1: \documentclass[12pt,preprint]{aastex}
2: %\documentclass[manuscript]{aastex}
3: \usepackage{natbib,graphics,graphicx}
4:
5: \slugcomment{12 January 2009 revision; accepted by AJ}
6:
7: \shorttitle{8-D Bayesian Quasar Selection}
8: \shortauthors{Richards et al.}
9:
10: \begin{document}
11:
12: \title{Eight-Dimensional Mid-Infrared/Optical Bayesian Quasar Selection}
13:
14: \author{
15: Gordon T. Richards,\altaffilmark{1,2}
16: Rajesh P. Deo,\altaffilmark{1}
17: Mark Lacy,\altaffilmark{3}
18: Adam D. Myers,\altaffilmark{4}
19: Robert C. Nichol,\altaffilmark{5}
20: Nadia L. Zakamska,\altaffilmark{6}
21: Robert J. Brunner,\altaffilmark{4}
22: W. N. Brandt,\altaffilmark{7}
23: Alexander G. Gray,\altaffilmark{8}
24: John K. Parejko,\altaffilmark{1}
25: Andrew Ptak,\altaffilmark{9}
26: Donald P. Schneider,\altaffilmark{7}
27: Lisa J. Storrie-Lombardi,\altaffilmark{3}
28: and Alexander S. Szalay\altaffilmark{9}
29: }
30:
31: \altaffiltext{1}{Department of Physics, Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104.}
32: \altaffiltext{2}{Alfred P. Sloan Research Fellow.}
33: \altaffiltext{3}{Spitzer Science Center, Caltech, Mail Code 220-6, Pasadena, CA 91125.}
34: \altaffiltext{4}{Department of Astronomy, University of Illinois at Urbana-Champaign, 1002 West Green Street, Urbana, IL 61801-3080.}
35: \altaffiltext{5}{Institute of Cosmology and Gravitation, Mercantile House, Hampshire Terrace, University of Portsmouth, Portsmouth, PO1 2EG, UK.}
36: \altaffiltext{6}{John N.\ Bahcall Fellow, Institute for Advanced Study, Einstein Drive, Princeton, NJ 08540.}
37: \altaffiltext{7}{Department of Astronomy and Astrophysics, The Pennsylvania State University, 525 Davey Laboratory, University Park, PA 16802.}
38: \altaffiltext{8}{College of Computing, Georgia Institute of Technology, 266 Ferst Drive, Atlanta, GA 30332.}
39: \altaffiltext{9}{Department of Physics and Astronomy, The Johns Hopkins University, 3400 North Charles Street, Baltimore, MD 21218-2686.}
40:
41: \begin{abstract}
42:
43: We explore the multidimensional, multiwavelength selection of quasars
44: from mid-IR (MIR) plus optical data, specifically from {\em
45: Spitzer}-IRAC and the Sloan Digital Sky Survey (SDSS).
46: Traditionally quasar selection relies on cuts in 2-D color space
47: despite the fact that most modern surveys (optical and infrared) are
48: done in more than 3 bandpasses. In this paper we apply modern
49: statistical techniques to combined {\em Spitzer} MIR and SDSS optical
50: data, allowing up to 8-D color selection of quasars. Using a Bayesian
51: selection method, we catalog 5546 quasar candidates to an $8.0\mu$m
52: depth of $56\mu$Jy over an area of $\sim24$ deg$^2$. Roughly 70\% of
53: these candidates are not identified by applying the same Bayesian
54: algorithm to 4-color SDSS optical data alone. The 8-D optical+MIR
55: selection on this data set recovers $97.7$\% of known type 1 quasars
56: in this area and greatly improves the effectiveness of identifying
57: $3.5<z<5$ quasars which are challenging to identify (without
58: considerable contamination) using MIR data alone. We demonstrate
59: that, even using only the two shortest wavelength IRAC bandpasses (3.6
60: and 4.5$\mu$m),
61: %that will be available during the upcoming {\em Spitzer} warm mission
62: it is possible to use our Bayesian techniques to select quasars with
63: 97\% completeness and as little as 10\% contamination (as compared to
64: $\sim$60\% contamination using colors cuts alone). We compute
65: photometric redshifts for our sample; comparison with known objects
66: suggests a photometric redshift accuracy of 93.6\% ($\Delta z\pm0.3$),
67: remaining roughly constant when the two reddest MIR bands are
68: excluded. Despite the fact that our methods are designed to find type
69: 1 (unobscured) quasars, as many as 1200 of the objects are type 2
70: (obscured) quasar candidates. Coupling deep optical imaging data,
71: %(e.g., the SDSS Southern Equatorial Stripe, $g \sim 24$ mag)
72: with deep mid-IR data
73: %2-band {\it Spitzer}-IRAC data
74: %of properly matched depth
75: could enable selection of quasars in significant numbers past the peak
76: of the quasar luminosity function (QLF) to at least $z\sim 4$. Such a
77: sample would constrain the shape of the QLF both above and below the
78: break luminosity ($L^*_Q$) and enable quasar clustering studies over
79: the largest range of redshift and luminosity to date, yielding
80: significant gains in our understanding of the physics of quasars and
81: their contribution to galaxy evolution.
82:
83:
84: %Thus, as the SDSS Southern Equatorial Stripe has optical flux limits
85: %($g\sim24$) comparable to those afforded by moderately deep IRAC data,
86: %coupling it 2-band {\em Spitzer}-IRAC data will enable quasars to be
87: %surveyed in significant numbers past the peak of the quasar luminosity
88: %function (QLF) to at least z$\sim$4.
89:
90:
91: \end{abstract}
92:
93: \keywords{catalogs --- quasars: general --- methods: statistical --- infrared: galaxies}
94:
95: \section{Introduction}
96:
97: Enormous progress has been made in the last decade on understanding
98: the nature of active galactic nuclei (AGNs) and their role in the
99: lifecycle of galaxies. The co-evolutionary behaviour of black holes
100: and spheroids, as implied by the $M-\sigma$ relation
101: \citep[e.g.,][]{fm00,gbb+00,tgb+02}, suggests that most massive
102: galaxies hosted an AGN at some time. Indeed, energy injection from
103: AGN through so-called ``feedback'' mechanisms
104: \citep[e.g.,][]{sr98,fab99,beg04,hhc+06} may be the linchpin
105: connecting the blue, star forming and massive red, dead elliptical
106: galaxies, whether through direct energy input
107: \citep[e.g.,][]{bbm+06,csw+06}, or as a somewhat more coincident
108: product of the main quenching mechanism (e.g., major mergers could
109: drive both AGN and quenching; \citealt{kh00,hhc+08}).
110:
111: Although feedback provides a key clue as to how galaxies evolve, there
112: is a great deal of degeneracy in quenching prescriptions
113: \citep[e.g.,][]{cdd+06}. A promising avenue is constraining feedback
114: models by examining the luminosity dependence to AGN clustering in
115: combination with the AGN luminosity function (e.g., \citealt{lhc+06}).
116: However, current large quasar surveys typically track only the peak of
117: the quasar luminosity function (QLF) at $z < 2$ and only probe the
118: brightest quasars at $z > 3$, for example the Sloan Digital Sky Survey
119: (SDSS; \citealt{yaa+00}) and the 2-Degree Field Quasar Survey (2QZ;
120: \citealt{csb+04}). Consequently, quasar clustering measurements
121: \citep[e.g.,][]{pn06,mbn+07,dsc+08} detect little luminosity
122: dependence at $z < 2$, and provide incomplete constraints at $z > 3$.
123:
124: If we wish to have a complete picture of the true role of feedback in
125: quenching star formation and black hole growth in galaxies, then an
126: important tool is to probe AGN (in statistically significant numbers)
127: to luminosities below $L^*_Q$ at high redshift and below host galaxy
128: depths at low redshift. In the long-term, the next generation of
129: survey facilities such as LSST \citep[e.g.,][]{lsst08}, Pan-STARRS
130: \citep{kab+02}, DES \citep{des05}, and VST/VISTA \citep{anp+07},
131: should produce sufficiently deep optical and near-IR photometry with
132: which to address this goal. In the near-term, any sufficiently
133: large-area survey of AGN that probes a wider luminosity range will
134: blaze important observational and theoretical trails. This is
135: particularly true of a large-area survey that contains the necessary
136: color baseline (typically mid-IR through optical/UV) with which to
137: simultaneously study AGN host galaxies at $z\lesssim1$ and to
138: characterize quasar photometric redshifts
139: \citep[e.g.,][]{rws+01,rbo+08} out to $z\sim3$ and beyond. As such,
140: herein we describe a novel method for efficient selection of AGN from
141: the combination of optical and mid-IR data that allows one to probe to
142: the depths (and areas) that are required to compile the sort of data
143: set that we have highlighted. As the existing overlap between SDSS
144: and mid-IR imaging
145: %(e.g., from {\em Spitzer}-IRAC during the
146: %``warm'' mission and from the upcoming
147: %WISE\footnote{http://www.astro.ucla.edu/~wright/WISE/} satellite)
148: grows, so too will our ability to collect statistically significant
149: samples of faint high-$z$ quasars that are needed to probe the
150: influence of quasar feedback on galaxy evolution.
151:
152: %Our results suggests that the {\em Spitzer} Warm Mission is by far the
153: %next best opportunity to compile a significant dataset with which to
154: %the study the AGN-galaxy connection.
155:
156: In principle, differentiating quasars from stars using mid-IR (MIR)
157: colors is straightforward. At MIR wavelengths, stars (excepting the
158: coolest brown dwarfs) are well-described by the Rayleigh-Jeans portion
159: of a blackbody spectrum ($f_{\nu}\propto\nu^2$), resulting in quite
160: blue colors, whereas quasars are much redder (steeper/softer;
161: $f_{\nu}\propto\nu^{0\; {\rm to}\; -2}$). For MIR colors from {\em
162: Spitzer} \citep{wrl+04}, it is not, in practice, quite this simple
163: because {\em Spitzer}'s (comparatively) large pixels ($1\farcs2$)
164: makes it difficult to distinguish point from extended sources, and
165: thus a simple color cut to select AGN suffers from considerable
166: contamination from quiescent galaxies. However, judicious use of
167: additional color cuts such as applied by \citet{lss+04} and
168: \citet{seg+04} can be used to select quasars (both obscured and
169: unobscured) with good efficiency and completeness for relatively
170: bright MIR sources. See \citealt{drp+08} for a recent review of AGN
171: selection techniques in the MIR.
172:
173: In this paper we explore ways to improve upon color selection of AGNs
174: using MIR colors. We particularly concentrate on 1) using
175: sophisticated Bayesian selection methods, to probe to fainter limits
176: (without additional contamination) than allowed by standard color
177: selection, 2) incorporating optical morphology information, and 3)
178: combining MIR and optical photometry, performing up to 8-D color
179: selection. The latter approach increases completeness to
180: high-redshift quasars (particularly $3.5<z<5.0$) where standard MIR
181: color cuts are incomplete (or heavily contaminated). In addition, we
182: will show that coupling deep optical and MIR data can overcome the
183: loss of the two longest wavelength IRAC bandpasses when {\em
184: Spitzer's} coolant is depleted. Indeed, selection of quasars (both
185: type 1 and type 2) and photometric redshift estimation for type 1
186: quasars suffer relatively little from this loss. Future coupling with
187: deep UV, near-IR, and far-IR data from {\em GALEX}, UKIDSS, {\em
188: Akari}, {\em WISE}, and {\em Herschel} will allow for further
189: improvements in both selection and photometric redshift estimation.
190:
191: We structure the paper as follows. \S~2 describes our sources of
192: data. We review our Bayesian selection algorithm in
193: \S~\ref{sec:overview}, where we also describe the training sets
194: designed for this selection. Application of our algorithm to combined
195: SDSS and {\em Spitzer} data sets is discussed in \S~\ref{sec:bayes}.
196: The resulting catalog is presented in \S~\ref{sec:cat}. Obscured
197: quasars are discussed in \S~\ref{sec:type2} and we present our
198: conclusions in \S~\ref{sec:conclusions}. Throughout this paper we
199: report photometry either in flux density (in Jy) or AB magnitudes
200: (denoted by square brackets). For the latter, {\em Spitzer}-IRAC
201: Channels 1-4 are given by $[3.6]$, $[4.5]$, $[5.8]$ and $[8.0]$, which
202: are the nominal wavelengths of the bandpasses in microns. The
203: conversion between AB and Vega ($[{\rm Vega}]-[AB]$) is taken to be to
204: be 2.779, 3.264, 3.748 and 4.382 mag for the four IRAC bandpasses.
205: For example $[3.6]-[4.5] ({\rm Vega}) = [3.6]-[4.5]({\rm AB}) +
206: 0.485$. Cosmology-dependent parameters are computed assuming
207: $H_o=70\,{\rm km\,s^{-1}\,Mpc^{-1}}$, $\Omega_m=0.3$ and
208: $\Omega_{\Lambda}=0.7$, in general agreement with the most recent WMAP
209: results \citep{dkn+08}. Unless otherwise specified, the term
210: ``quasar'' will refer to type 1 (broad-line) quasars/AGNs, regardless
211: of their luminosity.
