DESHIMA 2.0: A 200-400 GHz Ultra-wideband Integrated Superconducting Spectrometer

TL;DR

DESHIMA 2.0 tackles the need for compact, wide-band mm/sub-mm spectroscopy by implementing a broadband on-chip filterbank ISS with $339$ KIDs, NbTiN MS THz filters, and a leaky-wave antenna to achieve $200$-$400$ GHz coverage expressed as $Q_{filter} \approx 340 \pm 50$ across the band. The authors describe a lab setup that mirrors the telescope cabin to characterize THz filter responses, beam patterns, sensitivity, and a beam-pattern–based absolute frequency calibration, leveraging the KID multiplexing for simultaneous measurements. Key findings include a high filter yield ($>98\%$), aperture efficiency $\eta_{ap} \approx 0.70$, instrument efficiency $\eta_{inst} \approx 8\%$, and photon-noise-limited performance with NEP$_{inst} \approx 1.3\times 10^{-16}$ W/$\sqrt{\text{Hz}}$, plus a measured 1/f knee around $f_{knee} \approx 1.4$ Hz and verified far-field beam characteristics. The work demonstrates DESHIMA 2.0’s potential for scalable wide-band spectroscopic surveys and provides a foundation for improvements toward ISS-based large-format integral field units (IFUs) and complementary observing capabilities to existing interferometers.

Abstract

DESHIMA (Deep Spectroscopic HIgh-redshift MApper) is a broadband integrated superconducting spectrometer (ISS) for millimeter (mm) / sub-millimeter (sub-mm) wave astronomy based on Kinetic Inductance Detectors (KIDs). This paper describes characterization of DESHIMA 2.0 in laboratory settings. The instrument features NbTiN superconducting microstrip (MS) filters with low-loss a-SiC:H dielectric and an ultra-wideband leaky-wave antenna. A laboratory setup was designed, incorporating the cryostat housing cryogenic optics and ISS chip comprising 339 KIDs connected to MS filters tuned for (sub-)mm wave frequencies. Room-temperature mirrors on a hexapod stage allowed precise positioning and alignment of optical elements. The sky-position chopper was positioned on a motor-controlled stage for fine-tuned control over its position and alignment. Thanks to the multiplexing capability of KIDs, we could simultaneously measure multiple performance metrics across the entire frequency range. We showed that DESHIMA 2.0 achieved significant improvements in performance compared to its predecessor (DESHIMA 1.0): measured instantaneous frequency coverage was 200$-$400 GHz with a mean filter $Q_{filter}$ of $340 \pm 50$; instrument efficiency reached $\sim8$ \%, indicating 4 times wider band coverage and 4 times higher sensitivity. The yield rate for MS filters exceeded 98 \%. The estimated aperture efficiency from measured beam patterns agreed well with the designed value of approximately 70 \%. The telescope far-field beam patterns calculated from measured beam patterns also exhibited good agreement with design specifications. We also demonstrated validity of a new method of absolute frequency calibration using the data from beam pattern measurement.

Figures (10)

