What's New in DR9?

Prior data releases have not changed. Data from DR8 is available on this site, and data from DR7 is available from the original SDSS.

New spectra: Baryon Oscillation Spectroscopic Survey (BOSS)

SDSS Data Release 9 includes the first public release of spectroscopic data from the SDSS-III's Baryon Oscillation Spectroscopic Survey (BOSS), including 831,000 spectra of galaxies, quasars and stars over 3,300 square degrees.

BOSS is an ongoing survey whose primary science driver is to measure the baryon acoustic oscillation (BAO) feature in the clustering of galaxies and the Lyman-α forest. By the end of the survey, BOSS will measure spectra of 1.5 million galaxies to z~0.7, as well as for 160,000 quasars with redshifts between 2.2 and 3.5. Data Release 9 contains the first two years of BOSS spectroscopy (all BOSS observations between 2009 December 5 and 2011 July), about 20% of the final BOSS dataset.

DR9 contains 540,000 spectroscopically-confirmed galaxies (with a median redshift of 0.52), 100,000 quasars most with redshift z > 2.2, and 116,000 stars, all selected over 3300 deg² of sky.

New spectrographs

To prepare for BOSS observations, in 2009 the original SDSS spectrographs were completely revamped into the BOSS spectrographs. In particular:

The spectrographs are now fed by 1000 fibers of 2" entrance aperture, rather than 640 fibers of 3" aperture
The CCDs were replaced with new devices with higher throughput and smaller pixels
The gratings were replaced with volume-phase holographic gratings
Much of the optics of the cameras were recoated or replaced

These changes significantly improved the throughput of the spectrographs and increased the wavelength coverage to 3600-10,400 Å. The spectral resolution varies from ~ 1300 at 3600 Å to 3000 at 10,000 Å.

New target selection algorithms

The BOSS galaxy target selection algorithm is similar in spirit to the Luminous Red Galaxy target selection used in SDSS-I and II, in that it selects the reddest and most luminous objects at each redshift. However, the new algorithms target objects with a larger range of luminosities and colors than the old LRG algorithm. In addition, the BOSS target selection algorithms are focused on higher redshifts, aiming to measure the BAO signal at a median redshift of z~0.6.

The SDSS-I/II quasar target selection algorithm aimed for completeness at all redshifts below z=5.4. In contrast, the goal of BOSS is to measure the large-scale structure of the Lyman-α forest, and thus focuses on the redshift range between 2.2 and 3.5, with a magnitude limit of g=21.8. This range of redshifts and magnitudes is challenging since quasars with redshift z~2.7 have colors very similar to those of F stars, which means that the quasar spectroscopic sample has extensive contamination from stars.

In addition, about 3.5 percent of the fibers are ancillary targets, chosen to accommodate a variety of science goals, ranging from stars in the halo to distant quasars. Almost 30,000 objects have spectra repeated from SDSS-I/II; the typical such object has a spectroscopic S/N a factor of almost 2 higher in BOSS than in SDSS-I/II.

New processing pipelines

These spectroscopic data are processed by a pipeline that is a direct extension of that used by SDSS-I/II, and therefore the format of the spectroscopic data is very similar to the format in SDSS-I/II. Therefore, the tools to access spectra are similar to those used in prior data releases.

For galaxies, the BOSS redshift pipeline now includes the following Portsmouth and Wisconsin stellar population model algorithms, in addition to the MPA-JHU value-added catalogs released in DR8:

Portsmouth spectro-photometric stellar masses, ages and stellar formation rates based on the star-forming or passive stellar population models of Maraston et al. (2005), Maraston et al. (2009), computed from the best-fit SED model of Maraston et al. (2006).
Portsmouth Emission-line fluxes and equivalent widths, stellar and gas kinematics, based on the stellar population synthesis models of Maraston & Stromback (2011) and Thomas, Maraston & Johansson (2011) applied to BOSS spectra.
Wisconsin Stellar masses and velocity dispersions are derived from a BOSS spectrum principal component analysis of Chen et al. (2012) using the stellar population models of Bruzual & Charlot (2003).

