With the ability of capturing four thousands spectra in one exposure, the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) is a special quasi-meridian reflecting Schmidt telescope located in Xinglong Station of national Astronomical Observatory, China. It released the fifth spectral data this year, containing 8,171,443 star, 153,090 galaxy, 51,133 QSO and 642,178 unknown type, all of which are classified by the LAMOST 1D pipeline. This pipeline is used for spectral analysis, aiming to determine the spectral type and redshifts of the spectra observed by LAMOST by matching with spectral templates. Generally, the performance of the stellar classification greatly depends on the quality of templates. In this paper, we construct a new stellar template library, which is supposed to increase the number of types and the accuracy of the classification, for LAMOST based on the data from LAMOST DR5. All the 9 million spectra from LAMOST DR5 are participated in this construction experiment and they are gathered in 500 groups using k-means clustering method. Those group centers corresponding to spectra less than 1000 are abandoned at first. Then the weighted average spectrum (group center) is served as the template spectrum in each group. Initially, 417 centers are obtained. We visually inspect all template spectra and discard 181 centers due to low spectral quality or the similarity between different group centers. Furthermore, the types of the remained clustering centers are assigned by the subclass of spectra from LAMOST DR5. Meanwhile, 19 templates whose subclass are difficult to determine are also abandoned. Afterwards, we obtain a new template library containing 197 LAMOST template spectra with 82 different MK classes. Finally, the feasibility and accuracy of using this template for classification has been verified by comparing and analyzing the classification results of several control groups of data.
Link to PDF (may not be available yet): P1-6.pdf