1. High-ℓ CMB and the CBI Jonathan Sievers (CITA/UToronto) +CBI Collaboration

2. CMB at High-ℓ  Primary CMB fluctuations low past ℓ~2000  Signal expected to arise from secondary anisotropies – SZ galaxy clusters, point sources…  High power level seen in CBI (30 GHz, ℓ~2000), BIMA (30 GHz, ℓ~6000). ACBAR (150 GHz, ℓ~2000) sees level consistent with CBI if due to SZ (but also consistent with nothing?)  Observed level much(?) larger than expected, especially in light of WMAP3 low σ 8.  SZ clusters expected to be dominant (non-point source) component, but models suggest level is very sensitive to σ 8 (seventh power), preferably ~0.9-1, unlike 0.75 in WMAP3. ℓ? We don’t know. Might even be interesting. What is the dominant source of signal at high- ℓ? We don’t know. Might even be interesting.

3. Current Best CBI Spectrum (vs. Old) Bottom – old best TT spectrum. Top – current best spectrum. For most of ell range, errors down by ~40%. CBI in compact config. in polarization, so little high-ell data. Highest bin only ~10% smaller. NB: two binnings shown. Red/blue points *not* independent. Top panel – new spectrum! Will be published soon. Data on which current excess results based.

4. Current CBI+BIMA PS Fit CMB+Excess model to CBI data (using raw data, not a fit to the power spectrum). Red curve SPH simulation-based template (Bond et al.), blue curve analytic (Komatsu&Seljak, Spergel et al.). Red points latest CBI w/ finer binning. Black points latest BIMA. ACBAR green. Models extrapolated to BIMA points – not a fit. Differences between analytic/simulation templates of factor of 2, implies σ 8 model uncertainty of 0.08- 0.09. If CBI excess were due to unexpected source population, BIMA would see them. They don’t.

5. Actual σ 8 From Chains Do full parameter analysis with all CMB (including latest CBI, WMAP3). Inferred σ 8 for Bond et al. template is 1.00±0.1, for Komatsu & Seljak is 0.93±0.1. Komatsu & Seljak consistent with latest Chandra M-T relation (Vikhlinin et al.). Latest XMM M-T (Arnaud et al.) 50% higher than Chandra – σ 8 ~0.85? Subha Majumdar working on prediction. Errors assume Gaussian noise in PS only. Doesn’t include errors from non-Gaussianity of clusters, uncertainty in faint source counts (~35% increase) M-T SCATTER??

6. Sources w/ GBT Observing NVSS sources in CBI fields in 2-prong strategy. First: veto survey to measure all sources to see which ones matter at 30 GHz (large majority don’t). Second: deep obs. of faint NVSS sources to measure Ka counts. Are startbursts synchotron or Free-Free at 30? Requires GBT receiver to be stable for ~minute. White = NVSS sources in CBI fields. Green = observed by GBT as part of veto survey. Increases by 50-100% CBI data. Uncertain faint source flux at 30 GHz important source of uncertainty. When done with GBT observations, will nearly double significance of current CBI excess.

7. Very, Very Preliminary Optical-CMB If excess due to SZ, clusters in optical should correspond to holes in CBI maps. Have CFHT of CBI deep fields. Don’t have masses yet from RCS guys – only significance. Very noisy. Want mass to ~50%. (NB, only 2 clusters in highest bin, 5 in second- highest) Field galaxies should show positive correlation (some are radio sources). Explains low-significance positive correlation? Bin up clusters, measure CMB power vs. cluster “size”. SZ should be negative. Note - sources in low-significance clusters already corrected for statistically in spectrum.

8. This month: CBI2!  Upgrading CBI for better spectrum.  0.9m --> 1.4m dishes  Ground Shield (eventually)  Measure the excess  Much better SZ sensitivity.

9. CBI2 Forecast – 1 Year on CMB Black = CMB Red= CMB+current excess Magenta=CBI now Blue=CBI2 Green=thermal noise in CBI2 Slightly more pessimistic forecast gives 8% error on current excess. Should be able to get GBT follow-up observations.

10. Summary  Combined CBI dataset gives much better CMB power spectrum, modest improvement at highest ell.  CBI detects excess power at ℓ>2000 at 3σ. Will go to ~5σ (assuming level doesn’t change) with GBT data (most of which is in hand).  Working on CMB/Optical correlation, results should come soon.  CBI2 will do much better on high-ℓ excess. Important even with other expts. due to l-range, frequency. (new ACBAR will do same ℓ range at 150, SZA will do somewhat higher- ℓ.).  Observations w/ CBI2 start this month!

