key: cord-0190908-94877egg authors: Moncelsi, L.; Ade, P. A. R.; Ahmed, Z.; Amiri, M.; Barkats, D.; Thakur, R. Basu; Bischoff, C. A.; Bock, J. J.; Buza, V.; Cheshire, J.; Connors, J.; Cornelison, J.; Crumrine, M.; Cukierman, A.; Denison, E. V.; Dierickx, M.; Duband, L.; Eiben, M.; Fatigoni, S.; Filippini, J. P.; Goeckner-Wald, N.; Goldfinger, D. C.; Grayson, J.; Grimes, P.; Hall, G.; Halpern, M.; Harrison, S.; Henderson, S.; Hildebrandt, S. R.; Hilton, G. C.; Hubmayr, J.; Hui, H.; Irwin, K. D.; Kang, J.; Karkare, K. S.; Kefeli, S.; Kovac, J. M.; Kuo, C. L.; Lau, K.; Leitch, E. M.; Megerian, K. G.; Minutolo, L.; Nakato, Y.; Namikawa, T.; Nguyen, H. T.; O'Brient, R.; Palladino, S.; Precup, N.; Prouve, T.; Pryke, C.; Racine, B.; Reintsema, C. D.; Schillaci, A.; Schmitt, B. L.; Soliman, A.; Germaine, T. St.; Steinbach, B.; Sudiwala, R. V.; Thompson, K. L.; Tucker, C.; Turner, A. D.; Umilta, C.; Vieregg, A. G.; Wandui, A.; Weber, A. C.; Wiebe, D. V.; Willmert, J.; Wu, W. L. K.; Yang, E.; Yoon, K. W.; Young, E.; Yu, C.; Zeng, L.; Zhang, C.; Zhang, S. title: Receiver development for BICEP Array, a next-generation CMB polarimeter at the South Pole date: 2020-12-07 journal: nan DOI: nan sha: 9aa665fc5ea3225805ed5bbbb44ac118db49f3e8 doc_id: 190908 cord_uid: 94877egg A detection of curl-type ($B$-mode) polarization of the primary CMB would be direct evidence for the inflationary paradigm of the origin of the Universe. The BICEP/Keck Array (BK) program targets the degree angular scales, where the power from primordial $B$-mode polarization is expected to peak, with ever-increasing sensitivity and has published the most stringent constraints on inflation to date. BICEP Array (BA) is the Stage-3 instrument of the BK program and will comprise four BICEP3-class receivers observing at 30/40, 95, 150 and 220/270 GHz with a combined 32,000+ detectors; such wide frequency coverage is necessary for control of the Galactic foregrounds, which also produce degree-scale $B$-mode signal. The 30/40 GHz receiver is designed to constrain the synchrotron foreground and has begun observing at the South Pole in early 2020. By the end of a 3-year observing campaign, the full BICEP Array instrument is projected to reach $sigma_r$ between 0.002 and 0.004, depending on foreground complexity and degree of removal of $B$-modes due to gravitational lensing (delensing). This paper presents an overview of the design, measured on-sky performance and calibration of the first BA receiver. We also give a preview of the added complexity in the time-domain multiplexed readout of the 7,776-detector 150 GHz receiver. Measurements of the polarization of the Cosmic Microwave Background (CMB) provide crucial information to further our understanding of the early Universe. The Lambda cold dark matter (ΛCDM) model predicts an E-mode polarization pattern in the CMB at the level of a few µK as well as a B-mode polarization signal arising from gravitational lensing of E-modes by the large-scale structure of the Universe. Inflationary gravitational waves are predicted to produce an excess of B-mode power on degree angular scales that is scaled by the tensor-to-scalar ratio r. The value of r is directly tied to the energy scale of inflation. 1 While certain classes of inflation models predict extremely small values of r, a detection of primordial B-mode polarization would be direct evidence for the theory of inflation. However, polarized Galactic dust and synchrotron foregrounds hamper our ability to measure primordial gravitational waves and need to be disentangled from the degree-scale B-mode polarization signal by observing the sky at multiple frequencies with exquisite sensitivity. Additionally, the removal of B-mode power due to gravitational lensing (delensing) can lead to improved constraints on r, but it requires higher resolution observations. 