CMB Polarization from the South Pole: BICEP1, BICEP2, and Keck - - PowerPoint PPT Presentation

CMB Polarization from the South Pole: BICEP1, BICEP2, and Keck Array

Immanuel Buder for the BICEP1, BICEP2, and Keck Array Collaborations Harvard-Smithsonian Center for Astrophysics ibuder@cfa.harvard.edu CMB2013 Okinawa, Japan 2013-06-10

We need B-mode polarization measurements to go deeper on r

• SPT+WMAP7+BAO+H0: r < 0.11 (Story et al.,

2012)

• Planck+WMAP (pol.): r < 0.12 (Planck

Collaboration XXII, 2013)

• Theoretical limit from sample variance for

CMB temperature measurements: r < 0.1 (Knox & Turner, 1994)

We Have the Tools to Reach r = 0.01

• Statistical power (BICEP2 + Keck Array +

BICEP3 + …)

• Systematic control in analysis (being

demonstrated in BICEP1 3-year analysis)

Outline of This Talk

• BICEP and Keck Program Overview
• BICEP1 3-year Results
• BICEP2 Analysis Status
• Keck Array Status and Plans
• Conclusion

Overview of BICEP1, BICEP2, and Keck Array

• All designed for measuring ell ~ 100 B-mode

polarization

• All observe low-foreground field from South

Pole

• All have cold refracting telescopes: 26-cm

aperture (far field is close!)

• All have boresight rotation
• All have absorptive co-moving forebaffles and

stationary reflective ground shields

BICEP1, BICEP2, and Keck Collaborators

California Institute of Technology Harvard University JPL KIPAC Stanford University University of Minnesota Case Western Reserve University NIST University of British Columbia University of Toronto University of Chicago UCSD Wales Cardiff CEA Grenoble Thanks to National Science Foundation, W. M. Keck Foundation. Photo from Zak Staniszewski

BICEP1

• Observed 2006—2008
• Polarization-sensitive NTD

bolometers

• Feed-horn coupled
• 100/150 GHz dual-frequency

focal plane

• First results in 2009 (Chiang et

al.)

• Systematic error level at r ~ 0.1

(Takahashi et al.)

• Results in this talk based on 3-

year data

BICEP2 Scaled Up # of Detectors

• Observed 2010—2012
• Planar slot antenna

coupled TES

• 256 dual-polarization

pixels @ 150 GHz

• Time-domain SQUID mux

(x33)

• 10 times faster mapping

speed than BICEP1

Keck Array = 5 BICEP2's

• Switch to pulse tube coolers to pack 5

BICEP2-style telescopes on the DASI mount

• Currently all at 150 GHz, but plans include 100

and 220 GHz receivers in future

• Observations 2011—2016

BICEP1 3-year Results

• 3-year maps and sensitivity
• Analysis improvements since Chiang et al.

– Bandpower window function calculation – Likelihood model for bandpowers and r – Deprojection of beam systematic errors

• Power spectrum and r limit

BICEP1 3-year maps

Maps from Colin Bischoff

BICEP1 3-year 150-GHz polarization sensitivity: 500 nK-deg. for effective area 203 deg.2 NET: 54 uK*sqrt(s)

E-modes B-modes

RA (deg.) Dec (deg.)

• 5

5 μK 50

• 50
• 70
• 50

More accurate bandpower window function calculation

• Chiang et al. method included only the mask
• 3-year method incorporates filter and beam

– Higher ell resolution simulations to capture change

• f filter suppression within each bin

– Iteratively solve for filter suppression

• Doubles stat. error for lowest ell bin

ell 150 Bandpower Window Function Chiang et al. 3-year

• For r make quadratic

estimator and simulate at each r value

• For bandpowers, use

Hamimeche & Lewis (2008) approximation

• Replaces Bond,

Jaffe, Knox (2000) “offset lognormal” (OLN) approximation

Direct simulation-based likelihood from scaling r = 0.1 simulations and adding noise

Data release includes improved likelihood models

Quadratic Estimator Maximum Likelihood Offset lognormal has more bias and scatter

Deprojection of Beam Systematic Effects Reduced Error

• Make template map of CMB

temperature and its spatial derivatives

• Subtract the projection of these

templates on the real timestream data

• Suppressed the (previously)

largest systematic error by ~ 104 in power

6 types of beam imperfections can be deprojected—for BICEP1

• nly (a) is necessary

A-B gain mismatch systematic error reduced (Aikin et al., 2013 in prep.)

Deprojection of Beam Systematic Effects Reduced Error

• Make template map of CMB

temperature and its spatial derivatives

• Subtract the projection of these

templates on the real timestream data

• Suppressed the (previously)

largest systematic error by ~ 104 in power

6 types of beam imperfections can be deprojected—for BICEP1

• nly (a) is necessary

A-B gain mismatch systematic error reduced (Aikin et al., 2013 in prep.) from Takahashi et al. (2010)

BICEP1 Did Not Find B Modes

tensor- to- scalar ratio r < 0.70 (95% C.L.)

Power in CMB Polarization Angular Scale with r = 0.1 model BICEP1 BB Inset region (Barkats et al. 2013, in prep)

We added 50% more data. Why didn't the upper limit improve more?

• Additional data

fluctuated up (within statistical prediction)

• Previous bandpower

window function approximation underestimated first bin uncertainty

• Chiang et al. got a lucky

fluctuation of the OLN scatter

OLN (Chiang et al.) Chiang et al. found r < 0.72 (95%C.L.)

