The Atacama Cosmology Telescope: The polarization-sensitive ACTPol instrument
R.J. Thornton, P.A.R. Ade, S. Aiola, F. E. Angile, M. Amiri, J.A. Beall, D.T. Becker, H-M. Cho, S.K. Choi, P. Corlies, K.P. Coughlin, R. Datta, M.J. Devlin, S.R. Dicker, R. Dunner, J.W. Fowler, A.E. Fox, P.A. Gallardo, J. Gao, E. Grace, M. Halpern, M. Hasselfield, S.W. Henderson, G.C. Hilton, A.D. Hincks, S.P. Ho, J. Hubmayr, K.D. Irwin, J. Klein, B. Koopman, Dale Li, T. Louis, M. Lungu, L. Maurin, J. McMahon, C.D. Munson, S. Naess, F. Nati, L. Newburgh, J. Nibarger, M.D. Niemack, P. Niraula, M.R. Nolta, L.A. Page, C.G. Pappas, A. Schillaci, B.L. Schmitt, N. Sehgal, J.L. Sievers, S.M. Simon, S.T. Staggs, C. Tucker, M. Uehara, J. van Lanen, J.T. Ward, E.J. Wollack
PPreprint typeset using L A TEX style emulateapj v. 5/2/11
THE ATACAMA COSMOLOGY TELESCOPE: THE POLARIZATION-SENSITIVE ACTPOL INSTRUMENT
R. J. Thornton , P. A. R. Ade , S. Aiola , F. E. Angil`e , M. Amiri , J. A. Beall , D. T. Becker , H-M. Cho ,S. K. Choi , P. Corlies , K. P. Coughlin , R. Datta , M. J. Devlin , S. R. Dicker , R. D¨unner ,J. W. Fowler , A. E. Fox , P. A. Gallardo , J. Gao , E. Grace , M. Halpern , M. Hasselfield ,S. W. Henderson , G. C. Hilton , A. D. Hincks , S. P. Ho , J. Hubmayr , K. D. Irwin , J. Klein ,B. Koopman , Dale Li , T. Louis , M. Lungu , L. Maurin , J. McMahon , C. D. Munson , S. Naess , F. Nati ,L. Newburgh , J. Nibarger , M. D. Niemack , P. Niraula , M. R. Nolta , L. A. Page , C. G. Pappas ,A. Schillaci , B. L. Schmitt , N. Sehgal , J. L. Sievers S. M. Simon , S. T. Staggs , C. Tucker ,M. Uehara , J. van Lanen , J. T. Ward , E. J. Wollack ABSTRACTThe Atacama Cosmology Telescope (ACT) is designed to make high angular resolution measure-ments of anisotropies in the Cosmic Microwave Background (CMB) at millimeter wavelengths. Wedescribe ACTPol, an upgraded receiver for ACT, which uses feedhorn-coupled, polarization-sensitivedetector arrays, a 3 ◦ field of view, 100 mK cryogenics with continuous cooling, and meta materialanti-reflection coatings. ACTPol comprises three arrays with separate cryogenic optics: two arrays ata central frequency of 148 GHz and one array operating simultaneously at both 97 GHz and 148 GHz.The combined instrument sensitivity, angular resolution, and sky coverage are optimized for measuringangular power spectra, clusters via the thermal Sunyaev-Zel’dovich and kinetic Sunyaev-Zel’dovichsignals, and CMB lensing due to large scale structure. The receiver was commissioned with its first148 GHz array in 2013, observed with both 148 GHz arrays in 2014, and has recently completed itsfirst full season of operations with the full suite of three arrays. This paper provides an overview ofthe design and initial performance of the receiver and related systems. Subject headings:
Microwave Telescopes, CMB Observations Department of Physics, West Chester University of Pennsyl-vania, West Chester, PA 19383, USA Department of Physics and Astronomy, University of Penn-sylvania, Philadelphia, PA 19104, USA School of Physics and Astronomy, Cardiff University, TheParade, Cardiff, Wales, UK CF24 3AA Department of Physics and Astronomy, University of Pitts-burgh, Pittsburgh, PA 15260, USA Pittsburgh Particle Physics, Astrophysics, and CosmologyCenter, University of Pittsburgh, Pittsburgh PA 15260, USA Department of Physics and Astronomy, University of BritishColumbia, Vancouver, BC, Canada V6T 1Z4 NIST Quantum Sensors Group, 325 Broadway Mailcode817.03, Boulder, CO 80305, USA SLAC National Accelerator Laboratory, 2575 Sand HillRoad, Menlo Park, CA 94025 Joseph Henry Laboratories of Physics, Jadwin Hall, Prince-ton University, Princeton, NJ 08544, USA Department of Physics, Cornell University, Ithaca, NY14853, USA Department of Physics, University of Michigan Ann Arbor,MI 48109, USA Instituto de Astrof´ısica and Centro de Astro-Ingenier´ıa,Facultad de F´ısica, Pontificia Universidad Cat´olica de Chile, Av.Vicua Mackenna 4860, 7820436 Macul, Santiago, Chile Department of Astronomy and Astrophysics, The Pennsyl-vania State University, University Park, PA 16802, USA Institute for Gravitation and the Cosmos, The Pennsylva-nia State University, University Park, PA 16802, USA Pontificia Universit´a Gregoriana, Piazza della Pilotta 4,00187 Roma, Italy Department of Physics, Stanford University, Stanford, CA94305, USA UPMC Univ Paris 06, UMR7095, Institut d’Astrophysiquede Paris, F-75014, Paris, France Instituto deAstrof´ısica, Facultad de F´ısica, Pontificia Uni-versidad Cat´olica de Chile, Av. Vicua Mackenna 4860, 7820436Macul, Santiago, Chile Sub-Department of Astrophysics, University of Oxford, Ke-ble Road, Oxford, UK OX1 3RH Dunlap Institute for Astronomy and Astrophysics, Univer-sity of Toronto, Toronto, Ontario, Canada M5S 3H14 Sociedad Radiosky Asesoras de Ingenier´ıa Limitada Lin-coy´an 54, Depto 805 Concepci´on, Chile Canadian Institute for Theoretical Astrophysics, Universityof Toronto, Toronto, ON, Canada M5S 3H8 Physics and Astronomy Department, Stony Brook Univer-sity, Stony Brook, NY 11794-3800, USA Astrophysics and Cosmology Research Unit, School ofChemistry and Physics, University of KwaZulu-Natal, Dur-ban, South Africa National Institute for Theoretical Physics,KwaZulu-Natal, South Africa NASA/Goddard Space Flight Center, Greenbelt, MD20771, USA a r X i v : . [ a s t r o - ph . I M ] M a y R. Thornton et al. INTRODUCTION
Measurements of Cosmic Microwave Background(CMB) temperature anisotropies at scales from one ar-cminute to many degrees have placed precise constraintson the ΛCDM cosmological model. Recent results comefrom, e.g., the WMAP collaboration (Hinshaw et al.2013), Planck (Planck Collaboration et al. 2015), the At-acama Cosmology Telescope (ACT; Sievers et al. 2013),and the South Pole Telescope (Hou et al. 2014). ACT wascommissioned in 2008 with its first instrument, the Mil-limeter Bolometer Array Camera (MBAC; Swetz et al.2011). MBAC had three independent sets of optics at fre-quencies of 148 GHz, 218 GHz, and 277 GHz, and was notsensitive to polarization. In this paper we describe thepolarization-sensitive second generation camera, ACT-Pol.Measurements of CMB polarization provide indepen-dent constraints on cosmological parameters, and havelower foreground power at small scales because dustysources are relatively unpolarized. Polarization can beseparated into “curl-free” E-modes and “divergence-free”B-modes (Seljak & Zaldarriaga 1998; Kamionkowski &Kosowsky 1998). For (cid:96) (cid:38)
20, E-modes arise primar-ily from density perturbations in the early Universe. B-modes at (cid:96) (cid:38)