212:
213: \section{The Data}
214:
215: \subsection{Samples}
216: The selection methods described in this paper will make use of a
217: variety of data sets. The MIR data are drawn from publicly available
218: catalogs/images from all of the largest area {\em Spitzer} surveys
219: using the Infrared Array Camera (IRAC; \citealt{fha+04}) where there
220: exists overlap with optical data from the SDSS.
221: %The nominal wavelength centers of IRAC Channels 1-4 are $3.6\mu$m, $4.5\mu$m, $5.8\mu$m, and $8.0\mu$m,
222: %respectively.
223: %but clearly depend on the spectral energy distributions
224: %of the sources that they are measuring.
225: We include {\em Spitzer}-IRAC
226: data from the XFLS \citep{lacy05}; SWIRE \citep{lsr+03}, specifically
227: the ELAIS-N1, ELAIS-N2, and Lockman Hole fields; the NOAO-Bo\"{o}tes
228: area from the IRAC Shallow Survey \citep{esb+04}; and S-COSMOS
229: \citep{scosmos07}. The properties of these fields are summarized in
230: Table~\ref{tab:tab0}.
231: %These fields have J2000 (RA [deg], Dec [deg])
232: %centers of (259.5,59.5), (242.75,55.0), (249.2,41.029), (161.25,58.0),
233: %(150.62,2.21), and (218.02,34.28), respectively, covering
234: %approximately 4, 9.3, 4.2, 11.1, 8.5, and 2.0 deg$^2$.
235: It is important to note that these {\em Spitzer}-IRAC data sets have
236: very different depths, as indicated in Table~\ref{tab:tab0}.
237: %The shallowest data are from the XFLS which had
238: %60s of integration time, yielding depths of 20$\mu$Jy (77\%
239: %completeness) and 100$\mu$Jy (94\% completeness) at 3.6$\mu$m and
240: %8.0$\mu$m, respectively \citep{lsr+03}. With 90s of integration the
241: %NOAO-B\"{o}otes data from the IRAC Shallow Survey had 5-$\sigma$ image
242: %sensitivities of 6.4$\mu$Jy at 3.6$\mu$m and 56$\mu$Jy at 8.0$\mu$m
243: %\citep{esb+04}. The SWIRE data are somewhat deeper with 120s of
244: %exposure time and 5-$\sigma$ image sensitivities of 3.7$\mu$Jy and
245: %37.8$\mu$Jy; these correspond to 95\% completeness limits of 14$\mu$Jy
246: %and 56$\mu$Jy, respectively. Finally, the S-COSMOS images have 1200s
247: %of exposure, giving 5-$\sigma$ image sensitivities of 0.9$\mu$Jy and
248: %14.6$\mu$Jy at 3.6$\mu$m and 8.0$\mu$m, respectively
249: %\citep{scosmos07}.
250: As the SWIRE data represent the largest fraction of objects in our
251: analysis, we adopt their 95\% completeness limit of 56$\mu$Jy at
252: 8.0$\mu$m as the cutoff for our analysis. This choice excludes some
253: high signal-to-noise data from the COSMOS field and keeps lower
254: signal-to-noise data from the XFLS field, but is a good compromise for
255: common analysis of all the data sets. The specific IRAC catalogs used
256: are: XFLS (main\_4band.cat; \citealt{lacy05}), SWIRE ELAIS-N1, -N2,
257: and Lockman Hole (SWIRE2\_N1\_cat\_IRAC24\_16jun05.tbl,
258: SWIRE2\_N2\_cat\_IRAC24\_16jun05.tbl,
259: SWIRE2\_Lockman\_cat\_IRAC24\_10Nov05.tbl; Surace et al.\ 2005), and
260: COSMOS (COSMOS\_IRAC\_0407\_IRSA.tbl; \citealt{scosmos07}). At the
261: time of writing there were no publicly available MIR source catalogs
262: for the NOAO-Bo\"otes data\footnote{Now available at
263: http://data.spitzer.caltech.edu/popular/sdwfs/20081022\_enhanced/documentation/SDWFS\_DR1.html},
264: so we extracted photometry from the publicly available images when
265: constructing the quasar training set; see \S~\ref{sec:qsotrain}. In
266: all there exists $\sim32$ deg$^2$ of overlap between wide-area {\em
267: Spitzer} fields and SDSS, with another $\sim30$ deg$^2$ of SWIRE
268: data lacking SDSS coverage.
269:
270: Our primary goal herein is to perform
271: 8-dimensional\footnote{Technically, 8-color selection, but we prefer
272: the term dimension over color so as to be clear our method is
273: flexible enough to include information other than colors.} selection
274: of quasars (primarily type 1) using optical and MIR data sets. The
275: eight dimensions refer to the 8 unique colors afforded by SDSS $ugriz$
276: magnitudes \citep{fig+96} and {\em Spitzer}-IRAC Channel 1-4 flux
277: densities. In addition, we perform 6-D selection of quasars using all
278: 5 SDSS filters and the 2 short wavelength IRAC bandpasses (since the
279: long wavelength bandpasses will not be available during {\em
280: Spitzer's} warm mission after its coolant is exhausted;
281: \citealt{warm}). As such, we compile two combined MIR+optical data
282: sets. One includes all objects detected in both IRAC Channels 1 and 2
283: %(regardless of whether they are also detected in IRAC Channels 3 and 4)
284: that are matched to SDSS sources (95\% completeness at $g=22.2$,
285: $i=21.3$), for 6-D color classification. The second is a subset of
286: the first where the objects are additionally detected in both IRAC
287: Channels 3 and 4 (for 8-D color classification). IRAC upper limits
288: are not considered as our selection algorithm is currently not
289: equipped to handle them. Note, however, that the SDSS's use of asinh
290: magnitudes \citep{lgi+01} means that any object detected in one SDSS
291: bandpass will have meaningful magnitude measurements in all the other
292: bandpasses (even if they are nominally below the ``flux limit'').
293: Unless otherwise specified, all SDSS magnitudes herein are PSF
294: magnitudes that have been corrected for Galactic extinction according
295: to \citet{sfd98}.
296:
297: %All of the IRAC data sets are deep enough to enable 5-band optical
298: %detections of all SDSS type 1 quasars at the SDSS' spectroscopic
299: %limit.
300:
301: %The union of these catalogs yields 1,042,848 valid sources detected in
302: %both Channels 1 and 2 (with effective wavelengths of 3.6$\mu$m and
303: %4.5$\mu$m).
304:
305: \subsection{SDSS-{\em Spitzer} Matching}
306:
307: Matching the combined 2-band IRAC data to the SDSS optical imaging
308: catalog from the 6th SDSS data release (DR6; \citealt{sdssdr6}) with a
309: $2\arcsec$ matching radius (IRAC has $1\farcs2$ pixels and the median
310: SDSS seeing is $\sim1\farcs3$) yields 324,618 objects. Of these, 486
311: objects are duplicates (243 pairs) where more than one SDSS source
312: matched an IRAC source. To avoid any contamination, both objects in
313: all of these duplicates were rejected. Spot checking of a few
314: duplicates revealed that they tended to be galaxies that were
315: improperly deblended in the SDSS. The 324,132 matches obtained after
316: rejecting duplicates compose ``Sample A''. Further limiting Sample A
317: to objects detected in all 4 IRAC bands leaves 95,181 objects; we
318: refer to this sample hereafter as ``Sample B''. Sample B will be used
319: for our 8-D classifications (and in the construction of our training
320: sets). The MIR colors of Sample B are shown in
321: Figure~\ref{fig:iracstargal}.
322:
323: In Figure~\ref{fig:iracstargal}, the red contours/dots denote objects
324: classified as extended sources, while the blue contours/dots are point
325: sources. The extended vs.\ point-like classification is obtained on
326: the basis of SDSS optical data by comparing PSF magnitudes with measures of
327: extended flux for all bands in which the object is detected
328: \citep{slb+02}. For faint optical sources, star-galaxy separation
329: begins to break down and galaxies become significant ``stellar''
330: contaminants. At $i\sim20.8 \;(S_{8\mu \rm m}\sim115\mu {\rm Jy}$ at
331: $z=1.5$ for a type 1 quasars), roughly 10\% of SDSS point sources are
332: likely to actually be galaxies.
333:
334: The colors of the extended sources are concentrated in three clumps in
335: Figure~\ref{fig:iracstargal} (red contours/dots); understanding their
336: origins is important for optimal object classification. The spectral
337: energy distribution of extragalactic sources in the 1-8$\mu$m range is
338: composed of three main components: (1) The combined light of stellar
339: photospheres has a peak at 1.6$\mu$m \citep{saw02} and declines
340: according to the Rayleigh-Jeans law at longer wavelengths; (2)
341: star-formation activity in the galaxy results in polycyclic aromatic
342: hydrocarbon (PAH) emission from dust with strongest features at 3.3,
343: 6.2, 7.7 and 8.6$\mu$m; and (3) circumnuclear dust may be heated by
344: the central AGN resulting in continuum emission at MIR wavelengths,
345: which is typically well-represented by a power-law since the dust is
346: emitting at a wide range of temperatures. The relative contributions
347: of these components to the total spectrum determines the IRAC colors
348: of extragalactic objects\footnote{MIR colors can also be affected by
349: the 10$\mu$m silicate absorption/emission feature
350: \citep[e.g.,][]{hss+05} at $z\sim0$ if the feature is very broad
351: (IRAC channel 4 cuts off at $\sim9.5\mu$m).}. The colors of
352: star-dominated and PAH-dominated galaxies at a wide range of redshifts
353: (0--1.6) are quite distinct in the IRAC bands from the colors of the
354: thermal emission of circumnuclear dust allowing color separation of
355: AGNs from galaxies of all types. See \citet{bba+06} and
356: \citet{drp+08} for more details regarding how galaxies track through
357: MIR color space with redshift.
358:
359:
360: \subsection{Quasars}
361: \label{sec:qsotrain}
362:
363: In addition to matching the IRAC catalogs to the full SDSS optical
364: catalogs, we have also explicitly matched it to the 77,429
365: spectroscopically confirmed quasars from the SDSS-DR5 quasar catalog
366: \citet{shr+07}. As the density of these known quasars is much smaller
367: than the full SDSS object catalog, it is not too cumbersome to extract
368: IRAC photometry from the publicly available IRAC images of Bo\"{o}tes
369: field. While our Bo\"{o}tes data reduction was simplistic compared to
370: the XFLS, SWIRE, and COSMOS data reductions, our analysis is
371: substantiated by the lack of any difference in color as a function of
372: redshift for the quasars in the Bo\"{o}tes region.
373:
374: In all, matching the SDSS-DR5 quasar catalog to the IRAC data sets
375: (again using a $2\arcsec$ matching radius) resulted in 425 matches.
376: The relative IRAC and SDSS limits are such that all SDSS quasars are
377: detected in the MIR. The additional 166 quasars as compared with
378: \citet{rls+06} come from the COSMOS and Bo\"{o}tes fields. We
379: supplement these quasars with additional spectroscopically-confirmed
380: quasars with both {\em Spitzer}-IRAC and SDSS photometry from
381: \citet{lss+04}, \citet{pap+05}, and \citet{jfh+06}, where the latter
382: objects are $z\sim6$ quasars and the two former samples were
383: restricted to broad-line, type 1 AGNs. In all we have compiled 515
384: known type 1 quasars with both SDSS and IRAC detections; the MIR
385: colors of these quasars are shown in Figure~\ref{fig:iracstargal} as
386: green points.
387:
388: The SDSS-DR5 quasar catalog covers nearly all of the areas of overlap
389: between SDSS and the SWIRE fields. Therefore, with the exception of
390: several small-area fields and a part of the SWIRE/ELAIS-N1 area, these
391: objects represent essentially all of the spectroscopically confirmed
392: quasars that have been covered by both IRAC data and the final SDSS
393: data release (DR7). The next major opportunity to obtain a large
394: sample of spectroscopically confirmed quasars with MIR follow-up
395: observations is with {\em WISE} \citep{WISE}, scheduled to launch in
396: late 2009. {\em WISE}'s 120$\mu$Jy depth at 3.3$\mu$m will be
397: sufficiently sensitive to detect all of the $i<19.1$ quasars in the
398: SDSS catalog (but generally not the $z>3$ objects that extend as faint
399: as $i=20.2$) and will significantly increase the number of quasars
400: with MIR flux density measurements.
401: %The MIR colors
402: %of all of the known type 1 SDSS quasars are shown in
403: %Figure~\ref{fig:iracstargal} as green points.
404:
405: \iffalse
406: The IRAC bandpasses are generally refereed to as Channels 1 through 4
407: or as the 3.6$\mu$m, 4.5$\mu$m, 5.8$\mu$m, and 8.0$\mu$m bands,
408: respectively. For a quasar spectrum with MIR spectral index of
409: $\alpha_{\nu}=-1$ ($f_{\nu} \propto \nu^{\alpha}$), the effective
410: wavelengths of the IRAC bandpasses are actually closer to 3.52$\mu$m,
411: 4.46$\mu$m, 5.67$\mu$m, and 7.70$\mu$m.
