Telescopes and Instrumentation

APEX–SZ: The Atacama Pathfinder EXperiment Sunyaev–Zel’dovich Instrument

Daniel Schwan1 9 Materials Sciences Division, phenomena in great detail, for example, Rüdiger Kneissl2, 3 Lawrence Berkeley National Labo­ cluster formation and dynamics, cluster Peter Ade4 ratory, Berkeley, USA mergers and their associated shock Kaustuv Basu 5 10 Department of Physics, McGill waves (which are sites of cosmic ray Amy Bender 6 University, Montreal, Canada acceleration), the cosmic ratio of dark Frank Bertoldi 5 11 Department of Physics, University of matter to baryonic matter density, active Hans Böhringer7 Colorado, Boulder, USA galactic nuclei feedback, the evolution of Hsaio-Mei Cho 8 12 Department of Earth and Space in the ICM, heat transport pro- Gayoung Chon7 ­Sciences, Onsala Space Observatory, cesses, and instabilities (Sarazin, John Clarke1, 9 Chalmers University of Technology, 1988). Matt Dobbs10 Sweden Daniel Ferrusca1 13 Department of Physics, Columbia In addition, clusters offer a unique Daniel Flanigan1 ­University, New York, USA probe into the composition and evolu- Nils Halverson 6,11 14 Physics Division, Lawrence Berkeley tion of the Universe. They are the most William Holzapfel1 National Laboratory, Berkeley, USA massive gravitationally-collapsed objects Cathy Horellou12 15 Max--Institut für Radioastro­ and directly map large-scale structure Daniel Johansson12 nomie, Bonn, Germany over a large range of . Conse- Bradley Johnson1,13 16 I. Physikalisches Institut, Universität zu quently, measuring the evolution of the James Kennedy10 Köln, Germany cluster number density places powerful Zigmund Kermish1 constraints on the parameters that Matthias Klein 5 ­govern the growth of structure. These Trevor Lanting10 The APEX–SZ instrument was a milli­ parameters are, notably, the mass energy Adrian Lee1,14 metre-wave (150 GHz) cryogenic re­­ density and dark energy density of Martin Lueker1 ceiver for the APEX telescope designed the Universe, the dark energy equation of Jared Mehl1 to observe galaxy clusters via the state, and the matter power spectrum Karl Menten15 ­Sunyaev–Zel’dovich Effect (SZE). The normalisation (Carlstrom et al., 2002). Dirk Muders15 receiver contained a focal plane of Florian Pacaud 5 280 superconducting transition-edge The hot ICM is observable at millimetre- Thomas Plagge1 sensor bolometers equipped with a wavelengths via the SZE (Sunyaev & Christian Reichardt1 ­frequency-domain-multiplexed readout Zel’dovich, 1970). In the SZE, about 1% Paul Richards1 system, and it played a key role in the of cosmic microwave background (CMB) Reinhold Schaaf 5 introduction of these new, robust, and photons incident on the cluster are Peter Schilke16 scalable technologies. With 1-arcminute ­scattered by free intracluster electrons. Martin Sommer 5,15 resolution, the instrument had a higher The resulting signal distorts the CMB Helmuth Spieler 14 instantaneous sensitivity and covered a blackbody spectrum and is visible as a Carole Tucker 4 larger field of view (22 arcminutes) than temperature decrement (increment) at Axel Weiss15 earlier generations of SZE instruments. frequencies below (above) 217 GHz. The Benjamin Westbrook1 During its period of operation from 2007 SZE surface brightness is proportional Oliver Zahn1 to 2010, APEX–SZ was used to image to the integrated gas pressure along the over 40 clusters and map fields overlap- line of sight through the cluster. Since ping with external datasets. This paper the SZE is a scattering effect seen rela- 1 Department of Physics, University of briefly describes the instrument and tive to the CMB, and the scattered and California, Berkeley, USA data reduction procedure and presents unscattered photons are redshifted 2 ESO a cluster image gallery, as well as results together, the SZE signal is independent 3 Joint ALMA Observatory, Vitacura, for the , Abell 2204, Abell of the of the scattering cluster. Santiago, Chile 2163, and a power spectrum analysis in In contrast, X-ray and optical surface 4 School of Physics and , the XMM-LSS field. brightness dim with increasing redshift. ­Cardiff University, Wales, United The redshift independence makes the Kingdom SZE a uniquely sensitive method for dis- 5 Argelander Institute for Astronomy, Galaxy clusters covering and observing distant clusters. Bonn University, Germany 6 Center for Astrophysics and Space Galaxy clusters comprise Astronomy, Department of Astophysical and constituent galaxies with hot, low Instrument overview and Planetary Sciences, University of density ionised gas filling the space Colorado, Boulder, USA between the galaxies. This gas, known APEX–SZ observations were performed at 7 Max-Planck-Institut für extraterres­ as the (ICM), domi- 150 GHz (2 mm) from the 12-metre APEX trische Physik, Garching, Germany nates the baryonic mass in clusters. telescope (Güsten et al., 2006). Technical 8 National Institute of Standards and Thus, galaxy clusters are important labo- details of the receiver can be found in Technology, Boulder, USA ratories for studying various astrophysical Schwan et al. (2011). The telescope was