50 arise from the gravitational lensing ofthe CMB by large scale structure. At (cid:96) (cid:46) at 148 GHz, and the third po-larization array (“PA3”) operates at both 97 GHz and148 GHz (McMahon et al. 2012; Datta et al. 2014). The148 GHz principal frequency is driven by three consider-ations: its location in an atmospheric window betweenoxygen and water lines (Section 4.3), sensitivity to theSunyaev-Zel’dovich (SZ) effect and the CMB, and angu-lar resolution. At higher frequencies, resolution is betterbut atmospheric opacity increases; at lower frequencies,resolution is poorer but the atmosphere is more trans-parent. Each set of optics has a field of view (FOV) thatspans approximately 1 ◦ on the sky and illuminates anarray of corrugated feedhorns which, in turn, direct lightto pairs of transition edge sensor (TES) bolometers (onefor each orthogonal polarization). Each feedhorn in thePA3 array couples to four bolometers, with an orthogonalpair at 148 GHz and another pair at 97 GHz. Combined,there are 1279 feedhorns and 3068 detectors in the threearrays.This paper details the design and initial performance ofACTPol. First light with PA1 was achieved in June 2013.Both PA1 and PA2 operated in 2014. First light with allthree arrays installed was in February 2015. Section 2provides a brief review of the ACT site and telescope. Operating frequencies are based on effective CMB band centers(Table 1), where the 148 GHz nominal frequency is based on anaverage across all three arrays. The 146 GHz central frequencyfor PA1 in Naess et al. (2014) was based on pre-deployment labmeasurements made with a reduced Lyot stop.
Figure 1.
Photograph of ACT inside its stationary ground screen,which shields the telescope from ground emission. The secondarymirror is barely visible behind the inner, co-moving ground screen.For scale, from the ground to the top of the telescope is 12 m.
Our choice of bands and observing strategy are describedin Section 3. The details of the cryogenic instrument arediscussed in Sections 4 (optics), 5 (cryostat), and 6 (de-tectors). Section 7 outlines the relevant data acquisitionsystems, and Section 8 presents on-sky performance. OBSERVING SITE AND TELESCOPE DESIGN
The ACT site is located at an altitude of 5190 m on theslopes of Cerro Toco in the Atacama Desert of northernChile. At a latitude of 23 ◦ S, the telescope has access toover half of the sky. The high altitude and low precip-itable water vapor (PWV) at this location provide ex-cellent millimeter and submillimeter atmospheric trans-parency, with optimum observing between April and De-cember, when the weather is the driest. Further discus-sion of the site characterization during the ACT observ-ing seasons, including PWV, observing efficiency, and at-mospheric fluctuations, can be found in D¨unner (2009).The advantages of this location have attracted a numberof other millimeter experiments, including the AtacamaB-mode Search (Essinger-Hileman 2011), the POLAR-BEAR instrument (Kermish et al. 2012), and the Cos-mology Large Angular Scale Surveyor (Essinger-Hilemanet al. 2014).The telescope is a numerically optimized off-axis Gre-gorian design. The size of the 6-m primary was dictatedby the requirement for arcminute resolution. The pri-mary’s 5.2-m focal length allows for a compact arrange-ment and the ability for fast scanning. The major com-ponents of the telescope are shown in Figures 1 and 2.The primary and secondary mirrors are composed of 71and 11 adjustable aluminum panels, respectively. Themirrors are surrounded by an inner ground screen thatmoves with the telescope during scanning, and an outer,stationary ground screen. A climate-controlled receivercabin is situated underneath the primary and secondarymirrors. A more thorough discussion of the telescopeoptical design can be found in Swetz et al. (2011) andFowler et al. (2007). SKY COVERAGE AND SCAN STRATEGY
Two general survey strategies, “wide” and “deep,”have been implemented with ACTPol. The deep observ-ing strategy targets four ∼
70 deg regions that overlapCTPol Instrument 3 Receiver
Cabin Secondary structure Receiver
Elevation frame Receiver support structure
Primary structure
Figure 2.
Ray trace of ACT’s primary and secondary mirrors up to the entrance of the receiver. The major components of the telescopeupper structure are shown, except for the inner ground screen and part of the receiver cabin wall, which have been removed for clarity.The telescope is shown in its service position (where the receiver cabin floor is level), corresponding to a viewing elevation of 60 ◦ . with other rich multi-wavelength surveys. These deepfields are near the equator and spaced a few hours apart,enabling rising and setting observations of the differentregions to be interspersed. Observations of the deepfields were made in 2013 with the PA1 148 GHz array;a summary of these along with first results are presentedin Naess et al. (2014). The wide observing strategies im-plemented in 2014 and 2015 cover a few thousand squaredegrees and primarily overlap with the SDSS Baryon Os-cillation Spectroscopic Survey and other equatorial ob-servations. Southern regions with lower galactic fore-ground emission have also been observed, especially dur-ing the daytime when the Sun traverses the equatorialfields.To separate the CMB signal from drifts in the detec-tors and atmosphere, the entire upper telescope struc-ture, including primary, secondary, inner ground screen,and receiver, is scanned in azimuth at a fixed elevation tominimize the effects of changing airmass. Most scans areperformed both east (rising) and west (setting) so thatregions can be observed at a range of parallactic angles,thus allowing separation of instrumental and celestial po-larization, and improving the dynamic range of the mapsby providing cross-linking.The CMB fields are observed by scanning at1.5 deg/sec in azimuth with 0.4 sec turnarounds. ACTis typically operated in the elevation range of 40 ◦ to 55 ◦ for science observations, which gives it access to a signif-icant fraction of the sky. The duration of the scan gener-ally takes between 10 and 20 sec, depending on elevationand the target field. When the elevation is changed, the detector bias is modulated to recalibrate and check forany changes in the time constants (Section 5) due to thechange in sky load. CRYOGENIC OPTICS
Optics Overview
Three independent cryogenic optics tubes are used toreimage the Gregorian focus of the telescope onto thedetector arrays. These refractive optics are designed tomaximize the optical throughput and instrument sensi-tivity. Each optics tube uses three anti-reflection (AR)coated silicon lenses to reimage a ∼ ◦ diameter portionof the Gregorian focus. The primary elements of each setof camera optics are a cryostat window and IR blockingfilters (Tucker & Ade 2006) at ambient temperature; fol-lowed by a combination of blocking filters and low passedge (LPE) filters (Ade et al. 2006) at 40 K; the first lensand accompanying filters at 4 K; the Lyot stop, two morelenses, and additional low pass filters all at 1 K; and thefinal LPE filter and array package at 100 mK. A ray traceof the cold optics is shown in Figure 3.The size of each optics tube is limited by both the sizeof the cryostat (which had to fit in the existing receivercabin) as well as the the maximum diameter of the low-pass edge filters (Section 4.3). To minimize the size ofthe entrance optics, the receiver is positioned such thatthe Gregorian focus is located between the receiver win-dow and first lens. The Gregorian focus is not telecentric,which is a requirement for a large, flat feedhorn-coupleddetector array (see Hanany et al. 2013). To achieve atelecentric design, small offsets and tilts were incorpo- R. Thornton et al.rated into the three lenses. The final design is diffraction-limited across each focal plane. Lenses
