7 pages

A Kilopixel Array of TES Bolometers for ACT: Development, Testing, and First Light

of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
The Millimeter Bolometer Array Camera (MBAC) will be installed on the 6-meter Atacama Cosmology Telescope (ACT) in late 2007. For the first season of observations, MBAC will comprise a 145 GHz diffraction-limited, 1024-pixel, focal plane array of
  J Low Temp Phys (2008) 151: 690–696DOI 10.1007/s10909-008-9729-2 A Kilopixel Array of TES Bolometers for ACT:Development, Testing, and First Light M.D. Niemack  · Y. Zhao  · E. Wollack  · R. Thornton  · E.R. Switzer  · D.S. Swetz  · S.T. Staggs  · L. Page  · O. Stryzak  · H. Moseley  · T.A. Marriage  · M. Limon  · J.M. Lau  · J. Klein  · M. Kaul  · N. Jarosik  · K.D. Irwin  · A.D. Hincks  · G.C. Hilton  · M. Halpern  · J.W. Fowler  · R.P. Fisher  · R. Dünner  · W.B. Doriese  · S.R. Dicker  · M.J. Devlin  · J. Chervenak  · B. Burger  · E.S. Battistelli  · J. Appel  · M. Amiri  · C. Allen  · A.M. Aboobaker Received: 23 July 2007 / Accepted: 15 September 2007 / Published online: 24 January 2008© Springer Science+Business Media, LLC 2008 Abstract  The Millimeter Bolometer Array Camera (MBAC) will be installed on the6-meter Atacama Cosmology Telescope (ACT) in late 2007. For the first season of observations, MBAC will comprise a 145 GHz diffraction-limited, 1024-pixel, focalplane array of Transition Edge Sensor (TES) Bolometers. This will be the largest ar-ray of pop-up-detector bolometers ever fielded as well as one of the largest arrays of TES bolometers. We discuss the design specifications for the array and pre-assemblytesting procedures for the cryogenic components. We present dark measurements of the TES bolometer properties of numerous 32-pixel columns that have been assem-bled into the first kilopixel array for ACT, as well as optical measurements made withour 256-pixel prototype array, including first light measurements on ACT. M.D. Niemack (  ) · Y. Zhao · E.R. Switzer · S.T. Staggs · L. Page · O. Stryzak  · T.A. Marriage · J.M. Lau · N. Jarosik  · A.D. Hincks · J.W. Fowler · R.P. Fisher · J. AppelPrinceton University Physics Department, Princeton, NJ 08544, USAe-mail: mniemack@princeton.eduE. Wollack  · H. Moseley · J. Chervenak  · C. AllenNASA Goddard Space Flight Center, Greenbelt, MD 20771, USAR. Thornton · D.S. Swetz · M. Limon · J. Klein · M. Kaul · S.R. Dicker · M.J. DevlinUniversity of Pennsylvania Physics Department, Philadelphia, PA 19104, USAM. Halpern · B. Burger · E.S. Battistelli · M. AmiriUniv. of British Columbia Dept. of Physics and Astro., Vancouver, BC V6T 1Z1, CanadaK.D. Irwin · G.C. Hilton · W.B. DorieseNational Institute of Standards and Technology, Boulder, CO 80305, USAR. DünnerUniversidad Católica, Santiago, ChileA.M. AboobakerUniv. of Minnesota School of Physics and Astro., Minneapolis, MN 55455, USA  J Low Temp Phys (2008) 151: 690–696 691 Keywords  TES · Bolometer · Detector array · Astrophysics · Cosmology PACS  07.57.Kp · 95.55.Rg · 95.85.Bh · 95.55.Jz · 95.85.Fm · 98.80.Es 1 Introduction The Millimeter Bolometer Array Camera will measure the Cosmic Microwave Back-ground radiation (CMB) temperature anisotropies at small angular scales [1] on the6-meter ACT [2]. The construction of ACT was recently completed in Chile, and en-gineering observations have begun with our prototype instrument, the Column Cam-era (CCam) [3]. We are building three 1024-pixel arrays of Transition Edge Sensor(TES) bolometers for MBAC, which are optimized for use at 145, 217, and 265 GHz.These three frequencies will allow background and point source subtraction for CMBanalysis at high multipoles. In addition, they span the null of the Sunyaev-Zel’dovich(SZ) Effect—the Compton scattering of CMB photons off hot electrons in galaxyclusters. By measuring the SZ effect with these bands, we will generate an unbiasedgalaxy cluster catalog [1]. Combining this catalog with optical and X-ray measure-ments will allow us to measure cluster density as a function of redshift, which canbe used to constrain cosmological parameters, including the equation of state of thedark energy that dominates our universe.Our bolometer arrays are highly modular [4, 5], which enables us to pre-screen and choose the best components for each 32 pixel column module. Each module iscomprised of 5 primary components: a TES bolometer chip, a chip of shunt resistorsto voltage bias the TESs, a Nyquist inductor chip to band-limit the TES response, asuperconducting quantum interference device (SQUID) multiplexing chip, and a Sicircuit board with Al wiring to which all the other components are mounted [6]. Priorto assembly with detectors, the series resistance of the 32 resistors on each shunt chipandthe SQUIDcriticalcurrentsof the multiplexing(mux) chipsare characterizedin a 4 He dip probe. The pop-up detector [7] bolometers as well as the shunt resistor chipsare fabricated at NASA Goddard’s Detector Development Laboratory. The bolometerresponse is measured using a three-stage SQUID time-domain multiplexing (TDM)system [8, 9] developed at