A superconducting transition edge sensor (TES) bolometer operating in the spectral range from 0.1 THz to 3 THz was designed. It is especially intended for Fourier transform spectroscopy and features a higher dynamic range and a highly linear response at a similar response compared to commercially available silicon composite bolometers. The design is based on a thin film metal mesh absorber, a superconducting thermistor and Si3N4 membrane technology. A prototype was set up, characterized and successfully used in first applications.
© 2015 Optical Society of America
Due to a number of advantages over dispersive spectrometers, Fourier transform spectrometers (FTS) are the preferred instruments for optical spectroscopy in the MIR and FIR range . In radiometric applications often spectrally broad sources and / or optical components with wide spectral features are investigated with an FTS. The broad spectrum leads to an interferogram with a pronounced central peak and features a large dynamic range. Under these conditions detector nonlinearities lead to artifacts in the measured spectrum which can hardly be corrected . Therefore, a large linear range is required of detectors which are applied in radiometry with an FTS. In particular in the spectral range from 0.1 THz to 3 THz a lack of suitable detectors can be identified. In this spectral range usually pyroelectric detectors or silicon composite bolometers are applied for Fourier transform spectroscopy. The pyroelectric detectors are limited in their sensitivity due to their relatively high noise equivalent power. The silicon-composite bolometers have a limited linear range. Furthermore, to a large linear range also a high and uniform absorptance in the addressed spectral range is of interest. This would enable a more general use of the detector in broadband radiometric applications.
We address all these requirements with a specially designed superconducting transition edge sensor (TES) bolometer operating in the spectral range from 0.1 THz to 3 THz. The design is based on a thin film metal mesh absorber, a superconducting thermistor, Si3N4 membrane technology and a low-noise SQUID readout circuit.
In general a bolometer consists of an absorber for incident radiation, a thermistor to measure the change of the temperature of the absorber and a thermal link to a heat sink. In a composite design these elements can be optimized separately. Our bolometer is based on Si3N4 membrane technology and fabricated in lithographic processes. It features a microstructured metal absorber and a transition edge sensor (TES) as the thermistor. It is operated within a liquid helium cryostat, and the bath temperature of the heat sink is 4.2 K.
The bolometer is shown in Fig. 1. It has an overall size of (6.6 × 6.6) mm2. A solid frame of silicon (dark area in Fig. 1) which is coated with Si3N4 on both sides realizes the heat sink. The inner area of this frame (light area in Fig. 1) is an unsupported 1 μm thin Si3N4 membrane made by the chemical etching of the Si3N4 coated silicon from one side. The Si3N4 membrane acts as the support structure for the microstructured metal absorber and realizes the thermal link with the required weak thermal coupling between the absorber and the heat sink. The absorber with a dimension of (3.3 × 3.3) mm2 is the central gray area in Fig. 1. The design and characterization of the absorber is described in detail in Section 3.
The thermistor and its leads are placed on the same side of the membrane as the absorber. The thermistor itself with dimensions of (10 × 10) μm2 is located in the center of the device covering just a small fraction of the surface area. We use a TES  as a thermistor which is well established for radiation detection at cryogenic temperatures. A TES utilizes the very steep resistance transition between its superconducting and normal conducting state. The design and the characterization of the TES, its operating scheme and readout are described in Sections 4 and 5.
As an additional design feature the absorber is electrically contacted. This enables an electrical heating of the absorber for the characterization of the bolometer without incident radiation.
A spectrally flat and sufficiently high absorptance of optical radiation in the FIR and THz spectral range can be realized by metallic absorbers . Matching the sheet resistance to Z0/2 = 188.5 Ω/□ (“Ohms per square” is the common unit to describe a sheet resistance), which is half of the impedance of free space Z0 = (μ/∊)1/2 = 377 Ω, leads to a frequency independent maximum absorptance of 0.5 for the homogenous unsupported metal layer in the THz range [5, 6]. However, technological limits are reached by fabricating a high quality metal film with a sheet resistance of R = 188.5 Ω/□ made from materials which are accessible in standard thin film deposition. Due to the relatively low resistance of these metals at low temperatures, corresponding films would have a thickness of around 1 nm. This thickness is difficult to achieve with a good reproducibility. Therefore, our approach is to apply microstructured metal absorbers with a structural size much smaller than the shortest relevant wavelength. Two different absorber designs were investigated. Both are shown in Fig. 2. In Fig. 2(a) the design variant consisting of a metal grid with strip width w and pitch width p is shown. The second design variant shown in Fig. 2(b) has additional square loops inside each mesh with the same strip width w. Both designs are periodical with a pitch p.
