Automatic stellar spectral parameterization pipeline for LAMOST survey
aa r X i v : . [ a s t r o - ph . I M ] J u l Statistical Challenges in 21st Century CosmologyProceedings IAU Symposium No. 306, 2014Alan Heavens, Jean-Luc Starck & Alberto Krone-Martins, eds. c (cid:13) Automatic stellar spectral parameterizationpipeline for LAMOST survey
Yue Wu, Ali Luo, Bing Du, Yongheng Zhao and Hailong Yuan
Key Laboratory of Optical Astronomy, National Astronomical Observatories, ChineseAcademy of Sciences, Beijing 100012, Chinaemail: [email protected]
Abstract.
The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) projectperformed its five year formal survey since Sep. 2012, already fulfilled the pilot survey and the1 st two years general survey with an output - spectroscopic data archive containing about 3.5million observations. One of the scientific objectives of the project is for better understanding thestructure and evolution of the Milky Way. Thus, credible derivation of the physical propertiesof the stars plays a key role for the exploration. We developed and implemented the LAMOSTstellar parameter pipeline (LASP) which can automatically determine the fundamental stellaratmospheric parameters (effective temperature T eff , surface gravity log g , metallicity [Fe/H],radial velocity V r ) for late A, FGK type stars observed during the survey. An overview of theLASP, including the strategy, the algorithm and the process is presented in this work. Keywords. techniques: spectroscopic, methods: data analysis, stars: fundamental parameters
1. Introduction
LAMOST is a unique reflecting Schmidt telescope, with a large aperture and a widefield of view (Cui et al. 2012), sited at the Hebei province in China. It can simultaneouslyobserve 4000 celestial objects and obtain the spectra through the multi-fiber system.After the previous two years commissioning survey and one year pilot survey, LAMOSTaccomplished its 1 st two years general survey (Zhao et al. 2012) in the period betweenSep. 2012 and Jun. 2014. The LAMOST First Data Release - DR1 (Pilot and the 1 st year survey data) published in Aug. 2013, contains more than 2.2 million spectra, themajority of the targets are stars with a quantity of more than 1.9 million, others areGalaxy, QSO and UNKNOWN (low S/N). The whole DR2 (DR1 plus the 2 nd yearsurvey data, will be published at the end of 2014) sky coverage is demonstrated in theFig. 1. One of the two main components of the project is the LAMOST Experiment forGalactic Understanding and Exploration (LEGUE) which is focusing on the formationand evolution of the Milky Way (Deng et al. 2012). The LAMOST data processing andanalysis pipeline is integrated with three modulars: 2D pipeline (raw CCD data reduction,extraction, wavelength calibration), 1D pipeline (classification and the redshift estimationfor galaxy and QSO, Luo et al. 2012) and the LASP (Wu et al. 2011a).
