
1
1
1
and (the transmission line plus RF filter structure) considerably sharper, and it leads to a small down-shift of . A striking , result is that after annealing a remaining resistance below attributed to the transmission line plus RF filter structure, disappears. In Table I we list the various critical temperature values before and after annealing. For comparison we briefly describe a similar annealing experiment with HEBs, which are already close to optimum performance before annealing. These devices show excellent sensitivity, comparable to results reported before [7]. The decurve similar to the one vices also have a relatively sharp observed after annealing (see the inset in Fig. 5). Using these devices, the annealing gives either no change or improves the slightly. The resistive transireceiver noise temperature tion, before and after annealing, changes in the following way: the moves slightly downwards and the width of the transition changes slightly. On the other hand the transition associated with the contacts changes in a different way. The critical (due to the contacts) and (RF wiring) temperatures do not change, neither do the resistance values, except for the RF wiring which also turns superconducting. Apparently, annealing does not alter the contacts in these optimal devices, whereas in the poorly performing devices annealing improves the contacts to become comparable to the best devices. This is reminiscent of previous results [4], [6] demonstrating that cleaning and contact structure influence the performance of the mixer significantly. The resistive transition reflects changes in the devices, which carry over into the current-voltage characteristic in a complicated manner [13]–[16]. The latter is primarily controlled by . It is the resistance of the NbN itself characterized by striking that the pumped current-voltage characteristics hardly
402
IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 17, NO. 2, JUNE 2007
change suggesting that the curves of the NbN itself are curves which depend unchanged. However, the relevant also on the current and are needed for properly describing the are not available. It is worthwhile to mention that the lumped element model [5] developed for an ideal HEB is not able to explain our experimental results. The inconsistency illustrates that a more realistic model for HEB’s including all important ingredients such as superconducting transition, the contacts, temperature profile, and noise is needed. We note that progress has been made [17], however, the effect of the superconducting transition and the contacts remains absent. E. Ultimate Receiver Noise Temperature To determine the ultimate DSB receiver noise temperature of the device after annealing, we reduce the RF loss (1 dB) in the optics by using a Si lens coated with an antireflection layer of 29 m thick parylene C. In addition, we also add a metal mesh RF bandpass filter with an effective bandwidth of 200 GHz centered at 1.6 THz, mounted on the 4.2 K cold plate, which filters the hot/cold load power and thus reduces the direct detection effect [18] of broadband radiation. This effect, which can reduce the measured -factor, is particularly important in the case of the small area HEBs [18] requiring low LO power. The measured new receiver noise temperature is also plotted in Fig. 2. We observe that the receiver noise temperature decreases furof 740 K around the optimal ther, with the lowest operating point (0.4 mV). In this case the total optical loss is 3.87 dB. By flushing the signal path with dry gas, we measured a of 700 K (not shown in Fig. 2), corresponding . This value is identical to the record sensitivity reto 9 m) HEB ported in a spiral-antenna coupled large ( m mixer at the same frequency [3], [19]. IV. CONCLUSION In conclusion, we have demonstrated that annealing can reduce the receiver noise temperature of NbN HEB mixers. The analysis shows that the annealing reduces the intrinsic noise of the mixer by 37%, but has little effect on the mixer gain. The measured lowest DSB receiver noise temperature is 700 K at 1.63 THz, which is uncorrected for any optical loss. The lowest SSB mixer noise temperature, which is a fundamental property of the mixer, is estimated to be 590 K at the same frequency. before and after annealing, we attribute the Based on the improvement in sensitivity to the changes in the contact/NbN interface of the devices. We also found that despite the change curve, the optimally pumped current-voltage curves reof main the same. ACKNOWLEDGMENT We acknowledge R. Barends for stimulating discussions and W. J. Vreeling for his assistance during the measurements.
