The SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Quasar Target Selection

TL;DR

The paper presents the quasar targeting strategy for SDSS-IV/eBOSS, designed to deliver a homogeneous CORE sample for robust BAO measurements at z~1.5 and a Lyα quasar sample to probe BAO via the Lyα forest at z~2.5. CORE targets are selected through XDQSOz optical probabilities and WISE mid-IR colors, supplemented by a rare but powerful variability channel for Lyα targets; the approach balances uniformity with broader completeness to maximize scientific return. Results from the SEQUELS pilot demonstrate achievable densities (>58 deg^-2 CORE, with ~70 deg^-2 predicted) and a viable path to >500,000 new quasars, while extensive homogeneity tests show systematics can be modeled or masked to preserve clustering analyses. The study confirms eBOSS will significantly extend quasar-based cosmology, delivering precise BAO constraints and enabling a broad program of quasar science across SDSS footprints.

Abstract

As part of the Sloan Digital Sky Survey IV the extended Baryon Oscillation Spectroscopic Survey (eBOSS) will improve measurements of the cosmological distance scale by applying the Baryon Acoustic Oscillation (BAO) method to quasar samples. eBOSS will adopt two approaches to target quasars over 7500 sq. deg. First, a "CORE" quasar sample will combine optical selection in ugriz using a likelihood-based routine called XDQSOz, with a mid-IR-optical color-cut. eBOSS CORE selection (to g < 22 OR r < 22) should return ~ 70 quasars per sq. deg. at redshifts 0.9 < z < 2.2 and ~7 z > 2.1 quasars per sq. deg. Second, a selection based on variability in multi-epoch imaging from the Palomar Transient Factory should recover an additional ~3-4 z > 2.1 quasars per sq. deg. to g < 22.5. A linear model of how imaging systematics affect target density recovers the angular distribution of eBOSS CORE quasars over 96.7% (76.7%) of the SDSS North (South) Galactic Cap area. The eBOSS CORE quasar sample should thus be sufficiently dense and homogeneous over 0.9 < z < 2.2 to yield the first few-percent-level BAO constraint near z~1.5. eBOSS quasars at z > 2.1 will be used to improve BAO measurements in the Lyman-alpha Forest. Beyond its key cosmological goals, eBOSS should be the next-generation quasar survey, comprising > 500,000 new quasars and > 500,000 uniformly selected spectroscopically confirmed 0.9 < z < 2.2 quasars. At the conclusion of eBOSS, the SDSS will have provided unique spectra of over 800,000 quasars.

Figures (21)

• Figure 1: Flowchart depicting eBOSS quasar target selection. Red boxes represent sources of input information such as imaging (see §\ref{['sec:imaging']}) or catalogs of known objects. Black boxes depict cuts that are made to the input sources as part of the target selection algorithm (see §\ref{['sec:qts']}). Blue boxes depict output target selection bits (see §\ref{['sec:bits']}). The Boolean terms in purple describe how the four bits produced by matching to previous spectra are combined to set the DO_NOT_OBSERVE bit (see §\ref{['sec:donotobs']}). The dashed blue arrow indicates that QSO_REOBS targets are always reobserved, regardless of the value of DO_NOT_OBSERVE. The sample of known objects undergoes the CORE flag and magnitude cuts rather than the PTF magnitude cuts. Consequently, PTF selection could re-target previously known objects with bad IMAGE_STATUS and/or with $22 < g < 22.5$.
• Figure 2: The position of XDQSOz-selected PQSO($z > 0.9) > 0.2$ quasars in $ugriz$ optical color space (using PSF magnitudes). Black points depict $r < 19$PRIMARY point sources from a randomly chosen SDSS imaging run (5225). The $r < 19$ limit is chosen in order to illustrate the position of the stellar locus in SDSS filters; at fainter limits the locus widens considerably Bov11a. Spectroscopically confirmed PQSO($z > 0.9) > 0.2$ quasars from BOSS (DR10; squares) are plotted as a function of redshift, from $z=0.9$ to $z=4.15$ in bins of $\Delta z=0.65$. The error bars indicate the 1$\sigma$ scatter.
• Figure 3: The cumulative sky density of quasars and targets as a function of $z > 0.9$XDQSOz probability. The upper curves represent all quasars (red), and $0.9 < z < 2.2$ quasars (blue), from the CFHTLS-W3 test program. These curves yield an estimate of the completeness of eBOSS to quasars for various PQSO($z > 0.9)$ constraints. Grey contours illustrate the (Poisson) errors. The lowest curve represents all sources from SDSS imaging in the CFHTLS-W3 test region (magenta). This curve yields an estimate of the necessary fiber budget for eBOSS. A quantitative example of how to use the curves to predict quasar and target densities is provided in §\ref{['sec:xdqsozcuts']}. The vertical lines depict the adopted cut for eBOSS (after also applying an optical-IR color cut; see §\ref{['sec:combcuts']}), the cut for the eBOSS requirement of 58 deg$^{-2}$$0.9 < z < 2.2$ quasars, and the cut to assign $< 115$ deg$^{-2}$eBOSS fibers (the maximum assignable; see §\ref{['sec:SRD']}). All samples depicted have been limited to SDSSPRIMARY point sources with FIBER2MAG of $i > 17$ and de-extincted PSF magnitudes of $g < 22$ OR $r < 22$ (the initial cuts for the eBOSS CORE).
• Figure 4: The redshift distribution of spectroscopically confirmed quasars from the CFHTLS-W3 test program. The distributions that peak in the $0.9 < z < 2.2$ range are the redshift Probability Density Functions (PDFs). The distributions that climb to 1 near $z\sim3.5$ are cumulative. The distributions for three different cuts on the $z > 0.9$XDQSOz probability are depicted; PQSO($z > 0.9) > 0.8$ (orange, solid), PQSO($z > 0.9) > 0.5$ (blue, dotted), and PQSO($z > 0.9) > 0.2$ (green, dashed).
• Figure 5: The optical-IR cut (applied to PSF magnitudes) used to define eBOSS CORE quasar targets. The green line depicts the color cut in SDSS ($f_{g} + 0.8f_{r}+0.6f_{i})/2.4$ and WISE ($f_{W1} + 0.5f_{W2})/1.5$ stacks versus $g-i$ that was used to target quasars as part of the CFHTLS-W3 test program. Quasars of interest to eBOSS ($z~\hbox{$\sim$}\hbox{$<$}$$\,3.5$) generally occupy the region above this line; the stellar locus is a dense region in the lower part of the plot. Black points depict objects with anyeBOSS targeting bit set (see §\ref{['sec:bits']}) from a randomly chosen SDSS imaging run (5225) limited to $g < 22$. Spectroscopically confirmed quasars from BOSS (DR10; squares) are plotted as a function of redshift, from $z=0.9$ to $z=4.15$ in bins of $\Delta z=0.65$. The error bar indicates the 1$\sigma$ scatter.