412: \fi
413:
414: \iffalse
415: One of the advantages of matching {\em Spitzer}-IRAC data to SDSS data
416: is to gain morphology information that is not provided by IRAC's
417: $1\farcs7$ beam. However, the SDSS PSF degrades with magnitude and it
418: is important to understand the limitations of the SDSS data's
419: morphological reliability. Fortunately, with dozens on epochs of data
420: on SDSS Stripe 82 (the Southern Equatorial Stripe; $-60<RA<60$ and
421: $-1.25<Dec<1.25$) it is possible to robustly determine this
422: reliability. For single-epoch data, the SDSS morphology is set by
423: comparing the PSF magnitude to the model magnitude. If the difference
424: is larger than 0.15 magnitudes, it indicates that there is extended
425: flux unmodeled by the PSF and that the objects is likely a galaxy.
426: However, with using co-added images that achieve a much deeper
427: magnitude limit and better photometric reliability, one finds that a
428: PSF versus model magnitude difference of as little as 0.05 mag
429: provides good discrimination between point and extended sources.
430: Comparing the single- to multi-epoch data, we find that, at $g=21.5$,
431: better than 96\% of SDSS-classified point sources (single epoch) are
432: truly point sources. At that same limit more than 99\% of extended
433: sources are indeed extended. Thus, for cases where robust optical
434: morphology is needed, we will limit our sample to $g<21.5$. {\bf
435: Include more information? Plots as f(mag)?}
436: \fi
437:
438: In addition to type 1 quasars, our selection is potentially sensitive
439: to obscured (type 2) quasars. As such, it is important to know
440: where type 2 quasars lie in the MIR+optical color space that we
441: explore herein, so we have compiled a sample of type 2 quasars from
442: literature with measured redshifts (either photometric or
443: spectroscopic). These include samples from \citet[][291
444: objects]{zsk+03}; \citet[][887 objects]{rzs+08}; \citet[][6 objects
445: from XFLS]{lsp+07}; \citet[][type 2 objects from Table 3, 30 objects
446: from XFLS and SWIRE]{lps+07}; \citet[][one object at
447: $z=3.7$]{nhg+02}; \citet[][20 objects from their Table 2, type 2 and
448: unidentified objects (their type 9) with $\log(N_H) >
449: 21.8$]{mbh+02}; \citet[][one object at $z=3.288$]{smc+02};
450: \citet[][7 objects]{shl+06}; \citet[][125 objects]{pws+06};
451: \citet[][104 XMM-DS objects classified as AGN2 and starburst/AGN]{tpc+07};
452: \citet[][21 objects from SWIRE/NDWFS/FLS, their Table~1 with redshifts
453: from their Table~3]{ptm+07}; and \citet[][141 objects from
454: CDFS]{zmm+04}. We then matched this combined sample with all the
455: available IRAC catalogs (including and in addition to those above)
456: %FLS/XFLS,S-COSMOS,SWIRE catalogs (ELAIS-N1, ELAIS-N2,
457: %Lockman Hole, XMM-LSS and CDFS)
458: to extract {\em Spitzer}-IRAC photometry. Only objects detected at
459: both 3.6$\mu$m and 4.5$\mu$m were selected in the matched output.
460: This final output table was then matched to the SDSS DR6 catalog to
461: select objects with both SDSS and {\em Spitzer}-IRAC photometry. This
462: filtering/matching process was used to generate a final list of 43
463: type 2 quasars/candidates from literature with both SDSS and {\em
464: Spitzer}-IRAC photometry. The MIR colors these type 2 quasars are
465: given by open gray squares in Figure~\ref{fig:iracstargal}.
466:
467:
468: \section{Bayesian Selection Overview}
469: \label{sec:overview}
470:
471: Our selection method follows that of \citet{rng+04} and is detailed
472: therein. For multi-band imaging surveys (such as the SDSS) it is
473: possible to perform object classification beyond traditionally adopted
474: 2-D color cuts. In particular, if the parameter space of interest is
475: sufficiently populated by known objects, then these objects can be
476: used as a ``training set'' to derive classifications within the
477: parameter space \citep[e.g.,][]{htf01,rng+04,bbm+06,gzz08}. Our
478: algorithm of choice is based on kernel density estimation (KDE),
479: weighted by a Bayesian ``prior". The KDE aspect is that the
480: probability distribution function (pdf) that is used to evaluate the
481: classification is smoothed by some appropriate kernel function (e.g., a
482: Gaussian). Our algorithm uses two training sets, one that represents
483: the objects to be classified and one that represents everything else.
484: We compute the N-dimensional Euclidean color distance between some new
485: object that we wish to classify and each of the training set
486: objects. The new object is then classified based on how consistent its
487: colors are with each training set \citep[e.g.,][]{grn+05}. One
488: limitation of our current algorithm is that error information is not
489: explicitly included (cf., \citealt{pmh+07}), although it is {\em
490: implicitly} included by virtue of the training sets' inherent error
491: distribution.
492:
493: \iffalse
494: Our Bayesian selection method follows that of \citet{rng+04} where
495: technical details of the algorithm can be found, but the basics are
496: quite simple. Traditional quasar selection relies on simplistic
497: ``color cuts'', typically in 2-D. However, for the SDSS (and other
498: current and future multi-band imaging surveys) it is possible to
499: perform higher dimensional object classification that goes beyond 2-D
500: color cuts. In particular, if there are enough known objects (per
501: increment of parameter space) of the type that one is searching for,
502: then they can be used as a ``training set'' within the context of any
503: number of modern object classification algorithms
504: \citep[e.g.,][]{htf01,rng+04,bbm+06,gzz08}. Our algorithm of choice
505: is a non-parametric Bayesian classifier (NBC), based on kernel density
506: estimation (KDE). It is Bayesian in that it weights by a ``prior''.
507: It is non-parametric in that objects are classified as either ``in''
508: or ``out''. The KDE aspect refers to the fact that points are not
509: treated as discrete, but rather are smoothed by some appropriate
510: kernel function (e.g., a Gaussian) in building the probability
511: distribution function (pdf) that is used to evaluate the
512: classification.
513:
514: The essence of our algorithm is to define two training sets, one that
515: represents the type of objects that we are trying to find and one that
516: represents everything else. New objects are
517: %We then take some unknown object and
518: %compute the N-dimensional Euclidean color distance between the new
519: %object and each of the training set objects. The object is then
520: classified according to which of the training sets is most consistent
521: with the colors of the new object. The actual details are more
522: complicated and have largely to do with computing speed while dealing
523: with large data sets \citep[e.g.,][]{grn+05}. In short, the algorithm
524: is designed to
525: %not actually compute the color distances, but
526: only use as much information as is needed to uniquely classify each
527: new object (i.e., the code is fast because it performs non-parametric
528: instead of parametric classification). Some limitations of our
529: algorithm are that we currently do not include error information in
530: our analysis (cf., \citealt{pmh+07}) and only perform binary (quasar
531: or non-quasar) selection. While errors are not currently included
532: explicitly, they are included implicitly by virtue of the error
533: distribution inherent to the training sets. In the future we hope to
534: perform simultaneous star-galaxy-quasar classification.
535: \fi
536:
537: \iffalse
538: i.e. the code is fast because it uses geometric data structures to
539: eliminate points from consideration when possible while still
540: preserving the exact classification answers). This allows the method
541: to scale to millions of points tractably, unlike methods with similar
542: accuracy such as nonlinear support vector machines, which cannot scale
543: beyond tens of thousands of points. Because the method is
544: non-parametric, it is free of distributional assumptions and yields
545: state-of-the-art accuracy. Its Bayesian use of priors automatically
546: handles the common difficulty of training where one class (quasars)
547: has many fewer examples than the other (non-quasars), unlike most
548: common classification methods
549: \fi
550:
551: Our Bayesian selection algorithm is guided by four considerations.
552: First are the so-called ``wedge'' diagrams for MIR selection of AGNs.
553: These are regions of color space that tend to be occupied by AGNs.
554: \citet{lss+04} select AGNs using color cuts in non-adjacent bandpasses
555: to isolate AGNs, specifically $[3.6]-[5.8]$ and $[4.5]-[8.0]$; we
556: refer to these color cuts as the ``Lacy wedge''.
557: \citet{seg+04,ska+07} instead utilize adjacent bandpasses,
558: specifically $[3.6]-[4.5]$ and $[5.8]-[8.0]$; we refer to these color
559: cuts as the ``Stern wedge''. The exact color-cuts that describe these
560: regions have changed slightly over time as more {\em Spitzer} data
561: have been obtained. Thus we will refer to them in the abstract sense
562: throughout the paper, but we visually illustrate them in
563: Figure~\ref{fig:iracstargal}. The Lacy wedge region is given by the
564: dashed lines in the top left-hand panel and the Stern wedge region is
565: given by the dashed lines in the bottom left-hand panel.
566:
567: %In addition to point (blue) and extended (red) sources, we also plot
568: %known type 1 (green) and type 2 (gray) quasars in addition to known
569: %brown dwarfs (magenta; from \citealt{psb+06}).
570:
571: The second diagnostic considers the mean colors for stars and quasars
572: in the IRAC photometric system; Table~\ref{tab:tab1} gives both the
573: observed and theoretical (power-law approximated) values. For stars
574: the $[3.6]-[8.0]$ color is $\sim-1.7$, while for quasars it ranges
575: from $\sim0$ to $1.7$. Thus, well-measured point sources can be
576: grouped into stars and quasars in a straightforward manner, based on
577: MIR colors alone.
578:
579: Third, we consider the utility of having morphology information in
580: addition to photometry and the effects of over-reliance on morphology
581: as when star-galaxy separation fails to be robust.
582: Figure~\ref{fig:iracstargal} shows that AGNs are readily identified as
583: point sources with red MIR colors (even using only the two shortest
584: IRAC bandpasses). However, photometric errors complicate clean AGN
585: identification at faint limits. Using just the Lacy wedge as an
586: example, we demonstrate in Figure~\ref{fig:lacywedgemany} how using a
587: brighter MIR flux limit or morphology information can improve the
588: efficiency of MIR color selection.
589: %Inclusion of all 4-band
590: %IRAC-detected objects results in significant contamination of the
591: %color selection wedges.
592: Removing faint objects, saturated objects, and/or extended objects
593: significantly reduces contamination.
594:
595: Finally, we consider the color-redshift distribution for known
596: quasars. Six color combinations are shown in
597: Figure~\ref{fig:czplotnew} as a function of redshift for the known
598: quasar samples described above and for two template quasar spectra.
599: The Stern (dotted) and Lacy (dashed) wedge color cuts (see
600: Fig.~\ref{fig:iracstargal}) are included in order to show their
601: effects on completeness with redshift. Cuts in $[3.6]-[4.5]$ can
602: introduce incompleteness to $3.5<z<5.0$ quasars. The $z-[3.6]$ color
603: changes rapidly between redshift 0 and 2 due primarily to a minimum in
604: the SED, the so-called ``$1\mu$m inflection'' \citep[e.g.,][]{ewm+94}.
605:
606:
607: \subsection{Training Sets}
608: \label{sec:train}
609:
610: The core of our quasar training set is the 515 optically-selected,
611: spectroscopically-confirmed quasars that have existing {\em
612: Spitzer}-IRAC photometry as discussed in \S~\ref{sec:qsotrain};
613: this set is dominated by the 425 quasars from the SDSS-DR5 quasar
614: catalog. Due to some photometric errors (bad deblending, etc.)
615: objects with $[3.6]-[4.5]<-0.7$ or $[5.8]-[8.0]<-0.5$ are rejected.
616: As our primary goal is efficient N-dimensional selection of type 1
617: (broad-line) quasars, we have not explicitly included obscured or
618: intrinsically weak AGNs in the training set, although doing so would
619: be a logical next step for future investigation.
620:
621: Since the number of known quasars with optical and MIR photometry is
622: relatively small in comparison with the size of the sample that we
623: wish to classify, we further include IRAC 4-band matches to SDSS
624: sources that are highly likely to be quasars. Below we describe two
625: such classes of objects that are included. For these samples to be as
626: clean as possible, we require that the sources satisfy $56\mu{\rm
627: Jy}<f_{8.0\mu{\rm m}}<10{\rm mJy}$, and $i>14.0$ to exclude
628: saturated objects and to limit the samples to the approximate
629: 8.0$\mu$m 95\% completeness limit of the SWIRE data.
630:
631: First, given the power of combining morphology with MIR colors shown
632: in Figure~\ref{fig:lacywedgemany}, we identify point sources with red
633: (AGN-like) MIR colors as quasar candidates that are sufficiently
634: robust to be included in the training set. As SDSS star-galaxy
635: separation begins to break down at faint magnitudes ($r\sim21$;
636: \citealt{sjd+02}) with faint extended sources being more likely to be
637: classified as point-like, clean point source identification should be
638: restricted to $i\lesssim20.8$ (0.5 mag brighter than the nominal SDSS
639: $i$-band flux limit). Thus, point sources with $i<20.8$ are classified
640: as quasars if they have $[3.6]-[8.0]>0$, see Table~1. However, for
641: the XFLS data, the MIR photometric errors are large enough that a
642: simple cut in $[3.6]-[8.0]$ is insufficient to cleanly identify
643: quasars; for these objects we also require $[3.6]-[5.8]>-0.5$ and
644: $[4.5]-[8.0]>-0.5$ (akin to the Lacy wedge). Finally true point
645: source quasars do not typically have MIR colors outside of
646: $-0.4<[5.8]-[8.0]<1.4$ and $[3.6]-[4.5]<-0.6$, so we exclude such
647: objects to further limit any contamination by misclassified galaxies
648: and by normal stars.