The Messenger 147 – March 2012 7 Telescopes and Instrumentation Schwan D. et al., APEX–SZ: The APEX Sunyaev–Zel’dovich Instrument

dent on the absorber are measured by the TES mounted at its centre. The TES is a thin superconducting film with a tran­ sition temperature higher than the reser- voir temperature. A constant bias voltage is applied across the TES so that the total applied power (the electrical power dissipated in the TES plus optical power incident on the absorber) raises 1 mm the temperature of the sensor to pre- cisely the transition temperature. An increase (decrease) in the incident optical power produces a cancelling decrease (increase) in electrical power. Since the 25 mm voltage bias is held constant, the change in electrical power is measured as a change in current through the sensor. Figure 1. The TES bolometer array, with inset show- machined from a single aluminium block ing three spiderweb bolometers. Six identical trian- and mounted above the planar focal The bolometers are fabricated using gular sub-arrays are assembled into a single planar array, 133 mm in diameter, containing 330 bolome- ­surface. Each bolometer was centred standard photolithographic techniques. ters. Each bolometer has two leads which run from behind a cylindrical waveguide in the horn The absorber, the TES and the leads the spiderweb to bonding pads at the edge of its tri- array. There was a small gap between to bias the TES are deposited onto a 1 μm angle, visible as light traces between the bolometers. the horn array and the bolometers to pro- thick, low-stress silicon nitride (LSN) vide thermal isolation. Radiation not ­spiderweb membrane. The LSN spider- commissioned by the Max-Planck-Institut absorbed by a detector could leak radi- web is suspended by eight LSN legs für Radioastronomie, ESO and the ally to neighbouring detectors or be above a ~ 20 μm vacuum gap etched in ­Swedish Onsala Space Observatory for reflected back through the horns. This the silicon. The gap thermally isolates use with bolometric and heterodyne leakage resulted in a small, ~ 1%, optical the absorber from the silicon substrate. receivers. It is located near the ALMA site cross-talk between adjacent bolometers. The thermal conductance of the bo- at an altitude of 5107 metres on Llano lom­eter is set by a gold thermal link that de Chajnantor in the Atacama desert in The TES bolometers and frequency- runs from the TES to the thermal reser- Chile, which has exceptionally low water domain multiplexed (fMUX) readout sys- voir, parallel to the bias leads. The vapour conditions. In addition, the low tem used in the APEX–SZ focal plane APEX–SZ detector array is assembled latitude of the site (23 degrees south) have low noise, but require a cryogenic from six identical 55-element, triangular allows targeted observation of clusters system for their operation. The receiver subarrays (Figure 1). with rich multi-frequency datasets. Over- interior was cooled by a closed cycle lapping observations at different wave- pulse-tube cooler with cold stages at 60 K The APEX–SZ TES bolometers are read lengths enable a more complete study and 4 K. The mechanical cooler elimi- out with a SQUID (Superconducting and characterisation of clusters. nated the need for open reservoirs of liq- Quantum Interference Device) amplifier uid cryogens, greatly simplifying the fMUX, which allows several detectors to The telescope beam was coupled to the design and construction of the cryostat be read out by a single 4 K SQUID ampli- APEX receiver by a series of reimaging while reducing the cost and difficulty fier connected through a single pair of optics inside the Cassegrain cabin. The of remote observations. The bolometers wires (Lanting et al., 2004). Readout mul- full optical system had 58-arcsecond were further cooled to 280 mK by ti­plexing enables the use of large detector FWHM (Full Width at Half Maximum) a three-stage sorption fridge. arrays by greatly reducing the thermal beams, diffraction-limited performance, load on the low temperature stage, the over a 22-arcminute field of view (FoV). complexity of cold wiring and the system This large FoV enabled APEX–SZ to map Detectors and readout cost. The APEX–SZ system uses seven the large angular size of low redshift detectors read out by one SQUID ampli- ­clusters, where the SZE signal can extend Each APEX–SZ bolometer consists of fier in each multiplexer module. beyond a 10-arcminute radius. an optical absorber coupled to a TES that is connected to a thermal reservoir by In the fMUX system, the bolometers are Within the receiver, the APEX–SZ focal a weak thermal link. The absorbing ele- biased with alternating voltages at carrier plane was 133 mm across and consisted ment is a 3 mm diameter gold spiderweb frequencies (0.3–1 MHz) that are much of 330 feedhorn-coupled TES (transition- (Figure 1). The spiderweb shape absorbs higher than the bolometer thermal band- edge sensor) bolometers. Of these, millimetre-wave radiation, provides a width, so the bias deposits a constant 280 detectors were read out. The detec- small cross-section to cosmic rays and power on the sensor. Each bolometer in tors and readout system are described in minimises the heat capacity of the the module is biased at a different fre- the next section. The horn array was absorber. Changes in optical power inci- quency. As described above, changes in

