Results from the Nearby Supernova Factory R. C. Thomas for the SNfactory KITP: 2007-03-23 A/K/A (it s okay) Rollins Thomas Tom Rollins Roland Thomas
Collaboration LBL Yale LPNHE IPNL G. Aldering, PI C. Baltay P. Antilogus C. Buton S. Bailey N. Ellman J. Li Y. Copin S. Bongard D. Rabinowitz R. Pain E. Gangler M. Childress R. Scalzo R. Pereira G. Smadja S. Loken C. Wu P. Nugent U Chicago CRAL S. Perlmutter R. Kessler E. Pecontal R. C. Thomas G. Rigaudier B. Weaver Scientist/Faculty, Postdoc, Grad Student
SNfactory Science Goals Few hundred SN Ia light curves between 0.03 < z < 0.08 to reduce Dark Energy statistical uncertainty. like SNLS or ESSENCE JDEM with SNe Ia Linder 2006 SN Ia light curves and spectra in time series to study to enable the control of supernova systematic uncertainty. Analysis of the data set to improve our understanding of SN Ia physics and better calibrate them as standardized candles.
Motivation DETF For SN - detailed spectroscopic and photometric observations of 500 nearby supernovae to study the variety of peak explosion magnitudes and any associated observational signatures of effects of evolution, metallicity, or r e d d e n i n g, a s w e l l a s improvements in the system of photometric calibrations. Linder 2006
Motivation DETF For SN - detailed spectroscopic and photometric observations of 500 nearby supernovae to study the variety of peak explosion magnitudes and any associated observational signatures of effects of evolution, metallicity, or r e d d e n i n g, a s w e l l a s improvements in the system of photometric calibrations. BORING SNe Ia are also interesting physical phenomena! Linder 2006
Search Overview Palomar 1.2-m, QUEST-II camera (NEAT/QUEST surveys, piggybacked). Search pipeline at LBL for rapid (<24h) turnaround on large sky area. Processing/subtraction code from SCP with added infrastructure to scale up 100x or 1000x throughput. Reliance on distributed computing, distributed long term storage through NERSC facilities at LBL.
Sky Coverage Milky Way Followup Area Daytime NEAT Pointings Ecliptic Point+Track: 2500-3000 sq deg/month. Drift-scan: 100-200 sq deg/month.
Backseat Driver Search reference image new image subtraction reference image new image subtraction
Search Pipeline PALOMAR HPSS PDSF New Images Collected HPWREN Raw New Images Preprocessing Object Finding Astrometry Processed Images Reference Cache New/Ref Matching Less than 24 h turnaround is critical. Subtractions 30000-40000 (50GB) images/night. Subtractions Generated, Events Scored Candidates Scanned by Humans
Machine Learning Classifiers Candidates live in ~20 dimensional Boosted decision trees, trained S/N > A score space. by fakes. yes no motion < B FWHM < C Hyperplane cuts yes no no retain too much FWHM < D Reject yes Reject garbage. yes no Supernova Reject Supernova ~ 1000 candidates requiring human scanning. Reduced human scanning load to 100 candidates/day. Bailey, et al. 2007 (ApJ, submitted)
Machine Learning Classifiers Candidates live in ~20 dimensional score space. Boosted decision 10 0 trees, trained by fakes. 10-1 S/N > A Original Threshold Cuts Support Vector Machine Random Forest yes Boosted Tree no Hyperplane cuts retain too much garbage. False positive rate 10-2 10-3 yes 10-4 Supernova yes FWHM < D motion < B no Reject FWHM < C no yes Reject Supernova no Reject ~ 1000 candidates requiring human scanning. 10-5 Reduced human scanning load to 100/day. 0.0 0.2 0.4 0.6 0.8 1.0 True positive rate (signal efficiency) Bailey, et al. 2007 (ApJ, submitted)
Followup Overview LBL Subtractions Events (scored) IN2P3 (France) Postmortem Analysis Candidates (scanned, vetted) UH88/SNIFS Storage Resource Broker/Central SNIFS Processing 30% share of UH88 time 2-3 nights per week
C SNIFS Filter Wheel SuperNova Integral Field Spectrograph Spectroscopy channels Blue: 320-520 nm Red: 510-1100 nm P B 2k x 4k 2k x 4k 2k x 4k MLA FOV 6 x6 Calibration unit arc lamps continuum lamps Photometric channel ~ 9 x 9 FOV R 2k x 4k Microlens Array (15x15) automated acquisition guiding extinction monitoring
SNIFS SuperNova Integral Field Spectrograph C Filter Wheel P 2k x 4k 2k x 4k B 2k x 4k R 2k x 4k Microlens Array (15x15) extinction monitoring guiding
900-950 nm 350-400 nm Individual lenslet spectra mapped into datacubes. Quickie aperture extraction for realtime feedback.
Automated Spectroscopy Remote observe from anywhere via VNC. 2-3 day cadence, shifts done in France (daytime) UH observers have used SNIFS for comets, AGN and asteroids.
P Channel Acq Light Curves (also last chance to screen candidate) SNF20051003-004 20s V-band acquisition images Flux (arbitrary units) (s 0.96 -- SALT) useful cross-check or alternative calibration 0 10 20 30 40 50 60 MJD (+2453640)
SNIFS Productivity Spectroscopy vs Time 12th Sep 2006 Increase in slope at 19 May due to full night transition. 19th May 2006 1st Jan 2006 LIPS incident, POP motor problem responsible for flat Total spectro portions. 1st Jan 2005 14th Aug 2004 SNe Ia follow-up First 3 months of 2006 had the worst weather in decades on Mauna Kea.
