|
In studies of distant galaxies, the properties of the stellar populations--the age and metallicity--must be inferred from the integrated light of the entire population. This is done by modeling the spectral energy distributions (SEDs) of simple stellar populations, in which all of the stars are assumed to have the same age and metallicity. The results of such models (e.g., Bruzual & Charlot 2003, Maraston 2005) are used to interpret the star formation histories of galaxies, with implications for galaxy formation scenarios. One crucial component that must be input into stellar population models is a library of stellar spectra. These stellar spectra, which can be derived empirically or theoretically, are used to make synthetic galaxy SEDs and then compared to an observed galaxy's SED to constrain its properties. For the best results, the stellar library should have high signal-to-noise (S/N) data and cover a large range in stellar parameters (effective temperature, metallicity, surface gravity), as well as a range in stellar evolutionary phases. The Sloan Extension for Galactic Understand and Exploration (SEGUE) provides us with a rich dataset of multi-band photometry and medium resolution spectra of over 100,000 stars that can be used to produce such a library. The large sample size improves our chances of finding asymptotic giant branch (AGB) and blue horizontal branch (BHB) stars, which are few in number but luminous enough to contribute significantly to the integrated light. The uniform data reduction and flux calibration procedures should allow for the use of these spectra without too much further modification. The following are some preliminary analyses of the contents of the SEGUE survey. The stellar parameter coverage of survey targets is shown, using the results of a pre-existing reduction pipeline. The accuracy of this pipeline is evaluated through a comparison with high resolution results. Lastly, three specific populations of stars are examined: blue horizontal branch stars, hot subdwarfs, and cool giants.
The SEGUE Stellar Parameter Pipeline (SSPP, Lee et al. 2007) uses a number of methods to determine an estimate of the effective temperature Teff, surface gravity logg, and metallicity [Fe/H] of each target in the survey. The pipeline is most effective for stars in the temperature range 4500-7500K. The SEGUE sample is derived from 192 plates in the SDSS:
Each plate has 640 fibers, for a total of 122,880 targets; 53,116 targets have SSPP parameters (Teff≠-9999, logg≠-9.999, [Fe/H]≠-9.999) and S/N > 25, where S/N is the median signal-to-noise per pixel. These targets make up the sample examined here. Below are plots showing the coverage for eight different temperature bins, ordered from coolest (< 4500K) to hottest (> 7500K). These are number density plots, where the dark patches indicate a large concentration of objects. For bins in which there is only one object, the individual object is plotted. On the left are the positions of the stars in color space (g-i vs. u-g; click for larger version). To provide the same reference point in all of the temperature bins, the full sample of 53,116 is plotted as gray dots. As the temperature increases, the targets move from the reddest (upper right) to bluest (bottom left) part of the color-color diagram. On the right are the positions of the stars in logg-[Fe/H] space. (click for larger versions)  
 
 
 
 
 
 
  As in the analysis of Covey et al. (2007), the SEGUE targets typically fall on or near the stellar locus in the color-color diagram. In many of the temperature bins (for example, 5000 < Teff < 5500K and Teff > 7500K), a clear delineation between two populations is seen in the logg-[Fe/H] plots. Roughly, these are nearby metal-rich dwarfs (high [Fe/H], logg) in the disk and distant metal-poor giants (low [Fe/H], logg) in the halo, two populations which fall within the apparent magnitude range of the SEGUE survey (14 < r < 20). The above plots are qualitatively the same when a higher S/N cut is used (34,166 targets with S/N > 35).
An important check on the accuracy of the SSPP is to compare the results with those obtained in an independent way. Reliable "true" values of Teff, logg, and [Fe/H] can be determined using high resolution spectroscopy. This has been done for a selected sample of 114 stars, and the results are shown below. The large blue circles are the "true" values of logg and [Fe/H] determined from high resolution spectra, while the smaller red circles are the values derived from the pipeline. The full SEGUE sample of 53,116 is plotted as black dots for reference. (click to see larger versions)  
 
  Qualitatively, the pipeline parameters agree with the "true" values. In temperature bins where separate populations are visible (for example, 6000 < Teff < 6500) the discrepancies between the two methods are generally not large enough to move the object from one population to another.
