220 likes | 365 Views
An update on FERRE for SEGUE. C. Allende Prieto Instituto de Astrofísica de Canarias. Preview. Atmospheric parameters and abundances Automation Ferre then Ferre now The SSPP methods k24 and ki13 Recent developments for SDSS (SEGUE/BOSS) Tests Future improvements.
E N D
An update on FERRE for SEGUE C. Allende Prieto Instituto de Astrofísica de Canarias
Preview • Atmospheric parameters and abundances • Automation • Ferre then • Ferre now • The SSPP methods k24 and ki13 • Recent developments for SDSS (SEGUE/BOSS) • Tests • Future improvements
Atmospheric Parameters • Minimum set that we need to constrain for of carry out a spectral analysis • Usually Teff,logg,[Fe/H] • Sometimes , , [C/Fe], [/Fe] … • Frequently ignored but sometimes important vsini, B …
Teff • F = Teff4 • F R2 = f d2 • Can be directly determined from bolometric flux measurements f and angular diameters (2R/d) hard but spectacular progress recently • Photometry: model colors, IRFM • Spectroscopic: line excitation, Balmer lines • Spectrophotometric: model fluxes
logg • Gravitational field compresses the gas giving a nearly exponential density structure (pressure) • Hard to get with accuracy: the spectrum is only weakly sensitive to gravity • Photometry: ionization edges (Saha), molecular bands, or damping wings of strong metal lines • Spectroscopy: ionization balance (e.g. Fe/Fe+) or collisionally-dominated line wings • Stellar structure models (luminosity)
[Fe/H] • A big oversimplification • High sensitivity of the spectrum (can also be derived from photometry including blue/UV), but highly model dependent • Need many weak lines, good atomic data, good spectra, and a good model
Automation • Classical analysis methods can be coded in the computer • These will have limitations: need to reliably measure equivalent widths (EW) • Ultimately, the use of EW is related to simplify the calculations (scalar quantities instead of arrays) but is also somewhat blind, I.e. full spectral analysis preferred
Automation II • Optimization methods: local (gradient, Nelder-Mead…), global (metropolis, genetic algorithms…) • Projection methods (ANN, MATISSE, PCA, SVM…) • Bayesian methods • But many combinations possible • Spectral model can be calculated on the fly or interpolated • Issues are sometimes continuum normalization, complicated PSF, large number of dimensions, degeneracies
Massive surveys • RAVE (R~7500, 841-880 nm) • SDSS-SEGUE (R~1800, 380-910 nm) • SDSS-III APOGEE (R~30000, 1500-1700 nm) • Gaia RVS, BP/RP (R~11500, 847-874 nm) (R<50, 300-1000 nm) • Also: HERMES, HETDEX, Gaia-ESO, BigBOSS …
Ferre then • Development in the early (pre-SEGUE) SDSS times • Optimization with interpolation on a pre-computed grid • N-dimensional f90 code • Nelder-Mead algorithm (Nelder & Mead 1965) • Linear interpolation • Spectral library on memory
Ferre now • Multiple algorithms: Nelder-Mead (Nelder & Mead 1965), uobyqa (Powell 2002), Boender-Rinnooy Kan-Strougie-Timmer algorithm (1982) • Linear, quadratic, cubic spline interpolation • Spectral library on memory or disk • PCA compression • Handling of complex PSF w/o compression • Serial optimization for speed • OpenMP parallelization • Flexible: WD surveys, APOGEE, STELLA, Gaia-ESO…
SSPP methods: k24 and ki13 • These are just libraries of spectra. The same code is working behind the scenes • k24 considers both continuum-corrected spectra (~440-550 nm) and (g-r) photometry (Allende Prieto 2006) • ki13 considers only continuum-corrected spectra (~440-550 nm), but model fluxes include more physics (Lee et al. 2008)
Recent developments • model spectra: improved atomic and molecular linelists (mostly from Kurucz, upgraded to include Barklem VdW constants for metals), improved continuum opacities in the blue • LSF handling: Gaussian LSF (,fibid) • Higher-order interpolations • Speed
NGC 2420 • Chosen as testing data set (not sure why) • 2 Gyr old, [Fe/H]~ -0.4 open cluster • Took the two SEGUE 2078/2079 plates • Here only 2078-53378 will be discussed • Started with several grids inherited from MILES experiments: holes in several places • 3500-5000 K, 4750-6000 K, 6250-8000 K, 8500-11500 K, 12000-26000 K … • **WARNING** **DANGER** VERY PRELIMINARY RESULTS
NGC 2420 • 2 filter
NGC 2420 • Vrad filter
NGC 2420 • [Fe/H]
NGC 2420 • [Fe/H] F-type stars
NGC 2420 • [Fe/H] GF-type stars
NGC 2420 • Comparison with SSPP
Next • Add photometry • Add reddening • Fill the gaps (ongoing) • Normalizing the spectra • Abundances (see Emma’s talk)