SNIFS SN Ia Followup Historically we have supplemented our search with IAUC SNe. Total SNe followed by SNIFS 1st Sep 2006 1st May 2006 1st Jan 2006 Search productivity increase has reduced our time available for IAUC targets. 1st Jan 2005 SNe from others followed by SNIFS SNe from SNF followed by SNIFS
Screening and Followup SNIFS targets discovered by SNfactory SNIFS targets discovered by others
Time Series Coverage
Initial Phase Distribution SNe Ia discovered early receive most extensive followup 60 Number of SNe Ia 50 40 30 20! 1 SNIFS epoch! 6 SNIFS epoch! 10 SNIFS epoch Initial phase estimated by spectral feature match age accurate to ± 2 d at max 10 0-20 -10 0 10 20 30 40 50 Initial Phase
Redshift Distribution SNfactory sample has good overlap with local Smooth Hubble Flow 30 Number of SNe Ia 25 20 15 10! 1 SNIFS epoch! 6 SNIFS epoch! 10 SNIFS epoch Redshift estimated from spectral feature matches, accurate to ± 0.01. 5 0 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Redshift
Synthetic Photometry S/N 10 4 Supernova Signal-to-Noise 10 3 S/N 10 2 1% 1% 3% 3% 10 1 10 0 U (Mean = 198.2) B (Mean = 531.0) V (Mean = 404.7) R (Mean = 1000.0) I (Mean = 613.9) -10 0 10 20 30 40 50 Phase (days)
Synthesized Light Curve
Since August 2004 1200 spectra of individual supernovae. ~300 SNe with one spectrum or more. 2/3 from SNfactory search. 1/3 from other sources. TDC : Tête de cuvée Initial phase t < 0 Epochs SNe Ia Spectra Total TDC Total TDC 0.03 < z < 0.08 23 since May 2006 6 76 38 912 465 10 55 29 754 396
Current Challenges Increasing pre-maximum (t < -1 wk) purity. Rolling trigger search. Photometric screening aside from SNIFS. Increasing SNIFS throughput (12-14 per night). Not much room for operations improvements. Long-term schedule optimization SNIFS data pipeline -- primary issue is extraction. Host + Sky + Supernova -- blind source separation. Use of final reference subtraction.
SN 2005gj Relative F! + Const. SNIFS SNIFS Keck SNIFS Keck 11 d (Oct. 3.6) 64 d (Nov. 25.5) 71 d (Dec. 2.2) 71 d (Dec. 2.5) 133 d (Feb 2.2) 3000 4000 5000 6000 7000 8000 9000 10000 Observer Frame Wavelength (Å) V-band flux (ergs/cm /s/hz) 2 2 i-band flux (ergs/cm /s/hz) x1e-26 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 x1e-26 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 Normal SN template 1991T-like template SN 2005gj SNIFS SN 2005gj NEAT/QUEST SN 2002ic 0 20 40 60 80 100 Normal SN template 1991T-like template SN 2005gj SNIFS SN2005gj NEAT/QUEST SNIFS spectral decomposition 0 20 40 60 80 100 Days since explosion Aldering, et al., 2006
Progenitor/Host Host is low metallicity, 18 19 SN2005gj Host Model Z < 0.3 solar Starburst age is consistent with AB Magnitude 20 21 22 200 Myr Starburst Red Template 3 solar mass progenitor. 23 2000 4000 6000 8000 10000 Observed Wavelength (Å) 3 solar mass donor, loses 2 solar masses, WD already near Mch?
SN 2006D with SNIFS Thomas, et al. 2007
SN 2006D Acq LC Relative Flux 1.2 1.0 0.8 0.6 0.4 20s acq images provide preliminary V-band light curve 0.2 0.0 stretch 0.77-10 0 10 20 30 40 50 Lightcurve Phase
Comparisons
Synthetic Spectrum
How Sporadic is Carbon? Temperature/ionization? 17,000 km/s Explosion model parameters? Geometry? Small clumps of unburned material may always be present. Sparse distribution and small cross-section could account for sporadic observation. Polarization signature?
2007-2008 Arranged search/follow-up time to take the best 9 months of weather at both sites: April-December. Expect Palomar QUEST-II search to continue until at least Fall 2008 (end of Palomar+QUEST MOU). Expect UH88 SNIFS follow-up to be possible until at least Spring 2009 (UH88 demolition for PanSTARRS). Investigating additional imaging at Mt. Helmos 2.3-m. Investigating rolling trigger search at Palomar 1.2-m to reject old SNe, reduce SNIFS screening load, increase early-phase purity.
Want More Like This
2009 and Beyond? Sample size at end of 2008 ~ 200 SNe Ia. Move SNIFS to another telescope? Build a new SNIFS on another telescope? Use a different search camera? Use a different/additional search telescope? Partner with other searches coming online?
Conclusion The Nearby Supernova Factory has been discovering, and obtaining spectral time series of SNe Ia in the nearby smooth Hubble flow since late 2004. Writing some papers on interesting supernovae. We envision continuing operations until the end of 2008. We are investigating multiple options for continuing operations beyond that point. Look for light curves, spectra, and cosmology fits in the coming year.