649:
650: Second, in addition to the above red MIR point sources, we capitalize
651: on the work of \citet{lss+04} and \citet{seg+04} by including objects
652: that can be robustly identified as quasar candidates based on their
653: location in the MIR wedge diagrams. While selection using these
654: wedges is relatively clean at bright limits, the colors of quasars may
655: change with flux (indeed Fig.~\ref{fig:iracstargal} suggests that at
656: fainter limits the quasar contribution weakens with respect to the
657: galaxy contribution, making the MIR colors bluer on average), thus
658: additional constraints are needed to exclude contaminants. In
659: particular, we include in the quasar training set any {\em point
660: sources} that are in any of 1) the Lacy wedge, 2) the Stern
661: wedge, or 3) a modified Stern wedge (which excludes a region near
662: the galaxy locus, see \S~\ref{sec:wedge} below). We further include
663: {\em extended sources} that lie in {\em both} the Lacy wedge and our
664: modified Stern wedge. In addition to relatively robust MIR point
665: source candidates, these criteria select point sources near the
666: quasar/galaxy boundary in $[3.6]-[4.5]$ (between the original and
667: modified Stern wedges), but reject extended sources in this same
668: ``buffer zone'' where it may be difficult to distinguish true galactic
669: contaminants from host-dominated quasars.
670: %Nadia: show where these are. GTR: could, but that's not really
671: %point here, so I just cut this out.
672: % (e.g., type 2 quasars, which
673: %have quasar colors in the MIR, but galaxy colors and morphology in the
674: %optical).
675:
676: For these two MIR-selected training set populations, we have not
677: utilized any optical magnitude or color information since optical
678: quasar selection is known to be incomplete in certain redshift ranges
679: (e.g., $z\sim2.7$, \citealt{rsf+06}), and the inclusion of these
680: MIR-identified sources is our attempt to mitigate this effect.
681: However, roughly half of these objects
682: %have $u-g<0.6$ and $g-i<0.8$ and thus
683: are UV-excess sources that would be included using optical selection
684: alone. Similar to the $z\sim2.7$ redshift incompleteness in the
685: optical,
686: %we will see that
687: MIR-only selection is incomplete in a different redshift range
688: ($3.5<z<5$), thus the inclusion of optical-only selection objects in
689: the training set helps recover such objects in our higher dimensional
690: selection. Having a training set comprised of both optical-only
691: (the above 515 known quasars) and MIR-only identified sources aids in
692: the creation of a photometric quasar sample that is as complete as
693: possible at all redshifts. While no MIR-only or optical-only quasar
694: samples are fully complete at all redshifts, by combining the two and
695: then performing quasar selection simultaneously on both MIR and
696: optical colors, we overcome the limitations inherent to each method
697: separately.
698:
699: Due to photometric errors (bad deblending, etc.), a handful of the
700: quasar training set objects are outliers from the quasar
701: color-redshift distribution. We only retain objects that are in the
702: intersection of the following criteria: $[3.6]-[8.0]>-0.5$,
703: $[3.6]-[4.5]>-0.5$, $[4.5]-[5.8]>-0.7$, and $[5.8]-[8.0]>-0.5$. In
704: summary, the final quasar training set is the combination of 1) known
705: quasars, 2) red MIR objects with point-like optical morphologies, 3)
706: point/extended sources identified in two distinct MIR color wedges,
707: and 4) excluding some photometric error induced outliers and some
708: extended sources on the border between known quasars and galaxies.
709:
710: As was the case for our purely optical selection \citep{rng+04}, a
711: more difficult task is to define the non-quasar training set (here
712: both stars and normal galaxies) as we do not have a sufficiently large
713: (and representative) sample of spectroscopically confirmed objects.
714: While there are many thousands of SDSS spectra of galaxies and stars,
715: the galaxy sample extends only to $r=17.77$ and the star spectra cover
716: only specific regions of color parameter space. Furthermore,
717: optical spectra may not reveal the AGN nature of a galaxy as well as
718: the MIR colors would. Thus, we are essentially left with identifying
719: the leftovers from our quasar training set as the non-quasar training
720: set, with the exception of borderline objects that we exclude
721: from either training set. Objects in the non-quasar training set
722: satisfy the following conditions. First, they must not be in the
723: quasar training set. Second, they are rejected if they lie in the Lacy
724: wedge and also either the original or modified Stern wedges. The
725: remaining objects that meet our magnitude and flux limits constitute
726: the non-quasar training set.
727:
728: Figure~\ref{fig:iraccolorstrainlib} shows the MIR color distribution
729: of the objects in the quasar and non-quasar training sets. The SDSS
730: colors of the same objects are shown in
731: Figure~\ref{fig:sdsscolorstrainlib}. The training set has 53332
732: objects; 5627 labeled as quasars and 47705 as non-quasars. For our
733: Bayesian classification method, we need to estimate a ``prior''
734: that indicates the a priori probability of any given object in our
735: samples not being a quasar. We adopt the fraction determined from the
736: training sets as a reasonable value, specifically 89\%.
737:
738: %The full data set (including only cuts on MIR fluxes and optical
739: %magnitudes) has 53889 objects.
740:
741: %GTR: Has to come later
742: %Figures~\ref{fig:iraccolorstestcons}, ~\ref{fig:iraccolorstestlib},
743: %~\ref{fig:sdsscolorstestcons}, and ~\ref{fig:sdsscolorstestlib} are
744: %the same as the above four figures, but now showing the effect of
745: %self-classification.
746:
747:
748: \subsection{The Modified Stern Wedge Region}
749: \label{sec:wedge}
750:
751: In Figure~\ref{fig:wedgecomp} we compare and contrast the two primary
752: MIR color selection techniques in use today; see also \citet{drp+08}.
753: By exploring the color space occupied by objects selected by the Stern
754: wedge, but not the Lacy wedge, and vice versa, we hope to create a
755: more robust quasar training set from MIR colors. Examination of
756: Figure~\ref{fig:wedgecomp} reveals the following
757: \begin{itemize}
758: \item The Stern wedge is relatively clean (few objects are outside of the Lacy wedge).
759: \item The Lacy wedge is very clean to the 1\,mJy flux limit for which
760: it was defined but is quite contaminated at fainter MIR fluxes (it
761: contains many objects outside of the Stern wedge).
762: \item The Stern wedge's lack of contamination comes largely from its
763: cut on $[3.6]-[4.5]$. Although very effective over a wide range of
764: redshifts, this cut makes the Stern wedge incomplete to quasars with
765: $3.5 < z < 5.0$ (see Fig.~\ref{fig:czplotnew}).
766: \end{itemize}
767: Furthermore, we note that the Stern wedge's blue cut in $[5.8]-[8.0]$ excludes
768: objects that are in the Lacy wedge. Such objects are rejected
769: nominally to exclude high-redshift galaxies (see Fig.~1 in
770: \citealt{seg+04}); however, as these objects are found by
771: \citet{hjf+06} to be strong soft X-ray sources, the AGN population
772: likely crosses this dividing line.
773:
774: Indeed, during a 2008 June observing run on the Mayall 4-m telescope
775: at Kitt Peak National Observatory, we were able to make observations
776: of 9 randomly chosen sources blueward of the Stern wedge's
777: $[5.8]-[8.0]$ cut. Of the six sources redder than $[3.6]-[4.5]=-0.1$,
778: three are clearly quasars ($z=0.91, 1.61, 1.64$) and a 4th may also be
779: a quasar at low signal-to-noise; their MIR colors are shown in
780: Figure~\ref{fig:wedgecomp}.
781:
782: These objects are all in the region of strong soft X-ray sources (and
783: thus likely quasars) as indicated by \citet{hjf+06} and have two
784: similar features. All are flagged as color selected quasar candidates
785: \citep{rfn+02} in the SDSS database, but were too faint to be targeted
786: for spectroscopy.
787: %Such identifications normally
788: %should not be used for objects as faint as these since it is really
789: %meant for diagnostics of objects just beyond limiting spectroscopic
790: %magnitude (and thus with similar photometric errors), but is
791: %nevertheless instructive.
792: They also have very similar appearances in the optical, being
793: extended, but centrally concentrated bluish-white objects. When
794: coupled with MIR data, these characteristics may help to identify
795: additional quasars in the SDSS database that are nominally outside of
796: the Stern wedge, but that may nevertheless be AGNs.
797:
798: % See sdss_irac_qsos_9D.cat.rasort.match.kpno
799:
800: %[Check NED for non-Lacy, non-Stern]
801:
802: Thus, as mentioned above, we implement a ``modified Stern wedge'' that
803: is adjusted from the original Stern wedge as follows. We first make a
804: more conservative cut on the AGN/galaxy border in $[3.6]-[4.5]$ color.
805: Second, we allow objects bluer than $[5.8]-[8.0]<-0.07$ if they have
806: $[3.6]-[4.5]>-0.0626$ (the intersection of our modified $[3.6]-[4.5]$
807: and the original $[5.8]-[8.0]$ cuts). This modified Stern wedge is
808: defined by the following set of color cuts (see
809: Fig.~\ref{fig:wedgecomp})
810: \begin{eqnarray}
811: (([5.8]-[8.0])>-0.07 \; \&\& \; ([3.6]-[4.5])>0.18*([5.8]-[8.0])-0.05 \; \&\& \nonumber \\
812: ([3.6]-[4.5])>2.5*([5.8]-[8.0])-2.295) \;||\; \nonumber \\
813: ([3.6]-[4.5]>-0.0626 \; \&\& \; (-0.5<[5.8]-[8.0]\le-0.07)).
814: \end{eqnarray}
815: No modifications have been made to the Lacy wedge because coupling it
816: with the Stern wedge already removes spurious sources. We are not
817: suggesting that this modified wedge be used in place of the Stern
818: wedge for MIR AGN selection, rather we use it to minimizing
819: contamination from normal galaxies in our training sets, while
820: attempting to maximize completeness to quasars with $3.5<z<5$.
821:
822: %\section{Bayesian Selection}
823: \section{Application of the Algorithm}
824: \label{sec:bayes}
825:
826: Once the training sets are defined, we follow the techniques described
827: in \citet{rng+04} for utilizing these training sets to select objects
828: from a sample of data. We shall perform this selection for two sets
829: of color space: 8-D (MIR+optical) to attempt a more complete type 1
830: quasar selection than can be accomplished with MIR-only or
831: optical-only selection; and 6-D, anticipating the operation of {\em
832: Spitzer} post-cryogen, when only IRAC Channels 1 and 2 will be
833: operational. We have additionally attempted a 3-D MIR-only Bayesian
834: selection. Although this approach adds an extra color dimension that
835: is otherwise being wasted by the 2-D MIR wedge selection methods, we
836: find that it does not work significantly better (e.g., it is still
837: rather incomplete to $3.5<z<5.0$ quasars as is the Stern wedge) and do
838: not discuss it further here.
839:
840: In addition to a {\em prior} (as discussed above) our algorithm also
841: uses a leave-one-out cross-validation process to determine the optimal
842: bandwidths (kernel smoothing parameter) for the AGN and non-AGN
843: samples, respectively. These bandwidths are essentially the
844: resolution of the pdf, akin to the bin size for a histogram. See
845: \citet{rng+04} for details, but, in brief, the algorithm examines a
846: range of bandwidths and chooses the one that maximizes our
847: completeness to known quasars while minimizing contamination. The
848: adopted training set bandwidths notated as (star, quasar) were
849: $(0.2,0.11)$ and $(0.195,0.05)$ magnitudes, for the 8-D and 6-D
850: selection, respectively. The selection is reasonably robust to small
851: ($\sim0.05$ mag) changes from these values.
852:
853: \iffalse
854: \subsection{3-D}
855:
856: Traditional color-color selection has proven to work fairly well in
857: the MIR using the 4-band {\em Spitzer}-IRAC data
858: \citep{lss+04,seg+04}. Indeed, \citet{seg+04} quote better than 80\%
859: completeness with less than 20\% contamination to $S_{8\mu \rm
860: m}=76\mu {\rm Jy}$. However, MIR selection with {\em Spitzer}-IRAC
861: is generally done in 2-D color space when 3 unique colors are
862: available. Using instead our Bayesian algorithm in 3 dimensions, we
863: hope to improve both the completeness and efficiency of MIR AGN
864: selection. Here we specifically concentrate on unobscured, type 1
865: AGNs; however, our method also recovers some obscured AGNs as
866: well. A more explicit attempt to extract type 2 AGNs is left for
867: future work.
868:
869: We have applied our Bayesian selection algorithm to the MIR colors of
870: Sample A using the training sets, priors, and bandwidths described
871: above. As with the training sets, Sample A was first restricted to
872: $56\mu{\rm Jy}<f_{8.0\mu{\rm m}}<10{\rm mJy}$, and $i>13.0$ to exclude
873: saturated objects and to limit the sample to the approximate 8.0$\mu$m
874: 95\% completeness limit of the SWIRE data. In all, 5123 objects were
875: identified as AGN candidates, including one object as conservative
876: only, 800 as liberal only, and 4322 using both the conservative and
877: liberal selections.
