Optical selection of quasars: SDSS and LSST
Zeljko Ivezic, W. Niel Brandt, Xiaohui Fan, Chelsea L. MacLeod, Gordon T. Richards, Peter Yoachim
aa r X i v : . [ a s t r o - ph . I M ] D ec Multiwavelength AGN Surveys and StudiesProceedings IAU Symposium No. 304, 2014A. Mickaelian, F. Aharonian & D. Sanders, eds. c (cid:13) Optical selection of quasars: SDSS and LSST ˇZeljko Ivezi´c , W. Niel Brandt , Xiaohui Fan , Chelsea L. MacLeod ,Gordon T. Richards , and Peter Yoachim Department of Astronomy, University of Washington,Box 351580, Seattle, WA 98195-1580, USAemail: [email protected] Department of Astronomy and Astrophysics, The Pennsylvania State University,525 Davey Laboratory, University Park, PA 16802, USAemail: [email protected] Steward Observatory, University of Arizona,933 North Cherry Avenue, Tucson, AZ 85721, USAemail: [email protected] Department of Physics, U. S. Naval Academy,022 Chauvenet Hall, Annapolis, MD 21402, USAemail: [email protected] Department of Physics, Drexel University,3141 Chestnut Street, Philadelphia, PA 19104, USAemail: [email protected]
Abstract.
Over the last decade, quasar sample sizes have increased from several thousand toseveral hundred thousand, thanks mostly to SDSS imaging and spectroscopic surveys. LSST,the next-generation optical imaging survey, will provide hundreds of detections per object fora sample of more than ten million quasars with redshifts of up to about seven. We brieflyreview optical quasar selection techniques, with emphasis on methods based on colors, variabilityproperties and astrometric behavior.
Keywords. surveys, galaxies: active, quasars: general, stars: variables, stars: statistics
1. Introduction
The selection of large samples of active galactic nuclei, including quasars as their high-luminosity tail, is required in many astrophysical areas, such as galaxy evolution, blackhole growth, and the large-scale structure of the universe. The available quasar sampleshave increased by over two orders of magnitude in less than two decades and this rapidprogress is expected to continue (for example, simulations predict that LSST will delivera sample of about 10 million quasars; see Section 3.1). Here we briefly review opticalquasar selection methods based on SDSS data and discuss how they will be improvedwith the aid of more precise time-resolved photometry expected from LSST.
2. Finding quasars with SDSS
The similarity of unresolved quasars to stars in optical imaging surveys poses a diffi-culty in their identification. Quasars can be reliably identified by their spectra, and theSDSS provided a spectroscopically complete survey to i <
19 with ∼ ∼ α for-est analysis). These objects were selected as quasar candidates using colors measured1 ˇZeljko Ivezi´c et al. Figure 1.
The left panel illustrates the color-based selection method for quasar candidates.Essentially, the algorithm selects all sources whose colors place them outside the main stellarlocus, seen in the middle of the panel. Quasars with redshifts below 2.2 have distinctively blue u − g colors. The right panel shows the variation of volume number density of luminous quasars.The local maximum at about z = 2 . from the ugriz SDSS imaging data, and also using radio continuum 20 cm data fromthe FIRST survey (Becker et al. 1995). After SDSS collected enough multi-epoch imag-ing data, it was demonstrated that photometric variability-based selection is even moreefficient than color-based selection. These two selection methods are briefly summarizedhere. 2.1.
Color selection of quasar candidates
A small fraction of SDSS sources unresolved in imaging data were automatically targetedas quasar candidates for spectroscopic followup (Richards et al. 2002). The quasar tar-geting algorithm selects all point sources with 15 < i < . i is the apparent magnitudein the SDSS i band) whose colors place them outside the main stellar locus (it is a bitmore complicated than this – for details please see figure 1 in Richards et al. 2002). Forquasars at redshifts below about two, the most discriminatory color is the u − g color:compared to stars of the same visual color (e.g., the g − r color), quasars show an excessof ultraviolet flux in the u band and thus have bluer u − g colors (see the left panel inFigure 1); this method is often called the UV-excess selection. The colors of quasars athigher redshifts are typically significantly different from stellar colors. A thorough anal-ysis of the colors of quasars in the SDSS photometric system was presented in Richardset al. (2001). In addition, all unresolved sources from the same magnitude range thatare within 2 arcsec of a FIRST radio detection are also targeted (for analysis of theseradio quasars, see Kimball et al. 2011 and references therein). Some quasars were alsotargeted fortuitously via the algorithms for selecting galaxies (Strauss et al. 2002). Thecompleteness of the resulting quasar sample is above 90% (the confirmed fraction of allquasars within the adopted flux limits and within the surveyed area) and the efficiencyof color selection is about 65% (that is, about 35% of selected quasar candidates turnedout not to be quasars). This homogeneously selected sample spans a large redshift range(there are 56 quasars at redshifts beyond 5 in the Schneider et al. sample) and has en-abled numerous quasar studies. For example, the peak in the comoving volume number ptical selection of quasars Figure 2.
