The Supernova Legacy Survey (SNLS)

Transcription

1 The Supernova Legacy Survey (SNLS) Pierre Astier LPNHE IN2P3/CNRS Universités Paris VI&VII November 8, 2006

2 Supernovae and acceleration Distances to type Ia supernovae strongly favor a recent accelerated expansion of the Universe flat universe Knop et al 2003 d2a/dt2 = 0 today Riess et al 2004

3 Acceleration? We need a recent acceleration, but only recent. The expansion was decelerating in the past: Λ does the job. SNe Ia, Riess et al, 2004

4 Supernovae Ia Thermonuclear explosions of stars which appear to be reproducible Very luminous Can be identified Transient (rise ~ 20 days) Scarce (~1 /galaxy/millenium) Fluctuations of the peak luminosity : 40 % Can be improved to ~14 %

5 Hubble Diagram Magnitude (~2.5*Log(flux)+Constant) vs z Nearby sample Distant sample (Perlmutter et al 1999)

6 Measuring distances to SNe Ia peak flux Distant sample Nearby sample Sne Ia are observed to exhibit reproducible peak luminosities Dispersion ~ 40 % caused by luminosity variations. > Have to use luminosity indicators: decline rate (or light curve width) > fair time sampling of light curves color (i.e. ratio of fluxes in different bands) > measurement in several bands

7 Distances and cosmological parameters source r observer r(z) =(comobile) distance to a source at a redshift z. Source and observer are themselves comobile Messenger : light > ds = 0. With the Friedmann eq., How to measure cosmological distances? luminosity distance dl = (1+z) r(z) > observed flux of an object of known (or reproducible) luminosity angular distance da = r(z)/(1+z) > angle that sustains a known length Correlations of CMB anisotropies. Correlations of galaxies. Heidelberg 8/11/06

8 Degeneracies from distance data defines r(z) Matter Dark Energy E.O.S The expansion history depends on the sum of 3 terms. The equation of state enters in only one of them. > exact or quasi degeneracies 1) need to know Ωk (from C.M.B) 2) if w(z) is arbitrary, the expansion history (via r(z)) constrains a relation between ΩM and w(z), not both of them independently. 3) even assuming a constant w, there remain a strong (although not exact) degeneracy. > distance data alone does not fix unambiguously the E.O.S Curvature

9 Observing Dark Energy(!) Dark energy plays an important role in the recent universe (z <~1). Its effect decreases (vanishes?) with increasing z. Particularly sensitive methods (for z < ~1): Supernovae Ia Optical (and IR) telescopes, imaging and spectroscopy Figure of merit : number of SNe, z span Weak gravitational shear Optical telescopes, imaging Figure of merit : surveyed area on the sky (up to z~1) Baryon Acoustic Oscillations Optical telescopes, imaging and spectroscopy. Figure of merit : surveyed universe volume measures combinations of r(z) r(z) r(zlens,zsource) P(k;z) r(z), H(z) Ωmh2 (via zeq and csound)

10 Supernovae : some History Early days: Riess 1998 (10(+6) SNe Ia) Perlmutter 1999 (42 SNe Ia) => Acceleration of expansion Power spectrum of galaxies (2dF, SDSS) > ΩM Spergel 2003/2006, CMB (WMAP) > ΩT => Concordance model : Flat (+/ 0.02) universe dominated today by a Dark Energy compatible with a cosmological constant, ΩΛ~0.7 Later on: Sullivan 2002 Hubble diagrams by galaxy types (non evolution test) Tonry 2003 (+8) Barris 2003 (+23 SNe Ia, z<1) Knop 2003 (+11 SNe Ia measured with HST) Riess 2004: (+16 SNe Ia found and discovered with HST) up to z~1.6

11 A varying cosmological constant? (Λ or something else?) (ΩΜ, ΩΛ ) > (ΩΜ, Ωx First order expansion:,w) > (ΩΜ, 1 ΩΜ, w) Flatness constraint ρx ~ (1+z) 3(1+w) ΩΜ+ΩX=1 SCP, HiZ (1998)

12 More recent constraints on w SNe + flat universe + Large scale structures + CMB (Knop et al 2003) w= (stat) +/ 0.09 (sys) Similar constraint in Riess et al (2004) : w = (Knop et al 2003)

13 SNe Ia surveys: from workshops to factories Old observing way is a many step process: search: imaging at two epochs, ~3 weeks apart spectroscopy of candidates found Photometry of identified Ia's Drawbacks: Extremely vulnerable to bad weather poor yield of observations Many telescopes involved proposals/scheduling issues Photometric cross calibration issues Rolling search mode: Repeated imaging of the same fields Spectroscopy near peak Built in photometric follow up Bonuses: Mutiplex: many measurements/exposure Detection on a time sequence LC sampling independent of phase Imaging robust to bad weather Spectroscopy in service mode possible Only one imaging telescope to calibrate Deep stack at the end of the survey... Drawback: Imaging instrument failures...