2 The most stringent published constraint on the tensor-to-scalar ratio is r 0.05 < 0.06 at 95% confidence from Bicep/Keck Array data in conjunction with Planck and WMAP measurements 4 (see Figure 1 , which includes the most recent Planck points 3 ). Over the past 15 years, our experimental strategy of designing small-aperture, cryogenic, refracting wide-field telescopes observing from the South Pole has proven successful in probing the degree-scale polarization of the CMB (multipoles 35 < l < 300) with tight control of systematic errors. Bicep1 was the first polarimeter designed specifically to target the B-mode signal and operated from January 2006 through December 2008 with 49 orthogonal pairs of polarization-sensitive NTD bolometers observing at 100 and 150 GHz. Bicep2 observed the sky with 500 antenna-coupled transition-edge sensor (TES) bolometers at 150 GHz from 2010 to 2012, and reported an excess of B-mode power over the base lensed-ΛCDM model in the range 30 < l < 150, 5 which was later attributed to polarized dust emission. 6 The Keck Array consisted of five Bicep2 -like 260 mm aperture receivers and started observations at 150 GHz in 2012. The interchangeable Keck Array receivers allowed us to diversify the frequency coverage over the years: 95 GHz receivers were deployed in 2014, 220 GHz in 2015, and 270 GHz in 2017. The higher-throughput Bicep3 receiver replaced Bicep2 in its mount in 2015, with a 520 mm aperture and 2500 detectors all operating at 95 GHz for an on-sky instantaneous instrument sensitivity of 6.7 µK cmb √ s. 7 Figure 2 shows the progression of the Bicep/Keck Array program, to larger apertures, larger focal planes, and wider frequency coverage. The latest instrument in our program is Bicep Array, 8, 9 which adopts the same interchangeable concept used in Keck Array and is comprised of four Bicep3-class receivers in six frequency bands spanning from 30 to 270 GHz (Figure 3 ), which are all 27% wide. The lowest (30/40 GHz) and highest (220/270 GHz) frequency receivers are dual-band receivers each featuring a checker-board of focal plane modules with single-band detectors centered at the two adjacent bands (Figure 2 ). Splitting the atmospheric windows this way provides more spectral information on the Galactic synchrotron and dust emission, thus improving the constraints on the foreground model parameters. Scaling from previous on-sky performance, Table 1 shows the expected sensitivities and survey weights * of the 4 receivers, which amount to 32,000+ photon-noise limited polarization-sensitive detectors. At the end of 2019, the Keck Array telescope mount has been replaced by the new Bicep Array mount, 11 which provides bore-sight rotation about the optical axis, in addition to rotation in Azimuth and Elevation, following the Keck Array and Bicep3 designs. During the same Austral Summer, we also deployed the first 30/40 GHz Bicep Array receiver, equipped with a full complement of detector modules to fill the focal plane, which has now observed for an entire season. The other Bicep Array receivers will be installed in the new mount with a staged approach over the next few years † . In the meantime, we are continuing observations with a subset of the Keck Array receivers installed in open slots until they are filled by available Bicep Array receivers. The observations will continue to focus on the deep Bicep/Keck Array sky patch so that the new low-frequency data can be combined with existing multi-frequency observations. An updated project timeline is shown in Figure 12 . Bicep Array is largely based on the successful design of Bicep3. [12] [13] [14] [15] Each receiver is housed in a customdesigned vacuum cryostat 11 2.1 m tall and 0.9 m in diameter (excluding the additional envelopes of the multichannel readout electronics and the pulse-tube cooler). Figure 4 shows a cross-section of the receiver, along with the optical diagram for reference. The vacuum and 50 K shells are designed with a short base-stage, and long cylinders for the main section, which allow users to lift off the outer shells from the base plate without undoing the cabling, thermal joints and sub-Kelvin truss structure. The vacuum window is made out of HDPE for the 30/40 GHz receiver, while thinner solutions for the higher frequencies are actively being pursued. 