8 The Messenger 147 – March 2012 incident radiation modulate the current through the bolometer. This amplitude modulation translates the absorbed sig- nal spectrum to sidebands centred around the carrier bias frequency. Since the currents from the different bolometers are separated in frequency, they can be combined and transmitted to the SQUID amplifier with a single wire. Furthermore, the bolometers are connected through Bullet A2163A2204 series inductor–capacitor circuits, each of which is tuned to the appropriate bias frequency, so all bias carriers can also be fed through a single line. As a result, each multiplexer module requires only a single pair of wires from the low temperature stage to the 4 K stage to read out all the A1835 A1689 RXCJ1347 RXCJ0232 RXCJ0532 bolometers in the module.

Imaging techniques

With bolometer receivers operating near the performance limit set by statistical RXCJ0245 A520 RXCJ1206 MS1054 RXCJ2214 photon noise, the main challenge in long- integration, millimetre-wave observa- tions of spatially extended signals is the contamination from atmospheric noise. The emission from the turbulent atmos- phere carrying water vapour typically A2537 RXCJ1023 A3404A1300 A2813 exhibits a Kolmogorov power spectrum (Tatarskii, 1961), with power increasing steeply with increasing angular scale. It also varies with time relative to the celes- tial signal. The APEX–SZ observation ­pattern and data reduction techniques RXCJ0528 RXCJ2243 A2744A209XLSSC006 exploit these characteristics to mitigate atmospheric noise.

APEX–SZ used a circular drift scan pat- tern to observe clusters. In the circular drift scan, the telescope performs circu- lar scans which are stationary in azimuth RXCJ0516 RXCJ2337 A907 MACSJ1115MS0451 and elevation. After the source drifts across the array FoV, the telescope slews Figure 2. Signal-to-noise (S/N) maps of 28 clusters The procedure for removing atmospheric to the new source position. This circular in which the cluster signal map is divided by a noise noise also removes cluster signal, and estimate derived from jack-knife maps (a weighted drift scan pattern combines constant average of all scan maps in which a randomly the effect is especially pronounced low amplitude acceleration for efficiency selected half of the scan maps are multiplied by –1). for clusters of large angular size. We use with modulation of celestial signals The signal and noise maps are smoothed by con- a point source transfer function (PSTF) at timescales faster than typical atmos- volving with a 1-arcminute FWHM Gaussian function. to measure the effect of the instrument Each image is 15 arcminutes square. The Bullet pheric variations. map has a peak S/N roughly four times that of any beam and data filtering on the resulting other cluster and is shown with a different colour cluster map. Since the filtering process The data reduction algorithms remove table; all other clusters have the same S/N scale. is linear, the PSTF accurately encodes the the low frequency, large-scale atmos- distortion of the cluster signal. To gener- pheric noise using high-pass filters in the ate the PSTF, simulated point sources are time and spatial domains. A low-order spatial polynomial is fitted across the inserted into the real timestream and polynomial is fitted and subtracted from array for each time step. Figure 2 shows their spatial distortion is measured post- the data timestream for each channel. 28 cluster images measured with this processing. The transfer function encodes Additionally, a low-order two-dimensional procedure. information about the imaging quality. In