14 SNe Ia surveys: from workshops to factories (2) Rolling search is THE way to go for SNe surveys Three ongoing projects: ~8 deg, RI bands, 0.2<z<0.8, 5 years from SNLS@CFHT (within the CFHTLS) 4 deg2, griz bands, 0.2<z<1, 5 years from SNe in SDSS II 300 deg2, ugriz bands, z<~0.35, 3 years from fall Rolling searches become increasingly difficult as z decreases Requires very wide field imaging ~10 deg2 Large area > Large data volume. Many ground based wide field imaging projects are in the landscape: Pan Starss, DEC (@CTIO), LSST,... SCP

15 MegaCam at CFHT MegaCam: 36 CCDs 2k x 4.5k pixels 1 pixel = 0.185'' field of view : 1 deg2 1rst light at end of CFHT: diametre 3.6m Mauna Kea, Hawaii 4200 m <seeing> = 0.8''

16 The CFHT Legacy Survey CFHTLS: observation program using MegaCam over 5 ans, ~ 9 nights/lunation (i.e. ~ 500 nights) defined by French and Canadian communities started June Components: Very Wide (asteroids) 22 % Wide (weak shear) 33 % SN/Deep (Deeper and wider Hubble Deep Field) 45 % Deep/SN: 4 (one deg2) Megacam pointings spread on the sky. observed ~ 5 time a lunation in 4 bands g,r,i,z (400 > 1000 nm) > astro cinema : study of all variable objects. For SNe, observations combine detection and photometric followup

17 French Canadian Others Collaboration to discover, identify and measure SNe Ia in the CFHTLS(DEEP). About 40 persons. Spectroscopy : ~ 250 h/year on 8m class (!!) VLT (Europe): large programme 240 h over 2 ans. renewed. Gemini (US/UK/Can) : 60 h/semestre Keck (US): 30 h/year.

18 Supernova Legacy Survey: design Objectives: ~700 SNe Ia z<1.1 (~ 10 times previous studies) Observed in 4 visible bands (g,r,i,z) ~ SDSS Fair sampling of lightcurves (~ 5 points a month) Spectroscopic identification (of all 700 SNe) > 0.1 Justifications: High statistics also provide insight on systematics Photometry with a single telescope : beneficial for photometric calibration (thanks to repetition) helps when evaluating selection biases Multi band observations : mandatory to follow the same restframe spectral region at various z (B,V) z=0 <=> (g,r) z ~ 0.2 <=> (r,i) z ~ 0.5 <=> (i,z) z ~ 0.8 helps lightcurve modeling at z~0.4 with more than 2 bands, we have redundant measurements of distances

19 CFHTLS/Deep : Observing mode 40 nights/year for 5 years. Repeated observations every 4 night ( rolling search ), service mode 4 bands g,r,i,z 4 fields D1 XMM deep VIMOS SWIRE GALEX > Photometric data before objects are detected > Multiplexing : several SNe per field in a single exposure > Repeated calibration of field stars D2 Cosmos/ACS VIMOS SIRTF XMM D3 Groth strip Deep2 ACS D4 XMM deep

20 Detecting Supernovae New images (of the previous night) are subtracted off a reference image of the field (e.g. a stack of last year images) before subtraction one has to align images: geometrically photometrically PSF (bring to the same star shape). Detection of (positive) excesses (typically above 3 σ ) > Association of detections over nights/bands to reach ~ 8σ > Lightcurves are fit to a SNe Ia template to evaluate a Ia likelihood > Spectroscopy. Two pipelines search independently for cross checks

21 Spectroscopy Identification of SNe Ia Redshift (usually of the host galaxy) Detailed studies of a (small) sample of SNe Ia/II Telescopes VLT Large program (service) 240h in , idem Gemini : 60h/semestre (Howell 2005, astro ph/ ) Keck : 30h/y (spring semester)

22 Analysis for cosmology of the SNLS first year data sample August 2003 July 2004 Differential photometry Photometric calibration Fitting lightcurves Fitting cosmology Systematics 10' SNLS-03D4ag in the D4 Field

23 Differential photometry Fit galaxy(i,j) pixels on a stamp Fit a residual sky background on each image Fit a common position of SN to all images Robustify for bad/dead pixels. data model residuals weights Uses up to 500 images geometrically aligned One SN flux measurement per exposure Kernel