NIST, Boulder, where the Nyquist inductor chips are also fabricated. The Si circuit boards are fabricated at Princeton, where components aretested and built into arrays (Fig. 3b).In this paper we give a general description of the detector and parameter selectionfor these arrays. Then, we present dark measurements of bolometer properties forMBAC made in two rapid dip probe refrigerators [4]. We also present optical mea-surements of detectors in our prototype array made in CCam using the Multi-ChannelElectronics (MCE) [11] and conclude with our recent first light measurements withCCam on ACT. 2 Array Design: Detector Parameter Selection Among the variety of different measurements described here, three critical bolometerproperties determine the functionality and sensitivity of the detector array during ob-servations: the saturation power, the noise level, and the time constant, or frequency  692 J Low Temp Phys (2008) 151: 690–696 Table 1  Dark measurement results from eight of the detector columns with 32 pixels each that will gointo 145 GHz array. These measurements were acquired at  T  b ≈ 0 . 38 K, which is why the bias power,  P  J  o is somewhat lower than discussed in Sect. 2 T  c  R n  R sh  P  J  o  I  b  at 0.5  R n  Noise  at 10 HzAverage 0.512 K 32 m   0.74 m   6.5 pW 0.46 mA 4 . 1 × 10 − 17 Ws 1 / 2 Std. Dev. 0.026 K 5 m   0.10 m   1.1 pW 0.04 mA 1 . 3 × 10 − 17 Ws 1 / 2 response. The saturation power,  P  sat  , will determine whether the detectors can func-tion without saturating during observations. The noise level impacts the ratio of pho-ton noise to detector noise and affects the final sensitivity of the sky maps. The timeconstant,  f  3dB , determines how quickly the telescope can be scanned across the skywithout low-pass filtering small angular scale signals.The selection of detector parameters for the array has been a gradual processmotivated by a combination of theoretical calculations, measurements, and fab-rication recipe success. The non-multiplexed SQUID noise was measured to be ∼ 0.5µ  φ o / √  Hz [8],andwefoundthataliasingcausedbymultiplexingatourplannedrate, f  samp = 15 . 2 kHz,increasedtheSQUIDnoiselevelbyafactorof  ∼ 8.Thisdroveus to select a low TES normal resistance,  R n , of   ∼ 30 m   to ensure that the detec-tor current noise would dominate the TDM SQUID noise. The shunt resistance value, ∼ 0 . 7 m   R n , was selected to keep the TESs roughly voltage-biased without beingso small as to generate excess Joule heating in the shunt resistors that would warmup the bath temperature.The predicted loading in the 145 GHz band on the Atacama Plateau is  < 2 pWduring the observation season [6], which puts a lower limit on the acceptable  P  sat  .(Load curve measurements in Chile have recently confirmed this estimate.) Consid-eration of the measured scatter in detector parameters (Table 1), our desire to biasthe entire kilopixel detector array with only a few bias lines, and the results of timeconstant measurements (Sect. 4) drove us to choose a substantially higher  P  sat  . Asidefrom its dependence on the bath temperature,  P  sat   depends on the TES transitiontemperature,  T  c  and the thermal conductivity,  G , of the weak link to the bath. Thedependence of the latter two on fabrication recipes are determined by laboratory mea-surements [4]. For the 145 GHz array, the bolometers have four 5 µm wide × 1.1 µmthick  × 0.5 mm long Si legs connecting them to the ∼ 0 . 3 K bath and  T  c ≈ 0 . 51 K,resulting in  P  sat  ( 0 . 3 K ) ≈ 8 pW. 3 Dark Detector Measurements Each column of 32 bolometers is subjected to a series of tests prior to insertion intothe array. These tests are performed on every detector and include: measurements of the SQUID V- φ  curve through the SQUID feedback and detector bias lines; TES  T  c ;Johnson noise spectra to extract the shunt resistance,  R sh , and Nyquist inductance;loadcurvesforTESnormalresistance, R n ,saturationpower, P  sat   andbiascurrent, I  b ;multiple noise measurements on the transition to measure non-multiplexed noisespectra and check for anomalies (Table 1). Based on these measurements, columns  J Low Temp Phys (2008) 151: 690–696 693 Fig. 1  (Color online) Comparison of noise measurements ( solid  ) and aliasing estimates ( dashed  )pre-Nyquist inductors ( thin lines ) vs. post-Nyquist inductors ( bold lines ) at 0.2  R n  ( green —the top threecurves) and 0.65  R n  ( black  ). Thick   vertical arrows  near 6 Hz indicate the estimated reduction of in bandaliasing by adding the Nyquist inductor into the TES loop for the final array have typically had a 90% pixel yield. The few bad pixels on eachcolumn are caused by a variety of different failure modes, including: mechanicallybroken pixels, unresponsive or open stage 1 SQUIDs, electrical shorts on the TES orthe mux chip, and oscillating detectors at certain biases.Noise and complex impedance measurements of the bolometers indicate that ther-mally they comprise at least three elements with isolated heat capacities. Dünner hasled an effort to model this system. He found that the noise and impedance data canbe explained well by a model in which the TES and ion-implanted absorber heat ca-pacities are thermally isolated and connected to the bath through the bolometer Si.This seems appropriate since these bolometers have a small TES (50 µm × 50 µm)separated from a thin absorber layer by the relatively large (1.05 mm × 1.05 mm × 1.1 µm) bulk Si.The isolated heat capacities and the resulting increase in noise in the mid-frequency range (Fig. 1) should not pose a problem so long as the noise is preventedfrom aliasing into our low-frequency sky sampling band. Adding inductance into theTES loop reduces aliasing by decreasing the frequency of the  L/R  electrical polebelow the Nyquist sampling frequency,  f  Nyq  = f  samp / 2 = 7 . 