Due to the structuring of the absorber the absorptance is no longer completely spectrally uniform, and also the effective sheet resistance of a grid with a nearly uniform absorptance can differ from Z0/2 = 188.5 Ω/□. The reasons are resonances as in metallic mesh filters  and diffraction effects.
A set of samples of both absorber designs with varying design parameters was fabricated and tested for adequate absorptance and spectral uniformity. Variants with pitches ranging from p = 10 μm to p = 60 μm and linewidths of w = 3 μm and w = 6 μm were investigated. Furthermore, the sheet resistance of the metal strips was varied between R = 12 Ω/□ and R = 60 Ω/□ for both designs by varying the thickness of the metal layer. The size of the test structures were (12 × 12) mm2. With a spot size of around 6 mm, effects of the surrounding silicon substrate could be neglected.
The absorptances α of the structures under test were calculated from reflectance ρ and transmittance τ measurements by α = 1 − ρ − τ. The measurements were done with a vacuum FTS using a helium-cooled silicon composite bolometer and a room temperature pyroelectric D-LATGS detector in combination with a high-pressure mercury lamp. A measurement scheme with background correction was applied by measuring at two flux levels of the source .
Changing the design parameters affects the spectral behavior as well as the height of absorptance. This is exemplarily illustrated in Fig. 3 where the absorptance of the absorber designs with loops inside the grid is depicted. Their strip width w as well as in their pitches p were varied at constant sheet resistance R = 20 Ω/□.
Finally, one structure of each design was identified with an absorptance of adequate uniformity in the spectral range from 0.3 THz to 3 THz. Both structures have a strip width of 3 μm for the metal and a sheet resistance of R = 20 Ω/□. The structure without loops in the grid has a pitch of 10 μm, the structure with loops has a pitch of 35 μm. The absorptance of these structures is close to the theoretical maximum of 0.5 as shown in Fig. 4. The remaining spectral nonuniformity in the figure is caused by noise and is not related to any optical features.
A performance criterion of thermistors is the slope of its resistance with respect to temperature. The TES utilizes the steep resistive transition from its superconducting state to its normal conducting state. The TES thermistor used here is designed as a bilayer with layers of a superconducting metal (niobium) and a normal conducting metal (aluminum). The critical temperature Tc of the Nb-Al bilayer is reduced compared to a Nb layer due to the proximity effect . The leads to the TES are made from niobium. The temperature dependence of the resistance of the TES is shown in detail in Fig. 5. In Fig. 5(a) two superconducting transitions at 7.79 K of the TES and around 8.87 K of the niobium leads are clearly visible. While operating the bolometer the Nb leads remain superconducting while the TES is within its transition region.
The superconducting transition of the TES at 7.79 K has a width of 20 mK and is shown enlarged on the right-hand side of Fig. 5 in more detail.
5. Operation and readout circuit
At first glance the operation of the nonlinear TES in a narrow transition region between a superconducting and a normal conducting state might seem to be in contrast to the highly linear detector with a wide dynamic range we are aiming for. However, these requirements are addressed by a dedicated readout scheme. The TES operates in parallel with a shunt resistor Rshunt of 0.5 Ω in a strong voltage-biased condition as shown in the equivalent circuit diagram in Fig. 6. A voltage-biased TES is operated in negative electrothermal feedback (nETF) . Hereby, a highly linear response of the detector and an increased dynamic range are achieved. Furthermore, this operating scheme leads also to a reduced response time by several orders of magnitude below the intrinsic time constant  as the temperature of operation remains nearly constant. In contrast to semiconductor-based bolometers an operation in the kHz frequency range is possible .
Briefly, the TES is stabilized by the nETF as follows. The bias current Ibias is chosen accordingly to operate the TES at a steady state temperature T0 with a corresponding resistance R0 within the superconducting transition by dissipating Joule power in the TES. If optical power is incident on the bolometer, its temperature T and the corresponding resistance R(T) are increased resulting in an instantaneous decrease of the sensor current ITES and the dissipated Joule power. Hereby, the sum of the Joule power and the absorbed radiant power PR is kept constant and the TES remains nearly at T0.