16 The 4 K shell is divided into two lengthwise segments for ease of access during integration; the top optics section houses the objective lens and baffle rings designed to minimize reflections within the tube. The lower optics section holds the nylon filter and the field lens. The 4 K plate supports the fridge space and the subkelvin thermo-mechanical truss structure. The 50 K volume is supported by G-10 truss legs, providing robust structural support while maintaining low thermal conductivity between temperature stages. The truss legs for the 4 K and sub-Kelvin structures are instead made out of carbon fiber, owing to its superior room-temperature elastic modulus to low-temperature thermal conductivity ratio at ≤ 4 K. 17 All Bicep Array receivers (see Figure 4 ) use a Cryomech PT415 ‡ pulse tube with nominal cooling capacity of 40 W and 1.5 W at the 50 K and 4 K stages, respectively. The pulse tube is driven by a remote stepper motor that † Due to the COVID-19 pandemic, no crew will deploy at the end of 2020 and the instrument will remain unchanged. . Bicep Array optical diagram and cross-section of the cryogenic receiver. All optical components, except for the Zotefoam filters, are anti-reflection coated to minimize in-band reflections. The HDPE lenses, nylon filter, and aperture stop are precision-mounted in the receiver and cooled to 4 K to reduce detector loading. The focal plane assembly, which houses 12 detector modules, is cooled to 280 mK by a sorption fridge and surrounded by a superconducting Nb magnetic shield. The focal plane is actively temperature-controlled to achieve detector baseline stability over timescales of hours. The radially-symmetric receiver design allows well-matched beams for orthogonally polarized detectors at the focal plane. Table 3 ). The housekeeping (HK) and miscellaneous (M) entries are common to all Bicep Array receivers. More details on RF shielding are given in Section 2.6. is physically detached and electrically/mechanically isolated from the cold head. Residual mechanical vibrations induced by the fittings connecting the head and the motor are damped by a spring bellows connecting the cold head to the cryostat shell. We also use e-beam welded heat straps with flexible sections of ropelay stranded copper in order to minimize the rigidity and the transmission of vibration to the cryostat stages. The optical filtering follows the successful Bicep3 scheme. 12 Behind the 560 mm-aperture vacuum window, which lets > 100 W of thermal radiation into the receiver, a stack of 12 1/8"-thick layers of Zotefoam HD-30 § with 1/10" spacers provides infrared filtering, 18, 19 in conjunction with a 10 mm-thick IR-absorptive alumina filter, 20 which is installed at the top of the 50 K stage. The lower 4 K optics section holds an absorptive 7.45 mm-thick nylon filter, to prevent any short-wavelength radiation let through by the alumina filter above from reaching the focal plane. Finally a low-pass metal-mesh filter 21, 22 installed at 300 mK defines the upper edge of the band. Combining measurements with a thermal model, the total loading on the 50 K and 4 K stages with the receiver on-sky is estimated to be about ∼40 W and ∼0.5 W, respectively. Table 5 reports the median Bicep Array system temperatures during CMB observations in 2020. Sub-Kelvin cooling for the detectors is provided by a three-stage helium (4He/3He/3He) sorption fridge from CEA Grenoble, 23 with heat intercepts at 2 K (4He stage), 300 mK (intermediate cooler, or IC), and 250 mK (ultra cooler, or UC). With a design total loading of 70 and 15 µW at the 300 and 250 mK stages, respectively, the fridge has sufficient cooling capacity to maintain a 3-day uninterrupted observing schedule between thermal cycles. With the telescope on-sky during observations, we have achieved a minimum fridge hold time of 2.5 days, after which other scheduling constraints become relevant. The sub-Kelvin structure is located on top of the 4 K base plate, above the sorption fridge. It is separated into three thermal, "wedding-cake" carbon-fiber truss stages at 2 K, 300 mK, and 280 mK. Each stage provides radiative shielding and room for cable heat-sinking to the respective fridge stages, allowing low-loading environment for the focal plane and detectors. The detector modules and copper focal plane plate are heat-sunk to the fridge UC stage via a flexible high-purity copper-foil heat strap. The strap connects to the focal plane through a stainless steel block. The temperature of the strap is actively regulated, and the block serves as a passive low-pass thermal filter to attenuate thermal noise from the control circuit. The estimated conductive loading on the 2 K, IC, and UC stages is ∼178, ∼55, and 0.234 µW, respectively, from