The pipeline parameters can be used to provide some preliminary identifications of stars. One population for which this is possible is blue horizontal branch (BHB) stars, which are hot (> 7500K), luminous, and have low surface gravity. These stars may be important contributors to the ultraviolet (UV) light of distant galaxies with old stellar populations. Elliptical galaxies, in particular, have been seen to exhibit a large range of UV fluxes; the UV flux is thought to be a potential indicator of stellar age and/or metallicity (O'Connell 1999). BHBs make up one of the distinct populations in the number density plots above. For verification, a sample of 2852 confirmed BHB stars are compared to hot stars in the SEGUE sample. These are plotted below in color-color and logg-[Fe/H] space. (click for larger versions)   Nearly all of the members of the confirmed sample have SSPP effective temperatures Teff > 7500K, and fall into the less metal-rich, lower surface gravity clump shown above. The agreement improves for higher S/N targets (S/N > 50). The more metal-rich, higher surface gravity clump is likely made up of blue stragglers.
Another group that can be identified relatively easily is hot subdwarfs, sdO and sdB. Like BHBs, these stars may be important contributors to the UV light of old stellar populations. Since these stars have colors similar to those of white dwarfs (Harris et al. 2003), they are initially targeted in the survey as white dwarf candidates. Their spectral features, however, match those of template O and B stars. In SEGUE, these are simply objects that have parameter values PRIMTARGTYPE='WD' and SUBCLASS='O', 'OB' or 'B'. Using these selection criteria on our sample of 122,880 targets, we find 1222 WD, 22 sdO, 167 sdOB, and 117 sdB stars. For targets with S/N > 25, we find 162 WD, 22 sdO, 14 sdOB, and 69 sdB stars. Example spectra of the highest S/N objects are shown below (right), next to published spectra for comparison (Harris et al. 2003, left). The spectra are ordered by color, from bluest/hottest (top) to reddest/coolest (bottom). White Dwarfs have comparatively featureless spectra, characterized by strong, broad Balmer lines. (click to see larger version)   Hot Subdwarfs show additional features, including HeI and HeII lines, which are characteristic of B and O spectral types, respectively. (click to see larger version)   Lastly, one of the main goals in this analysis is to identify cool giant candidates. These stars are important because they can be very luminous at certain phases of their evolution; at those times they contribute significantly to the light of their host galaxies. The search for these stars is slightly more complicated than the methods described above. First, candidate cool stars must be identified using color selections. From this sample of candidates, the giants must be separated from the much more common dwarfs using luminosity sensitive spectral features. SPECTRAL INDICES  Additional information, including the effective temperature and metallicity of the star, may also be gleaned from a set of spectral indices described by Lépine et al. (2007). The spectral type of a star can be used as a proxy for its effective temperature, and this can be determined in a number of ways. Lépine et al. (2007) measure the strengths of two prominent CaH bands at 6829Å and 6975Å, and define the spectral type Sp as a function of the sum of these two bands. The stronger the CaH bands, the later the spectral type. Several other methods of determining the spectral type are built into the SSPP. One of these generates the Hammer Type, which is determined by comparing measured the strengths of spectral indices with those of known standards. The Hammer type is calibrated using solar metallicity dwarfs and is optimized for late type stars (K-M). A detailed description of the method is provided by Covey et al. (2007). Most of the objects in our sample of 650 candidates have HAMMERTYPE='M0' to 'M4'. Lépine et al. (2007) also define a parameter ζ, a function of both CaH and TiO, which is an indicator of metallicity for late-type dwarfs: a low value of ζ (weaker TiO at a given CaH strength) indicates low metallicity. This indicator is used to identify M-dwarfs of differing metallicities: subdwarfs, extreme-subdwarfs, and ultra-subdwarfs. If the measurements of ζ are reliable they can be used to find relative metallicities to determine the range of metallicities present in our sample. COLOR SELECTION: 0.65 < r-i < 3.0, g-r > 0.3
