The Cornell Atlas of Spitzer Spectra (CASSIS) and recent advances in the extraction of complex sources V. Lebouteiller (CEA & Cornell University) D.J. Barry, G.C. Sloan, H.W.W. Spoon, D.W. Weedman, J.R. Houck (Cornell University) ADA7 - May 2012
Spitzer/IRS Infrared Spectrograph (IRS; Houck et al. 2004) on board Spitzer (0.85m; Werner et al. 2004). MIR range. Low-resolution mode R=λ/Δλ=60-127, 5-38µm, high-resolution mode R=600, 10-37µm Cryogenic mission ended in May, 2009 16 000 observations in low-resolution, 9 000 in high-resolution Interstellar and circumstellar dust emission Ionic emission-lines, [NeII], [NeIII], Huα,... Star-formation rate tracers, chemical abundances Active Galactic Nuclei, [OIV], [NeV] Molecule emission/absorption, H2, C2H2, PAH, CO2,... Bernard-Salas et al. (2010)
The means and needs of IR spectroscopy Herschel/(PACS, SPIRE): 50-670µm, R-40-5000. End of mission closing in (early 2013) Airborne observatory SOFIA/FORCAST: 5-14µm, less coverage and less sensitive than IRS. Other instruments complementary in the resolution/sensitivity space To come, JWST/MIRI (2018 TBD): 5-28µm. Small slit width, not so well adapted to z < 0.1-0.2 sources To come, SPICA (2018?), 5-210µm / CCAT (ground, 25m), 200-2200µm IRS left a legacy serving as basis and complement for present and future IR missions. Redefine science goals? Need 1: Spectroscopic atlas of MIR spectra for the simplest cases. Publication-ready Enables massive analysis + easy access for the IR and non-ir communities Need 2: specific algorithms for the complex/faint sources new science inside!
Spectral extractions Integrate flux (regular), for point-like and extended sources Weigh by the PSF (optimal), for point-like Super-sampled PSF mandatory. Empirical is better because of the complex optical path spectral order 1 spatial peak-up image Regularized image construction applied to 1D (Pinheiro da Silva et al. 2006 for COROT) Iterative algorithm on point-like source scans along the aperture wavelength order 2 peak-up image
Iterative process undersampled spectra = (downsampling matrix) (geometric transform, i.e., shift) (super-sampled PSF) + noise µ controls the convergence speed, λ controls the regularization (from a theoretical PSF or first-guess) W (source shift) depends on the source finder, itself depending on the super-sampled PSF Not symmetrical! Intrapixel responsivity matters when sampling is low
Brand new science with a dead instrument Larger S/N (1.5-2) Meant for point-like sources but can quantify deviation from point-like => source extent Multiple/blended source extraction, complex background through a simple multiple linear regression For a given wavelength
SMART (Higdon et al. 2004; Lebouteiller et al. 2010) IDL package for data reduction and analysis NGC1365 Starburst disk & AGN nucleus
Extraction pipeline CASSIS pipeline (Lebouteiller et al. 2011) 12 000 spectra automatically extracted and publication ready Focuses on extraction and background removal. Start from individual exposure images corrected for optical artifacts Multi-step decision tree based mostly on diagnostics from optimal extraction background subtraction methods (contaminating source check) combine image exposures only if pointing accuracy ok extraction method and flux calibration based on source spatial extent (deviation from point-like) First version released last September, new version released in July, testing phase More than just a repository
Massive data analysis Local access offered All products, including spectra as SQL tables. IDL & Python script templates provided Coming soon: online SQL queries through the portal with syntax simplification Automatic feature measurements in pre-defined samples Spectral stacking VO format also provided A class of (galaxies, PNe, HII regions,...) may show very different spectral shapes A spectral feature may appear in very different object types
Massive data analysis Particularity of IR-submm spectra: many entangled features (dust continuum, molecular emission/absorption bands, emission lines) Spectral comparison and matching is a key driver for science finding all sources showing a given feature (e.g., fullerenes) building samples from a set of spectral and observational criteria Spectral matching based on a line and/or the continuum. Very naive approach so far. Any suggestions for patternmatching welcome!
Application to the extra-galactic sample Wealth of spectral features => large potential for IR spectroscopic determinations of redshifts. 1st step: cross-correlation with (optical, HI) spectroscopic redshifts (2200 sources) 2000 more sources: redshifts using IXcorr Test sample: list of CASSIS spectra with known redshifts Spectral templates Publication of the results for future surveys Wavelength
Projects Current & Astrophysics Data Analysis Program (NASA) funding applications PCA/NNMF methods for spectral shapes in IR luminous galaxies (PI. Farrah; University of Virginia) CASSIS/SDSS cross-correlation. Extinction, AGN diagnostics, star-formation calibration (PI. Hao; Shangai) Spectroscopic redshift machine (PI. Spoon; Cornell) Serendipitous source catalog (PI. Lebouteiller; CEA) => mostly background galaxies Dust and obscuration in the most luminous objects in the Universe (PI. Weedman; Cornell). WISE, SDSS, SWIFT High spectral resolution optimal extraction atlas (PI. Sloan; Cornell) Spectral feature measurements for extragalactic sample (PI. Spoon; Cornell)
Summary IR detectors and spectra are fun :) New extraction techniques revealed new science cases application to future IR missions with similar design concepts (JWST/MIRI/LRS) MIR spectroscopic atlas variety of IR spectra makes it mandatory to come up with comparison tools