878:
879: %awk '$32>0 || $33>0 {print $0}' sdss_irac_qsos_479D.zphot.match.dat | wc
880:
881: The MIR and optical color distributions of the objects are shown in
882: Figures~\ref{fig:iraccolorstest4Dclass} and
883: ~\ref{fig:sdsscolorstest4Dclass}, respectively. We find that the vast
884: majority of the objects are isolated to the Stern wedge, even
885: considering our attempts to be more inclusive. Thus, unfortunately,
886: even extending standard 2-D color selection to a 3-D Bayesian analysis
887: does not seem to allow robust selection of $3.5<z<5.0$ quasars.
888: However, we do find a small cluster of objects somewhat bluer than the
889: Stern wedge in $([5.8]-[8.0])$, specifically near $[3.6]-[4.5]=0.2,
890: [5.8]-[8.0]=0.2$, which are consistent with the soft X-ray sources in
891: \citet{hjf+06}. Essentially all of the 3-D Bayesian selected objects
892: lie in the Lacy wedge, but only populate a fraction of that parameter
893: space, consistent with our expectation that the Lacy wedge should
894: suffer contamination at fluxes fainter than its defining limits.
895:
896: The upper left-hand panel of Figure~\ref{fig:zhisttrainall} shows the
897: redshift dependence of the fraction of objects recovered by our 3-D
898: MIR-only Bayesian color selection algorithm. Overall the completeness
899: to spectroscopically confirmed type 1 quasars is 95.7\%. However, as
900: with the Stern wedge, the sample is most incomplete over $3.5<z<5.0$,
901: with the liberal training set performing somewhat better over
902: $2.5<z<3.5$, as designed.
903: \fi
904:
905: \subsection{8-D}
906:
907: Quasar selection is essentially an algorithm that identifies outliers
908: from the stellar locus in the optical or the galaxy locus in the MIR.
909: Thus, one might expect that combining optical and MIR photometry
910: together will yield more robust quasar selection than optical or MIR
911: photometry alone as objects only need to be outliers in one dimension.
912: In particular, our desire is to combine the 5 SDSS and all 4 IRAC
913: bandpasses together to recover the $2.5<z<3.0$ quasars lost by optical
914: selection due to stellar locus contamination and the $3.5<z<5.0$
915: quasars lost by MIR selection due to galaxy locus contamination.
916:
917: We have applied our Bayesian selection method to the 8 unique colors
918: afforded by SDSS plus IRAC photometry in Sample B. As with the
919: training sets, we limit Sample B to $56\mu{\rm Jy}<f_{8.0\mu{\rm
920: m}}<10{\rm mJy}$, and $i>14.0$, which reduces the sample from
921: 95,181 objects to 52,659 objects. In all, 5468 quasar candidates were
922: found.
923:
924: The MIR and optical color distributions of these 8-D selected objects
925: are shown in Figures~\ref{fig:iraccolorstest79Dclass} and
926: \ref{fig:sdsscolorstest79Dclass}, respectively. Here we have added
927: theoretical power-law colors to the MIR color-color plots, which
928: demonstrates that the most robust objects tend to have power-law
929: colors in the MIR \citep[e.g.,][]{drp+08}.
930: %\footnote{Deviations from power-laws
931: % are seen in the Stern wedge due to the use of adjacent bandpasses
932: % where the small-wavelength structure can dominate the overall shape
933: % of the MIR spectral energy distribution.}
934: The primary differences
935: between these and 2-D MIR wedge selection are that 541 objects outside
936: of the Stern wedge are now selected and the left (blue) part of the
937: Lacy wedge is more represented. The redshift completeness (to known
938: type 1 quasars) is given in the left-hand panel of
939: Figure~\ref{fig:zhisttrainall}. As compared to MIR-only selection
940: using the Stern wedge (right-hand panel), it is clear that that 8-D
941: MIR+optical selection performs better over $3.5<z<5$. Overall the
942: completeness to spectroscopically confirmed type 1 quasars is 97.7\%.
943: This high completeness suggests that it should be possible to attempt
944: to classify fainter sources and still maintain a reasonably high
945: completeness --- although clearly photometric errors limit how
946: faint the method can be applied.
947:
948: %XXX are DR6 photometric quasars.
949:
950:
951: %awk '$28>0 || $29>0 {print $0}' sdss_irac_qsos_479D.zphot.match.dat | wc
952:
953: \subsection{6-D}
954:
955: These results from our 8-D Bayesian selection are promising; however,
956: once {\em Spitzer} has exhausted its coolant, it will no longer be
957: able to observe in the two longest IRAC bands. Thus an interesting
958: question is how well our methods will work when only IRAC Channels 1
959: and 2 are operational.
960:
961: As such, we have also performed a 6-D Bayesian selection using the 5
962: SDSS bandpasses and only the 2 shortest wavelength IRAC bandpasses.
963: Instead of using Sample A here (which includes object rejected from
964: Sample B due to lack of Channel 3/4 detections), we instead have only
965: considered the same 52,659 magnitude/flux-limited sources from Sample
966: B above, simply ignoring the color information from Channels 3/4. We
967: have made this choice in order to provide the most direct and unbiased
968: comparison between 8-D and 6-D selection. However, we emphasize that,
969: similarly limiting sample A to unsaturated sources and the SWIRE 95\%
970: completeness limit at 3.6$\mu$m (Table~\ref{tab:tab0}) [allowing
971: non-detections in Channels 3/4], we can potentially apply our 6-D
972: algorithm to over 290,000 sources.
973:
974: Overall our 6-D algorithm selects 5222 objects and is not
975: significantly worse than 8-D, with a completeness of 97.1\% with
976: respect to the type 1 quasars in the training set. The completeness
977: as a function of redshift (red line in the left-hand panel of
978: Fig.~\ref{fig:zhisttrainall}) is consistent with that for 8-D
979: selection. The drop in $z>5.5$ completeness is not statistically
980: significant given the small number of objects considered. In all, 324
981: (6\%) objects that were selected by the 8-D algorithm are not 6-D
982: candidates. There are also 78 (1\%) 6-D selected sources that are not
983: 8-D candidates. Figures~\ref{fig:iraccolorstest79Dclass} and
984: \ref{fig:sdsscolorstest79Dclass} show the locations of objects
985: selected by both the 8-D and 6-D algorithms and also those selected by
986: only one.
987:
988:
989: \subsection{Comparison of Selection Methods}
990:
991: Table~\ref{tab:tab2} shows a comparison between our Bayesian selection
992: methods, the Lacy and Stern wedges, and a simple $[3.6]-[4.5]>-0.1$
993: color cut. For bright flux limits ($f_{8.0\mu{\rm m}}>1\,{\rm mJy}$)
994: our Bayesian method agrees well with the Lacy and Stern wedges, with
995: the $[3.6]-[4.5]>-0.1$ cut being about 50\% contaminated. At
996: fainter limits the Lacy wedge is seen to be contaminated by inactive
997: galaxies, the $[3.6]-[4.5]$ cut somewhat less so. Of 8-D Bayesian
998: selected quasars candidates, 99\% are within the Lacy wedge and 90\%
999: are within the Stern wedge; thus we expect the Bayesian sample to be
1000: quite clean.
1001:
1002: In terms of completeness, Figure~\ref{fig:zhisttrainall} demonstrates
1003: that the Lacy wedge and either of our 8-D or 6-D Bayesian algorithms
1004: are quite complete to type 1 quasars at nearly all redshifts. The
1005: comparison sample is the set of spectroscopically-confirmed type 1
1006: quasars that were used to construct our quasar training set. Thus,
1007: Figure~\ref{fig:zhisttrainall} represents the self-selection
1008: completeness. As an additional check, we find that our algorithm
1009: recovers all 51 of the type 1 quasars cataloged by \citet{tim+07} in
1010: the COSMOS field (to $i<21.3$ and $f_{8.0\mu{\rm m}}<56\,\mu$Jy). The
1011: Lacy wedge is much more contaminated (Table~\ref{tab:tab2}) than our
1012: Bayesian-selected samples. The Stern wedge would appear to be the
1013: least contaminated, but is also the least complete, particularly for
1014: $3.5<z<5.0$. A simple $[3.6]-[4.5]$ color cut is somewhat more
1015: complete than Stern wedge selection, but suffers from considerable
1016: contamination. While a $[3.6]-[8.0]$ color cut might be expected to
1017: be cleaner than $[3.6]-[4.5]$, the presence of PAH features at $8\mu$m
1018: causes contamination from low-redshift PAH-dominated galaxies that
1019: overwhelms the loss of stellar contaminants due to the longer
1020: baseline.
1021:
1022: A particularly interesting question is how well our MIR+optical
1023: selection performs above optical-only selection algorithm. For this
1024: comparison, we utilize our optical-only Bayesian-selected quasar
1025: catalog \citep{rmg+09}, which includes unresolved SDSS quasar
1026: candidates to the nominal SDSS magnitude limit of $i<21.3$. We find
1027: that 1702 of 2426 (70\%) of our optical+MIR selected targets are also
1028: selected by the optical-only algorithm (for $i<21.3$ point sources).
1029: However, our MIR+optical selection benefits by adding 2117 extended
1030: sources and 1003 points sources with $i>21.3$ (i.e., nominally fainter
1031: than the SDSS magnitude limit). Thus, using the same SDSS dataset as
1032: a basis, our MIR+optical selection improves quasar selection by a
1033: factor of 3.25 (5546/1702).
1034:
1035: It is instructive to consider the future of MIR-based selection of
1036: quasars in general, even without the benefit of our Bayesian selection
1037: method. Indeed, while the wedge selection methods rely on IRAC
1038: Channels 3 and 4, it is actually Channels 1 and 2 that are most needed
1039: --- in terms of separating AGNs from stars and normal galaxies. As
1040: can be seen in Figure~\ref{fig:iracstargal}, in the absence of
1041: accurate morphological classification, quasars selected with a
1042: $[3.6]-[4.5]$ cut will have less contamination from galaxies than a
1043: $[3.6]-[8.0]$ cut, yet be similarly effective in removing stars.
1044: Using a $[3.6]-[4.5]$ color cut alone, we find that for
1045: $[3.6]-[4.5]>0.1$, 91\% of known quasars are recovered. Relaxing the
1046: cut to $[3.6]-[4.5]>-0.1$ recovers 98\% of quasars, although the
1047: missing objects are predominantly high-$z$. However, we must also
1048: consider the level of contamination. Figure~\ref{fig:s34compeff}
1049: shows the tradeoff between (type 1) completeness and contamination.
1050: Here we have made the simplifying assumption that any object lying
1051: outside {\em either} the Lacy and Stern wedges are not quasars and
1052: that objects lying {\em inside both} the Lacy and Stern wedges are
1053: quasars. For $[3.6]-[4.5]>-0.1$ the contamination fraction is nearly
1054: 60\%, though we caution that this number depends significantly on the
1055: flux limit. More importantly, if we desire to recover all of the
1056: high-$z$ quasars, we must instead use $[3.6]-[4.5]>-0.4$, which has
1057: over 85\% contamination. On the other hand, our 6-D Bayesian
1058: selection (which also uses only Channels 1 and 2) is 97.1\% complete
1059: to known type 1 quasars (including high-$z$), with only 10\%
1060: contamination, using the same criteria.
1061:
1062: %Fraction of 6-D sources outside of wedges
1063: %awk '$32+$33>0 && $26==1&&$27==1 {print $0}' sdss_irac_qsos_479D.zphot.match.dat | wc
1064: % 4986 209412 1452075
1065:
1066: %awk '$32+$33>0 && ($26==0||$27==0) {print $0}' sdss_irac_qsos_479D.zphot.match.dat | wc
1067: % 137 5754 39870
1068:
1069: %137/(4986+137) = 3%
1070:
1071: \section{Catalog}
1072: \label{sec:cat}
1073:
1074: Our catalog is presented in Tables~\ref{tab:cat} and \ref{tab:tab3},
1075: where Table~\ref{tab:cat} describes the columns in
1076: Table~\ref{tab:tab3}. For the sake of completeness, we also tabulate
1077: the 593 objects that were not selected by our Bayesian algorithms but
1078: that otherwise meet our flux/magnitude criteria and are in {\em both}
1079: the Lacy and Stern wedges; these objects are given in
1080: Table~\ref{tab:tab4}. The numbering of objects is common to
1081: Tables~\ref{tab:tab3} and \ref{tab:tab4} and are sorted by right
1082: ascension.
1083:
1084: The first 25 columns in the data tables merely repeat the publicly
1085: available optical and MIR information on these sources, see
1086: \citet{aaa+08} and the references above for more information.
1087: Columns 26--30 deal with object selection as discussed in
1088: \S~\ref{sec:bayes}. Columns 31--38 give photometric redshift
1089: information as discussed in the next section. Columns 39-41 provide
1090: information on previous identification of these objects, whether
1091: photometric \citep{rmg+09}, or spectroscopic (DR5x, \citealt{shr+07};
1092: DR6x, \citealt{aaa+08}; T07x, \citealt{tim+07}; P06x;
1093: \citealt{pap+05}). For the spectroscopic identifications we have simply
1094: repeated the classifications from the indicated references.