24 Example We process independently the 4 bands z=0.35 Arbitrary units g i r z By product: averaged galaxy images

25 Photometric calibration Relies on repeated observations of Landolt standard stars. Calibration in Landolt (Vega) magnitudes because nearby SNe are calibrated this way Produces calibrated star catalogs in the CFHTLS Deep fields, in natural Megacam magnitudes. Comparison of synthetic and observed color terms (Megacam/Landolt & Megacam SDSS 2.5m) g r i z Landolt stars spectrophotometric synth. mags Zero 0.01 (0.03 in z) Repeatability better than 0.01 (0.015 in z)

26 Known relations between Sne photometry observables Brighter slower Brighter bluer stretch: time scale parameter of the (B) lightcurve, corrected for (1+z) or decline rate: decrease of flux at 15 (RF) days from max + Correlation between U B color and stretch (at fixed B V) => enables to reduce brightness scatter to ~13 % (0.13 mag) color B V (rest frame) at peak.

27 Fitting Lightcurves Spectral Adaptive Lightcurve Template Model of the SNe Ia Spectral Energy Density as a function of: phase (date w.r.t the B maximum) lambda (SN restframe wavelength) stretch (dilatation factor of the time axis in B) color =E(B V) at the B maximum Guy et al, astro ph/ ,a&a Tuned on a set of very nearby Sne (not necessarily in the Hubble flow) U B V R SNe Lightcurves for several values of stretch between 0.7 and 1.1 Cardelli Extinction law A A(B) for E(B V)=0.1

28 Examples SNLS 04D3fk z=0.358 All bands can be fitted, because they lie within U to R restframe. g, r, and i roughly match UBV restframe. Four parameters: tmax,m*b, s, c SNLS 04D3gz z=0.91 g' and r' bands are too blue in the restframe to be fitted (no model!). r and i roughly correspond to restframe B and V bands

29 First year SNSL sample (July 2003 > July 2004) 142 acquired spectras : 20 Type II SNe 9 AGN/QSO 4 SN Ib/c 91 SNe Ia -10 were missing images without SN when we froze the sample - 6 only have one band (lost). 75 usable events 2 events don't match a SNe Ia lightcurve 2 events are strong outliers to the Hubble Diagram

30 Hubble Diagramme Magnitude in a restframe filter (sliding for observer): Cosmological fit: Measurement uncertainty. Nearby Sne from the literature Cosmological parameters. Intrinsic Dispersion (ajusted so that chi2 = ndof). Depends on the quality of the distance estimator. where Correlations of luminosity with stretch and color. Parameters M, α et β fitted together with cosmology and marginalized over.

31 Confidence Contours w= 1 ΩT=1 < Λ BAO: Baryon Acoustic Oscillations (Eisenstein et al 2005, SDSS) 68.3, 95.5 et 99.7% CL

32 Identified Sources of systematic uncertainty photometry, photometric calibration, modeling of the bandpasses Evolution: of the composition of the progenitor (white dwarf) of the environment. > study of properties as a function of host galaxy > comparisons of distant and nearby SNe Empirical modeling of lightcurves : Restframe region used : fit (B,V) > (U,B) at large redshift > tests on intermediate z SNe Selection biases: When close to the sensitivity limit, only the brightest objects are measured. => bluest and/or slowest SNe > simulation of the detection pipeline contamination: SN II, SN Ib/c are sometimes difficult to reject > cross check spectral identifications with photometry extinction by intergalactic dust: > Quasar colors as a function of redshift. gravitational lensing : broadens peak magnitude distribution (but preserves the average)

33 Photometric calibration uncertainties Use of color color relations offset : calibration itself slope : sensitive to difference of filter sets Landolt stars spectrophotometric synth. mags Comparison of measured relations with synthetic one (from the assumed instrument bandpasses) tests the central wavelengths of the simulated instruments. Offsets : conservatively defined to 0.01 (0.03 in z) Slopes : uncertainties translate to ~1.5 nm in central wavelengths.

34 From magnitudes to fluxes SNe cosmology requires ratio of fluxes measured in different observer spectral bands Magnitudes provide ratio of fluxes measured in the same band. Hence magnitudes have to be converted into fluxes which requires the spectrum of standard stars. Vega spectrum known to ~ 1% (Hayes 1985, Bohlin 2004) Vega: HST/Ground (Bohlin 2004) SNe cosmology forecasts usually assume ~1 % systematic uncertainty of relative (distant/nearby = red/blue) flux scales. This is realistic but may become pessimistic. Could we calibrate instruments against lab standards rather than sky standards? Essence has such a project underway (@CTIO) SNLS is in the feasibility study phase.