The Messenger 147 – March 2012 9 Telescopes and Instrumentation Schwan D. et al., APEX–SZ: The APEX Sunyaev–Zel’dovich Instrument

¦ࣳ  Figure 3. APEX–SZ al. (2009). The cluster map of maps of two clusters. is produced by applying a procedure  Top: Map of the Bullet cluster (from Halverson similar to the CLEAN algorithm used in radio interferometry to deconvolve the ࣳ  et al., 2009). The circle

 in the lower left corner signal image and PSTF. The procedure, represents the 85-arc- detailed by Nord et al. (2009), allowed   second FWHM map res-

) ࣳ olution, which is the the first direct reconstruction of the den- ! M result of the instrument sity and temperature profiles in a cluster HN

 ", beam, data reduction without a parametric model.

§* ­filter, and -1 arcminute  ࣳ  FWHM Gaussian

#DBKHM @S smoothing applied to the map. The contour Cluster gas constraints from the SZE and  intervals are 100 μK . ࣳ CMB X-ray The + marker indicates  the centroid position of the best-fit elliptical Joint SZE and X-ray observations allow  -model. The * marker strong constraints to be placed on the G L R L R L R β        indicates the position of cluster gas. The surface brightness of the 1HFGS RBDMRHNM) a bright, dust-obscured, SZE signal is proportional to the electron lensed galaxy, which ­ density and the temperature of the intra- 29$CDBQDLDMSƅ*  is detected at higher ",! ­frequencies (e.g., at cluster gas. X-ray emission from the gas l  l  l  l     350 GHz by LABOCA; is dominated by thermal , Johansson et al., 2010), which is roughly proportional to the but does not signifi- l  l l  l l  l  cantly contaminate the square of the density and the X-ray cool- ,)XRQ measurement at 150 ing function, itself a weak function of l  GHz. Bottom: Map of temperature. Multi-frequency data enable Abell 2163 (from Nord et detailed measurements of the cluster al., 2009). The APEX–SZ map, deconvolved from gas, including determination of the gas l  the effects of the trans- temperature and mass, as well as their fer function, is overlaid distributions. with XMM-Newton X-ray l  contours in units of 10–13 erg s–1 cm–2 arcmin–2. Halverson et al. (2009) fit the intracluster gas of the Bullet cluster to an elliptical & isothermal -profile (Cavaliere & Fusco- #$ l   β 

"  Femiano, 1978) convolved with the PSTF. 

#$ A Markov-chain Monte Carlo approach l    is used to find the maximum likelihood in parameter space with an X-ray-derived prior on β. The analysis yields a core l  radius rc = 142 +/– 18 arcseconds, an axial ratio 0.889 +/– 0.072, and a central l  /$7l29AD@L temperature decrement –771 +/– 71 μKCMB (in units of temperature referred to a               source at the CMB ­temperature), in­cluding 1 #$& a +/– 5.5 % flux calibration uncertainty. With a map of projected electron density particular, negative sidelobes indicate ­filtering on the sky signal. The Bullet clus- from Chandra data, the SZE-derived gas missing spatial information on larger ter is a system of two merging clusters (electron) ­temperature is Te = 10.8 +/– 0.9 angular scales. This is very similar to an (mass ratio ~ 1:10), which shows an offset keV. This temperature is lower than some interferometric synthesised beam, where between the gas and dark matter (Clowe previously reported X-ray temperatures, the inner portion of the Fourier plane, et al., 2006; also Figure 5). The map which could be biased towards hot, com- corresponding to very short baselines, is of Figure 3 is produced by masking a cir- pact regions. On the other hand, the often not sampled. In both cases, the cular region centred on the cluster source gas mass fraction derived from the SZE observation process is designed so that prior to fitting for the low-order polyno- map (under the approximation of hydro- the most relevant information is most mial filters, then applying the filters to the static equilibrium), is in good agreement densely sampled. entire dataset. The masking procedure with estimates from X-rays. limits attenuation of the cluster extended Figure 3 shows maps of the Bullet cluster emission at the expense of increased Nord et al. (2009) and Basu et al. (2010) and Abell 2163 made using two different map noise. The observations and data further explore cluster physics with SZE schemes to mitigate the effect of noise processing are detailed by Halverson et data combined with X-rays. Nord et al.