6 kHz. Due to the com-plexity of the bolometer model, we determined the inductance experimentally using aNIST multi- L  chip with values between  L = 0 . 1–1.4 µH. The noise on each bolome-ter was measured at multiple bias points before and after adding the inductor (Fig. 1).The noise level after aliasing,  N  a (f)  where  f   is a frequency between 0 Hz- f  Nyq ,was estimated before and after adding the inductor by folding the measured powerspectral density (PSD),  N  m (f) , about  f  Nyq :  N  a (f) =  ki = 0 N  m ( 2 if  Nyq + f( − 1 ) i ). We sum over  i  =  0  →  k  folds, where  k  is an integer that meets the condition: N  m ( 2 kf  Nyq  + f( − 1 ) k )  N  a (f)  for  k  and integers greater than  k . Aliasing esti-mates in Fig. 1 were summed to  k = 4, or 38 kHz.To select the optimal inductance, we found the maximum decrease of in-bandaliased noise that occurred after introducing the inductor. Above the optimal induc-tance, detectors tend to either have an increase in low-frequency noise or be driven  694 J Low Temp Phys (2008) 151: 690–696 into oscillations. Due to the scatter in our detector properties, we chose an inductance15% below the optimal value, resulting in a total circuit inductance of   ∼ 0 . 75 µH.This has been a successful approach in that we often find a single detector in each32-pixel column that is driven into an unstable resonant state at some biases, but thereis rarely more than one per column. Wirebonds are disconnected from detectors thatexhibit resonant behavior to prevent oscillations from contaminating the signals onneighboring detectors.We are in the process of testing a Si coupling layer design that optimizes theoptical impedance of the bolometers to match free space. The layer resembles anevanescently coupled anti-reflection coating. Calculations indicate that such a layercan yield close to 80% radiation efficiency. 4 Optical Time Constant Measurements Our observation strategy is to scan the entire telescope in azimuth 5 ◦  peak-to-peak [2]. The angular velocity of this scan imposes a lower limit on the detector f  3dB  to prevent filtering of the beams at high frequencies [3]. Measurements of thedetector time constants of the bolometers in the prototype 8 × 32 array were madeby chopping a 300 K Ecosorb source in front of a small aperture near a focus of theCCam dewar at numerous frequencies. A 9% transmissive neutral density filter wasin the dewar optical path to prevent saturation of the bolometers. All measurementswere made in a multiplexed mode using the MCE [11] for data acquisition.The data were analyzed by Fourier transforming the time streams, integrating overthe peak response to the chopper and subtracting the background PSD. Bolometer f  3dB  was calculated from a single-pole fit to the frequency response. Load curvesacquired before and after each series of measurements confirmed system stabilityand provided detector parameters at the 10-20 biases studied.Based on an isothermal TES bolometer model in the low inductance limit, wewould expect [10] f  3dB = G 2 πC  1 + ( 1 − R sh /R o )α I  ( 1 + β I   + R sh /R o )GT  o P  J  o  , where  P  J  o  is the Joule power applied to the TES and  R o  is the TES resistance(parameter definitions follow the conventions of Irwin and Hilton [10].) The low-inductancelimitisanacceptableapproximation,despiteouruseof Nyquistinductors,because the  L/R  electrical time constant is roughly an order of magnitude smallerthan the optical time constants. With other parameters held constant,  f  3dB  is propor-tional to  P  J  o . Measurements of two columns of prototype detectors and one columnof detectors for the first array follow this trend (Fig. 2). The bolometers also haveroughly constant  f  3dB  between 25%–75% of   R n , indicating changes in the parame-ters  C ,  α I  , and  β I   do not have a strong impact on  f  3dB  in this regime. These mea-surements were made in the laboratory at elevated bath temperatures and loading;however, by combining this function with  G  measurements, we can estimate  f  3dB under observing conditions. We will study time constant stability during observationsby both scanning quickly across point sources and measuring bolometer responses tosquare wave steps on the detector bias line using the MCE [11].
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!