1095:
1096: In all there are 5546 objects cataloged in Table~\ref{tab:cat}. Note
1097: that this number is similar to the number of objects in our quasar
1098: training set. This similarity is a result of our using the same flux
1099: limit for both our training and test sets, while we could have
1100: performed 6-D selection to much fainter limits (nearly 6$\times$ as
1101: many objects). As we consider our work here to be a
1102: ``proof-of-concept'', we save fainter 6-D quasar selection as an
1103: exercise for the future after the current catalog has been more fully
1104: validated with spectroscopic observations.
1105:
1106: Thus our procedure has essentially thrown out some wedge-selected
1107: quasar candidates and has included some objects outside of the quasar
1108: color wedges in the MIR. That this is a worthwhile process can be
1109: seen by noting that quasars make up 92\% of the known objects in among
1110: our Bayesian selected targets (Table~\ref{tab:tab3}), while the
1111: fraction of quasars among our rejected targets (Table~\ref{tab:tab4})
1112: is only 29\%. In addition, our Bayesian algorithm was shown to be
1113: more robust than MIR-only color section over $3.5<z<5.0$
1114: (Fig.~\ref{fig:zhisttrainall}), and is less contaminated at fainter
1115: flux limits (Table~\ref{tab:tab2}). These properties will be
1116: particularly beneficial to the deeper census that can be performed
1117: with objects detected in only the 2 bluest IRAC bandpasses.
1118:
1119:
1120: % 534 of 581 qsos and 6 of 21 qsos
1121:
1122: \subsection{Photometric Redshifts}
1123:
1124: While the errors on the IRAC flux densities ($\sim10$\% for IRAC [but
1125: see \citealt{hcs+08}] vs.\ $\sim2$\% for SDSS) are generally too
1126: large to permit accurate MIR-only photometric redshift estimation, the
1127: combination of SDSS and IRAC photometry allows for considerable
1128: leverage in estimating redshifts of AGNs. We have updated the
1129: algorithm described by \citet{rws+01} and \citet{wrs+04} to operate on
1130: any number of color dimensions. In essence, quasar photometric
1131: redshift estimation relies on the small, but distinct, color changes
1132: produced as broad emission lines move through photometric bandpasses.
1133: At $z\gtrsim2.3$, Lyman-$\alpha$ forest absorption mimics the Balmer
1134: break that is so useful in reducing galaxy photometric redshift
1135: estimates to the 2\% level \citep[e.g.,][]{ccs+95}. However, quasar
1136: emission lines are generally strong enough that they also produce
1137: measurable features in the color-redshift relations that can be used
1138: to estimate photometric redshifts (photo-$z$'s). The mean SDSS colors
1139: as a function of redshift were shown most recently by \citet{shr+07},
1140: while the redshift dependence of IRAC colors for SDSS quasars was
1141: given by \citet{hpp+04} and \citet[][Fig.~3]{rls+06}; see also
1142: Figure~\ref{fig:czplotnew}.
1143:
1144: Using the IRAC-detected, spectroscopically-confirmed quasars noted
1145: above, we have updated the quasar color-redshift relations and
1146: computed the photometric redshifts for all of our quasar candidates.
1147: In addition, Table~\ref{tab:cat} provides not only the most likely
1148: photometric redshifts, but also a range and the probability that the
1149: actual redshift is within that range. We compute photometric
1150: redshifts both using both 8 colors and 6 colors (dropping the 2
1151: reddest IRAC bandpasses to simulate the {\em Spitzer} warm mission
1152: data). In the left-hand panel of Figure~\ref{fig:zzplot}, we show
1153: both the 8-D and 6-D photometric redshifts versus spectroscopic
1154: redshifts. The 6-color photo-z's are not significantly worse than the
1155: 8-color photo-z's. This is perhaps not surprising given that most of
1156: the information that comes from adding the IRAC bandpasses is provided
1157: by the $z-[3.6]$ color which, due to the 1 micron inflection, spans an
1158: impressive 2 magnitudes over $0<z<2$ (see the bottom right-hand panel
1159: in Fig.~\ref{fig:czplotnew}) and serves to break most of the redshift
1160: degeneracies seen when computing photometric redshifts from SDSS
1161: photometry alone. Within $\pm 0.3$ in redshift, we find that both the
1162: 8-color and 6-color photo-z's are accurate 93.6\% of the time. In
1163: addition, the majority of the photo-z's are considerably more accurate
1164: than $\pm0.3$; 82--84\% are accurate to $\pm 0.1$ in redshift. This
1165: accuracy is illustrated in the right-hand panel of
1166: Figure~\ref{fig:zzplot}, where we show a histogram of the fractional
1167: redshift errors. Most of the outliers are fainter objects with
1168: $i>19.1$, but even for those the fraction of 6-color photoz's accurate
1169: to $\pm0.3$ is 92\%.
1170:
1171: In principle, any deep optical imaging data can be used to determine
1172: photometric redshifts, but in practice the SDSS filter set is nearly
1173: ideal for photometric redshifts of quasars due to the more ``top-hat''
1174: nature of the SDSS filters than the traditional
1175: Johnson-Morgan/Kron-Cousins filters. We emphasize that
1176: our goal here is primarily to find type 1 quasars for which our
1177: photometric redshift algorithm should work quite well. However, at
1178: faint flux limits and/or larger host galaxy contribution, our
1179: templates fail to yield accurate photo-z's --- necessitating more
1180: careful photometric redshift techniques. For example, if the
1181: multi-wavelength coverage is large enough (e.g., UV, optical, near-IR,
1182: and MIR), it is also possible to determine photometric redshifts
1183: through SED template fitting \citep[e.g.,][]{bba+06,rbo+08,shi+08} which
1184: also enables simultaneous photometric redshift estimation for inactive
1185: galaxies and type 2 quasars.
1186:
1187:
1188: \subsection{Bulk Properties}
1189:
1190: With the aid of photometric redshifts, we can examine other properties
1191: of the catalog. Figure~\ref{fig:nmi} shows the number counts of our
1192: sample as compared to the SDSS-DR3 \citep{rsf+06} and 2QZ results
1193: \citep{csb+04}. Two extreme cuts on our catalog are shown to give the
1194: reader the range of possible values as objects get fainter and
1195: classification (and redshift estimation) become less robust. The
1196: number counts are just slightly above those of the SDSS spectroscopic
1197: sample for relatively bright $z<2.2$ sources, which is consistent with
1198: the fact that fully or partially obscured quasars in our sample have
1199: not been corrected for internal extinction (and thus will be shifted
1200: to fainter bins). As such, even in the presence of a significant
1201: population of obscured sources, we would not expect the number counts
1202: as a function of observed magnitude to be much higher than for the
1203: SDSS spectroscopic sample in this range. For $3<z<5$ quasars, we see
1204: a marked increase in our sample as compared to the SDSS spectroscopic
1205: sample. This is likely a combination of the inclusion of obscured
1206: quasars that are intrinsically brighter and also due to dust reddened
1207: quasars having a greater tendency to (incorrectly) have higher
1208: photometric redshifts (since high-$z$ quasars have redder colors at
1209: short wavelengths).
1210:
1211: %For high-$z$, we estimate that, for our {\em Spitzer}+SDSS selection,
1212: %one needs of order XX deg$^2$ of imaging with {\em
1213: % Spitzer}-IRAC to yield errors as small at $i=21$ as SDSS alone has
1214: %at $i=20$.
1215:
1216: Figure~\ref{fig:zmplot} shows the distribution of our objects in the
1217: absolute magnitude vs.\ redshift plane, and demonstrates how this
1218: sample can be used to help break luminosity-redshift degeneracies
1219: inherent to any flux-limited survey. For example, $z\sim3.5$ quasars
1220: can now be compared to $z\sim2.5$ quasars at the same luminosity.
1221: Note that our $f_{8.0\mu{\rm m}}<56\,\mu$Jy restriction corresponds
1222: roughly to 15--20$\,\mu$Jy at 3.6$\mu$m, which is about a factor of 3
1223: shallower than (5-$\sigma$) SWIRE-depth (120s) IRAC data, thus there
1224: is room for further improvements in dynamic range. Similarly, the
1225: dashed line shows the improvement that can be expected from using the
1226: SDSS southern equatorial stripe (``Stripe 82''), which has up to 100
1227: epochs of SDSS imaging data, yielding a co-added flux limit of
1228: $i\sim23$ \citep{alm+06}. For luminous sources ($M_i\lesssim-23$) the
1229: central engine dominates over the host galaxy, but for less luminous
1230: sources (with $z\lesssim2.5$) we will have to account for a more
1231: significant host galaxy contribution. The need for more dynamic range
1232: in luminosity is illustrated by the dotted black line which traces the
1233: "break" luminosity in the quasar luminosity function (QLF) using the
1234: multi-wavelength bolometric determination of \citet{hrh07}.
1235: %Combining deep optical and MIR data allows for a powerful dataset with
1236: %dynamic range in luminosity that is normally only obtained for IR or
1237: %X-ray selected samples, but with greater ability to cover large areas
1238: %of sky (Stripe 82 covers nearly 300 deg$^2$). Indeed,
1239: As discussed in the introduction, the ability to determine the
1240: luminosity dependence of quasars clustering (particularly at $z>2.5$),
1241: and the slope evolution of the faint end of the quasar luminosity
1242: function are among the most important near-future constraints on
1243: feedback models of galaxy evolution. Our selection algorithm will
1244: help to enable the creation of a quasar sample that can be used the
1245: address these issues.
1246:
1247: \section{Obscured Quasars}
1248: \label{sec:type2}
1249:
1250: %\subsection{Type 2 Quasars}
1251:
1252: Note that, while morphology was used in the creation of the training
1253: sets, no optical morphology information is used in the actual Bayesian
1254: selection. Thus we expect that our selection includes some type 2
1255: quasars in addition to the type 1 quasars. This simply reflects the
1256: fact that
1257: %quasar selection is essentially an outlier algorithm and
1258: type 2 quasars have similar colors in the MIR as type 1 quasars. This
1259: is true for redshifts low enough ($z\la 2$) that the IRAC bands probe
1260: thermal emission of circumnuclear dust. At higher redshifts, as
1261: rest-frame optical and NIR emission moves into the IRAC passbands, MIR
1262: colors of type 2 quasars deviate significantly from those of type 1
1263: quasars; our procedure is not expected to be sensitive to such
1264: objects.
1265:
1266: As \citet{hjf+07} have shown, type 1 and type 2 quasars are relatively
1267: well separated in optical-to-MIR flux ratio; we examine this
1268: distribution for our sources. Figure~\ref{fig:hickoxall} shows the
1269: $r-[4.5]$ vs.\ $[4.5]$ color-mag distribution of our Bayesian-selected
1270: quasar candidates with known type 1 (green) and type 2 (gray) quasars.
1271: While \citet{hjf+07} use luminosity in their plots, our photometric
1272: redshifts will be in error for type 2 quasars, so we have made the
1273: plot in flux units. This choice has no effect on the vertical axis
1274: which is used by \citet{hjf+07} to separate type 1 and type 2 quasars
1275: (modulo the change in units used herein), but it does cause a
1276: different distribution along the horizontal axis. Point sources
1277: (blue) and extended sources (red) have well-separated mean values that
1278: are bifurcated along $r-[4.5]$, similar to the type 1/2 dividing line
1279: advocated by \citet{hjf+07}. This comparison suggests that our sample
1280: contains a significant number of type 2 quasars --- despite the fact
1281: that we have only attempted type 1 selection ({\em and have required
1282: matching to the relatively shallow single-epoch SDSS photometric
1283: catalog}).
1284:
1285: While our sample is certainly not complete to type 2 quasars, we
1286: note that type 2 quasars are more likely to have extended morphologies
1287: and that Figure~\ref{fig:hickoxall} shows a ``valley'' between point
1288: (peak $r-[4.5]\sim2$) and extended (peak $r-[4.5]\sim3$) morphology
1289: quasar candidates at a color of $r-[4.5]\sim2.5$. As such, the most
1290: logical place to look for type 2 quasars among our quasars candidates
1291: would be those extended sources with $r-[4.5]>2.5$ of which there are
1292: 1252 in all. In all 28 of 47 (60\%) known type 2 quasars that we
1293: recover are extended sources with $r-[4.5]>2.5$. Comparing to the
1294: type 2 quasars cataloged by \citet{tim+07} in the COSMOS field, our
1295: algorithm recovers 20 of 66 (30\%) with $i<21.3$ and $f_{8.0\mu{\rm
1296: m}}<56\,\mu$Jy.
1297:
1298: However, this dividing line in $r-[4.5]$ is not absolute and
1299: the optical magnitude and redshift play a significant role in the
1300: location of objects in this diagram. Nevertheless, it seems likely
1301: that $r-[4.5]$ can be used as a crude diagnostic of type 1 vs.\ type 2
1302: AGNs as suggested by \citet{hjf+07}. On the other hand, point sources
1303: with $r-[4.5]\le2.5$ are quite likely to be type 1 quasars and there
1304: are 2536 such objects in our catalog.
1305:
1306:
1307: \iffalse
1308: [MAS: Your operational definition of a type 2 object (extended objects
1309: with AGN colors) wasn't 100% convincing.]
