Weak Lensing Cluster Survey in CFHTLS Wide Fields

1. Weak Lensing Cluster Survey in CFHTLS Wide Fields HuanYuan Shan Work done in Marseilles Collaborators: J.P. Kneib, C. Tao, O. Ilbert, M. Limousin, K. Thanjavur, Z.H. Fan…

2. Outline • Backgrounds • Shear measurement • Calibration of shear measurement pipeline • Mass reconstruction and candidate clusters • Optical/X-ray Counterparts • Lens tomography • Conclusions

3. Conclusions 1. We develop a new method to do shear measurement pipeline calibration 2. Considering the noise effects, we get peak counts in the CFHTLS fields with prediction from a LCDM Universe 3. The good agreement on the clusters redshift and velocity dispersion implies no evidence of selection bias compared to these other techniques 4. We also derive an empirical relation between the cluster mass and the galaxy velocity dispersion, which is in reasonable agreement with the prediction of N-body LCDM simulations

4. Galaxy Clusters Observe clusters: Optical, X-ray, SZ effects, Lensing Why gravitational lensing? • Independent of cluster mass model, purely geometrical method • Independent of the cluster physics • Independent of the early Universe physics

5. Weak Gravitational Lensing - Weak distortions caused by LSS of the universe: common/weak - Clusters of galaxies are important targets for weak-lensing studies Individual Statistical abundances mass distribution cosmological probes Clowe et al. 2006 Hamana et al. 2004

6. MegaCam: 36 CCDs (0-35) - 1 sq. degree CFHTLS-Wide: T0006 release - 171 sq.degree - multi-band: u', g', r', i', z' - Density: ng ~ 10-15 arcmin-2 - 22<mag<26 (i' band)

7. g = g1 + ig2 Shear Measurement • Ellipticity of galaxy • e = e(intrinsic) + 2 g • Estimate shear g by averaging over many galaxies • Cosmic shear: ~1% distortions

8. This is hard, why? • Galaxies are not circles or ellipses (random noise) • Galaxy orientations may align during formation • Telescope and atmosphere convolve image - point spread function (PSF) - spatially varying - time varying • Partial and patchy sky coverage • We do not have galaxy distances

9. Shear Measurement Pipelines Our shear measurement pipelines are based on KSBf90 • Objects detection • Star-Galaxy separation • Measure and model PSF • Correct galaxy shapes Shear Measurement Methods only differ in part 4.

10. KSB methods - The core requires the measurement of the quadrupole moments of each observed galaxy image weighted by a Gassian function. - The only currently widely used method

11. Objects detection hfindpeaks (cyan) vs. SExtractor (red)

13. Star selection - rhvs. mag & max vs. mag - SNR >100 & |e|<1 - Stars number vs. SNR Galaxy selection - r>1.05*rpsf - SNR>10 - 22< mag< 24.5 - |e|<1, d>3 arcsec

14. Measure and model PSF - KSB+ : Polynomial model Rational function (Van Waerbeke et al. 2005)

15. Left: Projection of the stellar ellipticities before and after PSF anisotropy correction. Right: PSF ellipticity variation with half light radius before and after PSF anisotropy correction.

16. Correct galaxy shapes KSB method: - Large noise exists on P - Fit P (mag, rh)

17. The corrected ellipticities of galaxies ~ 0 for all galaxy sizes, all magnitudes and all SNR. We bin and average the shape catalogue by size, magnitude and SNR

18. Calibration of Shear Measurement Pipelines - Simulations: STEP (Heymans et al. 2006; Massey et al. 2007) & Great08 (Bridle et al. 2009; 2010) Disadvantage: Shear and PSF do not vary across an image - Real Data: ACS-COSMOS vs. CFHTLS-Deep 2 Space-based vs. Ground-based, Cover the same sky area - Matched shear: P fit Without fit: large noise P (rh) vs. P (mag, rh)

19. - Real Data: r' band vs. i' band data

20. Mass Inversion - Fourier transform: with KS inversion (Kaiser & Squires 1993) Left: simulated convergence map (Vale & White 2003) Right: Shear map superimposed on the convergence map CFHT-Wide (W1): Gaussian smoothing (1 arcmin)

21. Random noise • Without intrinsic alignments, the noise correlation can be written as - Galaxy number density - Dispersion - Filter function • Wide and deep survey works on the first two • Optimal filter function

22. - Gaussian filter

23. Noise after Gaussian smoothing: Signal-to-noise ratio: The galaxy number density of CFHTLS-Wide: Much lower than the Deep fields: We choose SNR>3.5/4 peaks as candidate clusters

24. The probability distribution function (PDF) of the peak height. The symmetric bimodal: positive peaks and negative peaks (troughs).

25. The cumulative number of positive and negative peaks: Data vs. analytical model (Fan et al. 2010)

26. Optical/X-ray counterparts WL: 139/39 positive peaks with SNR>3.5/4 I. Compare with optical detection (Thanjavur et al. 2009) - BCG vs. WL peaks 75 candidate clusters with SNR>3.5 (efficiency ~ 50%) II. Compare with X-ray detection (Adami et al. 2010) - 6 sq. deg, XMM-LSS - 66 spectroscopically confirmed clusters (0.05 < z < 1.5) - 53 X-ray clusters are withinW1 field 22 counterparts The offset between X-ray and weak lensing center is much bigger than the offset between optical and weak lensing measurements.

28. The convergence SNR map for the w1+2+3 field

29. Lens tomography - SIS & NFW mass model: - Chi-square fitting analysis: with the error - Photometric redshift of source galaxies (Arnouts et al. 2010)

30. Lens tomography redshift vs. photometric redshift Left: WL & Optical counterparts Right: WL & X-ray counterparts

31. Matched velocity dispersion: X-ray vs. WL method

32. Scaling relations between velocity dispersion and weak lensing mass. Left:  vs. Mvir Right:  vs. M200 (N-body simulation)

33. Conclusions 1. We develop a new method to do shear measurement pipeline calibration 2. Considering the noise effects, we get peak counts in the CFHTLS fields with prediction from a LCDM Universe 3. The good agreement on the clusters redshift and velocity dispersion implies no evidence of selection bias compared to these other techniques 4. We also derive an empirical relation between the cluster mass and the galaxy velocity dispersion, which is in reasonable agreement with the prediction of N-body LCDM simulations