• Figure 1: (a) Picture of the ASTE telescope, whose main reflector has a diameter of 10 m. It is located at 4860 m altitude on Pampa la Bola, the Atacama desert, Chile. (b) Setup inside the ASTE cabin. The location of the sky-position chopper is $\sim180$ cm above from the warm mirrors. The red and green arrows correspond to two different beam positions (Position 1 and 2) shown in (c). (c) Schematic drawing of inside of the chopper. Two beams from Position 1 and 2 are separated by 100.46 mm, and reflected by mirrors with tilt angle of 45.267$^\circ$ to create 234 arcsec pointing separation in the sky.
• Figure 2: (a) Cross section and overview of the DESHIMA cryostat with explanations of components. (b) Picture of Si vacuum window with laser-ablated AR structures. (c) Total transmittance of Si window, thermal shaders, and IR blocking filters along the cryogenic optical throughput. Four kinds of IR blocking filters: 1.5 THz low-pass filter (LPF), 650 GHz LPF, 550 GHz LPF, and 450 GHz LPF are installed in the optical throughput. (d) Picture of the DESHIMA 2.0 spectrometer (ISS) chip.
• Figure 3: (a) Photograph of the DESHIMA 2.0 spectrometer chip with its holder. The white rectangle indicates the spectrometer chip, which is close-up in (c). (b) Cross section around the Si lens and the antenna (not to scale). (c) Micrograph of the spectrometer chip. The location of the leaky-wave antenna, the signal line, the wideband couplers, the THz filters (filterbank), the signal termination, dark KIDs and the readout line are indicated. The Perminex spacers of $10~\mu$m height around the antenna are also shown. (d) Zoomed-in micrograph of the leaky-wave antenna on the SiN membrane. The antenna width is $500~\mu$m, and it has a CPW based quarter-wave impedance matching structure that eventually connects to the NbTiN CPW signal line with a geometry of $2-2-2~\mu$m. (e) Colored scanning electron microscope (SEM) image of the wideband coupler. The coupler length is $\sim28~\mu$m, weakly coupled to the MS signal line of $1.1~\mu$m width. The coupling strength is designed to be $-29 \pm 1$ dB over the frequency range of DESHIMA 2.0. (f) Colored SEM image of the THz filter (image taken from Thoen2022). The filter coupler length varies from $\sim25 - 50~\mu$m depending on the resonance frequency of the filter. The distance between the filter and the signal line is designed to be 300 nm. In the filterbank, the spectral channels are placed in decreasing order of frequency from 400 GHz to 200 GHz. (g) Schematic drawing of a pair of MS THz filter and KID, showing how each component is connected or coupled. KID consists of a short-ended NbTiN MS line with a width of 1.5 $\mu$m that couples to the filter, NbTiN-Al hybrid CPW line with a geometry of $2-2-2~\mu$m, and open-ended NbTiN CPW line with a geometry of $5-5-5~\mu$m. Al bridges are used to electrically connect the NbTiN MS line, the NbTiN-Al hybrid CPW line and the NbTiN CPW line of the KID. The open-end of the KID is capacitively coupled to the CPW readout line with an Al bridge. Not displayed in the drawing, Al bridges are also used to balance the ground planes of the readout CPW line by placing them in a random spacing. (h) Cross section of the connection between the MS line and the Al line of the hybrid CPW (not to scale). The corresponding location is indicated by the red dashed line in (g).
• Figure 4: Rendered CAD drawing of the laboratory setup with the inset picture of the laboratory setup. The red lines correspond each components of the instrument in the lab picture to the rendered drawing. The optical path from the sky chopper to the cryostat is shown as the yellow arrow. The definition of two measurement planes (MP1 and 2) are indicated as yellow dashed lines.
• Figure 5: (a) Measured response of the DESHIMA 2.0 spectral channels, $|S_{31}|^2$. Because they are half-wavelength filters, the second resonances from the lower frequency range (around $200-220$ GHz) are also visible above 400 GHz. The black points represent $|S_{31}|^2_{max}$ with error bars, which are combination of errors from a Lorentzian fit and the design value of $\eta_{wb, bf}$. The blue shaded box shows the expected $|S_{31}|^2_{max}$ from a numerical filterbank model. (b) THz filter responses normalized in the frequency domain (colored curves) and averaged shape (black curve). The red dashed curve shows the result from a Lorentzian fitting to the averaged shape. The obtained $Q_{filter}$ from the fit is $\sim 330$. From the peaks at the normalized frequency of around 0.975 and 1.025 of the averaged shape, we can also estimate a cross-talk level of $\sim 1$ % that is mainly caused by the cross talk by neighbour KIDs in the microwave frequency domain. (c) The obtained $Q_{filter}$ from the fitting for all THz filters as a function of $F_{filter}$ (black points). The green solid line and shaded area displays the averaged $Q_{filter}$ and its standard deviation, respectively. (d) The 2-dimensional distribution of normalized $(F_{filter}, f_r)$. The black points are design, and the red crosses show the measured data, showing good correspondence between them. (e) The black points in the left panel shows calculated frequency spacing of the THz filters ($\Delta F_{filter} / F_{filter} = (F_{filter}^{i+1} - F_{filter}^{i}) / F_{filter}^i$) as a function of $F_{filter}$, while the blue line indicates the design value of 0.02 ($= 1/500$). The right panel displays the distribution of the frequency spacing with the value of standard deviation ($\sigma$) obtained from a Gaussian fit. The result of the fitting is also shown as the red curve in the right panel. (f) The black points in the left panel shows calculated frequency-scattering of KIDs ($|f_r^{measured} - f_r^{design}|/f_r^{design}$) as a function of $f_r$. The right panel displays the distribution of the frequency-scattering with the value of standard deviation ($\sigma$) obtained from a Gaussian fit. The result of the fitting is also shown as the red curve in the right panel.