Structure function for quasar variability measured in the SDSS g band and in theobserver’s frame. The small dots with error bars represent the SDSS measurements, and thedata points at ∆ t > density of luminous quasars (dominated by type I objects) is now reliably and preciselydetermined (see the right panel in Figure 1), and the luminosity functions of quasarsand AGN galaxies (selected using emission line strengths) appear mutually consistentdespite grossly different selection procedures (Hao et al. 2005). We note that severalhundred candidate type II quasars (high-luminosity analogs of type 2 Seyfert galaxies)were found in the SDSS spectroscopic survey of galaxies (Zakamska et al. 2003, 2004).The relatively simplistic color-based selection algorithm employed by the original SDSSspectroscopic target selection pipeline has been significantly improved using modern datamining and machine learning methods. For example, Richards et al. (2009) introduced akernel density estimator and a non-parametric Bayesian classification method, and Bovyet al. (2012) introduced a Gaussian mixture model to recognize quasar candidates. Thesemethods have yielded large samples of candidates (of order a million) and with improvedcompleteness and efficiency tradeoff; Richards et al. (2009) reported efficiency of up to97% while maintaining fairly high completeness levels above 70%. More details aboutthe performance of these modified selection algorithms are available in Pˆaris et al. (2012)and Ross et al. (2012). 2.2. Variability selection of quasars
Quasars are variable sources with optical amplitudes of several tenths of a magnitude fortime scales longer than a few months, and this behavior can be used for their selection(Hawkins & Veron 1995; Ivezi´c et al. 2003). Sesar et al. (2007) showed using SDSSStripe 82 data (a ∼
300 deg equatorial region imaged about 60 times) that practicallyall quasars spectroscopically confirmed by SDSS are also variable in SDSS imaging data,and Koz lowski et al. (2010) demonstrated that variability can be used to separate quasarsfrom most variable stars even in the dense stellar environments of the Magellanic Clouds.A number of studies used SDSS light curves for close to 10,000 quasars from Stripe 82 ˇZeljko Ivezi´c et al. Figure 3.
Illustration of improvements in variability-based selection due to added time-scaleinformation. The solid line in the left panel shows the maximum completeness as a function ofefficiency, for quasars from SDSS Stripe 82 region with i <
19, achieved when using only theamplitude of the structure function for short time lags (ˆ σ = SF ∞ / √ τ ). The dashed line showshow efficiency improves when also including a cut in characteristic time scale ( τ ). The dottedline shows results when using the τ information alone. The right panel shows the distributionof spectroscopically confirmed quasars (solid contours, enclosing 40%, 60%, 75%, 85%, and 90%of data points) and stars (dashed contours) in the time scale ( τ ) vs. asymptotic variability(SF ∞ ) diagram for objects from SDSS Stripe 82 with i <
19. The dotted lines represent lines ofconstant ˆ σ . The two thick dashed lines correspond to ˆ σ = 10 − . and 10 − . mag yr − / . Thegray region represents the contaminating (stellar) region when selecting sources with ˆ σ < − . mag yr − / . When imposing a lower limit at τ = 10 . = 36 . to quantify the structure function for quasar variability (Ivezi´c et al. 2004; MacLeod etal. 2010; Schmidt et al. 2010; Butler & Bloom 2011; Palanque-Delabrouille et al. 2011).The structure function as a function of time lag ∆ t , SF(∆ t ), as defined in recent quasarstudies, is equal to the standard deviation of the distribution of the magnitude difference m ( t ) − m ( t ) evaluated at many different times t and t , such that time lag ∆ t = t − t (and divided by √ ∝ (∆ t ) α , then PSD ∝ /f (1+2 α ) .The power-law index α is a good discriminator between variable stars and quasars † (Schmidt et al. 2010; Butler & Bloom 2011). The key insight is that for time lags belowa year or so quasars have a much steeper structure function ( α ∼ . − .