Silicon was chosen for the lens material due to its highthermal conductivity, high index of refraction (n = 3.4),and low loss at our wavelengths. The high index of re-fraction necessitates the use of AR coatings. ACT previ-ously used Cirlex coatings (Lau et al. 2006), but they in-curred an estimated 15% net efficiency reduction (Swetzet al. 2011). For ACTPol, we created “meta material”AR coatings produced by removing some of the siliconto controlled depths from the surfaces of each lens atsub-wavelength scales using a custom three-axis silicondicing saw, creating layers of square pillars. The result-ing coating has a coefficient of thermal expansion match-ing that of the rest of the lens. Lenses based on a two-layer design (Figure 4a) are used in both the PA1 andPA2 optics. Simulations showed that the resulting coat-ing has low-reflection ( < ◦ with low cross-polarization (Datta et al. 2013).Figure 4b shows measurements verifying the simulatedperformance. The meta material method was extendedto a three-layer pillar design for the wider bandwidthof PA3. The measured performance agrees with predic-tions, and both PA1/2 and PA3 are consistent with sub-percent level reflections (Datta et al. 2015). Filters and Bandpasses
Each optics tube has its own circular, 6.4 mm-thickwindow made of ultra-high molecular weight polyeth-lene with an expanded Teflon AR coating. The PA1 andPA2 windows are 32 cm in diameter and the PA3 windowis 34 cm in diameter. Although thinner windows wouldhave reduced in-band loading, the increased deformationwould have interfered with the blocking filters immedi-ately behind them (Figure 3). There are IR blocking fil-
Array LPE Lens 3 2X LPE Lyot stop Lens 2 Lens 1 300K filters 40K filters 4K filters
Figure 3.
Ray trace of the cold optics. The upper trace showsthe PA3 (multichroic) optical path and the lower trace shows thePA1 path. The PA2 optical path is a mirror image to that of PA1and has been removed for clarity. The constituent elements aredescribed in the text.
120 130 140 150 160 170−40−35−30−25−20−15
Frequency (GHz)
Target Band
TM polarizationTE polarization measurementsimulation
Figure 4. (Left) Isometric view of a two-layer antireflection coat-ing showing how the material removal process creates “pillars” onthe lens surface. For scale, the pillar pitch is 450 µ m. (Right) Com-parison between simulated and measured reflectance of a two-layercoating on one side of a flat silicon sample. ters (Tucker & Ade 2006) at 300 K, 40 K, and 4 K that re-flect high-frequency out-of-band radiation to reduce theoptical load, in particular, on the poorly thermally con-ducting LPE filters. These LPE capacitive mesh filters(Ade et al. 2006) are used at 40K, 4K, 1K, and 100 mKto limit loading on successive stages. Filter sets witha range of cutoff frequencies allow suppression of out-of-band leaks from individual filters. For PA1/2, these LPEfilters are also used to define the upper edge of the band.The lower edge of the PA1/2 band is set by a waveguidecutoff at the end of each feedhorn (Section 5.1). The PA3bands are set by on-chip filters (Section 5.2). Figure 5.
Plot of the ACTPol bandpasses for all three arrays(shown with arbitrary scaling) superimposed on the atmosphericbrightness displayed for a range of PWVs. The bandpasses shownare an average of 21 detectors for PA1, 84 detectors for PA2, 18detectors for PA3 97 GHz, and 28 detectors for PA3 148 GHz. Thepredominant atmospheric features are the 60 GHz and 117 GHzoxygen lines and the 183 GHz water line. The loading due towater increases with the atmospheric water vapor content whilethe oxygen emission remains relatively constant. Because of itsproximity to the 183 GHz line, the 148 GHz band is significantlymore sensitive to weather variation than the 97 GHz band.The atmospheric transmission was generated with the ALMAAtmospheric Transmission Modeling (AATM) code, which is arepackaging of the ATM code described in Pardo et al. (2001).
Each filter’s frequency response at room temperaturewas measured with a Fourier transform spectrometer(FTS). After the filters were installed in the receiver,CTPol Instrument 5additional FTS measurements were made on the fullycooled system before shipping to the site, as were testswith thick-grill filters (e.g., Timusk & Richards 1981) tocheck for high frequency leaks. A final set of FTS mea-surements was performed after the receiver was installedon the telescope so that bandpasses could be taken in thesame configuration as science data acquisition, and areshown in Figure 5.Using each array’s measured bandpass, we follow themethod of Page et al. (2003) for calculating the effectivecentral frequency for broadband sources, as well as theCMB and SZ effect. The effect of the varying sourcespectra on the band center is to shift it slightly. Theresults are given in Table 1. ARRAYS
ACTPol has three superconducting focal plane arrays,each consisting of silicon micro-machined feedhorns thatdirect light to matched, photon-limited bolometric detec-tor arrays (Yoon et al. 2009). The feedhorns, detectors,and SQUID multiplexing components are fabricated atNIST. PA1 and PA2 each has 512 feeds, which couple to1024 bolometers (one per linear polarization) and oper-ate at 148 GHz; PA3 has 255 feeds that couple to 1020bolometers (one per linear polarization per frequency)and operate simultaneously at 97 and 148 GHz. Eacharray is approximately 15 cm in diameter and is locatedbehind a 100 mK LPE filter (Figure 3).
Feedhorns
The feedhorns impedance-match radiation from freespace to the detector wafers and their dimensions areoptimized for a suitable bandwidth (Britton et al. 2010;Nibarger et al. 2012). The feedhorn arrays are assem-bled from stacks of silicon wafers with micro-machinedcircular apertures that correspond to individual corru-gations (Figure 6). The individual silicon wafer plateletsare etched, RF sputter coated with a Ti/Cu layer on bothsides, stacked, aligned, and finally copper and gold elec-troplated to form a close-packed feedhorn array. This ap-proach preserves the advantages of corrugated feedhornssuch as low sidelobes, low cross-polarization, and wide-band performance while reducing the difficulty of build-ing such a large array using traditional techniques like di-rect machining or electroforming individual metal feeds.While similar arrays can be made from metal platelets,difficulties associated with weight, thermal mass, and dif-ferential thermal contraction are avoided with the sil-icon platelet array concept. Furthermore, tolerancesachievable with optical lithography and silicon micro-machining make for extremely good array uniformity.The feedhorn profiles for all three arrays contain sectionsto define the low-frequency cutoffs of the detector band-passes. The wider band of PA3 motivated a section ofring-loaded platelets (Takeichi et al. 1971) for that array(see Figure 6 and McMahon et al. 2012).
