The SDSS-III Baryon Oscillation Spectroscopic Survey: Quasar Target Selection for Data Release Nine

TL;DR

This paper documents the evolution of quasar target selection for SDSS-III BOSS during its first two years, detailing CORE (uniform, single-epoch SDSS photometry) and BONUS (supplementary data-driven targets) strategies. It compares KDE, Likelihood, NN, and XDQSO methods, showing XDQSO becoming CORE and NN-Combinator underpinning BONUS, with ancillary UKIDSS and GALEX data enhancing yields. The Year One results include 11,263 new $z>2.2$ quasars over 878 deg^2 and a global mean targeting efficiency around a few tens of percent, informing the DR9 quasar catalog and BAO Lyα forest studies. The work highlights the critical balance between maximizing high-redshift quasar densities for cosmology and maintaining a uniformly selected CORE sample for statistical quasar science, while outlining future prospects such as variability-based selection and multi-wavelength data assimilation.

Abstract

The SDSS-III Baryon Oscillation Spectroscopic Survey (BOSS), a five-year spectroscopic survey of 10,000 deg^2, achieved first light in late 2009. One of the key goals of BOSS is to measure the signature of baryon acoustic oscillations in the distribution of Ly-alpha absorption from the spectra of a sample of ~150,000 z>2.2 quasars. Along with measuring the angular diameter distance at z\approx2.5, BOSS will provide the first direct measurement of the expansion rate of the Universe at z > 2. One of the biggest challenges in achieving this goal is an efficient target selection algorithm for quasars over 2.2 < z < 3.5, where their colors overlap those of stars. During the first year of the BOSS survey, quasar target selection methods were developed and tested to meet the requirement of delivering at least 15 quasars deg^-2 in this redshift range, out of 40 targets deg^-2. To achieve these surface densities, the magnitude limit of the quasar targets was set at g <= 22.0 or r<=21.85. While detection of the BAO signature in the Ly-alpha absorption in quasar spectra does not require a uniform target selection, many other astrophysical studies do. We therefore defined a uniformly-selected subsample of 20 targets deg^-2, for which the selection efficiency is just over 50%. This "CORE" subsample will be fixed for Years Two through Five of the survey. In this paper we describe the evolution and implementation of the BOSS quasar target selection algorithms during the first two years of BOSS operations. We analyze the spectra obtained during the first year. 11,263 new z>2.2 quasars were spectroscopically confirmed by BOSS. Our current algorithms select an average of 15 z > 2.2 quasars deg^-2 from 40 targets deg^-2 using single-epoch SDSS imaging. Multi-epoch optical data and data at other wavelengths can further improve the efficiency and completeness of BOSS quasar target selection. [Abridged]

Figures (26)

• Figure 1: Color-color diagrams of point sources drawn from 7 deg$^{2}$ (the BOSS spectrograph field of view) in the SDSS photometric database. (Left) 2,400 objects with $18.0<g<19.0$, and (Right) 7,000 objects with $21.0<g<22.0$. Most of the objects shown are stars; low-redshift ($z < 2.2$) quasars lie preferentially in the region $u-g<0.6, g-r>0$ where very few stars are found. At $z > 2.2$, quasars become systematically redder in $u-g$ as the Ly$\alpha$ forest moves into the $u$-band and Ly$\alpha$ emission moves into $g$. At $z \sim 2.7$, quasars have colors similar to those of blue horizontal branch (BHB) stars. The larger photometric errors at faint magnitudes broaden the stellar locus considerably (especially in the $u$-band for the reddest stars, which gives rise to the spread at $g-r \sim 1.5$), illustrating the challenges involved in selecting faint objects by their colors. Tracks for the quasar locus, as presented in Bovy et al. (2011b, in prep.) are also shown, with the corresponding redshift given by the color-bar legend. Approximate surface densities are quoted, and stellar classifications are given as a guide.
• Figure 2: Flowchart for the BOSS quasar target selection, as implemented from the beginning of the second year of BOSS observations. The various broad categories of targets, including CORE, BONUS, KNOWN objects, and those detected by the FIRST survey, are indicated, and are described in detail in Section \ref{['sec:methods']}; SUPPZ refers to a small number of lower-redshift objects targeted to study the effects of metal line absorption (§ \ref{['sec:chunks14to19']}). The flowchart for the first year of BOSS target selection is given in Appendix \ref{['sec:flowchart_YearOne']}. The CORE sample is fixed for DR9 and the remainder of the BOSS. Objects which satisfy the XDQSO probability cut of $P(XDQSO)>0.424$ are selected as CORE, and the QSO_CORE_MAIN target flag bit is set. CORE selection is based on single-epoch SDSS photometry, but other selections use multi-epoch photometry where it is available (e.g., in regions where SDSS imaging stripes overlap).
• Figure 3: Redshift versus $(u-g)$ color for BOSS FIRST quasar targets. Objects from the BOSS commissioning were either targeted by FIRST, and also a optical selection, (black) crosses, or, they were targeted only as FIRST sources, (red) squares. These early findings inspired our $(u-g)>0.4$ cut to minimize contamination from $z < 2.2$ quasars.
• Figure 4: The BOSS quasar target surface density in Equatorial coordinates in the NGC, from a run of the BOSS QTS with a selection made by combining the three Year One methods, KDE, Likelihood and NN, in such a way that the average target density over the full given NGC area was $\sim60$ quasar targets deg$^{-2}$. The color indicates the local number density of targets per square degree. The tidal stream of the Sagittarius dwarf spheroidal galaxy is prominent in the region $180^\circ<\alpha<240^\circ$, and $0^\circ< \delta <+15^\circ$. The white lines show the "Blind Test Area", described in § \ref{['sec:blind_test']}.
• Figure 5: The targeting footprint for the SDSS-III:BOSS Ly$\alpha$ forest/Quasar Survey. The various chunks are indicated by different colors. Chunks 16, 17 and 18 lie within the footprint of Chunk 15. The full targeting footprint is 10,200 deg$^{2}$, with a total of $\approx$430,000 tiled targets. Roughly $\sim$150,000 of these targets will have spectra by the end of Year Two observations. The global Year One quasar target density is 60.4 targets deg$^{-2}$, and the mean target density over all chunks shown is 47.9 targets deg$^{-2}$. The dashed line is at Galactic latitude $b =25^{\circ}$.