BICEP2 Analysis Status

• Map and sensitivity: 16 uK*sqrt(s) NET
• Analysis improvements

– E/B Separation – Matrix-based analysis toward map release

• Systematic error investigation example: beam

differential pointing

3-Year Map More Sensitive than Anything Before

Sensitivity to Q/U = 119 nK*deg. and effective area of 388 sq. deg.

Map from Angiola Orlando RA (deg.) Dec. (deg.) uK

E-modes

16 uK*sqrt(s) NET

Using “pureB” Estimator to Reduce E/B Mixing Effect

• Using K. Smith (Phys.Rev.D74:083002)
• Evaluating analysis choices (e.g. apodization)
• Exploring simulation approach to filtering

effects

Plot from Sarah Kernasovskiy

• “Standard” pseudo-Cl

can reach pureB performance

• Can improve PureB

performance

B-modes preserved E-mode leakage suppressed

Simulation-based deprojection of E modes can further improve E/B separation

# of deprojected E-mode realizations Standard estimator noise PureB estimator noise Plots from Kirit Karkare Ell ~ 100 power

Direction of longer-term future analysis is matrix-based

• Npix < 105
• Plan to release maps, reobserving matrix,

covariance matrices (signal, noise), beam profiles, and bandpass

• Enables joint analysis, optimal E/B separation
• Reobserving matrix calculated for BICEP2

Figure from Jamie Tolan

Characterize Beams with Artificial Sources

• Mast and flat mirrors

allow far-field measurement

• Main beam shape

beyond circularly symmetric Gaussian

• Sidelobes
• Polarization angles and

efficiencies

Most important beam effect so far is differential pointing within a pair

Difference beam of A and B polarizations A-B offset for all detectors (x20) Figure: Randol Aikin

Strategy for differential pointing

• Measured for each detector
• Deproject in analysis
• Simulate residual systematic error after

deprojection

• Understand in lab measurements
• Make improved detectors—we are upgrading

Keck Array with improved antennas

Keck Array Status

• Maps and sensitivity: 20 uK*sqrt(s) NET

[2011] → 11 [2012] → 9.5 [2013]

• Improvements for 2013 season
• Beam investigation example: sidelobes and

forebaffle loading

• Future plan

Keck Array Status

• Maps and sensitivity: 20 uK*sqrt(s) NET

[2011] → 11 [2012] → 9.5 [2013]

• Improvements for 2013 season
• Beam investigation example: sidelobes and

forebaffle loading

• Future plan

Keck Array 1-year 150-GHz Q/U sensitivity: 170 nK-deg. for effective area 397 deg.2

Keck 2012 maps are as deep as 2 years of BICEP2

Maps from Sarah Kernasovskiy

We're improving Keck every year

• 2011: 3 receivers at 150

GHz (2 with HWP)

• 2012: 5 receivers at 150

GHz (no HWP)

• 2013: Replaced 2.25

focal planes to improve sensitivity and A-B mismatch

Improvement of A-B mismatch Figure: Chin Lin Wong BICEP2 Keck 2013

Figure from Roger O'Brient

Dramatically improved A/B pointing mismatch

Keck is losing significant sensitivity to forebaffle loading

• Forebaffle on/off test found 0.5 pW (4 K)

change in load power

• Possible 8% NET improvement (confirmed

with reflective forebaffle measurement)

Figure from Sarah Kernasovskiy Change in loading (pW) In-lab Forebaffle simulator

Keck far-sidelobe beam mapping found something related

• Large angle beam maps

with forebaffles off

• Arc/ring sidelobe features

found 20~30 deg. from main beam in many detectors

• Believed to be due to

reflections from incompletely blackened telescope walls

• Terminate at warm baffle

→ excess loading

Beam map without forebaffles

Sidelobe Treatment Plan

• In-lab verification of cause
• Improve telescopes at Pole

for next season (probably with additional baffles at 4 K)

Figure from Samuel Harrison

Keck Array is going to get even better

• Observation funded through 2016
• Upgrade up to 3 focal planes for 2014 season
• Considering switching some to 100/220 GHz

– Mature 100-GHz design proven by SPIDER – 220-GHz in development

Summary and Conclusions

• BICEP1 3-year data release soon with

improved analysis tools

• BICEP2 analysis and systematic

characterization well underway

• Keck Array is more sensitive every year and

will soon get multi-frequency coverage

• We will reach the r = 0.01 level (and maybe

the first B-mode detection?!) soon

• Continuation of this program: BICEP3 talk

later this conference

Extra Slides

Planck Results Make Inflation B Mode Searches Even More Important

“The situation suggests a new view of future data opportuni-

• ties. At present, the three new problems arise because of con-

flicts between prediction and observation at the 2−3σ level. Fu- ture data can amplify or diffuse them. Detecting tensor modes and pushing the limits on non-Gaussianity further downward would ease the problems. On the other hand, not detecting ten- sor modes or detecting non-Gaussianity would each represent yet another threat which, combined with the three problems identified in this paper, spell doom for the inflationary paradigm and encourage consideration of alternatives.”

• -Ijjas, Steinhardt, and Loeb (2013)

Jackknives are the First Check for Systematic Errors

• Detection power comparable to statistical error
• 8 standard jackknives (time divisions, scan

direction, channel divisions, pointing-based)

• Check X^2 and sum of bandpower deviations