Detectors
Behind the feedhorns, radiation is coupled onto thepolarimeters via planar orthomode transducers (OMT).The detectors are fabricated on monolithic three-inch sil-icon wafers in two different varieties: a hexagonal (“hex”)wafer and a semi-hexagonal (“semihex”) wafer. Each hex ! loaded ! corrugations Figure 6. (Left) Design of a single horn-coupled multichroic po-larimeter, consisting of a broad-band ring-loaded corrugated feed-horn and a planar detector array. A broad-band Ortho-ModeTransducer (OMT), shown in magenta, separates the incoming ra-diation according to linear polarization. (Right) Photograph ofa cross-section of a PA3 test feedhorn. There are a total of 25gold-plated silicon wafers, five of which are ring-loaded to form abroad-band impedance matching transition between the corrugatedinput waveguide and the round output guide. wafer has 127 pixels in the 148 GHz design and 61 pixelsin the dichroic design, while the semihexes have 47 and24 pixels, respectively. Each hex and semihex section isactually composed of four silicon wafers: a top, waveg-uide interface wafer, an OMT and detector wafer, anda two-piece quarter-wave backshort wafer. A full ACT-Pol array is assembled from three hex wafers and threesemihex wafers.Changes in radiation power are detected using super-conducting TES bolometers (Lee et al. 1996; Irwin &Hilton 2005). When a TES is appropriately voltage bi-ased, negative electrothermal feedback maintains it atthe transition temperature, T c , under a wide range of ob-serving conditions. Its steep resistance-vs-temperaturecurve transduces temperature fluctuations into currentfluctuations, which are read out by the use of super-conducting quantum interference device (SQUID) am-plifiers using a time-domain multiplexing architecture(Reintsema et al. 2003) and room temperature electron-ics (Battistelli et al. 2008).Photographs of a hex wafer, an ACTPol pixel, anda TES bolometer for the 148 GHz band are shown inFigure 7. Opposing pairs of OMT antenna probes sep-arate the orthogonal polarization signals. The signalstravel through a coplanar waveguide to microstrip tran-sition, which is impedance-matched to reduce loss, toNb microstrip lines, where they exit. On a PA3 dichroicpixel, diplexers comprised of two separate five pole res-onant stub band-pass filters separate the radiation into75-110 GHz and 125-170 GHz pass-bands. The signalsfrom opposite OMT probes within a single sub-band arethen combined, using a hybrid tee to reject high ordermodes, and the desired signal is routed to a TES islandbolometer. Each PA3 pixel has four bolometers for thetwo linear polarization signals at each frequency. De-tails about the pixel design can be found in Datta et al.(2014). We define the detector optical efficiency η as theratio of the power detected by the bolometer’s TES tothe optical power incident on the OMT. Both signal re-flections and signal loss can cause η <
1. We originally R. Thornton et al.
Array PA1 PA2 PA3 (lower) PA3 (upper)
Effective bandwidth ± ± ± ± Effective band centers: compact sources
Synchrotron 143.9 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Effective band centers: diffuse sources
Synchrotron 145.3 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Conversion factors δ T CMB / δ T RJ ± ± ± ± δ W/ δ T RJ (pW/K) 1.41 ± ± ± ± RJ ( µ W / Jy) 7640 ±
300 8160 ±
290 5740 ±
300 6160 ± CMB ( µ W / Jy) 13160 ±
510 14070 ±
490 9720 ±
520 7820 ± Beam solid angles (nsr) ± ± ±
11 560 ± Table 1
Effective central frequencies for broadband compact and diffuse sources. The values are obtained by averaging over individual detectorsand based on the measured response of the receiver on the telescope. Assumed spectral indices: synchrotron emission ( α = − . α = − . α = 2 . α = 3 . ! ! OMT
TES
Nb microstrip
Lossy Au meander
MoCu bilayer
SiN legs μ m TES bolometer
Figure 7. (Left) An ACTPol 148 GHz hex wafer containing 127 pixels and 254 TESes. Each array consists of three hex wafers and threesemihex wafers for a total of 522 pixels. (Middle) A single ACTPol pixel. The incident radiation couples to each detector via the centralOMT antenna (the different fin shape than that in Figure 6 is a characteristic of the single-band versus multichroic design). Niobiummicrostrip lines carry the power to one of the two bolometer islands on the edges. (Right) An ACTPol TES bolometer. The power isdeposited on the island as heat through the lossy Au meander. The island is thermally isolated from the bath, connected only throughfour SiN legs which also carry the TES bias and signal lines. The electrical power dissipated in the voltage biased MoCu superconducting“bilayer” monitors the power deposited in the lossy Au meander. projected η ≤ .
76 and η ≤ .
66 for the 97 and 148 GHzbands, respectively. Table 2 shows median measured val-ues for the PA3 wafers, indicating the achieved dielectricloss tangents were slightly better than the conservativeprojections.For all three arrays, the Nb microstrip lines termi-nate at a pair of TES islands, where the power is de-posited as heat through lossy gold meanders. Each is-land is connected to the bulk silicon through four SiN legs of length 61 µ m and widths that range from 14 µ mto 53 µ m, depending on the wafer. These legs carry theTES bias/CMB signal lines onto the island and deter-mine the thermal conductivity G to the thermal bath.The superconducting element of the ACTPol TES de-sign is a MoCu “bilayer” with a T c tuned to approxi-mately 150 mK through the choice of the molybdenumand copper geometry and thickness. In the design of theTES, the choice of T c and G represents a balance be-CTPol Instrument 7 Hex Wafer FlexWiring ChipInterface Chip Mux Chip Readout column Row selects (SQ1 bias lines)Flex
Hex Semihex Wafer clamping and heatsinkingPCB aluminum shieldsCable heatsinkingFlex circuitry
Figure 8. (Left) An assembled hex wafer viewed edge-on. A hex is read out using two PCBs, each serving four readout columns. Thecircuitry of each column includes a mux chip, an interface chip, and a wiring chip. The wiring chip provides superconducting 90 ◦ bendrouting from the flex wire bonds to the interface chip wire bonds. Connections to the PCB for the SQ1 bias and feedback lines, SQ2 biasand feedback lines, and detector bias lines are made via aluminum wire-bonds. The detector bias lines are carried from the PCB to thewafer via the folded flex which is wire-bonded at either end. (Right) A fully assembled ACTPol array (PA2). The hex and semihex wafersare sitting on the feedhorn array in the center with the readout PCBs arranged vertically around the edge. The feedhorn apertures arepointed down in this photo. tween minimizing the thermal noise and increasing themaximum operational optical loading power, P sat . Toincrease the stability of the TES through increased heatcapacity to slow the detector response, a region of PdAuis coupled to the TES bilayer. A summary of the detectorcharacteristics and parameter measurements is shown inTable 2.The response of the ACTPol detectors to a delta func-tion signal diminishes with time due to the electro-thermal time constant, τ , of the TES architecture. Themedian time constants for the three arrays are 1.9 ms,1.8 ms, and 1.3 ms. With a scan speed of 1.5 deg/s, thesevalues are equivalent to 3 dB points at multipoles (cid:96) be-tween roughly 30,000 and 40,000. The time constants areloading dependent. We find that f = 1 / πτ typicallyvaries by less than 20 Hz/pW.Median array sensitivities for 2015 estimated bycalibration from Uranus are: ∼ µ K √ s for PA1, ∼ µ K √ s for PA2, ∼ µ K √ s for PA 3 (97 GHz),and ∼ µ K √ s for PA3 (148 GHz). These sensitivityestimates are for PWV/sin(alt)= 1 mm (because the de-tectors are photon-noise limited, their sensitivities aredependent on the level of optical loading), and repre-sent the array white noise level, evaluated near 20 Hz.The white noise level of final maps is higher by ∼ Readout