heat-sinking of cables and the mechanical truss structure, but not including radiative contributions (see Table 2 ). Following the same concept as in previous small-aperture telescopes in the Bicep/Keck Array series, the Bicep Array receivers are simple diffraction-limited two-lens refractors, which provide a telecentric and flat focal surface with minimum aberrations over a wide field of view (see Figure 4 for the optical diagram). The telescope has a mean f -ratio of f /1.57 at the center of the focal plane, which has a diameter of 475.8 mm. The HDPE lenses are 650 mm in diameter, with a clear aperture of 550 mm and a field of view of 29.6 degrees. The lenses and nylon filter are cooled to 4 K to minimize loading on the detectors and are anti-reflection coated to minimize in-band reflections. The 4 K space between the objective and the field lens is lined with baffle rings coated with Eccosorb ® HR-10 microwave absorber to suppress far-sidelobes reflections. The same Eccosorb ® is also used to define the 550 mm optical stop for the system, which is located just above (sky side) the objective lens. The detector beams are designed to terminate on the cold stop with an edge taper ranging from -10 to -13 dB with respect to the main lobe, depending on the frequency. Following the approach of previous optical designs in the Bicep/Keck Array telescope series, we take advantage of the excellent directivity provided by the uniform illumination of the planar phased-array antennas and maintain a conservative edge taper at the aperture stop. 26 The current material for the vacuum window is 1"-thick HDPE at 30/40 GHz, where HDPE has negligible in-band emission. At higher frequencies, thinner windows are being pursued, which will reduce the in-band optical loading and thereby significantly improve mapping speed. 16 The Bicep Array receivers are equipped with state-of-the-art dual-polarization antenna-coupled TES arrays, fabricated on 150 mm diameter silicon wafers. 26, 27 The detector arrays are based on planar in-phase combined slot antennas that provide higher pixel packing density than feedhorns, given a chosen edge taper ¶ . The readout architecture is based on Time-Domain Multiplexing (TDM 28, 29 ) , developed by the University of British Columbia and NIST, with two stages of SQUIDs driven using a Multi-Channel Electronics (MCE) system that provides SQUID/TES bias voltages, controls SQUID addressing, and digitizes detector signals. The readout chain is based on numerous previous successful experiments, 30, 31 with the detector modules at 280 mK carrying the first-stage SQUIDs on a PCB located behind the λ/4 backshort (see Figure 5 ), which then connect to a custom focal plane distribution board and through superconducting NbTi wires to a stage of Series Array SQUIDs at 4 K (see Figure 6 ). In the design of these PCBs and the architecture of the cables, we paid particular attention to pairing signal and return in order to minimize cross-talk among channels. Moreover, the SQUIDs and TES detectors are carefully shielded from sources of magnetic fields, including the Earth's field (see Section 2.5 for more details). Twelve detector modules are tiled onto the focal plane, each containing 32 to 2048 detectors, depending on the observing frequency (see Table 1 ). The module is based on the successful Bicep3 design, 14 except the detectors are now fabricated on 6" silicon wafers, instead of the previous 4" wafers to expedite lithographic fabrication with higher detector count per module. The first-stage SQUID readout multiplexing chips are housed in a FR4 circuit board inside the module with superconducting tin-copper traces (Bicep3 and Bicep Array 30/40 GHz used aluminum traces on alumina substrates instead). The housing is constructed with superconducting niobium and aluminum, which, along with a high-µ A4K sheet inside the module, are designed to achieve high magnetic shielding performance (Section 2.5). Figure 5 shows the exploded view of the module design at 150 GHz. Electromagnetic interactions between the edge antennas and the metal frame around the module can cause polarized beam mismatch on the edge pixels, resulting in potential temperature-to-polarization leakage in the CMB maps. We designed corrugated walls to minimize this effect. The 30/40 GHz module implements a novel broadband corrugation design that mitigates the differential ellipticity caused by the metal frame over 57% bandwidth, from 25 GHz to 45 GHz. 32, 33 The first 30/40 GHz receiver (BA1) in its final configuration will be populated with 6 modules at each frequency in a checker-board arrangement. The design for the two frequencies is similar, with appropriate dimensions of the antenna scaled accordingly to frequency. 