10 The Messenger 147 – March 2012 MFTK@QCHRS@MBD@QBLHMTSDR      QDK@WDC BNNKBNQD

LDQFDQCHRSTQADC

Q     L  5B D5 JD J   Q * D 3 



          QQ QJOB 

Figure 4. Derived radial profiles of cluster gas proper- non-parametric estimates for the total are the electron temperature and density, ties. Left: De-projected temperature for Abell 2163 mass and the gas mass, giving a gas respectively) for Abell 2204 and Abell with 1σ uncertainties (from Nord et al., 2009). The mass fraction profile, which again is con- 2163 shows that the entropy in the centre vertical arrow labelled “r200” marks the aproproximate virial radius (i.e., outer boundary) of the cluster. sistent with X-ray analyses. Nord et al. of a relaxed cluster is much lower than in Right: Entropy derived from APEX–SZ and X-ray data (2009) also combine the APEX–SZ SZE a violently disturbed merging cluster and for the relaxed cluster Abell 2204 (red circles) and the measurements of Abell 2163 at 150 GHz follows a power law (Figure 4). This result merging cluster Abell 2163 (blue squares). with 345 GHz measurements from the LABOCA camera on APEX and published (2009) derive gas density and tempera- measurements at other frequencies to Figure 5. Left: Multi-wavelength overlay (from ture profiles for Abell 2163 via the Abel place constraints on the peculiar velocity ­Halverson et al., 2009) on the Bullet cluster X-ray (XMM-Newton) colour image with SZE contours integral in Silk & White (1978). The derived of the cluster. Basu et al. (2010) present in white (100 μKCMB level intervals) and the weak- temperature profile shown in Figure 4, an analysis of the relaxed cluster Abell lensing surface mass density reconstruction con- representative of the bulk of the gas, is in 2204, where the assumption of spherical tours in green. Right: B, V, R colour image of the excellent agreement with the high gas- symmetry should be especially valid. central region of RXCJ0232.2-4420 (z = 0.28) with SZE contours in cyan (–90, –110, –130, –150, temperature estimates from X-ray data. A comparison of the derived entropy pro- –170 μKCMB) and smoothed weak-lensing conver- –2/3 The profiles have been converted to yield files (entropy is Te n , where Te and n gence contours in red (0.10, 0.11, 0.12, 0.13, 0.14).

–55°52ࣳ 1ࣳ = 253 kpc

54ࣳ ) 00 20

(J 56ࣳ

at ion

Declin 58ࣳ

00ࣳ

06h59m00s 58m30s 58m00s Right Ascension (J2000)

The Messenger 147 – March 2012 11 Telescopes and Instrumentation Schwan D. et al., APEX–SZ: The APEX Sunyaev–Zel’dovich Instrument

+ 1.4 2 –1 agrees with previous X-ray studies, but band powers) or 1.7– 1.3 Jy sr . Assuming A major aspect of the recent and future is derived independently without X-ray that the same sources are responsible work is a multi-wavelength analysis of spectroscopy. for the power fluctuations measured by clusters imaged with APEX–SZ (Figure 5). BLAST at 600 GHz, the spectral index For this reason, an optical follow-up + 0.4 of the source population is 2.6– 0.2. The ­programme has been conducted using Power spectrum analysis ­Atacama Large Millimeter/submillimeter the Wide Field Imager (WFI) at the 2.2- Array (ALMA) will resolve remaining ques- metre MPG/ESO telescope. Deep B-, V- Reichardt et al. (2009) use an SZE tions about contaminating sources by and R-band images obtained during image taken with APEX–SZ within the providing the missing statistical informa- 42 nights, together with archival data XMM-LSS field (0.8 square degrees, tion on population counts over all rele- from WFI and Subaru Suprime-Cam, pro- ­centred on the medium mass cluster vant wavelengths and by providing the vide excellent weak-lensing data for