The time-domain multiplexing (TDM) readout schemeemploys three stages of SQUID amplification. The mul-tiplexing of the readout has the advantage of reducingthe wiring requirements to limit the thermal load on thecold stage. The current through each TES couples to theinput coil of a corresponding SQUID amplifier, called thefirst-stage SQUID (SQ1). A set of 32 SQ1s are coupledto one second stage SQUID (SQ2) through a summingcoil. Each SQ2 and its corresponding SQ1s are housedon a multiplexing chip (“MUX11c”). The multiplexingreadout is achieved through rapid sequential biasing of each of these 32 SQUIDs. The final stage of amplifi-cation is accomplished by a set of SQUID series arrays(SA). Twisted pair NbTi cables carry the SQ1, SQ2, anddetector lines from the five SA boards housed on the 1 Kstage to nine other printed circuit boards (PCBs) on the100 mK stage, which hold the multiplexing and otherreadout chips. Each detector is voltage biased onto itstransition using an ∼ µ Ω shunt resistor housed on theinterface chip, which also contains a 600 nH “Nyquist”inductor designed to band limit the response.The detector bias lines are carried from the readoutPCBs onto the detector wafers through flexible circuitryconsisting of a Kapton base and superconducting alu-minum traces connected on both ends by aluminum wire-bonds. Both PA1 and PA2 as well as the hex wafers inPA3 are assembled with two layers of 200 micron pitchflexible circuitry made by Tech-Etch. For the PA3 semi-hexes, single-layer, 100 micron pitch flex fabricated bythe ACT collaboration is used (Pappas 2015). The bias-ing and multiplexing are handled by the Multi-ChannelElectronics (MCE) crate (Battistelli et al. 2008; Has-selfield 2013). From the MCE box, a set of five 100-pincables carry the bias and readout lines for the SQUIDsand detectors to the SA boards. The multiplexing andsampling rates are discussed in Section 7.The total achieved readout yield for each arrays, wherethe yield is based on detectors with functional IV curves,is 67% for PA1, 77% for PA2, and 83% for PA3. Thenumber of detectors used for CMB analysis is typicallysomewhat less than this (e.g., about 500 detectors forPA1), depending on observing conditions and the num-ber of detectors that can be successfully biased. The fab-rication yield of the detector wafers was high, with manywafers having perfect physical yield of the bolometers.Most detector loss originated from problematic SQUIDbiasing and readout lines, each of which corresponds to32 TESes, and individual detector opens and shorts inthe flexible circuitry. Through improvements in assem-bly protocols, higher yield was achieved with each succes-
45 Aldrin Rd, Plymouth, MA 02360
R. Thornton et al.
Array Wafer Type T c (mK) P sat (pW) G (pW/K) η det (%) PA1 W10 H 143 ± ± ±
30 59 ± ± ± ±
24 61 ± ± ± ±
33 62 ± ± ± ±
25 47 ± ± ± ±
24 18 ± ± ± ±
19 21 ± ± ± ±
16 60 ± ± ± ±
48 65 ± ± ± ±
111 62 ± ± ± ±
55 51 ± ±
11 12 ± ±
43 51 ± ± ± ±
138 51 ± ± ± ±
13 83 ± ± ± ±
17 88 ± ± ± ±
16 73 ± ± ± ±
16 75 ± ± ± ±
15 47 ± ± ± ±
15 37 ± ± ± ±
10 71 ± ± ± ±
13 79 ± ± ± ±
13 70 ± ± ± ± ± ± ± ±
11 59 ± ± ± ±
10 71 ± Table 2
The measured average ACTPol detector parameter values by wafer. The “Wafer” column gives unique identifiers related to the fabricationgeneration. The “Type” column distinguishes hex (H) and semihex (SH) wafers; for PA3, the additional designator A (B) distinguishesthe 97 GHz (148 GHz) devices. We tabulate the mean critical temperatures ( T c ); the mean bias powers to bring the detectors to 90% oftheir normal resistances when the bath temperature is 80 mK in the absence of significant optical loading ( P sat ); the mean thermalconductances from the detector islands to the bath ( G ); and the mean detector efficiencies ( η det ). For each device, we estimate theefficiency as the fraction of the optical power incident on the OMT which is detected by the TES. Thus the efficiency accounts for lossesdue to absorption and reflection in the OMT and microstrip on the pixel. For each quantity, the error listed is the standard deviation ofthe distribution of devices for that wafer (or wafer/frequency subset for PA3). These parameters are derived from a combination ofmeasurements in the laboratory and on the telescope. More detail can be found in Grace (2015), Ho et al. (in press), and Pappas (2016). sive array. Figure 8 shows part of the readout electronicsarchitecture as well as a fully assembled array package. CRYOSTAT
The cryostat is a custom aluminum (primarily al-loy 6061) structure fabricated by Precision Cryogenics. Shown in Figure 9, it is a cylinder 1.5 m in length and1.1 m in diameter. The size was limited by what could fitinside the existing telescope receiver cabin. The designwas motivated, in part, by the success of the MBACcryostat (Swetz 2009; Swetz et al. 2011). The frontplate of the shell serves as the optical bench to whichall of the cold optics and radiation shields are sequen-tially mounted. The 40 K optics plate, which supports afilter stack (Figure 3), is attached to the cryostat frontplate via a G10 fiberglass cylinder. The 4 K optics plateis in turn attached to the 40 K optics plate via a secondG10 fiberglass cylinder. The 4 K optics plate is madeout of alloy 1100 aluminum for increased thermal con-ductivity and acts as the primary support for all of theoptics tubes. Nested 40 K and 4 K aluminum radiationshields surround the optics and cryogenics at those tem-peratures. Welded to the 40 K radiation shield are high-purity aluminum strips for improved heat removal fromthe 40 K filters.
Cryogenics The remote site location makes the use of non-recycledliquid cryogens both difficult and expensive. Thus, everyeffort was made to include closed-cycle cooling systemswherever possible. A single Cryomech PT410 pulse tuberefrigerator is responsible for cooling the optics and cryo-stat structures attached to the 40 K and 4 K stages (Fig-ure 10). These include the 40 K and 4 K cold plates, radi-ation shields, 40 K filters, and the 4 K components (firstlens, baffles, filters) contained within the upper halves ofall three optics tubes (Section 6.2). To minimize vibra-tions from the pulse tubes and scan turnarounds, acous-tically deadened copper braid and flexible copper sheetsare used to attach the PT410 cold stages to the rest ofthe internal cryostat structures (Figure 9). The receiveris located near the the telescope azimuthal axis, whichalso reduces scan-induced vibrations.All remaining components are cooled below 4 K using apulse tube (Cryomech PT407) - backed custom He – Hedilution refrigerator (DR; Shvarts et al. 2014) manufac-tured by Janis Research Corporation. The 1 K stageor “still” of the DR is responsible for cooling the backend of the optics tubes: the Lyot stop, the second andthird lenses, and filters. The 100 mK stage or “mixingchamber” of the DR sets the bath temperature for allthree detector arrays and can continuously supply over100 µ W of cooling power at 100 mK, making the DR
225 Wildwood Ave, Woburn, MA 01801
CTPol Instrument 9
Cryostat Pulse TubeDilutionRefrigerator (DR)
40K – 3K G10
Suspension 4K Copper Tower
Window
PA2 Optics Tube Vacuum Shell 4K Cold Plate300K – 40K G10Suspension
DR Pulse Tube PA1 Optics TubePA1
PA2 PA3Front plate
Figure 9.