5) than mostvariable stars. In other words, compared to their variability at a time lag of, say, one year,quasars vary by 1-2 orders of magnitude less at time lags below a month or so – this isnot true for the overwhelming majority of variable stars. For UV-excess selected objects,variability-based methods that utilize α select quasars with a completeness of 90% and apurity of 95% (Schmidt et al. 2010). Furthermore, this performance level is maintainedin the redshift range 2 . < z < † This parameter is also a good discriminator of various models for the origin of quasarvariability, see Kawaguchi et al. (1998). ptical selection of quasars τ , and the asymptotic (attime scales longer than τ ) root-mean-square variability, SF ∞ . Alternatively, the model(as well as its modifications) can be described via a covariance matrix (see the contribu-tion in these Proceedings by Ivezi´c & MacLeod). The predicted structure function for thedamped random walk process is SF(∆ t ) = SF ∞ [1 − exp( − ∆ t/τ )] / . At small time lags,SF(∆ t ) ∝ ∆ t / , and thus a damped random walk is equivalent to an ordinary randomwalk for ∆ t < τ (for a random walk, PSD ∝ /f ; the “damped” aspect manifests itselfas a flat PSD for ∆ t > τ ).The time span of SDSS data from Stripe 82 is sufficiently long to constrain τ for themajority of the ∼ only variability-based constraints (see the left panel in Figure 3), and 2%stellar contamination when a UV-excess constraint is added (see Figure 12 in MacLeodet al. 2011), while maintaining a completeness of 90%.
3. Finding quasars with LSST
The last decade has seen fascinating observational progress in optical imaging surveys.The SDSS dataset is currently being greatly extended by the ongoing surveys such asPan-STARRS (Kaiser et al. 2010) and the Dark Energy Survey (Flaugher 2008). TheLarge Synoptic Survey Telescope (LSST; for a brief overview see Ivezi´c et al. 2008) is themost ambitious survey currently planned in the visible band. LSST will extend the faintlimit of SDSS by about 5 magnitudes and will have unique survey capability in the fainttime domain. In particular, LSST will revolutionize our understanding of the growthof supermassive black holes with cosmic time, AGN fueling mechanisms, the detailedphysics of accretion disks, the contribution of AGN feedback to galaxy evolution, thecosmic dark ages, and gravitational lensing (for a detailed discussion, see Chapter 10 inthe LSST Science Book, Abell et al. 2009).While no massive spectroscopic followup of quasar candidates will be attempted as partof the LSST project, the time-resolved aspect of LSST photometric and astrometric datawill enable significant improvements in the completeness and efficiency of resulting quasarsamples compared to single-epoch measurements. We first briefly describe anticipatedLSST surveys and then discuss how these data will be used to construct quasar sampleswith up to about 10 million objects.3.1.
Brief overview of anticipated LSST surveys . The LSST design is driven by four main science themes: probing dark energy and darkmatter, taking an inventory of the Solar System, exploring the transient optical sky, andmapping the Milky Way. LSST will be a large, wide-field ground-based system designedto obtain multiple images covering the sky that is visible from Cerro Pach´on in NorthernChile. The project is scheduled to have first light around 2019 and the beginning ofsurvey operations by 2021. ˇZeljko Ivezi´c et al.The current baseline design, with an 8.4m (6.5m effective) primary mirror, a 9.6 deg field of view, and a 3.2 Gigapixel camera, will allow about 10,000 deg of sky to becovered using pairs of 15-second exposures in two photometric bands every three nightson average, with typical 5 σ depth for point sources of r ∼ .
5. The system is designedto yield high image quality as well as superb astrometric and photometric accuracy. Thesurvey area will include 30,000 deg with δ < +34 . ◦ , and will be imaged multiple timesin six bands, ugrizy , covering the wavelength range 320–1050 nm. About 90% of theobserving time will be devoted to a deep-wide-fast survey mode which will observe an18,000 deg region over 800 times (summed over all six bands) during the anticipated10 years of operations, and yield a coadded map to r ∼ .
5. These data will resultin databases including about 20 billion galaxies and a similar number of stars, and willserve the majority of science programs. The remaining 10% of the observing time will beallocated to special programs such as a Very Deep and Fast time-domain survey. Moredetails about various science programs that will be enabled by LSST data can be foundin the LSST Science Book (Abell et al. 2009).3.2.