1310:
1311: [NZ: For type 2 quasars, because of the requirement for the point-like
1312: morphology in the optical, there might be a bias toward qso2 with
1313: high IR/optical ratios (because those that live in luminous galaxies
1314: or have significant extended NL emission will be more likely to be
1315: classified as extended).]
1316:
1317: [ML: Include Palomar spectroscopy? Two type 2 candidates were
1318: confirmed (both high-$z$), one type 2 has $r-[4.5]=1.25$ and one
1319: reddened type 1 candidate was confirmed as such but has
1320: $r-[4.5]=3.6$.]
1321: \fi
1322:
1323:
1324: %\subsection{Reddened Type 1 Quasars}
1325:
1326: In addition to quasars whose central engines are fully obscured in the
1327: optical, there also exist quasars that are simply heavily reddened,
1328: but still exhibit broad-line emission features that are characteristic
1329: of type 1 quasars. Predictions of the size of this population range
1330: from $\sim15$\% \citep{rhv+03} to $\sim60$\% or more
1331: \citep[e.g.,][]{whb+03,ghw+07}. Most recently \citet{mhw+08} have
1332: used UKIDSS\footnote{http://www.ukidss.org/} data to argue that the
1333: fraction of type 1 quasars missing from $i$-band selected surveys
1334: (i.e., SDSS) is $\sim$30\%. Recent work suggests that some of this
1335: dust reddening may come from the host galaxy
1336: \citep[e.g.,][]{mgt98,smh+07,dck+07,pwh+08} rather than the putative
1337: dusty torus. Thus a complete census is needed to fully understand the
1338: demographics of black hole fueling (and how it affects galaxy
1339: evolution), and these quasars represent a important, but under-studied
1340: population. As with type 2 quasars, our emphasis on unobscured type 1
1341: quasars should allow for more complete selection of quasars that have
1342: been extincted or reddened out of purely optically selected samples.
1343: Indeed, we are able to recover 6 of the 7 reddened type 1 quasars in
1344: the XFLS area cataloged by \citet{lps+06}.
1345:
1346: % Add Wilkes reference?
1347:
1348: % [GTR: I think that the 2 others
1349: % don't meet the IR flux limits, but I should check.]
1350:
1351:
1352:
1353: %[GTR: Mark -- feel free to add to this discussion, particularly any of
1354: % the stuff with Martinez-Sansigre. As there's not really a catalog
1355: % of ``reddened'' quasars, I thought it was hard to say much more
1356: % here.]
1357:
1358:
1359:
1360: \iffalse
1361: The above plots/diagnostics are for the sample of objects matched to
1362: SDSS photometry. However, as this selection does not actually make
1363: use of any optical information, and the MIR sample is deeper than the
1364: optical for a typical type 1 quasar SED, we also consider the full
1365: IRAC 4-band sample. Figure~99 shows the same for all 4-band IRAC
1366: detections in the XFLS, SWIRE, and COSMOS data sets to a flux limit of
1367: $S_{8\mu \rm m}=40\mu {\rm Jy}$. No matching to SDSS was done in
1368: order to probe as faint as possible -- at the expense of the loss of
1369: morphology information.
1370:
1371: Table~99 is similar to Table~\ref{tab:tab2}, but for the deeper 4-band
1372: IRAC sample without SDSS matching. Again the Lacy and $[3.6]-[4.5]$
1373: cuts are contaminated at the faint end. Of the 16,254 quasar
1374: candidates selected by our Bayesian method, 16175, 13996, and 16214
1375: are selected by the Lacy, Stern, and $[3.6]-[4.5]$ cut, respectively.
1376: Thus, again, we expect the Bayesian sample to be relatively clean.
1377: \fi
1378:
1379: %\section{Warm Mission}
1380:
1381: %[NEED TO UPDATE]
1382:
1383: \section{Conclusions}
1384: \label{sec:conclusions}
1385:
1386: In this paper we present a method to select type 1 quasars from a
1387: combination of optical and MIR photometric data. The method is based
1388: on Bayesian analysis techniques in multi-dimensional MIR+optical color
1389: space. We demonstrate that our algorithm presents a significant
1390: improvement over MIR-only and optical-only selection procedures. Both
1391: the completeness of selection (i.e., the percentage of true quasars
1392: recovered by the method) and the robustness of selection (i.e., the
1393: percentage of objects recovered that are quasars rather than
1394: contaminants) are increased. In all, we catalog 5546 quasar
1395: candidates detected in all four bands of IRAC in $\sim 24$ deg$^2$
1396: ($>200$ deg$^{-2}$), yielding a factor of $\sim20$ increase in density
1397: compared to that of SDSS spectroscopic quasar catalog. Relaxing our
1398: requirement for detections in IRAC Channels 3 and 4 would increase the
1399: quasar density by more than a factor of 5 ($\sim1000$\,deg$^{-2}$).
1400:
1401: Comparison with existing samples shows that the catalog is more than
1402: 95\% complete to known type 1 quasars at all redshifts. By combining
1403: the 5 SDSS and the 4 IRAC bandpasses, we recover the $2.5<z<3.0$
1404: quasars lost by optical selection due to contamination by stars and
1405: the $3.5<z<5.0$ quasars lost by the MIR selection due to contamination
1406: by star-forming galaxies. Furthermore, we find that combining optical
1407: and MIR data allows selection of quasar candidates to much fainter
1408: fluxes than those afforded by the MIR cuts currently in use in the
1409: literature. At the same time, working in the optical+MIR color-color
1410: space greatly helps with rejecting contaminants (stars and inactive
1411: galaxies) in an efficient manner.
1412:
1413: Inclusion of MIR data significantly improves photometric redshift
1414: estimation for type 1 quasars. The fraction of quasars with redshifts
1415: within 0.3 of the true values increases from $\sim 80\%$ for the
1416: optical-only photo-$z$ to $\sim94$\% for optical+MIR photo-$z$. Much of
1417: this improvement is due to a rapid change of $z-[3.6]$ color of
1418: quasars as a function of redshift at $z\la 2$.
1419:
1420: We demonstrate that removing the two longest wavelength IRAC channels
1421: has little detrimental effect on the selection procedure and on the
1422: quality of photometric redshift estimates. Therefore our method can
1423: be successfully used on data collected during the warm extension of
1424: the {\it Spitzer} mission.
1425: %For example, an IRAC survey with
1426: %SWIRE-depth imaging ($f_{3.6\mu{\rm m}}>4\mu{\rm Jy}$) would be
1427: %well-matched to the deep optical catalogs from SDSS Stripe 82 or the
1428: %CFHT Legacy Survey ``Wide''
1429: %fields\footnote{http://www.cfht.hawaii.edu/Science/CFHLS/cfhtlsgoals.html}.
1430: %limits of SDSS Stripe 82 ($i\sim23.3$) for a quasar spectral energy
1431: %distribution.
1432: %Indeed, SWIRE depth imaging is nearly perfectly matched to the deeper
1433: %magnitude limits of SDSS Stripe 82 (which covers $\sim300\,$deg$^2$
1434: %over $-60^{\circ}<\alpha_{\rm J2000}<60^{\circ}$ and
1435: %$-1.25^{\circ}<\delta_{\rm J2000}<1.25^{\circ}$) for a quasar
1436: %spectral energy distributions.
1437: %An IRAC-mapped region of SDSS Stripe 82 (which covers
1438: %$\sim300\,$deg$^2$ over $-60^{\circ}<\alpha<60^{\circ}$ and
1439: %$-1.25^{\circ}<\delta<1.25^{\circ}$)
1440: A wide-area sample of overlapping deep optical and MIR data would make
1441: groundbreaking contributions to our understanding of quasar feedback
1442: and the evolution of galaxies by breaking the redshift-luminosity
1443: degeneracy inherent to current quasars surveys (Fig~\ref{fig:zmplot}).
1444:
1445: Although the algorithm was primarily designed to be complete for
1446: selection of type 1 quasars, it is also sensitive to at least some
1447: type 2 quasars. This is possible because MIR colors of low-redshift
1448: quasars are dominated by the thermal emission of circumnuclear dust
1449: and are therefore similar for type 1 and type 2 objects. We estimate
1450: that as many as 1200 of our quasar candidates are type 2 quasars.
1451: Although our procedure is not complete for type 2 quasars, our work
1452: lays the foundation for identification of type 2 quasars using modern
1453: statistical methods.
1454:
1455: \iffalse
1456: While MIR selection of AGNs has enabled great progress in AGN
1457: identification in recent years (particularly with regard to obscured
1458: quasars), we have shown that using modern Bayesian selection
1459: techniques on the combination of optical and MIR data allows for more
1460: complete and more efficient quasar selection to even deeper limits
1461: than is afforded by the use of MIR color cuts alone. Although we
1462: concentrate on type 1 (unobscured) quasars, our method is also
1463: sensitive to type 2 (obscured) quasars and lays the groundwork for
1464: improvement in future identification of type 2 quasars using modern
1465: statistical methods. In addition to quasar selection, we showed that
1466: coupling optical and MIR data together allow for significant
1467: improvements in estimating redshifts for type 1 quasars from
1468: photometry alone. With sufficient multiwavelength coverage over large
1469: areas, in the near future it should be possible to extend this
1470: accuracy to type 2 quasars.
1471:
1472: In all we catalog 5546 optical+MIR quasar candidates in just $\sim24$
1473: deg$^2$ (>200 deg$^{-2}$), yielding a far greater density than from SDSS
1474: data alone ($\sim10$ deg$^2$), despite using only the single-epoch
1475: SDSS imaging data. Comparison with existing samples shows that the
1476: catalog is no less than 90\% complete to type 1 quasars. In addition,
1477: as many as 1200 of our quasar candidates are expected to be type 2
1478: quasars. Inclusion of MIR photometry significantly improves
1479: photometric redshift estimation for broad-line quasars from $\sim80$\%
1480: to nearly $95$\%.
1481:
1482: Indeed, the combination of optical and MIR data is sufficiently
1483: powerful that quasar selection and photometric redshift identification
1484: suffer little from the loss of {\em Spitzer's} longest wavelength
1485: channels. We note that the SWIRE depth imaging is nearly perfectly
1486: matched to the magnitude limits of SDSS Stripe 82 (which covers
1487: $\sim300$deg$^2$ over $-60<{\rm RA(deg)}<60$ and $-1.25<{\rm Dec
1488: (deg)}<1.25$ for quasars spectral energy distributions. In addition
1489: to a significant amount of multi-wavelength data in the UV, NIR, mm,
1490: and radio bandpasses, the multiplexed nature of the Stripe 82 data
1491: allows uniquely powerful investigations leveraged from variability,
1492: proper motion, and astrometric redshifts. Thus we look forward to the
1493: contributions that an IRAC-mapped region of SDSS Stripe 82 would make
1494: to our understanding of quasar feedback and the evolution of galaxies
1495: by breaking the redshift-luminosity degeneracy inherent to current
1496: quasars surveys.
1497: \fi
1498:
1499: \iffalse
1500: Figure~\ref{fig:limitssedwarm} compares the flux limits of a 120s
1501: depth SWIRE-like {\em Spitzer} warm mission $S_{3.6\mu \rm m}=6\mu
1502: {\rm Jy}$ to the flux limits of the UKIDSS survey, the SDSS main
1503: survey, SDSS Stripe 82, and GALEX. For such a depth in the MIR, SDSS
1504: Stripe 82 is more than deep enough to recover type 1 quasars at all
1505: redshifts, including quasars reddened by as much at $E(B-V)=0.4$ at
1506: $z=1$ (and even more reddened at higher redshift) and will detect type
1507: 2 quasars to somewhat shallower MIR limits. For those objects with
1508: depth comparable to the SDSS main survey in the optical, UKIDSS and
1509: {\em GALEX} detections will be available, allowing nearly perfect
1510: photometric redshift determination (for $0.9<z<2.4$ UKIDSS is
1511: particularly powerful as it traces the H$\alpha$ emission line).
1512:
1513: [ADM: Be more specific here.]
1514:
1515: Covering an area of only a few hundred square degrees will allow for a
1516: very powerful MIR+optical AGN survey. Specifically, one that can
1517: break the long-standing L-z degeneracy and tackle the issue of the
1518: luminosity dependence of quasar clustering \citep{lhc+06}, and the
1519: nature of black hole feeding at $z>3$ \citep{hlh+07}. [GTR: More
1520: here. Maybe justify an area. Basically include some stuff that we
1521: can put into the ADIOS proposal.]
1522:
1523: [Stripe 82 is bounded by $-60<{\rm RA(deg)}<60$ and $-1.25<{\rm
1524: Dec}<1.25$ and includes over 60 repeat observations.]
1525:
1526: Write Some. See Abstract for now.
1527: \fi
1528:
1529:
1530: \iffalse
1531: [ML: One thing we did find with the "Lacy" wedge was that it was somewhat
1532: incomplete at z<0.3 if the objects had strong PAHs (i.e. were composite
1533: AGN/starbursts with red [4.5]-[8.0]) so that might be worth mentioning (I
1534: think it's in the Lacy et al. 2007 AJ paper), presumably the Stern wedge
1535: is similarly affected in [5.8]-[8.0].]