Model of the as-built cryostat. For scale, the length of the cryostat is 1.5 m. The PA3 optics tube and most of the radiationshields have been removed for clarity. A combination of flexible copper sheets and copper braid are used to reduce vibrational couplingbetween the pulse tubes and internal cryostat components. an excellent choice when compared to more conventional100 mK adiabatic demagnetization refrigerators due tothe latter’s limited cooling power. The DR’s lower basetemperature (thereby improved detector sensitivity) andcontinuous run-time also out-perform the He adsorptionfridges used in the MBAC cryostat.Cooling the cryostat to base temperature typicallytakes 14 days with all three sets of optics installed. Thisis due in large part to the considerable thermal mass con-tained within its optical components, focal-plane arrays,and many low-temperature thermal interfaces. The vastmajority of the cool-down ( >
13 days) is the time re-quired for all components to reach the base temperatureof the pulse tube stages. During this process, the 1 Kand 100 mK components are connected to the 4 K stageof the DR pulse tube via a mechanical heat switch. Oncethe pulse tube base temperatures are reached, the heatswitch is disengaged and the He – He mixture is allowedto condense while being circulated within the DR insertto complete the cool-down. An additional 4 – 5 hours areneeded for the DR mixing chamber to drop below 100 mKonce this final step has been initiated. Throughout thecool-down, as well as during normal operations, the DRis monitored via an ethernet link connected to its gas-handling system computer inside the receiver cabin ofthe telescope.The lowest temperature reached by each cryogenicstage depends on a number of operational and environ-mental factors, including exterior temperature, telescopeelevation, scanning motion, and the detector read-outelectronics. Table 3 lists the measured base tempera-tures and estimated thermal loading for each cryogenic stage during optimal cooling conditions – note that thetemperature of the coldest stage (the DR mixing cham-ber) may be up to 13 mK warmer during typical observ-ing operations (telescope in motion and detector read-out powered on). Since a lower bath temperature resultsin higher detector saturation powers (and thus a largerdynamic range over atmospheric loading conditions (seeSection 5), gold-plated high-purity (99.999%) annealedcopper links were used to make thermal connections be-tween the DR and the focal-plane arrays to minimizetemperature gradients. During observations, the typicalarray temperature with all three sets of optics installedranged from 100 to 115 mK.
Cryogenic Stage Temperature LoadPT410 1st Stage 34.1 ± ± ± ±
30 mWDR Still 644 ± ± ± ± µ W Table 3
Measured base temperatures and thermal loads of the receivercryogenic stages with three sets of optics during optimal coolingconditions (detector electronics powered down and telescope atrest).
Optical Support Structures
Each of the three sets of optics is contained in a cylin-drical optics tube. To minimize weight, aluminum wasgenerally used to mount optical elements between 300 Kand 1K. For 100 mK assemblies, where conductivity is0 R. Thornton et al. and IR blocking filter stacks 1K contact 100 mK contact Central thermal bus tower40K filter plate4K baffle tube4K cold plate 4K-1K carbon fiber suspensionLens 1 Lens 2 Lens 3Cryostat front plate 1K-100mK carbon fiber suspensionDouble layer of magnetic shielding 1K radiation shieldArray moduleG10 wedge
Figure 10.
Cutaway view of the PA3 optics tube showing the internal optics, mechanical structures, magnetic shielding, and cold straps. critical and where aluminum is superconducting (result-ing in reduced thermal conductivity), oxygen-free high-conductivity copper (OFHC) was generally used. An-other reason for using mostly copper below ≈ product designed primarily for shielding from elec-tromagnetic inteference. Most of the cylindrical struc-tures comprising each optics tube are mounted perpen-dicular to the cold plates. Individual lens and array tiltswere achieved through custom angled mounts that holdeach component at the proper angle with respect to eachoptics tube.The first lens, nominally at 4 K, is supported from the4 K cold plate via an aluminum tube, inside of which aremachined steps that precisely locate baffles. The bafflesare blackened with a mixture of 2850 FT Stycast, loadedwith 5 - 7 % carbon lampblack by weight, so that the min-imum thickness exceeds 1 mm and is textured. The Lyotstop, second and third lenses, and surrounding low-passfilters are all maintained near 1 K. To reduce the load onthe DR, these structures are supported from the 4 K coldplate using a custom-made carbon fiber tube. Attachedto this 1 K framework is a second, re-entrant carbon fibersuspension that is 0.6 mm thick for thermally isolatingthe 100 mK array package from the 1 K components. Magnetic Shielding
The SQUID multiplexers and amplifiers, as well as theTESes themselves, are sensitive to changing magneticfields and therefore require magnetic shielding from bothEarth’s field and AC fields associated with the telescopemotion. To complement shielding around the SA mod-ules and at the PCBs, additional magnetic shielding wasprovided for the collection of SQUIDS for each array byenclosing the lower half of each optics tube (Figure 10)with Amumetal 4 K (A4K). This is a proprietary mate-rial with a high nickel concentration and prepared with aheat treatment process to give it a high magnetic perme-ability at cryogenic temperatures. To achieve the max-imum attenuation, the thickest available A4K (1.5 mm)was used. Each optics tube has two layers of shieldingseparated by approximately 6 mm since, given sufficientspacing, using additional layers approaches the limit ofmultiplicative increases in the field attenuation. Receiver Alignment to Telescope
The receiver is supported and aligned by an aluminumframe that was designed to interface with the receivercabin structure (Figure 2). Plain (sliding) phosphorbronze bearings support the receiver and allow it to betranslated in three directions and rotated in two direc-tions. The positions of the telescope optics and receiverare measured using a VSTARS photogrammetry systemfrom Geodetic Systems Inc. (GSI). Retroreflective tar-gets are placed at the location of the alignment actua-tors on each panel of the primary and secondary mirrorsas well as at fiducial points on the front surface of the A trademark of Vacuumschmelze GmbH in Hanau, Germany.Local Distributor: Amuneal Manufacturing Corporation, 4737Darrah St., Philadelphia, PA 19124, USA, [email protected],(800)-755-9843. CTPol Instrument 11cryostat. Several auxiliary targets are placed around theperimeter of each mirror and on the inner ground screen.The target 3D positions are solved for using GSI propri-ety software. After post-processing with in-house soft-ware and actuator adjustment, the mirror surfaces arealigned to better than 27 µ m RMS during nighttime ob-serving and the reciever is positioned to within 1 mm ofits optimal location. DATA ACQUISITION
Overview
Both science data and housekeeping data are recorded.The science data consist of detector output from thethree arrays, with data from each written to a separatelocal acquisition computer. A housekeeping computerlogs data on ambient and cryogenic temperatures, tele-scope encoders and motors, etc. Due to the dusty envi-ronment and low atmospheric pressure, all of the harddrives at the telescope site are solid state. Data fromeach acquisition machine are transmitted via radio linkto a RAID server at our base near San Pedro de Atacama,where they are merged together into a single data prod-uct. The data are ultimately archived in North America.