The operation point is determined by the bias current only. For the further operation of the TES a value of 0.7 Rn = 4.97 Ω with a corresponding current Ibias = 10.026 mA was chosen. This value is ten times higher than Rshunt and ensures a stable voltage-biased condition.
The change in current is read out with a SQUID-based low-noise current amplifier operating in flux-locked loop (FLL) mode . The amplifier is operated by our high-speed FLL electronics XXF-1 [12, 13, 14].
6. Optical setup
The TES bolometer is operated in a helium bath cryostat. All internal components are gold-plated as shown in Fig. 7. Radiation enters the cryostat through a wedged diamond window. It then passes a 100 cm−1 low pass filter mounted on a cooled filter wheel. A Winston cone concentrates the radiation onto the absorber of the bolometer. The bolometer and its SQUID readout are mounted on a copper block shown in detail on the right-hand side of Fig. 7. Thermal radiation from the surroundings of the bolometer is blocked by a radiation shield shown in the background on the left-hand side of Fig. 7. Special care was taken that radiation which is reflected from or transmitted through the bolometer does not hit the absorber twice. To this end, the radiation shield is blackened on the inside by an epoxy (Stycast 2850 FT) gritted with silicon carbide with grain size F60. It has been verified that the reflectance of this blackened surface is well below one percent in the FIR spectral range of interest. Furthermore, parallel surfaces on the mounting components were avoided to prevent standing waves. A copper plate is mounted in one position of the wheel to provide a dark enclosure with low thermal background for test measurements, i.e. the electrical characterization.
7.1. Noise equivalent power
The frequently used figure of merit to describe the sensitivity of thermal detectors is the noise equivalent power (NEP), defining an incident power required to obtain a signal equal to the noise of the detector in a one Hz bandwidth (S/N = 1). To determine the NEP, the bolometer was operated at its typical operating point at 0.7 Rn = 4.97 Ω in a dark and helium-cooled environment: the cooled cryostat with the incident radiation blocked by the copper plate in the filter wheel. The output signal of the SQUID electronics was analyzed by a spectrum analyzer. The spectral noise power density shown in Fig. 8 was obtained by a multiplication of the measured current noise density of the TES bolometer with a responsivity of 203 A/W. The responsivity was calculated according to  by the reciprocal of ITESR0 which is the zero-frequency responsivity valid for strong voltage-biased nETF. Ignoring the well-known 1/ f noise of the SQUID amplifier, a plateau between 10 Hz and 100 Hz can be identified. In this regime an electrical NEP of 3.8 × 10−13 W/Hz1/2 was determined which is denoted by a red line.
The NEP can be compared to photon noise resulting from a thermal background. The photon noise is given as the square root of the photon flux absorbed by the detector. It is calculated from the photon flux of the background given by the radiance L(ν, T) according to Planck’s law multiplied by the relevant geometry parameters.
Parameters for the calculations  were an optical throughput of AΩ = 1.2 × 10−5 sr m2, a transmittance of τ = 0.1 of the optical system, an absorptance of the bolometer of α = 0.5, an emissivity of ε = 1 and a temperature of 300 K for the thermal background. The bandwidth was chosen to be Δν = 1 cm−1.
In Fig. 9 the NEP of the bolometer is represented by the horizontal green line. The black line represents the photon noise of the thermal background.
7.2. Determination of the linear response range
In order to determine the linear response range of the bolometer, a source with a power level tunable over several orders of magnitude has been used. PTB operates an electron storage ring, the Metrology Light Source (MLS). The MLS provides broadband THz radiation tunable in its power over 11 orders of magnitude depending on the ring current and the mode of operation [16, 17]. Two different modes of operation were used. In normal mode (incoherent mode) synchrotron radiation could be provided with a maximum THz radiation power in the order of 1 μW. In addition the MLS can be operated in the low-α mode where coherent synchrotron radiation (CSR) with a maximum THz radiation power in the mW range can be provided. For the determination of the linear response range the MLS was operated at five different power levels: in normal mode at ring currents of 10 mA, 20 mA, 70 mA and 140 mA and in low-α mode at 10 mA. To sensitively vary the optical power incident on the bolometer within these five levels, we used a combination of two wire grid polarizers in our setup. Visible and infrared radiation was blocked with a 3 THz long pass filter. The setup is operated under air and is shown in Fig. 10. It consists of three off-axis parabolic mirrors (OAP1 to OAP3) for imaging, three plane mirrors (M1 to M3), two wire grid polarizers (P1 and P2) and the THz filter F. Mirror M3 was mounted on a translation stage to switch between two beam paths of equal length to irradiate the TES bolometer or a reference detector for comparison. The reference detector THz30 from SLT  has been calibrated for its spectral power responsivity at the THz calibration facility of PTB [19, 20].
