English | 简体中文

You are here

Spectra Analyse

LAMOST 1D-Pipeline is designed to classify the spectra of 2D-Pipeline output and to measure their redshifts (or radial velocity). A reliable technique, which is develop by Glazebrook et al.(Glazebrook 1998), for automatically determining galaxy redshift is becoming increasingly important. This method generalizes the cross-correlation approach by replacing the individual templates with a simultaneous linear combination of orthogonal templates. This effectively eliminates the mismatch between templates and data and has the potential to provide for the possibility of good error estimation. The method, which is called “PCAZ”, is based upon the use of principal component analysis to make the general linear problem amenable to efficient computation. This method proves more robust in the low S/N regime than independent cross- correlation and has greater potential for very high success rates in upcoming very large redshift surveys. It is also appropriate for the QSO and stellar spectra. Our 1D-Pipeline is based on the specBS pipeline which is used to analysis SDSS spectra as one of 1D pipelines. This analysis code carry out ¬χ2 fits of the spectra to templates in wavelength space (in spirit of Glazebrook et al.1998), fitting spectra with linear combinations of eigen-spectra and low-order polynomials. A set of carefully zero-pointed SDSS spectra of a variety of spectral types, which contains galaxy, QSO and stars (subdivision to child type), are used to construct our templates.

The accuracy of our method, like any other analysis, is affected by signal-to-noise, wavelength calibration, and flux calibration. Relatively high accuracy redshift and correct spectral type are obtained from good quality spectra, which have accuracy wavelength and better flux calibration, especially high signal-to-noise. For spectra with low quality, our templates matching method cannot ensure the correct of our classification and redshift estimate, but we can give a indicator which present the confidence of the redshift measurement. We employed the keywords ZWARNING and ZCONF in the FITS headers to express the confidence of redshift (as in SpecBS). The ZWARNING is a flag which shows all warnings during the spectra reduction, such as bad fiber, bad sky fiber, or bad fitting. If the flag ZWARNING is OK, then the larger ZCONF, the more confident the redshift. Meanwhile, the signal-to-noise is a direct index of spectra quality. The Pilot Survey data release needs a strategy for final spectral type and final redshift of LAMOST spectra for astronomical users. Base on the above, we divide our data for several parts. For the spectra with S/N>5, and ZCONF>10%, we are inclined to apply 1D-Pipeline result of spectral type and redshift. The low confidence measurement or low signal-to-noise LAMOST spectra, human checking will provide a spectral type of the spectra, and we will recalculate the redshift of those low quality spectra. Some of them would be considered useless and abandoned after checked by human eyes. A flag named “specflag” would be used to explain the status of final redshift. After compare the classification results of 1D-Pipeline with SDSS DR8 spectral type, our pipeline shows a well performance for the good quality part of spectra, a 96% correction is obtained from spectra with a threshold of 10% for ZCONF and 10 for the r-band signal-to-noise. For the correct classified spectra, we compare the redshift estimate with SDSS. A precision with a sigma of 0.0001 for galaxy, 0.01 for QSO and 13km/s for star spectra shows the capability of the code.