Making Cosmologyʼs Best Standard Candles Even Better The Nearby Supernova Factory and Spectrophotometric Observations of SNe Ia Stephen Bailey LPNHE, Paris for the Nearby Supernova Factory Deciphering the Universe through Spectroscopy Potsdam 22 September 2009 G. Aldering 2, P. Antilogus 1, C. Aragon 2, S.B. 1, C. Baltay 3, S. Bongard 1, C. Buton 4, M. Childress 2, N. Chotard 4, Y. Copin 4, D. Fouchez 6, E. Gangler 4, M. Kowalski 7, S. Loken 2, P. Nugent 2, K. Paesch 7, R. Pain 1, E. Pecontal 5, R. Pereira 4, S. Perlmutter 2, D. Rabinowitz 3, G. Rigaudier 5, P. Ripoche 1, K. Runge 2, R. Scalzo 3, G. Smadja 4, H. Swift 2, C. Tao 6, R.C. Thomas 2, C. Wu 1, J. Zylberberg 2 1 LPNHE (Paris), 2 LBL (Berkeley), 3 Yale (New Haven), 4 IPNL (Lyon), 5 CRAL (Lyon), 6 CPPM (Marsaille), 4 Universität Bonn
Overview SNe Ia were original method to discover dark energy Power comes from the ability to standardize their luminosities Better standardization = Better cosmology constraints This talk: new ways to standardize them with more accuracy Outline Background Cosmology measurements with SNe Ia The Nearby Supernova Factory Classic methods to standardize SNe Ia Spectral flux ratios Other spectral metrics Conclusions 2
Cosmology with Luminosity Distance dl(z) is a function of the cosmology Flux = L 4πd 2 L 5 log 10 dl + const MLCS2k2 fitted distance modulus (mag) 46 44 42 40 38 36 34 33 nearby (JRK07) 103 SDSS-II (this paper) 56 ESSENCE (WV07) 62 SNLS (Astier06) 34 HST (Riess07) Kessler et al (SDSS) 2009 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 redshift (actually a fit to a different Hubble Diagram dataset, but a prettier plot) 3
Type Ia Supernovae What you need to measure luminosity distance dl: Very bright objects Uniformly bright objects Type Ia Supernova Thermonuclear runaway when white-dwarf accretes to Chandrasekhar mass Uniform starting conditions = (almost) uniform luminosity Can be as bright as host 4
Standardizable Candles Variations in peak magnitude correlate with other observables: Color c: bluer = brighter Shape x1: broader = brighter Correct observed peak mag using x1, c mb mb + αx1 - βc Color: E(B-V) color mb: Peak magnitude in restframe B-band Shape x1: Lightcurve shape/width parameter 5
Standardizable Candles Broader = Brighter Correct for stretch Peaks agree within ~15% S. Perlmutter, Physics Today, April 2003, p. 54 6
SN Cosmology Fitting Peak mag shape & color m B M + αx 1 βc 5 log 10 [d L (z; θ)] Normalization Standardization factors Luminosity distance depends on cosmology parameters θ: Ωm, ΩΛ, w,... Best fit with dark energy No dark energy Hicken et al 2009 7
SN Cosmology Fitting Hicken et al 2009 Best fit with dark energy No dark energy 1 2 Cosmology constraints come from comparison of nearby and distant supernovae Why weʼre the Nearby Supernova Factory: currently low-z sample is limiting factor Better standardization = Better cosmology constraints The focus of this talk 8
Nearby Supernova Factory 1. Discover Palomar Nightly NERSC Search ended Sept 2008; >1000 SNe discovered in 28 months of searching 185 followed in detail, 0.02 < z < 0.09 Ref New New-Ref = ~10-7 of the area observed per night 3. Analyze 2. Observe SNIFS UH 2.2-m Every 2-3 nights Custom, unique spectrometer designed for nearby SN obs 9
SuperNova Integral Field Spectrometer (SNIFS) Photometric Channel Microlens array to two channel spectrograph 15x15 = 225 spectra R channel: Galaxy + Sky Extinction monitoring, calibration Acquisition, Guiding SN + Galaxy + Sky Pick-off Prism at SN loc Sky 9.4ʼ x 9.4ʼ FOV; 0.14 /pix Every obs: flux calibrated spectra, 320 520, 510 1000 nm coverage 6 x 6 FOV; 0.4 /spaxel On UH 2.2m on Mauna Kea; SNfactory uses every 2-3 nights for ~9 months/year Hard work... SN 10 12