Data acquisition for ACTPol in the 2013 and 2014 sea-sons was largely identical to that of MBAC, so here weonly summarize the acquisition and flow of data, and re-fer the reader to Swetz et al. (2011) for details. In the2015 season some systems were modified, and we outlinethose in Section 7.3The MCE bolometer readout has a raw sampling rateof 50 MHz. Using TDM, with a row-switching rate of500 kHz and 33 row selects, an entire array of detectorsis sampled at 15.15 kHz. The signal is then anti-aliasfiltered with a four-pole, low-pass, Butterworth designwith a 120 Hz cutoff. The data are downsampled to 399Hz before transmission by optical fiber to the data ac-quisition computer for recording to disk. Each detectorarray has its own MCE, read out by and recorded on itsown computer. Synchronization of the MCEs is achievedby means of a ‘sync box’ that provides a common clockfor the 15.15 kHz sampling, as well as an integer ‘synccounter’ that increments at the 399 Hz recording rate.The data for each MCE are tagged with the sync counterso that they can be merged downstream.The same sync counter is used to synchronize thereadout of the azimuth and elevation encoders with thebolometer channels. Our housekeeping data acquisitioncomputer contains a custom-built PCI card with an on-board FPGA that receives the sync counter from thesync box via an RS485 connection (clock plus NRZ). Re-ception of a new sync word generates an interrupt onthe PCI bus, which alerts a device driver to poll the az-imuth and elevation encoders using a Heidenhain IK220PCI card in the same computer. The driver then bun-dles the sync counter with the encoder readings and theabsolute time—the latter is accurate to < USB-6218 DAQs; althoughthis readout is asynchronous with the the sync box, wehave measured that they are normally aligned to < µ s precision. Having this merged dataproduct means that mapmaking software need not searchfor disparate files and can work with fully-synchronizeddata.The presence of new data files is registered in a MySQLdatabase, and a series of software daemons that commu-nicate with this database copies files to portable harddisks for transport to Santiago and North America. Thesoftware tracks all file copies and ensures that redun-dant copies exist before space is freed on site computers.This is true for merged files as well as raw, unmergedfiles which we keep for full redundancy. One can alsorequest that files be copied over the internet rather thanby transport drives; our internet bandwidth is too small,however, to forego the use of transport drives. Readout Systems for 2015
Modifications to our housekeeping readout were imple-mented before the 2015 season of observations to pre-pare for the installation of ambient-temperature half-waveplates (HWP), with the goal of significantly reduc-ing atmosheric and instrument 1/f noise (e.g., Kusakaet al. 2014).The HWP encoders will be read out by a DNA-PPC8system, which consists of a CPU interfaced to a lowerdata acquisition layer. The latter is customizable by in-stalling various boards provided by the same vendor in the unit’s six available slots. Reading our high-speedserial sync data would be challenging with this device, sowe have built a new module for the sync box that trans-lates the serial RS-485 sync counter to a parallel signallatching at the 399 Hz data acquisition rate. This paral-lel sync counter is read by a DNA-DIO-403 card in theDNA-PPC8 and any data acquired by the unit can betagged with the sync counter.We took advantage of our ability to interface the DNA-PPC8 with our sync box to update our azimuth and en-coder readout as well, as the interrupt-generating schemedescribed above (Section 7.2) was quite demanding onour housekeeping computer, causing various complica-tions. The new readout uses a Heidenhain EIB 741 unit, ON-SKY BEAMS AND POLARIZATION
Beams
The telescope beams are characterized with observa-tions of planets. Saturn is used to align and focus thesecondary mirror at the start of each season, but ob-servations of Uranus are ultimately used for beam char-acterization because its brightness is low enough to notsaturate the detectors.A single planet observation is achieved by scanning thetelescope back and forth in azimuth, at fixed elevation,while the planet rises or sets through an array’s field ofview. The resulting time-ordered data are reduced to asingle map of the celestial sky, centered on the planet, foreach detector array and frequency band. This mappingprocess relies on knowing the relative pointings of thedetectors on the sky, and on an accurate calibration ofeach detector’s response to the source.The detector positions are determined from Saturn andUranus observations by fitting a model for the source tothe time-stream data directly. A single average point-ing template is computed for each season and array, andused for all subsequent planet mapping. The relative po-sitions of the detectors are constrained at roughly thearcsecond level, and thus the pointing template uncer-tainty contributes negligibly to beam degradation.The detector calibrations are determined indepen-dently for each mapped observation by comparing theamplitude of each detector’s response to the commonmode from atmospheric emission.Each planet map provides a measure of the telescope’s“instantaneous” beam, averaged over all responsive de-tectors in each array and band. The average instan-taneous beams are shown in Figure 11, and some basicproperties are provided in Table 8.1. The effective pointspread function of the telescope in the CMB survey mapsdiffers from the instantaneous beam primarily due to theimpact of pointing variance, which acts as a low-passspatial filter, and due to the way in which sky rotationsymmetrizes the beam by stacking up observations at avariety of parallactic angles. For the interpretation ofCMB survey maps, the beam’s harmonic transform andits covariance are determined following the procedure de-scribed in Hasselfield et al. (2013).
Polarization Angles
The ability to separate polarized intensity into E andB components depends upon how accurately detector po-larization angles are known. For each hex and semihexwafer, a feed horn couples to an OMT pair having oneof two possible orientations, which is set by lithographyduring fabrication. When this is combined with both theorientation of the semi-hex and hex wafers within an in-dividual array as well as the orientation of each array as a whole into the cryostat, a roughly equal distribution ofdetector angles, covering every 15 degrees, is produced.Figure 12 shows these resulting detector polarization an-gles as seen on the sky.Apart from the physical OMT orientations, however,there is also a polarization rotation introduced by theACTPol optical chain as seen on the sky. The optics-induced rotations, as well as the detailed mapping fromthe focal plane position to projected angle on the sky,are determined using a polarization-sensitive ray tracethrough a model of the telescope in the optical designsoftware CODE V. Given a point on the sky, the soft-ware calculates the polarization state across the entrancepupil diameter. This polarization state, averaged acrossthe pupil and propagated to the focal plane, produces thepolarization output for a single point on the focal plane.Given a perfectly polarized input state of known an-gle, the difference between the output polarization stateand the input state results in the polarization rotation.The resulting polarization rotation output is producedat twenty five points on the focal plane for each of thethree ACTPol arrays. This rotation on the focal planeis then fit to a simple 2D quadratic model, which is inturn used to produce the polarization angle rotation atthe location of each feed horn in all three arrays. Theoptics-induced rotations have a range of approximately3 ◦ . While this is too small to be seen in Figure 12, it istoo large to be ignored in the anaylsis.Thus far the measured polarization angles from theCrab Nebula and minimizing the EB spectrum (Naesset al. 2014) are consistent with the calculated polariza-tion angles based on the optics design and detector lay-out. Further analyses will presented in future papers,e.g. Koopman et al. (2016). CONCLUSION
We have presented the design and performance ofthe ACTPol instrument. The new receiver pro-vides a number of improvements to the ACT experi-ment: polarization-sensitivity, continuous 100 mK cool-ing, meta material AR coatings, and multichroic arraytechnology. ACTPol acquired three seasons of data from2013 - 2016. First analyses of Seasons 1 and 2 are pre-sented in Naess et al. (2014) and Louis et al. (in prep).We thank AMEC/Dynamic Structures/Empire fortheir work on the initial telescope construction and KukaRobotics for their continued support on the motion con-trol system. Vladimir Shvarts and the rest of the JanisResearch Corporation were essential for the successfulintegration of the dilution refrigerator into the receiver.We are very grateful for Bill Dix, Glen Atkinson, and therest of the Princeton Physics Department Machine Shop,and Harold Borders and Jeff Hancock at the University ofPennsylvania Physics Department Machine Shop. JesseTreu served as a management consultant for the latterhalf of the project.ACT is on the Chajnantor Science preserve which wasmade possible by the Chilean Comisi´on Nacional de In-vestigaci´on Cient´ıfica y Tecnol´ogica. We are grateful forthe assistance we received at various times from the ABS, CTPol Instrument 13
PA3 97 GHz PA3 148 GHz
PA1 148 GHz
PA2 148 GHz R e l a t i v e i n t e n s i t y ( d B ) Figure 11.