In the low-α (CSR) mode of operation the optical power incident on the TES bolometer was determined with the calibrated broadband THz detector THz30. This pyroelectric detector operates in a spectral range comparable to the bolometer. Due to its NEP of 2 × 10−7 W/Hz1/2 it can be used for absolute optical power measurement in low-α mode only. The absolute radiant power was measured for different angle settings of the tunable polarizers corresponding to different levels of attenuation of the synchrotron radiation. Afterwards the output of the TES bolometer was recorded for the same angular settings of the polarizers. The results are shown in Fig. 11(b) as light blue circles. The overall uncertainty of this comparison is given by the uncertainty of the calibrated detector and the repeatability of the measurements. The overall relative standard uncertainty is 14 % at 0.05 mW and 6 % at 0.3 mW of incident power. A linear fit of the data in the named range gives a zero offset and a power responsivity of 28.4 V/mW. The fit and the uncertainties are shown in Fig. 11(b).
In the normal mode of operation the calibrated detector THz30 was not sensitive enough to determine the radiant power. But in this mode of operation the radiant power is known to be proportional to the ring current. Furthermore, the attenuation due to the polarizers is calculable from their angles of rotation. At each of the four ring currents of the MLS (140 mA, 70 mA, 20 mA, 10 mA) one polarizer was rotated to a set of different angles and the output of the TES bolometer was recorded. The response of the TES bolometer for different power levels according to the four different ring currents and the attenuation of the polarizers is shown in Fig. 11(a). These measured response values scale linearly with the ring current and are, furthermore, proportional to the radiant power which is well visible in the plot. The slight waviness of the measured data compared to the unscaled linear slope plotted in the 140 mA data is of the order of magnitude of the fluctuation of the atmospheric background and not significant. Consequently, the data can be scaled with the ring current and one common proportionality factor can be found for all ring currents to match the power responsivity determined in the high power range.
Figure 12 shows the complete results for the four ring currents in normal mode, the results obtained in low-α mode and the slope fitted to the low-α mode data. The consistency of all measurements and the linear response of the bolometer over at least four orders of magnitude of THz radiant power is clearly visible. In these measurements the linear response was limited by the thermal and atmospheric background at about 30 nW, well visible by the curvature in the low power region, and by the saturation of the bolometer only. The bolometer starts to be saturated at a radiant power of around 0.3 mW. At 0.35 mW the output signal differs 8.6 % and at 0.38 mW 16 % from the linear extrapolation.
8. Metrological application
In order to demonstrate the applicability of the TES bolometer for quantitative spectroscopy, three transmittance measurements in a vacuum Fourier transform spectrometer VERTEX 80v were performed. Two THz long pass filters with a cut-on at around 100 cm−1 were measured individually and, additionally combined in a stack.
The transmittances of the two filters measured individually are shown as gray lines in Fig. 13. A multiplication of these two transmittances gives a calculated transmittance of the combination of the two filters shown by the blue line. For comparison the measured transmittance of the two filters combined in the stack is shown as a red curve.
The calculated and measured transmittance of the stack agree very well and demonstrate the quality of the bolometer. Each result is an average of 128 scans with a resolution of 1 cm−1 and a complete recording time of seven minutes. A high-pressure mercury lamp was used as a source and three different beam splitters were used to span the spectral range from 105 cm−1 to 10 cm−1. Furthermore, the standard deviation of a series of three repeated measurements of one of the single filters is shown in Fig. 13. The three measurements (each of seven minutes duration) were recorded over 70 minutes. The standard deviation is well below 10−3 over the whole spectral range of interest demonstrating the very low type-A uncertainty of the bolometer and the low drift of the whole system consisting of the bolometer and the spectrometer.
A TES bolometer developed for quantitative Fourier transform spectroscopy in the spectral range from 0.1 THz to 3 THz was designed, built, characterized and applied. A unique feature is its thin film metal mesh absorber with a high and spectrally uniform absorptance in the above spectral range. As a sensor a superconducting thermistor with SQUID readout was used. It shows a linear response over at least four orders of magnitude of radiant power up to 300 μW and an NEP of 3.8 × 10−13 W/Hz1/2. Its applicability for metrological measurements was demonstrated in a first test.