We identify cool giant candidates based on the strengths of the spectral indices of Kirkpatrick et al. (1991). Below, right, is a reproduction of their Figure 7 using the values measured for the SEGUE targets. The spectral type Sp is the index defined by Lépine et al. (2007); this index is examined in more detail below. The numerical value of Sp corresponds to the M subclass (1 for M1); negative numbers denote K subclasses (-1 for K7, -2 for K5). The red symbols are those that satisfy the quantitative criteria Ratio B/Ratio A > 1.0, Ratio B/Ratio C > 1.03, Ratio D/Ratio A > 1.2, and Ratio D/Ratio C > 1.2, i.e., in these objects, all four spectral features indicate that they are giants.  The spectra of 6 candidate giants that fulfill these criteria are shown below (left) with non-candidates shown for comparison (right). As expected from the measured ratios, giants have little or no NaI and strong CaII, while dwarfs have stronger NaI and weaker CaII. (click to see larger version)   Next, we look at the spectral type Sp and metallicity indicator ζ defined by Lépine et al. (2007). For our sample, the results of the Lépine method and the Hammer type (from SSPP) show rough agreement (below, left; click to see larger version). The giants are represented by the red symbols and appear to have systematically earlier spectral types according to the Lépine method. This discrepancy occurs because giants have weaker CaH absorption, the strength of which is represented by Ratio A, one of the features defined by Kirkpatrick et al. (1991). For giants, the calculated spectral type may not be correct on an absolute scale but is useful in terms of assigning relative spectral types (and relative temperatures) to a set of giants.   While relative temperatures may be obtained through spectral indices, relative metallicities appear to be less attainable. The measurement of the metallicity indicator ζ defined by Lépine breaks down when the CaH absorption is very weak, i.e., in earlier-type stars and giants. The fractional error in ζ is noticeably larger for stars earlier than M (Sp < 0; above, right; click for larger version). All of the giants (red crosses) in our sample are earlier than M (Sp < 0), so ζ does not appear to be a reliable measure of the metallicity in these cases. The errors are obtained for each of the indices (CaH2, CaH3, and TiO5, etc.) by generating one hundred realizations of the same spectrum that fall within the errors determined by the SDSS reduction pipeline. The indices are measured for each of the one hundred realizations. The error on an index measurement is the standard deviation of this set of one hundred measurements. Errors on functions of the indices--the metallicity index &zeta and the spectral type Sp--are measured in the same way; they are not derived from the measured errors on the indices. The plots below show the fractional errors for each index as a function of spectral type Sp. The red crosses indicate the giants in the sample. (click for larger versions)     The fractional errors of the individual indices do not appear to vary significantly as a function of spectral type. The errors on Sp and &zeta, which are functions of the indices, however, are systematically higher for earlier spectral types (Sp < 0, bottom left). This is much more pronounced for errors in &zeta (Sp < 0, above). Below are four color-color plots showing the 650 targets in this sample; the giant candidates are shown in red (click for larger version). Infrared (JHK) colors are obtained through the 2MASS Point Source Catalog by matching the right ascension and declination to within one arcsecond. Most of the giant candidates are color outliers in infrared. The blue lines indicate different proposed color cuts. The boundaries drawn in the g-r vs. r-i plot (bottom left) are the ones used in this sample.  COLOR SELECTION: g-r > 1.3, u-g > 1.8 + 0.9(g-r)
The same criteria based on the Kirkpatrick indices are applied to this sample to yield 33 giant candidates. Their Figure 7 is reproduced below (click for larger version). The giant candidates are shown in red.  The highest S/N candidates are shown below, ordered by their Hammer type (click for larger version).  Below are four color-color plots showing the 55 targets in this sample; the giant candidates are shown in red (click for larger version). The blue lines indicate different proposed color cuts. The boundaries drawn in the u-g vs. g-r plot (bottom right) are the ones used in this sample.  REFERENCES |