REFERENCES [1] E. M. Gershenzon, G. N. Gol’tsman, I. G. Gogidze, A. I. Eliantev, B. S. Karasik, and A. D. Semenov, “Millimeter and submillimeter range mixer based on electron heating of superconducting films in the resistive state,” Sov. Phys. Superconductivity, vol. 3, p. 1582, 1990. [2] A. D. Semenov, H.-W. Hübers, J. Schubert, G. N. Gol’tsman, A. I. Elantiev, B. M. Voronov, and E. M. Gershenzon, “Design and performance of the lattice-cooled hot-electron terahertz mixer,” J. Appl. Phys., vol. 88, no. 11, pp. 6758–6767, 2000. [3] S. Cherednichenko, P. Khosropanah, E. Kollberg, M. Kroug, and H. Merkel, “Terahertz superconducting hot-electron bolometer mixers,” Physica C, vol. 372–376, pp. 407–415, 2002. [4] J. J. A. Baselmans, M. Hajenius, J. R. Gao, T. M. Klapwijk, P. A. J. de Korte, B. Voronov, and G. Gol’tsman, “Doubling of sensitivity and bandwidth in phonon cooled hot electron bolometer mixers,” Appl. Phys. Lett., vol. 84, pp. 1958–1960, 2004. [5] B. S. Karasik and I. Elantiev, “Noise temperature limit of a superconducting hot-electron bolometer mixer,” Appl. Phys. Lett., vol. 68, pp. 853–855, 1996. [6] M. Hajenius, J. J. A. Baselmans, J. R. Gao, T. M. Klapwijk, P. A. J. de Korte, B. Voronov, and G. Gol’tsman, “Low noise NbN superconducting hot electron bolometer mixers at 1.9 and 2.5 terahertz,” Supercond. Sci. Technol., vol. 17, pp. S224–S228, 2004. [7] J. J. A. Baselmans, M. Hajenius, J. R. Gao, P. A. J. de Korte, T. M. Klapwijk, B. Voronov, and G. Gol’tsman, “Doubling of sensitivity and bandwidth in phonon cooled hot electron bolometer mixers,” in Proc. of SPIE, Glasgow, UK, 2004, vol. 5498, p. 168. [8] Z. Q. Yang, M. Hajenius, J. J. A. Baselmans, J. R. Gao, B. Voronov, and G. N. Gol’tsman, “Reduced noise in NbN hot electron bolometer mixers by annealing,” Supercond. Sci. Technol, vol. 19, pp. L9–L12, 2006, (Rapid Com.). [9] J. R. Gao, M. Hajenius, F. D. Tichelaar, B. Voronov, E. Grishina, T. M. Klapwijk, G. Gol’tsman, and C. A. Zorman, “Can NbN films on 3CSiC/ Si change the IF bandwidth of hot electron bolometer mixers?,” in Proc. of the 17th International Symposium on Space Terahertz Technology, Paris, France, 2006, (in press). [10] A. R. Kerr, “Suggestions for revised definitions of noise quantities, including quantum effects,” IEEE Trans. Microwave Theory and Technology, vol. 47, no. 3, pp. 325–329, 1999. [11] H. Ekström, B. S. Karasik, E. L. Kollberg, and K. S. Yngvesson, “Conversion gain and noise of niobium superconducting hot_electron mixers,” IEEE Trans. Microwave Theory and Technology, vol. 43, pp. 938–947, 1995. [12] M. Hajenius, R. Barends, J. R. Gao, T. M. Klapwijk, J. J. A. Baselmans, A. Baryshev, B. Voronov, and G. Gol’tsman, “Local resistivity and the current-voltage characteristics of Hot Electron Bolometer mixers,” IEEE Trans. Applied Superconductivity, vol. 15, pp. 495–498, 2005. [13] D. Wilms Floet, E. Miedema, T. M. Klapwijk, and J. R. Gao, “Hotspot mixing: A framework for heterodyne mixing in superconducting hot electron bolometers,” Appl. Phys. Lett., vol. 74, pp. 433–435, 1999. [14] H. Merkel, P. Khosropanah, P. Yagoubov, and E. Kollberg, “A hot spot mixer model for superconducting phonon–cooled HEB’s far above the quasiparticle bandgap,” in Proc. of the 10th International Symposium on Space Terahertz Technology, Charlottesville, USA, 1999, p. 592. [15] T. M. Klapwijk, R. Barends, J. R. Gao, M. Hajenius, and J. J. A. Baselmans, “Improved superconducting HEB devices for the THz range,” in Proc. of SPIE, Glasgow, UK, 2004, vol. 5498, p. 129. [16] R. Barends, M. Hajenius, J. R. Gao, and T. M. Klapwijk, “Current induced vortex unbinding in bolometer mixers,” Appl. Phys. Lett., vol. 87, p. 263506, 2005. [17] E. L. Kollberg and K. S. Yngvesson, “Quantum-noise theory for terahertz hot electron bolometers,” IEEE Trans. Microwave Theory and Techniques, vol. 54, p. 2077, 2006. [18] J. J. A. Baselmans, A. Baryshev, S. F. Reker, M. Hajenius, J. R. Gao, T. M. Klapwijk, B. Voronov, and G. Gol’tsman, “Direct detection effect in small volume hot electron bolometer mixers,” Appl. Phys. Lett., vol. 86, pp. 163503–163505, 2005. [19] M. Kroug, S. Cherendnichenko, H. Merkel, E. Kollberg, B. Voronov, G. Gol’tsman, H.-W. Hübers, and H. Richter, “NbN hot electron bolometric mixers for terahertz receivers,” IEEE Trans. Applied Superconductivity, vol. 11, pp. 962–965, 2001.