34 Due to the COVID-19 pandemic, no crew will deploy at the end of 2020 and the instrument will observe the sky for two full seasons in the configuration it was first deployed at the end of 2019, with the following breakdown of modules: 4 at 30 GHz, 7 at 40 GHz, plus one prototype dual-band 30/40 GHz module with 16 dichroic optical pixels (64 detectors; Shiu et al. in prep. ‖ ). The design of the two Bicep Array receivers at 95 and 150 GHz is at an advanced level of development and is based on the same TDM architecture and TES detector arrays. Table 3 shows the multiplexing factors and cabling arrangement for each receiver. We are currently in the process of validating the 150 GHz module design, ¶ See https://cmb-s4.org/wiki/images/2019_10_16_CMBS4_DetectorTechUpdate_V01.pdf#page=6. ‖ https://microdevices.jpl.nasa.gov/interactivepdfs/index.html?pdf=2020-MDL-Annual-report.pdf, page 61. which allows an 18x18 pixel tile (see Figure 5 ) to be read out. The large number of MCEs needed at 150 GHz required a relatively minor mechanical redesign of the MCE boxes in order to fit within the space constraints imposed by the telescope mount. In the context of CMB-S4, 19 our team has developed ambitious preliminary designs of TDM modules that can read up to 1,920 detectors, without requiring superconducting flex cables. The last Bicep Array receiver at 220/270 GHz will eventually feature over 20,000 detectors, and will be read out using RF-multiplexing schemes, which promise higher multiplexing factors than TDM. One possible implementation is the microwave SQUID readout (µMux 35 ) system that has been already been operated with our current TES detectors for a year as an on-sky demonstrator. 36 In this system, the TES detectors are unchanged and are multiplexed by coupling each TES to its own superconducting microwave resonator through an RF-SQUID. Alternatively, Thermal Kinetic Inductance Detectors (TKID 37, 38 ) offer on-wafer RF multiplexing to greatly simplify wafer hybridization, and are currently being developed and tested within the collaboration. 50K and 4K shields (combined) 0.24% 0.52% Detector module (as a whole) 0.12% 0.20% Table 4 . Simulated magnetic shielding performance in Bicep Array, listed as separate sub-system contributions. The SQUID/TES readout system is known to be susceptible to external magnetic fields 39, 40 and Bicep Array's magnetic shielding architecture is designed to minimize the level of spurious signals generated as the telescope moves along the Earth's magnetic field lines. We studied multiple configurations using COMSOL Multi-Physics software, 41 which allowed us to simulate the Meissner behavior of a superconducting material (see Figure 7) . We selected an architecture with a 1 mm-thick cylindrical A4K high magnetic permeability (µ) sheet wrapped around the 50 K cryostat shell, combined with a 2 mm-thick Niobium superconductive flared cup at 300 mK, which provide a shielding factor of 200 or better at the focal plane. Slotted mounting holes on the 50 K shield allow sufficient thermal expansion mismatch between the aluminum and A4K. The sub-Kelvin Niobium shield required a continuous base in addition to the flared cylinder to enhance the shielding performance. The Niobium base is not electrically connected to the copper base plate to avoid potential superconductor-metal interactions at the contact surface. The TESs and the SQUIDs are then further shielded inside the focal plane modules, each of which features a superconducting aluminum detector frame, a superconducting Niobium SQUID box and a 0.020"-thick sheet of high-µ A4K material to provide a combined shielding factor of about 500 or better. Simulation shows the biggest field leakage comes from the slots for the readout cables. An additional Niobium sheet (not shown in Figure 7 for simplicity), with cable slots that are offset with respect