Stellar Parameters: Improvements to SSPP

For stars, a new version of the SEGUE Stellar Parameters Pipeline (SSPP) has been run, with improved temperature estimates for cool stars (T_eff < 5000 K) thanks to InfraRed Flux Method (IRFM) implemented in the SSPP.

A grid of synthetic spectra with microturbulence depending on surface gravity gives improved estimates of metallicity for metal-rich stars ([Fe/H] > -0.5). In comparison with values from DR8, the mean DR9 T_eff are larger by ~60 K, the mean log g values are lower by 0.2 dex, and the [Fe/H] have not changed significantly. The new SSPP code has been run on all stars in the SEGUE-1 and SEGUE-2 surveys, as well as those in the SDSS-I/II Legacy survey. However, the new SSPP code has not been run on stars that were observed as part of the BOSS survey.

New SEGUE Value-Added Catalogs

SEGUE has released three important value-added catalogs as part of DR9. First, the SSPP automated measurements of [α/Fe] from Lee et al. 2011 are now available. Second, we have released the target selection weights from Schlesinger et al. 2012. This value-added catalog explains the different observational biases in the SEGUE G- and K-dwarf sample and how to correct for them. Finally, SEGUE also provides a catalog of Duplicate Spectra.

Improvements to Astrometric Calibration

The reprocessing of the SDSS imaging data in DR8 unfortunately introduced several errors in the astrometric calibration. These have been corrected in DR9, and all relevant files updated. The principal improvement is fixing a systematic 250 mas error in coordinates, in regions northwards of declination +41 degrees. We now also properly apply color terms to the astrometry, the principal effect of which is to improve proper motion estimates for a number of objects. SDSS Object IDs have not changed as a result as a the improved astrometry, although the IAU identifiers (SDSSJ...) may have changed.

Significant Changes to Data From DR8 and DR7

The spectra are now available in the Science Archive Server (SAS) spectra search in a packaged file, one per spectrum. The spectra were created by co-adding individual 15-minute exposures. In DR9, those individual exposures are also available as FITS files for both BOSS and SDSS spectrograph data (see the Spectra per-object files download page for instructions).
As of DR9, the imaging mask tables are now available in the Catalog Archive Server (CAS). These mask tables track the geometrical regions surrounding bright stars and their bleed trails, as well as the seeing in each subregion of each field. The Navigate tool allows you to see masks in Navigate tool, and why you might want to see the masks
Starting in DR9, multi-epoch image information is now included in the sweep files of the photometry, including coadded catalog fluxes and variability information from multiple observations of the same object.
In DR8, we had changed the default algorithm for matching imaging and spectra to match each spectrum to the object contributing most to the r-band light at the location of the spectrum (a flux-based match), rather than simply searching for the object with the closest center (a position-based match). For DR8, bestObjID gave the flux-based match and origObjID gave the position-based match. For DR9, we have changed the default algorithm back to position-based matching, so that bestObjID now gives the position-based match (as in DR7 and previous) whereas fluxObjID gives the flux-based match. Very few objects are affected by this change, but some spectra have bestObjID that now differ from their DR8 value. See the matching algorithm documentation for a more detailed description.
In DR8 and later, several changes to the format were made relative to DR7 and earlier:
- The isophotal parameters in the photometry were dropped, since we consider them untrustworthy in most cases.
- Because DR9 contains target lists of a number of sorts (Legacy, SEGUE, and BOSS, as well as a number of special programs), we no longer have a target flag defined for every object. Thus, features such as the SkyServer Navigate tool no longer will show every potential spectroscopic target, only those with actual spectra.
- In particular, the primTarget and secTarget flags are not reset for the new photometric reductions in DR8 and later, and are not given for each photometric object. In addition, in the spectroscopic catalogs these flags are deprecated because their meanings are overloaded, as describe in the spectroscopic targeting documentation.
- The objID values for photometric objects changed between DR7 and DR8 and later, because the new photometric reductions mean that new objects have appeared, and others have disappeared, and the rerun number is different.
- The specClass spectroscopic classification in DR7 has been replaced by class and subclass.