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el Slides: Rui Pereira 11
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el Slides: Rui Pereira 11
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el Slides: Rui Pereira 11
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el Slides: Rui Pereira 11
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el Slides: Rui Pereira 11
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el Slides: Rui Pereira 11
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el Slides: Rui Pereira 11
Spectrophotometry From Spectra to Lightcurves synthetic photometry of SN2005el One spectrum per point / night Synthesizable in any filter Lightcurves + spectral features Slides: Rui Pereira 11
Motivations for Spectrophotometry S-corrections K-corrections Relative Flux (F! ) Kowalski et al 2008 3000 4000 5000 6000 7000 8000 9000 10000 Wavelength (Å) Info for models etc. vs 12
Two Classic Corrections Classic corrections Δμorig ½=0.85-0.2 0.0 0.2 0.4 0.6 color Color: Bluer = Brighter Lightcurve shape: Broader = Brighter ~40% ~16 20% scatter Can we do better with spectral info? Search correlations of features with residuals ½= 0.68 Δμorig - βc 40% 16% -4-3 -2-1 0 1 2 3 x1 13
Previous Spectral Metrics Absorption Ratios e.g. RSi 1 4 Pseudo-Equivalent Widths / fractional absorption area e.g. EW(SiII 6355) Flux Ratios e.g. RSiS 2 Feature Velocities e.g. vsi 3 14
Generalized Flux Ratios Spectra sorted by SALT color Normalized flux + offset 25 20 15 10 5 0 SNF20080514-002 SN2007bd SNF20070727-016 SNF20080516-022 SNF20070806-026 SNF20070712-003 SNF20070531-011 SNF20080512-010 SNF20080510-001 SNF20080626-002 SNF20070630-006 SNF20080822-005 SNF20070810-004 SN2005hc SNF20070912-000 SN2006dm SNF20070424-003 SNF20080522-011 SNF20060908-004 SN2007kk SNF20070717-003 SNF20070803-005 SNF20061021-003 SNF20070818-001 SNF20061111-002 SNF20071021-000 SNF20080803-000 SNF20080714-008 SNF20071015-000 4000 5000 6000 7000 8000 Wavelength [A] Consider all flux ratio combos, not just ratios of known peaks Search for correlations with uncorrected Hubble residuals SNfactory spectra Flux calibrated to standard stars Smooth Hubble flow minimal peculiar velocity or cosmo uncertainties Within ±2.5 days of peak brightness Training and Validation Datasets Search with training set (28 SNe) Cross check w/ validation set (30 SNe) Minimizes bias and confirms results 15
Training Set ρ = 0.94 Flux Ratio Correlations Bailey et al 2009, A&A Letters, arxiv 0905.0340 Lower diagonal: Decolor spectra before forming ratios Statistically Significant Develop method and pick ratios based upon training sample Then look at validation sample 16
Training Set ρ = 0.94 Flux Ratio Correlations Bailey et al 2009, A&A Letters, arxiv 0905.0340 Validation Set ρ = 0.96 Lower diagonal: Decolor spectra before forming ratios Statistically Significant Correlations Develop method and pick ratios Stronger based upon than training color or sample stretch Selected only from training sample Then look at validation sample Confirmed by validation sample 16
Nearby Hubble Diagram Uncorrected σ = 0.40 mag 17
Nearby Hubble Diagram SALT2 corrects 0.40 0.16 mag What if we fit with R643/442 instead? SALT2 µb = (mb M) Uncorrected + αx1 - βc σ = σ 0.161 = 0.40 mag mag σcore = 0.156 mag 17
Nearby Hubble Diagram SALT2 µb = (mb M) Uncorrected + αx1 - βc σ = σ 0.161 = 0.40 mag mag σcore = 0.156 mag SALT2 corrects 0.40 0.16 mag What if we fit with R643/442 instead? Flux Ratios µb = (mb M) + γr σ = 0.128 mag σcore = 0.108 mag Flux Ratios standardize SNe Ia better than x1 and c combined Bailey et al 2009 Accepted by A&A Letters arxiv: 0905.0340 Hubble Residuals Sample R642/443 x1, c Training 0.130 0.154 Validation 0.134 0.171 All 0.128 0.161 17