to those in the SQUID box, is mounted on the back side of the module Niobium housing to minimize the leakage. Table 4 shows the simulated magnetic shielding performance of the receiver and module, listed separately. The combined system shielding performance is less straightforward to simulate due to interaction among the shields, and is beyond the scope of this paper. We have performed direct measurements of the system to validate the COMSOL simulations by applying an external field that is modulated at a known frequency, including the use of cryogenic flux gate sensors installed at the location of the SQUIDs and TES. The results of these measurements will be published in a dedicated study (Schillaci, Basu Thakur et al, in prep.) Figure 7 . Bicep Array magnetic shielding. Top left and center: Comsol simulation of the residual total magnetic flux density when a purely axial 100 µT external magnetic DC field is applied. Two horizontal slices of the receiver are shown, at the SQUID and TES planes, respectively. Color bar units are in µT. Bottom left: same as above but for a vertical slice at the detector module. Top Right: A 1 mm thick, high-µ A4K sheet is wrapped around the 50 K stage. A superconducting niobium flared cup shield is mounted at the 300 mK stage, surrounding the focal plane. Bottom right: The modules provide an additional level of shielding for the SQUIDs and the detectors. The various components made out of different materials are highlighted in color for clarity, and their respective labels are colored accordingly. Following the successful electromagnetic interference (EMI) mitigation strategy of previous small-aperture telescopes in the Bicep/Keck Array series, 42 the detectors and SQUID readout chain are enclosed in a radio frequency (RF) shield. The RF shield begins on the top of the focal plane plate and extends down to the 4 K insert volume. Circular clamps at each temperature stage hold an aluminized Mylar shroud that drapes over the "wedding-cake" truss structure (a photo showing this has been published in a previous conference proceeding 9 ). The continuous Mylar sheet is used to create a conductive path that surrounds the stages at different temperatures without thermally linking them (see Table 2 for an estimate of the thermal loading due to the Mylar). The ring on the stage at the bottom of the truss connects to the aluminum walls of the fridge volume and the 4 K base plate, which then connects to the optics tube. Filter connectors (micro-D saver PI filtered, model no. C48-00063-01 * * ) at the 4 K base plate protect the cold electronics from RF interference picked up in wiring outside the cryostat. The 30/40 GHz BA1 cryostat was successfully integrated with the new Bicep Array mount and cooled to base temperatures in January 2020. Table 5 reports the median system temperatures in BA1 during CMB observations in 2020, demonstrating adequate cryogenic performance achieved already during the first season. The exact focal plane configuration that was deployed is described in Section 2.4 and will remain unchanged for two full observing seasons due to the COVID-19 pandemic. An extensive detector calibration campaign was conducted over the first several months of 2020, by a combination of the Austral Summer crew and winter-over Nathan Precup. Figure 8 shows FTS spectra for all the single-band detectors in BA1, demonstrating well-behaved band-passes at both frequencies and high end-to-end optical efficiency from the combination of detectors and optical train. Another major check of the optical system is provided by taking beam maps with a source in the far-field of the telescope. 43, 44 Figure 9 shows band-averaged far-field beam map (FFBM) composites, obtained by coadding all per-detector FFBM composites in a frequency band (not including the prototype dual-band detectors mentioned in Section 2.4). Per-detector FFBM composites are made by coadding all individual FFBMs for a detector, which are taken at different boresight angles. The band-averaged composites demonstrate in-focus imaging and FWHM beam sizes in line with the design values of 76 and 57 arcmin at 30 and 40 GHz, respectively (see Table 1 ). Finally, Figure 10 Bicep Array focuses its CMB observations on the same sky patch as Bicep3, spanning right ascension −60 deg