11. CBI (+ACBAR?+BIMA) Excess  SZ clusters should contribute to the CMB power spectrum in a frequency-dependent way.  Signal level very sensitive to σ 8 – roughly σ 8 7 Ω M 2.  CBI currently detects excess at 4.1σ (vs. primary CMB) in overall level, fold in uncertainty due to faint point source contributions, goes to 2.9σ.  Currently observing radio point sources with GBT. Gets us more data, plus better knowledge of faint sources (currently 50% uncertainty).  CBI excess wants σ 8 ~1.  BIMA also detects power at a level consistent with CBI if SZ. ACBAR has suggestion of detection, consistent with CBI+BIMA (150 GHz gets ¼ power of 30 GHz) - but also pure CMB.

12. How to Proceed?  If excess is from clusters, should be optical-radio correlation. Signal clear since clusters are negative in radio, unlike everything else in the sky. We have obtained CFHT images of CBI deep fields, doing correlation now.  Better source observations. 30 GHz faint source counts uncertain at ~50%. Leads to 25% uncertainty in excess level. GBT 30 GHz system working – we are observing faint sources in NVSS to nail down 1.4-30 spectral index distribution.  Look at clusters in more detail. CCCP (PI Henk Hoekstra) a program to do weak lensing of ~50 X-Ray bright clusters. CBI will get SZ of ~20 – will help with prediction of SZ amplitude.  And, of course, BETTER DATA! CBI being upgraded – 1.4m dishes being installed as we speak. CBI2 will be 5-20 times as efficient at excess/cluster observations as original CBI. Should have much better spectrum in a year.

13. Weak Lensing vs. SZ/X-Ray Masses Simple CBI2 forecast. Assume 130 uK error (differencing – worst case) or 80 uK (no differencing) errors. Fold in errors on X- Ray temp., see what errors on mass come out. Does not include shape uncertainties, point sources… Assumes same shape for everybody. (~10 total nights observing time on CBI2) (noise added vertically, but not horizontally)

14. Some Clusters Already Have observed ~15 z<0.1 clusters with CBI, of order 5 of CCCP sample. Image of 1 night on A2163. Fit isothermal-β, fixing shape parameters from X- Ray. Get same model SZ decrement (10%) as OVRO, which is sensitive to ~10 times smaller scales. So for this case, isothermal-β a pretty good fit.

15. CBI 2000+2001, WMAP, ACBAR, BIMA Readhead et al. ApJ, 609, 498 (2004) SZESecondary CMBPrimary +Boom03; Acbar05: very nice TT, Oct05. parameters & new excess analysis as SZ  8 7.82 +-.11 =.85 +-.05 CMBall+LSS  8 Kuo et al. (2005, in prep)

16. What I Still Did Last Night Split vs. redshift. SZ stronger at higher z? Maybe just a selection effect. Need to get better estimates of cluster masses…

17. What I Did This Morning New spectrum with GBT non-detections subtracted (not projected), noise term added. Errors improve by ~15%. Should get another 10- 15% better when all of the sources are measured. Absolutely critical: measurement of faint (~1 mJy) 1.4 GHz sources. They are a different population, may have different 30 GHz properties. Uncertainty in excess evenly split between spectrum/source terms.

18. Can Tune Obs. Plan after 1 Month Optimal observing strategy depends on signal level. After 1-month, should have a much better idea of level Will be able tune obs. plan during the run, which is nice.

19. The Cosmic Background Imager  13 90-cm Cassegrain antennas 78 baselines78 baselines  6-meter platform Baselines 1m – 5.51mBaselines 1m – 5.51m  10 1 GHz channels 26-36 GHz HEMT amplifiers (NRAO)HEMT amplifiers (NRAO) Cryogenic 6K, Tsys 25 KCryogenic 6K, Tsys 25 K  Single polarization (R or L) Polarizers from U. ChicagoPolarizers from U. Chicago  Analog correlators 780 complex correlators780 complex correlators  Field-of-view 44 arcmin Image noise 4 mJy/bm 900sImage noise 4 mJy/bm 900s  Resolution 4.5 – 10 arcmin ℓ) (Currently strongest detection of high-ℓ power)