35 3 bands: compatibility of colors (UBV at rest) U3 = U(measured) U(guessed from B and V) z 0 U B V 0.3 g' r' i' 0.7 r' i' z' Nearby SNLS σ = UBV relations of SNe Ia seem reproductible Redundancy of distances (SNLS première) Robustness and small uncertainty of SALT k corrections Color fluctuations are a small source of uncertainties.

36 Correlations : distant (z<0.8) vs nearby SNe Brighter slower Blue: nearby SNe Black: SNLS SNe Brighter bluer Stretch, color and relations with luminosity are essentially compatible between nearby and distant events.

37 LC rise time (nearby vs distant) Compare distant and nearby early LC shape (in B) Nearby : / 0.2 Distant (SNLS) : / 0.2 (stat) +/ 0.2 (syst) (Conley et al, 2006)

38 Selection (Malmquist) bias from simulated SNe - Generate Mock SNe from observed stretch and color distributions, and constant rate - Account for correlations with luminosity and tune limiting mag to reproduce observed peak i'(ab) and z distributions. black: SNLS data red: Simulations Biais on the distance modulus : z=0.8 z=1. Impact on Ωm (flat ΛCDM): nearby SNe ± SNLS SNe ± 0.010

39 Grey dust Colors of SDSS quasars (as a function of z) provide a limit on the dust column density as a function of Rv δ(ωm)<0.025, δ(w)<0.048 (for Rv<12, Ostman 2005) Gravitational lensing. Weak lensing : asymetric distribution of residuals, but conserves the average flux. Strong lensing : net loss of light when multiple images Stringent limits on multiple images from the CLASS survey (Myers et al 2003) δ(ωm)<0.005, δ(w)<0.01 Contamination from other SNe types (Ib/c mainly) ~0.5 SNIb/c expected within the sample

40 Systematics summary Photometric calibration dominates => Improvement expected from the observation of secondary standards with Megacam + observation of field stars of nearby SNe Uncertainty on the selection bias: will decrease for SNLS => Will soon be limited by the uncertainty of selection bias of nearby SNe

41 SNLS: cosmological results Flat ΛCDM cosmology, Flat ΩM,w cosmology: Combined with Baryon Acoustic Oscillations(Eisenstein et al, SDSS, 2005): Confirms acceleration of expansion with 71 new distant SNe Ia. Results fully compatible with Λ Distance estimator accounts for color. Systematics can still improve (especially photometric calibration) Same result obtained by the Canadian teams using independent software

42 Supernovae and WMAP3 WMAP & SNLS (flat universe) w (cst) = WMAP + 2dfGRS + SDSS + SNLS+GOODS: Ωk and w Spergel et al (2006) (double counting of nearby SNe?)

43 SNe+BAO:Short term forecasts for (constant) w (SNLS Collab., 2005) Expected realistic statistical improvements of the (ΩM,w) constraints. Nearby SNe Distant SNe with current σ(ωm) SNfactory SDSS SNe SNLS SNe inf BAO accuracy σ(w0) BAO x 2 σ(ωm) (4000 >8000 deg2) σ(w0) 2008?

44 Current work/short term plan Improve photometric calibration improve our knowledge of instrument non uniformity transfer primary or secondary standards to the SNe fields cross calibrate SDSS and SNLS SNe fields. lab calibration (to get at least the instrument response functions) Improve distance estimator build a stretch dependent spectral template. extend wavelength coverage below 300 nm. Improve statistics underway... between bad winters and earth quakes. Improve cosmology? heavily depends on an improved nearby sample.

45 Calibration: camera (non)uniformity Improved measurement of reponse non uniformity using large dithers on dense stellar fields. intensity non uniformity color non uniformity one CCD dec RA Precision well below 1% (N. Regnault et al, in prep)

46 SNLS status by Sept Public list of candidates: => ~300 SNe Ia/Ia?

47 SNLS ~2.5 years Hubble Diagram Current plan: process 3 years of data improved calibration minor changes in algorithms use SALT2 for LC fits get galaxy types (from colors)... y r a n i m i l e pr

48 Supernova empirical modeling The SALT2 approach Concept : parametrize the SNe Ia spectra sequence using photometric and spectroscopic available data use PCA to describe SNe variability no a priori parameters, rediscover the stretch derive model uncertainty. dive in the far UV. z=0.91 J.Guy & the SNLS, in prep.

49 LC fitting : from SALT to SALT2 J. Guy et al. (in prep) Same cosmo, σ(w) reduced by ~ 15%

50 Summary / conclusions SNLS first year data set yields ~ 0.1 (or slightly better) Systematics : ~ Likely to go down. The collaboration is working to publish a three year sample. Improvements: Calibration. Distance estimator. Host galaxy typing.... Nearby sample has become too small and deserves a specific effort SDSS (SNe) is likely to become a key player. w 0.05 not unlikely by the end of the survey (2008)