XLSSUJ022145.2-034614, with 12 μKCMB ability to directly correct SZE images for 35 galaxy clusters at redshift 0.15–1 for root mean square noise at the centre) their point source contamination. the study of SZE and weak-lensing scal- for statistical analysis of the temperature ing relations. Our multi-frequency cluster fluctuations from undetected clusters data provide a unique sample to study and other sources. This type of fluctua- Future directions cluster mass scaling laws and baryon tion analysis applied to large datasets physics in unprecedented detail. places strong constraints on σ8, the nor- APEX–SZ was a pioneering instrument for malisation of the matter density power arcminute resolution SZE cluster observa- spectrum on the scale of roughly 10 Mpc. tions, advancing bolometer and readout References The power spectrum of the integrated technology and applying multi-wavelength Arnold, K. et al. 2010, SPIE, 77411E SZE from all clusters along the line of data analysis to constrain cluster gas Basu, K. et al. 2010, A&A, 519, A29 7 sight scales as σ8. The APEX–SZ image physics. The TES bolometer array with Bender, A. et al. 2012, in preparation was the most sensitive fluctuation meas- fMUX readout has been adopted for dedi- Carlstrom, J. E., Holder, G. & Reese, E. D. 2002, urement at 150 GHz at the time, with cated instruments with larger arrays, Ann Rev Astron. Astrophys., 40, 643 Cavaliere, A. & Fusco-Femiano, R. 1978, A&A, the resulting constraint on the matter notably the South Pole ­Telescope, with 70, 677 fluctuation amplitude beingσ 8 < 1.18 at 960 detectors, and ­POLARBEAR (Arnold Clowe, D. et al. 2006, ApJL, 648, L109 95 % confidence. Determination of the et al., 2010), a Berkeley CMB polarisation Güsten, R. et al. 2006, A&A, 454, L13 constraint requires consideration of con- experiment, with over 1200 detectors. Halverson, N. W. et al. 2009, ApJ, 701, 42 Johansson, D. et al. 2010, A&A, 514, A77 tamination of the SZE signal by other Lanting, T. M. et al. 2004, Nuclear Instruments and populations, such as radio sources or APEX–SZ took over 800 hours of scien- Methods in Physics Research A, 520, 548 dusty submillimetre (sub-mm) galaxies, tific data during its four years of opera- Nord, M. et al. 2009, A&A, 506, 623 which can be correlated with clusters as tions, targeting over 40 X-ray selected Reichardt, C. L. et al. 2009, ApJ, 694, 1200 Sarazin, C. L. 1988, X-ray emission from clusters of members or via gravitational lensing. clusters in a wide range of mass and red- galaxies (Cambridge: Cambridge University Press) The Reichardt et al. (2009) analysis esti- shift. Over 30 of these clusters were Schwan, D. et al. 2011, Review of Scientific mates the contribution of sub-mm galax- detected with high significance. The Instruments, 82, 091301 ies to the power spectrum at 150 GHz analy­sis of the full dataset is still ongoing Silk, J. & White, S. D. M. 1978, ApJL, 226, L103 + 0.9 –5 2 Sunyaev, R. A. & Zel’dovich, Y. B. 1970, Comments to be Cl = 1.1 – 0.8 × 10 μK (the ampli- (Bender et al., 2012). These observations on Astrophysics and Space Physics, 2, 66 tude of the angular power spectrum are being used to study the scaling of Tatarskii, V. I., ed. 1961, Wave Propagation in a for spherical harmonic multipole index l in the SZE signal with cluster mass (hydro- Turbulent Medium (New York: Dover) the range 3000–10 000 assuming flat dynamical and weak lensing).

Composite image of the complex merger Abell 2744 (redshift 0.308) formed by com­ bining visible light images from NASA/ESA Hubble Space Telescope and VLT with an image of the hot intracluster gas (pink) from NASA Chandra X-ray Observatory data. A reconstruction of the dark mat- ter in the cluster is overlaid (blue). See Science

NASA, ESA, Coe ESO, & D. CXC (STScI)/J. Merten (Heidelberg/Bologna) Release eso1120 for full details.

12 The Messenger 147 – March 2012