2. Data and Strategy
The LAMOST spectra are with a resolving power of R ∼ A wavelength range. The consecutive data processing system of LAMOST is organizedwith a central versioned main data base (MDB) which is supported by MySQL. Theobserved data, the conventional observing information as well as all the intermediateand final output of each data reduction and analysis modular are all archived in thisMDB, the published spectroscopic data are in the format of fits file. According to the1 Y. Wu et al.moon status, we approximately divided the survey into three parts: dark night, brightnight, and instrumental test night. The current selecting criteria for the LASP input are:‘final class’ is STAR and ‘final subclass’ is late A or FGK type, with g band S / N > / N >
3. Methods and Procedure
CFI method . The name CFI is an abbreviation of ‘correlation function interpolation’(Du et al. 2012). Provided that the observed flux vector is O , and the synthetic modelflux vector is S , theoretically, the best-fit pair is that cos < O , S > = 1. So it searchs forthe best-fit by maximizing the value of cos < O , S > as functions of T eff , log g and [Fe/H],where cos < O , S > = ( O · S ) / ( | O |×| S | ) which is referred to as correlation coefficient. Forthe selection of the synthetic spectral girds, CFI employed Kurucz spectrum synthesiscode based on the ATLAS9 stellar atmosphere model (Castelli & Kurucz 2003). ULySS method . ULySS (Koleva et al. 2009, Wu et al. 2011b) determines the atmo-spheric parameters via minimizing the χ between the observation and the model spec-tra which are generated by an interpolator built based on the ELODIE stellar library(Prugniel & Soubiran 2001,2007). The model is:Obs( λ ) = P n ( λ ) × G( v , σ ) ⊗ TGM( T eff , log g , [Fe / H] , λ ), where Obs( λ ) is the observedspectrum sampled in log λ , P n ( λ ) a series of Legendre polynomials of degree n, andG( v , σ ) a Gaussian broadening function parameterized by the residual velocity v , andthe dispersion σ . The multiplicative polynomial is meant to absorb errors in the fluxcalibration, Galactic extinction or any other source affecting the shape of the spectrum.The TGM function is an interpolator of the ELODIE library, it consists of polynomialexpansions of each wavelength element in powers of log( T eff ), log g , [Fe/H] and f( σ ) (afunction of the rotational broadening parameterized by σ ). Three sets of polynomials aredefined for three temperature ranges (roughly matching OBA, FGK, and M types) withimportant overlap between each other where they are linearly interpolated. The derived Figure 1.
The LAMOST DR2 sky coverage. utomatic stellar parameterization pipeline for LAMOST
Procedure . Based on the algorithms CFI and ULySS, by fitting the spectra, LASPexecuted in two stages to effectively derive the stellar parameters. Since the LAMOSTspectral flux calibration is relative, in the 1 st stage we measure the original observations,in the 2 nd stage, we measure the normalized spectra. Considering the low instrumentalresponse in both edges, as well as for optimizing the computing time and the storagespace, CFI and ULySS respectively select [3850, 5500] ˚ A and [4100, 5700] ˚ A as the fittingwindow. In each stage, firstly, we utilize CFI to quickly get a set of initial coarse estima-tions, after that, by using CFI results as a guessed starting point, we adopt ULySS toobtain more accurate and credible measurements as the final output. CFI can rapidly se-lect the appropriate stellar template fitting range, this advantage helps effectively reducethe computational quantity and time for ULySS by using less guessed starting grid.
4. Conclusions
The data procession phase of the large survey project comprises three important tasks:validation, calibration and in-mission software development. LASP is being improved andwill continue producing an enormous set of parameters for various stars. Thus criticalassessment of its measurements is significant for the realization of the project’s scientificgoal. In one of the independent external validations, by comparing the common hundredsof stars between LAMOST DR1 and the PASTEL Catalog (Soubiran et al. 2010), weobtain precisions of 110 K, 0.19 dex, 0.11 dex and 4.91 km/s for T eff , log g , [Fe/H] and V r respectively in the specified temperature range (Gao et al. 2014). A systemic compre-hensive validation and calibration works according to different spectral types and SNRshave already been carried out, the results will be published in a separate work.With the LAMOST DR1, a catalogue of stellar parameters which contains more than1 million stars was simultaneous published, it has been characterized as the largest onein the world so far. For the spectra acquired during the 2 nd year survey, LASP alreadypreliminarily determined ∼ Acknowledgements . The authors thank the grants (Nos. 11103031, 11273026, 61273248,11178021) from NSFC and the National Key Basic Research of China 2014CB845700.
References
Castelli, F. & Kurucz, R. L. 2003,
IAUS , 210, A20Cui, X. Q., Zhao, Y. H., Chu, Y. Q, et al. 2012,
RAA , 12, 1197Deng, L. C., Newberg, H. J., Liu, C., et al. 2012,
RAA , 12, 735Du, B., Luo, A. L., Zhang, J. N., Wu, Y., et al. 2012,
SPIE , 8451, 37Gao, H., Zhang, H. W., Liu, X. W., et al. 2014,