Color and variability selection of quasars
The existing selection algorithms based on colors and photometric variability developedwith the aid of SDSS data will be applicable to LSST data, too. Although LSST willnot obtain simultaneous multi-band photometry like SDSS did, the averaging of manyobservations (ranging from about 50-60 in the u band to 180-190 in the r and i bands) willresult in sufficiently precise color measurements to recognize easily color offsets from themain stellar locus. The addition of variability information will boost the sample efficiencyto levels comparable to those obtained for spectroscopic samples. Two additional selectionmethods, enabled by the multi-epoch LSST imaging and described below, will furtherimprove the resulting samples.Detailed simulations of the quasar luminosity function and light curves, the LSST ob-serving cadence, and the LSST photometric error distribution (Abell et al. 2009; MacLeodet al. 2011; Palanque-Delabrouille et al. 2013) show that the LSST quasar sample willinclude close to 10 million objects, and will be complete for M < −
23 objects (a formalabsolute magnitude definition cutoff of quasars) to redshifts beyond 3. Notably, LSSTwill discover about 1000 quasars with redshifts beyond 7 (using the z -band drop-outtechnique) which will represent a valuable sample for studying the epoch of reionization.3.3. Additional constraints: proper motion and differential chromatic refraction
In addition to photometry, astrometric measurements can help to distinguish quasarsfrom stars. First, measurable proper motion will reject about 2/3 of all stars, even beforeany color or photometric variability criteria are applied. LSST proper motion errors willbe 0.5 mas/yr for sources with r = 23 and 1.0 mas/yr for r = 24 (Ivezi´c et al. 2008).Simulations of the proper motion distribution for Milky Way stars (for model description,see Ivezi´c, Beers & Juri´c 2012) indicate that over ∼
80% of stars with r <
23 and over ∼
70% of stars with r <
24 will have proper motions three times larger than expectedmeasurement errors (with very little dependence on Galactic coordinates). These highfractions of rejecteable stars are also expected for various stellar subpopulations, such asbrown dwarfs (contaminants for very high-redshift quasar candidates) and white dwarfs(contaminants of z < . ptical selection of quasars u and g bands, and between the g and r bands, can be up to about 20 mas (dependingon redshift), even for moderate airmass ( ∼ r = 22, about 5 mas for objects with r ∼
23, and 10-15 mas for objects with r ∼ References
Abell, P.A., Allison, J., Anderson, S.F., et al. 2009, ArXiv:
Becker, R.H., White, R.L. & Helfand, D.J. 1995,
ApJ , 450, 559Bovy, J., Myers, A.D., Hennawi, J., et al. 2012,
ApJ , 749, 41Butler, N.R. & Bloom, J.S. 2011 AJ , 141, 93Flaugher, B. 2008, In A Decade of Dark Energy: Spring Symposium, Proceedings of the confer-ences held May 5-8, 2008 in Baltimore, Maryland. (USA). Ed. by N. Pirzkal & H. Ferguson.
Hao, L., Strauss, M.A., Fan, X., et al. 2005, AJ , 129, 1795Hawkins, M.R.S. & Veron, P. 1995, MNRAS , 275, 1102Ivezi´c, ˇZ., Lupton, R.H., Johnston, D.E. et al. 2003, ArXiv: astro-ph/0310566
Ivezi´c, ˇZ., Lupton, R.H., Juri´c, M. et al. 2004, ArXiv: astro-ph/0404487
Ivezi´c, ˇZ., Tyson, J.A., Acosta, E., et al. 2008, ArXiv:
Ivezi´c, ˇZ., Beers, T.C. & Juri´c, M. 2012,
ARAA , 50, 251Kaczmarczik, M.C., Richards, G.T., Mehta, S.S. & Schlegel, D.J. 2009, AJ , 138, 19Kaiser, N., Burgett, W., Chambers, K., et al. 2010, Society of Photo-Optical InstrumentationEngineers (SPIE) Conference Series , vol. 7733Kawaguchi, T., Mineshige, S., Umemura, M., & Turner, E. L. 1998,
ApJ , 504, 671Kelly, B.C., Bechtold, J. & Siemiginowska, A. 2009,
ApJ , 698, 895Kimball, A., Ivezi´c, ˇZ., Wiita, P.J. & Schneider, D.P. 2011, AJ , 141, 182Koz lowski, S., Kochanek, C.S., Udalski, A., et al. 2010, ApJ , 708, 927MacLeod, C.L., Ivezi´c, ˇZ., Kochanek, C.S., et al. 2010,
ApJ , 721, 1014MacLeod, C.L., Brooks, K., Ivezi´c, ˇZ., et al. 2011,
ApJ , 728, 26MacLeod, C.L., Ivezi´c, ˇZ., Sesar, B., et al. 2012,
ApJ , 753, 106Palanque-Delabrouille, N., Y´eche, Ch., Myers, A.D., et al. 2011,
A&A , 530, 122Palanque-Delabrouille, N., Magneville, Ch., Y´eche, Ch., et al. 2013,
A&A , 551, 29Pˆaris, I., Petitjean, P., Aubourg, ´E., et al. 2012,
A&A , 548, 66Richards, G.T., Fan, X., Schneider, D.P., et al. 2001, AJ , 121, 2308 (see also arXiv:1311.4870)Richards, G.T., Fan, X., Newberg, H.J., et al. 2002, AJ , 123, 2945Richards, G.T., Strauss, M.A., Fan, X., et al. 2006, AJ , 131, 2766Richards, G.T., Myers, A.D., Gray, A.G., et al. 2009, ApJS , 180, 67Ross, N.P., Myers, A.D., Sheldon, E.S., et al. 2012