The authors would like to thank A. Steiger and R. Müller for providing the calibrated reference detector THz30 and helpful discussions. Also the valuable support of A. Höhl and the MLS team during the measurement of the linear response is acknowledged.
References and links
1. J. Chamberlain, Principles of Interferometric Spectroscopy (John Wiley & Sons Ltd, 1979).
3. K. Irwin and G. Hilton, “Transition-edge sensors,” in Cryogenic Particle Detection, vol. 99 of Top. Appl. Phys., C. Enss, ed. (Springer, 2005).
4. J. M. Gildemeister, A. T. Lee, and P. L. Richards, “A fully lithographed voltage-biased superconducting spider-web bolometer,” Appl. Phys. Lett. 74, 868–870 (1999). [CrossRef]
5. W. Woltersdorff, “Über die optischen Konstanten dünner Metallschichten im langwelligen Ultrarot,” Z. Phys. 91, 230–252 (1934). [CrossRef]
6. C. Hilsum, “Infrared absorption of thin metal films,” J. Opt. Soc. Am. 44, 188 (1954). [CrossRef]
7. R. Ulrich, “Far-infrared properties of metallic mesh and its complementary structure,” Infrared Phys. 7, 37–55 (1967). [CrossRef]
8. M. Kehrt, R. Müller, A. Steiger, and C. Monte, “Background corrected transmittance and reflectance measurements in the FIR,” in Proceedings of 38th International Conference on Infrared, Millimeter, and Terahertz waves (IRMMW-THz) (IEEE, 2013).
9. A. T. Lee, P. L. Richards, S. W. Nam, B. Cabrera, and K. D. Irwin, “A superconducting bolometer with strong electrothermal feedback,” Appl. Phys. Lett. 69, 1801–1803 (1996). [CrossRef]
10. M. Kehrt, J. Beyer, C. Monte, and J. Hollandt, “Design and characterization of a TES bolometer for Fourier transform spectroscopy in the THz range,” in Proceedings of 39th International Conference on Infrared, Millimeter, and Terahertz waves (IRMMW-THz) (IEEE, 2014).
11. J. Clarke and A. I. Braginski, The SQUID Handbook: Fundamentals and Technology of SQUIDs and SQUID Systems (Wiley-VCH, 2005).
12. D. Drung, “High-Tc and low-Tc dc SQUID electronics,” Supercond. Sci. Technol. 16, 1320 (2003). [CrossRef]
13. D. Drung, C. Assmann, J. Beyer, A. Kirste, M. Peters, F. Ruede, and T. Schurig, “Highly sensitive and easy-to-use squid sensors,” IEEE Transactions on Applied Superconductivity 17, 699–704 (2007). [CrossRef]
14. Magnicon GmbH, “XXF-1 – ultra-high bandwidth dc SQUID electronics,” http://www.magnicon.com/fileadmin/download/datasheets/Magnicon_XXF-1.pdf.
15. P. L. Richards, “Bolometers for infrared and millimeter waves,” Appl. Phys. 76, 1–24 (1994). [CrossRef]
16. R. Klein, G. Brandt, R. Fliegauf, A. Hoehl, R. Müller, R. Thornagel, G. Ulm, M. Abo-Bakr, J. Feikes, M. v. Hartrott, K. Holldack, and G. Wüstefeld, “Operation of the metrology light source as a primary radiation source standard,” Phys. Rev. ST Accel. Beams 11, 110701 (2008). [CrossRef]
17. J. Feikes, M. von Hartrott, M. Ries, P. Schmid, G. Wüstefeld, A. Hoehl, R. Klein, R. Müller, and G. Ulm, “Metrology Light Source: The first electron storage ring optimized for generating coherent THz radiation,” Phys. Rev. ST Accel. Beams 14, 030705 (2011). [CrossRef]
18. R. Müller, W. Bohmeyer, M. Kehrt, K. Lange, C. Monte, and A. Steiger, “Novel detectors for traceable THz power measurements,” J. Infrared Millim. Terahertz Waves 35, 659–670 (2014). [CrossRef]
19. A. Steiger, B. Gutschwager, M. Kehrt, C. Monte, R. Müller, and J. Hollandt, “Optical methods for power measurement of terahertz radiation,” Opt. Express 18, 21804–21814 (2010). [CrossRef] [PubMed]