CMB polarization observations with the POLAR and COMPASS experiments Christopher O’Dell Observational Cosmology Lab University of Wisconsin-Madison

Josh Gundersen, 2001 POLAR Upper Limit Reionization Peak Z re ~ 30

Zaldarriaga, 2001 No Reionization. P rms ~ 0.5  K

  = 0.3, Z re ~ 30 Zaldarriaga, 2001 Reionization at z ~ 30 P rms ~ 3.2  K

Polarization Observations of Large Angular Regions (POLAR) A. de Oliveira-Costa (UPenn) J. Gundersen (Miami) B. Keating (Caltech) S. Klawikowski (UW-Madison) C. O’Dell (UW-Madison) L. Piccirillo (Cardiff) N. Stebor (UW-Madison) D. Swetz (UW-Madison) M. Tegmark (UPenn) P. Timbie (UW-Madison)

Ground Screens The Spinning Correlation Polarimeter ExEx EyEy GxGx GyGy OMT Multiplier E N x y CMB  0°, 180° Phase Shifter TP-x TP-y

POLAR Main Features Corrugated conical feed horn achieves 7° beam with very low sidelobes. HEMT amplifiers (25 K noise temperature, NET ~ 800  K sec -1/2 ) Commercial Cryocooler (no liquid cryogens). Frequency bands: GHz, 3 sub-bands. Clean, simple design: no lenses or mirrors no magnetic or moving parts (excepting overall rotation)

Outer Ground Screens Inner Ground Screen Outer Ground Screen Clam-shell Dome POLAR Site: Pine Bluff, WI

POLAR Scan Strategy Q and U at ~ 20 pixels on the sky, on a 7° ring at declination 43°

(30%) (3%) Josh Gundersen, 1999 Polarized Foregrounds Power Spectra

Calibration Typically calibrate with wire grid, giving a signal = T hot - T cold ~ 250 K (90% pol) We needed a much smaller signal (both in power and fractional polarization).

Solution: Replace Grid with Thin Dielectric Sheet, calculate reflection properties using simple Fresnel equations. Calibration Signal ~ 12 K (5% pol) O’Dell, Swetz, & Timbie, accepted in IEEE Trans. Mic. Th. Tech., 2002 Calibration Typically calibrate with wire grid, giving a signal = T hot - T cold ~ 250 K (90% pol) We needed a much smaller signal (both in power and fractional polarization).

Stability and Sensitivity

Q and U Time Stream for Entire Season

Offset Removal Certain matrix operations can remove sensitivity to specific “modes” in a map. There exist several formalisms for removing unwanted modes. We apply this formalism to each channel and “submap” in our cleaned data set. Then combine these “de-offsetted” submaps into final channel maps. For Each Submap/Channel: Bond, Jaffe, Knox 1998 Tegmark, 1998 Simply add Constraint Matrix to  : (simple offset removal)

Final Combined Q, U Maps

Flat Band-Power Model : E, B Spectra have constant power at all scales, characterized by variances (T E 2, T B 2 ) Full 2D Likelihood:Prior Constraint that T B = 0

Cosmic Microwave Polarization at Small Scales (COMPASS) G. Dall’Oglio Rome III P. Farese UCSB J. Gundersen U. Miami B. Keating CalTech S. Klawikowski UW L. Knox UC-Davis A. Levy UCSB P. Lubin UCSB C. O’Dell UW A. Peel UC-Davis L. Piccirillo Cardiff J. Ruhl UCSB Z. Staniszewski UW P. Timbie UW

COMPASS Facts Same radiometer as POLAR (26-36 GHz) 2.6 meter on-axis reflector, 20’ beam Scanned 1° and 1.6° dia disk at NCP. ~ 30 pixels (20’ ea) at 1.6° scan. Season 2001 ~180 usable hours (U). Season 2002: currently observing Q. Calibration on Tau A (6.6% pol’d) Effelsberg 100m companion survey at 32 GHz 1°1° 1.6° NCP

Polarized Calibration on the Tau A radio source:

COMPASS Current Status Pointing is known with good accuracy, ± 4.3’ Az, ± 1.6’ El. Noise is well behaved, with N.E.T. ~ 600  K · sec 1/2 Analysis of 2001 data (U) nearing completion. Jack-knife tests show map consistency, calibration well-understood. Addition of 2002 data (Q) will help nail down systematics, foregrounds, etc. Addition of 90 GHz (W-band) system will allow us to probe different frequency as well as smaller angular scales (7’ beam). Proposal in for HEMT array with ~ 20 pixels, will drastically increase sensitivity.

POLAR Radiometer

Atmospheric Emission at Sea Level Broad Line Emission Possible O 2 Zeeman splitting by earth’s B-field Negligible Unpolarized in absence of external fields Will add to system noise, and affect required integration time. O2O2 H2OH2O

Ground-Binned Data

Multi-Frequency Maps

Corrugated Conical Feed Horn at Ka-band (26-40 GHz) Sidelobe Rejection > 60 dB at 90 degrees.

Reflectance Test Chamber Testing the Dielectric Sheet Reflectance

R TE R TM Results Testing the Dielectric Sheet Reflectance

408 MHz Haslam Map (convolved with 7° beam) 100  m DIRBE Map (convolved with 7° beam)

Data Quality Assessment & Cuts

Observations and Data Selection March 1 st – May 31 st, 2000

Mapmaking Problem : Re-construct the map from your data that is closest to the true, underlying map. Solution : “Minimum Variance Mapmaking” (not simply weighted averaging in the presence of 1/f noise) Data Vector True Sky Map Noise Vector Pointing Matrix

Mapmaking The underlying map Data Vector Problem : Construct such that is minimized. In the Presence of 1/f noise, this is not the simple weighted- averaging solution! “Minimum Variance Mapmaking” Full Solution is characterized by map and covariance matrix: Covariance Matrix

Rotation-Ordered Data (“ROD”) We form a time stream of Q’s and U’s, tagged on the sky. Do this for all channels.

Offsets Binned on the Sky Channel J2i

Mathematics for Offset Removal Certain matrix operations can remove sensitivity to specific `modes’ in a map. There exists at several formalisms for removing unwanted modes. We apply this formalism to each channel and “submap” in our cleaned data set. The combine the de-offsetted submaps into final channel maps. For Each Submap/Channel:

Likelihood Analysis Bayes’ Theorem Likelihood Function where is a set of parameters defining the theory, and Full Covariance Matrix depends on theory and beam shape depends on noise properties of data

E-Mode Limits B-Mode Limits

Extrapolated Synchrotron at dec 43º, smoothed with a 7º beam. All extrapolations assume  = -2.8 from to 5.0 GHz. Constraints on Large-Scale Synchrotron Polarization