Hubble Residuals Bailey et al 2009, A&A Letters, arxiv 0905.0340 Single parameter correction: Better at correcting red and peculiar SNe σ = 0.13 Combined with color: σ = 0.12 Traditional method (SALT2) σ = 0.16 σ = 0.16 0.12: statistically equivalent to having 1.8x as many SNe Better for oddballs: better systematics control 18
Literature SNe Comparison Bailey et al 2009, A&A Letters, arxiv 0905.0340 Literature SNe from Matheson, with photometry from Jha and Hicken Overall, supports our results within the resolution of the data One outlier (99cl) known to be unusual: Very heavily reddened Time variable sodium absorption Very low RV value 19
Related Work: vsi and Color Bright Magnitude Dim Brighter = Bluer but what slope? Blue Color Red Slope of color correction related to Si velocity vsi Separating high/normal vsi significantly improves scatter (0.178 0.125 mag) 99cl is in high vsi set X. Wang et al. 2009 ApJ Letters, arxiv:0906.1616 Improved distances to Type Ia Supernovae with Two Spectroscopic Subclasses vsi 20
Traditional corrections applied to the luminosity are stretch and color, with # and $ tuned to minimize the residuals to the cosmological fit to the data.! Color cut Classic Metric Studies corr M = M α x1 β etc.) c b Complete study underway of classic metrics (RSib,+EW(4000), Ability to standardize SNe Ia! Spectral correction can be applied in a same way Here, example of EWSiII4000 to show the power o Covariance with each other and stretch and color the with indicators. Example: EW(SiII 4000) Hubble Residual se 5 6 7 3 1 Color cut c& c& Correction None c & x1 EWSiII None c & x1 EWSiII 4000 4000 EWSiII4000"s correlations with Color Cut Hubble residuals and x1 increase after the color cut (see table).! RMS 0.406 0.161 0.164 0.217 0.153 0.123 nmad 0.264 0.159 0.177 0.243 0.139 0.148 Standard deviation and normalized median absolute deviation. EWSiII4000 is independant of color and a good proxy for x1.! After EW(Si color IIcut, EWSiII4000 4000) + Color is an excellent candidate to estimate part of the SN competitive with intrinsic x1 + Color Ia variability and replace the x1 parameter. (cp Bronder EW alone) Nicolas=Chotard R642/443 F (642 nm)/f (443 nm) SNfactory Corrected Preliminary Hubble residuals EW(SiII 4000) Please do not reproduce without asking; we may have updated results... 21
K-correctionless Hubble Diagram Synthesize photometry on a redshift-dependent filter-set One filter integrates the same spectral range on all SNe Minimize systematic errors due to the light curve fitter spectral model (SALT2) Normalized flux 1.0 0.8 0.6 0.4 0.2 z = 0.03 B SNf V SNf R SNf z=0.03 Normalized flux 1.0 0.8 0.6 0.4 0.2 z = 0.08 B SNf V SNf R SNf z=0.08 10 0.0 1.0 0.4 0.5 0.6 0.7 0.8 0.9 1.0 10 4 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Wavelength (Å) 10 4 (74 SNe) standard SALT2 no K-correction Δσ 2 0.0 σcore (mag) 0.158 0.138 0.077 SNe Ia σint (mag) 0.131 0.110 0.071 SNfactory Preliminary Please do not reproduce without asking; we may have updated results... Rui Pereira 22
Deciphering the Universe through Spectroscopy Entering new era of understanding SNe Ia Driven by spectral measurements Standard methods: 8 10% accuracy on distance Flux Ratios can calibrate to ~6% using a single spectrum First spectral method with robust improvements over standard methods Bailey et al 2009, A&A Letters, arxiv 0905.0340 vsi grouping improves standard corrections to similar level Better statistical power, better systematics control Need high-z programs to match Julien Guy (SuperNova Legacy Survey): We donʼt need more supernovae, we need better supernovae Spectral measurements are providing that 23