Beam maps for the ACTPol arrays, showing the response to a point source in a coordinate system such that North is parallelto increasing altitude and West is parallel to increasing azimuth. These maps include the data from all responsive detectors, averaged over81, 86, 39, and 37 observations for the PA1, PA2, PA3/148, and PA3/97 arrays, respectively. The white contour lines denote where theresponse falls to half of its peak value.Array – Band Solid Angle Elliptical FWHM modelMajor axis Minor axis Major angle(arcmin ) (arcmin) (arcmin) ( ◦ )PA1 – 148 GHz 2 . ± .
05 1.37 1.32 -49PA2 – 148 GHz 2 . ± .
03 1.33 1.31 56PA3 – 148 GHz 3 . ± .
13 1.58 1.33 -3PA3 – 97 GHz 6 . ± .
22 2.13 2.00 -2
Table 4
Beam parameters, as derived from the maps in Figure 11. The elliptical FWHM model parameters are fit to points in the map whichshow a response within 10% of the half-maximum level. FWHM major angle is measured from the North, increasing towards the East.
ALMA, APEX, ASTE, and POLARBEAR groups. ThePWV data come from the public APEX weather site.Field operations were based at the Don Esteban (oper-ated by Astro-Norte) and RadioSky facilities. This workwas supported by the U.S. National Science Foundationthrough awards AST-0408698 and AST-0965625. Fund-ing was also provided by Princeton University, the Uni-versity of Pennsylvania, Cornell University, the Wilkin-son Fund and the Mishrahi Gift. The developmentof multichroic detectors and lenses was supported byNASA grants NNX13AE56G and NNX14AB58G. CMacknowledges support from NASA grant NNX12AM32H.BS, BK, CM, EG, KC, JW, and SMS received fundingfrom NASA Space Technology Research Fellowships. RDthanks CONICYT for grants FONDECYT 1141113 andAnillo ACT-1417.This paper includes contributions from a United States government agency and is not subject to copyright.
REFERENCESAde, P. A. R., Pisano, G., Tucker, C., & Weaver, S. 2006, inSociety of Photo-Optical Instrumentation Engineers (SPIE)Conference Series, Vol. 6275, Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference Series, 62750UBattistelli, E. S., Amiri, M., Burger, B., et al. 2008, Journal ofLow Temperature Physics, 151, 908Britton, J. W., Nibarger, J. P., Yoon, K. W., et al. 2010,arXiv:1007.0805Datta, R., Munson, C. D., Niemack, M. D., et al. 2013,Appl. Opt., 52, 8747Datta, R., Hubmayr, J., Munson, C., et al. 2014, Journal of LowTemperature Physics, arXiv:1401.8029Datta, R., Austermann, J., Beall, J. A., et al. 2015, ArXive-prints, arXiv:1510.07797
Figure 12.
Plot of relative detector positions and polarization angles as seen on the sky. PA1 is in the lower right, PA2 the lower left,and PA3 on top. Each hex/semihex wafer has OMT pairs in two different orientations, which results in six different detector polarizationangles in each array. There is a slightly higher number of physical detectors than there are available readout lines for, resulting in theregion of seemingly “missing” detectors in each array.D¨unner, R. 2009, Ph.D. Thesis, Pontificia Universidad Catolicade ChileEssinger-Hileman, T. 2011Essinger-Hileman, T., Ali, A., Amiri, M., et al. 2014, in Society ofPhoto-Optical Instrumentation Engineers (SPIE) ConferenceSeries, Vol. 9153, Society of Photo-Optical InstrumentationEngineers (SPIE) Conference Series, 1Fowler, J. W., Niemack, M. D., Dicker, S. R., et al. 2007, Appl.Opt., 46, 3444Grace, E. 2015, PhD thesis, Princeton UniversityHanany, S., Niemack, M. D., & Page, L. 2013, CMB Telescopesand Optical Systems, ed. T. D. Oswalt & I. S. McLean, 431Hasselfield, M. 2013, Ph.D. Thesis, The University of BritishColumbiaHasselfield, M., Moodley, K., Bond, J. R., et al. 2013, ApJS, 209,17Hinshaw, G., Larson, D., Komatsu, E., et al. 2013, ApJS, 208, 19Ho, S., Pappas, C., Austerman, J., et al. in press, Journal of LowTemperature PhysicsHou, Z., Reichardt, C. L., Story, K. T., et al. 2014, ApJ, 782, 74 Irwin, K., & Hilton, G. 2005, Cryogenic Particle Detection:Transition-Edge Sensors chapter (Springer)Kamionkowski, M., & Kosowsky, A. 1998, Phys. Rev. D, 57, 685Kermish, Z. D., Ade, P., Anthony, A., et al. 2012, in Society ofPhoto-Optical Instrumentation Engineers (SPIE) ConferenceSeries, Vol. 8452, Society of Photo-Optical InstrumentationEngineers (SPIE) Conference Series, 1Koopman, B., Austerman, J., Beall, J., et al. 2016 (SPIE)Kusaka, A., Essinger-Hileman, T., Appel, J. W., et al. 2014,Review of Scientific Instruments, 85, 024501Lau, J., Fowler, J., Marriage, T., et al. 2006, Appl. Opt., 45, 3746Lee, A. T., Richards, P. L., Nam, S. W., Cabrera, B., & Irwin,K. D. 1996, Applied Physics Letters, 69, 1801Louis, T., Grace, E., Hasselfield, M., et al. in prepMcMahon, J., Beall, J., Becker, D., et al. 2012, Journal of LowTemperature Physics, 167, 879Naess, S., Hasselfield, M., McMahon, J., et al. 2014, JCAP, 10, 7Nibarger, J. P., Beall, J. A., Becker, D., et al. 2012, Journal ofLow Temperature Physics, 167, 522