1536: \fi
1537:
1538:
1539: \acknowledgments
1540:
1541: GTR acknowledges support from an Alfred P. Sloan Research Fellowship,
1542: a Gordon and Betty Moore Fellowship in Data Intensive Sciences, and
1543: NASA grants NNX06AE52G and 07-ADP07-0035. We thank Michael Strauss
1544: for critical review of the paper, and Ryan Riegel for help with the
1545: non-parametric classificaiton algorithm. Funding for the SDSS and
1546: SDSS-II has been provided by the Alfred P. Sloan Foundation, the
1547: Participating Institutions, the National Science Foundation, the
1548: U.S. Department of Energy, the National Aeronautics and Space
1549: Administration, the Japanese Monbukagakusho, the Max Planck Society,
1550: and the Higher Education Funding Council for England. The SDSS is
1551: managed by the Astrophysical Research Consortium for the Participating
1552: Institutions. The Participating Institutions are the American Museum
1553: of Natural History, Astrophysical Institute Potsdam, University of
1554: Basel, Cambridge University, Case Western Reserve University,
1555: University of Chicago, Drexel University, Fermilab, the Institute for
1556: Advanced Study, the Japan Participation Group, Johns Hopkins
1557: University, the Joint Institute for Nuclear Astrophysics, the Kavli
1558: Institute for Particle Astrophysics and Cosmology, the Korean
1559: Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos
1560: National Laboratory, the Max-Planck-Institute for Astronomy (MPIA),
1561: the Max-Planck-Institute for Astrophysics (MPA), New Mexico State
1562: University, Ohio State University, University of Pittsburgh,
1563: University of Portsmouth, Princeton University, the United States
1564: Naval Observatory, and the University of Washington.
1565:
1566: This work is based [in part] on archival data obtained with the
1567: Spitzer Space Telescope, which is operated by the Jet Propulsion
1568: Laboratory, California Institute of Technology under a contract with
1569: NASA. Support for this work was provided by an award issued by
1570: JPL/Caltech (\#1290740). Part of this work is based on observations
1571: obtained with {\em XMM-Newton}, an ESA science mission with
1572: instruments and contributions directly funded by ESA Member States and
1573: the USA (NASA). Some data presented here were obtained at KPNO, a
1574: division of the NOAO, which is operated by AURA under cooperative
1575: agreement with NSF.
1576:
1577: %% To help institutions obtain information on the effectiveness of their
1578: %% telescopes, the AAS Journals has created a group of keywords for telescope
1579: %% facilities. A common set of keywords will make these types of searches
1580: %% significantly easier and more accurate. In addition, they will also be
1581: %% useful in linking papers together which utilize the same telescopes
1582: %% within the framework of the National Virtual Observatory.
1583: %% See the AASTeX Web site at http://www.journals.uchicago.edu/AAS/AASTeX
1584: %% for information on obtaining the facility keywords.
1585:
1586: %% After the acknowledgments section, use the following syntax and the
1587: %% \facility{} macro to list the keywords of facilities used in the research
1588: %% for the paper. Each keyword will be checked against the master list during
1589: %% copy editing. Individual instruments or configurations can be provided
1590: %% in parentheses, after the keyword, but they will not be verified.
1591:
1592: {\it Facilities:} \facility{Sloan, Spitzer, XMM, Mayall}.
1593:
1594: \appendix
1595:
1596: \section{X-ray Observations}
1597:
1598: We have reduced data from {\em XMM-Newton} on 2 fields in the XFLS
1599: area, which is contained in this catalog (Table~\ref{tab:tab3}). As
1600: the programs that these fields were part of were not completed, the
1601: data has not appeared elsewhere, but they are nevertheless useful and
1602: we catalog them in Table~\ref{tab:xmm}. Five of the 24 detected
1603: sources are quasar candidates in our catalog and 12 of the sources
1604: have known spectroscopic redshifts. We have used the XAssist package
1605: \citep{pg03} to reduce the data from these observations. As this is a
1606: fully automated processing routine, it is possible that more accurate
1607: results could be obtained with more careful data reduction; however,
1608: the XAssist results are more than suitable for our purposes here given
1609: the incomplete nature of the observations. In addition to the
1610: coordinates and exposure times (ks), Table~\ref{tab:xmm} gives the
1611: total flux, soft- and hard-band counts (background corrected),
1612: hardness ratio (h-s)/(h+s) and its error. We also denote which X-ray
1613: sources appear in our quasar catalog (by the catalog ID number) along
1614: with any previous identifications (see the references in
1615: \S~\ref{sec:cat}, with two additional objects matched to
1616: \citealt{lacy05} and \citealt{fms+06}).
1617:
1618: %\bibliography{sdsstech,flsqsos,richards_shortnames,new,rpd}
1619: %\bibliographystyle{apj3}
1620:
1621: \begin{thebibliography}{93}
1622: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
1623: \expandafter\ifx\csname url\endcsname\relax
1624: \def\url#1{{\tt #1}}\fi
1625: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
1626: \providecommand{\eprint}[2][]{\url{#2}}
1627:
1628: \bibitem[\protect\astroncite{{Adelman-McCarthy} et~al.}{2007}]{sdssdr6}
1629: {Adelman-McCarthy}, J.~K., et~al. 2007, ArXiv e-prints, 707, \eprint{0707.3413}
1630:
1631: \bibitem[\protect\astroncite{{Adelman-McCarthy} et~al.}{2008}]{aaa+08}
1632: --- 2008, \apjs, 175, 297, \eprint{arXiv:0707.3413}
1633:
1634: \bibitem[\protect\astroncite{{Annis} et~al.}{2006}]{alm+06}
1635: {Annis}, J.~T., et~al. 2006, in Bulletin of the American Astronomical Society,
1636: vol.~38 of {\em Bulletin of the American Astronomical Society\/}, 1197--+
1637:
1638: \bibitem[\protect\astroncite{{Arnaboldi} et~al.}{2007}]{anp+07}
1639: {Arnaboldi}, M., {Neeser}, M.~J., {Parker}, L.~C., {Rosati}, P., {Lombardi},
1640: M., {Dietrich}, J.~P., \& {Hummel}, W. 2007, The Messenger, 127, 28
1641:
1642: \bibitem[\protect\astroncite{{Ball} et~al.}{2006}]{bbm+06}
1643: {Ball}, N.~M., {Brunner}, R.~J., {Myers}, A.~D., \& {Tcheng}, D. 2006, \apj,
1644: 650, 497, \eprint{arXiv:astro-ph/0606541}
1645:
1646: \bibitem[\protect\astroncite{{Begelman}}{2004}]{beg04}
1647: {Begelman}, M.~C. 2004, in Coevolution of Black Holes and Galaxies, ed. L.~C.
1648: {Ho}, 374--+
1649:
1650: \bibitem[\protect\astroncite{{Brodwin} et~al.}{2006}]{bba+06}
1651: {Brodwin}, M., et~al. 2006, \apj, 651, 791, \eprint{arXiv:astro-ph/0607450}
1652:
1653: \bibitem[\protect\astroncite{{Cattaneo} et~al.}{2006}]{cdd+06}
1654: {Cattaneo}, A., {Dekel}, A., {Devriendt}, J., {Guiderdoni}, B., \& {Blaizot},
1655: J. 2006, \mnras, 370, 1651, \eprint{arXiv:astro-ph/0601295}
1656:
1657: \bibitem[\protect\astroncite{{Connolly} et~al.}{1995}]{ccs+95}
1658: {Connolly}, A.~J., {Csabai}, I., {Szalay}, A.~S., {Koo}, D.~C., {Kron}, R.~G.,
1659: \& {Munn}, J.~A. 1995, \aj, 110, 2655, \eprint{arXiv:astro-ph/9508100}
1660:
1661: \bibitem[\protect\astroncite{{Croom} et~al.}{2004}]{csb+04}
1662: {Croom}, S.~M., {Smith}, R.~J., {Boyle}, B.~J., {Shanks}, T., {Miller}, L.,
1663: {Outram}, P.~J., \& {Loaring}, N.~S. 2004, \mnras, 349, 1397,
1664: \eprint{arXiv:astro-ph/0403040}
1665:
1666: \bibitem[\protect\astroncite{{Croton} et~al.}{2006}]{csw+06}
1667: {Croton}, D.~J., et~al. 2006, \mnras, 365, 11, \eprint{arXiv:astro-ph/0508046}
1668:
1669: \bibitem[\protect\astroncite{{da {\^A}ngela} et~al.}{2008}]{dsc+08}
1670: {da {\^A}ngela}, J., et~al. 2008, \mnras, 383, 565,
1671: \eprint{arXiv:astro-ph/0612401}
1672:
1673: \bibitem[\protect\astroncite{{Deo} et~al.}{2007}]{dck+07}
1674: {Deo}, R.~P., {Crenshaw}, D.~M., {Kraemer}, S.~B., {Dietrich}, M., {Elitzur},
1675: M., {Teplitz}, H., \& {Turner}, T.~J. 2007, \apj, 671, 124,
1676: \eprint{arXiv:0709.3076}
1677:
1678: \bibitem[\protect\astroncite{{Donley} et~al.}{2008}]{drp+08}
1679: {Donley}, J.~L., {Rieke}, G.~H., {Perez-Gonzalez}, P.~G., \& {Barro}, G. 2008,
1680: ArXiv e-prints, 806, \eprint{0806.4610}
1681:
1682: \bibitem[\protect\astroncite{{Dunkley} et~al.}{2008}]{dkn+08}
1683: {Dunkley}, J., et~al. 2008, ArXiv e-prints, 803, \eprint{0803.0586}
1684:
1685: \bibitem[\protect\astroncite{{Eisenhardt} et~al.}{2004}]{esb+04}
1686: {Eisenhardt}, P.~R., et~al. 2004, \apjs, 154, 48
1687:
1688: \bibitem[\protect\astroncite{{Elvis} et~al.}{1994}]{ewm+94}
1689: {Elvis}, M., et~al. 1994, \apjs, 95, 1
1690:
1691: \bibitem[\protect\astroncite{{Fabian}}{1999}]{fab99}
1692: {Fabian}, A.~C. 1999, \mnras, 308, L39, \eprint{arXiv:astro-ph/9908064}
1693:
1694: \bibitem[\protect\astroncite{{Fadda} et al.}{2006}]{fms+06}
1695: Fadda, D., et al.\ 2006, \aj, 131, 2859
1696:
1697: \bibitem[\protect\astroncite{{Fazio} et~al.}{2004}]{fha+04}
1698: {Fazio}, G.~G., et~al. 2004, \apjs, 154, 10
1699:
1700: \bibitem[\protect\astroncite{{Ferrarese} \& {Merritt}}{2000}]{fm00}
1701: {Ferrarese}, L. \& {Merritt}, D. 2000, \apjl, 539, L9,
1702: \eprint{arXiv:astro-ph/0006053}
1703:
1704: \bibitem[\protect\astroncite{{Fukugita} et~al.}{1996}]{fig+96}
1705: {Fukugita}, M., {Ichikawa}, T., {Gunn}, J.~E., {Doi}, M., {Shimasaku}, K., \&
1706: {Schneider}, D.~P. 1996, \aj, 111, 1748
1707:
1708: \bibitem[\protect\astroncite{{Gao} et~al.}{2008}]{gzz08}
1709: {Gao}, D., {Zhang}, Y.-X., \& {Zhao}, Y.-H. 2008, \mnras, 386, 1417
1710:
1711: \bibitem[\protect\astroncite{{Gebhardt} et~al.}{2000}]{gbb+00}
1712: {Gebhardt}, K., et~al. 2000, \apjl, 539, L13, \eprint{arXiv:astro-ph/0006289}
1713:
1714: \bibitem[\protect\astroncite{{Glikman} et~al.}{2006}]{ghw06}
1715: {Glikman}, E., {Helfand}, D.~J., \& {White}, R.~L. 2006, \apj, 640, 579,
1716: \eprint{arXiv:astro-ph/0511640}
1717:
1718: \bibitem[\protect\astroncite{{Glikman} et~al.}{2007}]{ghw+07}
1719: {Glikman}, E., {Helfand}, D.~J., {White}, R.~L., {Becker}, R.~H., {Gregg},
1720: M.~D., \& {Lacy}, M. 2007, \apj, 667, 673, \eprint{arXiv:0706.3222}
1721:
1722: \bibitem[\protect\astroncite{{Gray} et~al.}{2005}]{grn+05}
1723: {Gray}, A.~G., {Richards}, G., {Nichol}, R., {Brunner}, R., \& {Moore}, A.~W.
1724: 2005, in Proceedings of PhyStat 05 (Statistical Problems in Particle Physics,
1725: Astrophysics and Cosmology)
1726:
1727: \bibitem[\protect\astroncite{{Hao} et al.}{2005}]{hss+05}
1728: Hao, L., et al.\ 2005, \apjl, 625, L75 \eprint{arXiv:astro-ph/0504423}
1729:
1730: \bibitem[\protect\astroncite{{Hastie} et~al.}{2001}]{htf01}
1731: {Hastie}, T., {Tibshiani}, R., \& {Friedman}, J. 2001 (Springer-Verlag: New
1732: York)
1733:
1734: \bibitem[\protect\astroncite{{Hatziminaoglou} et~al.}{2005}]{hpp+04}
1735: {Hatziminaoglou}, E., et~al. 2005, \aj, 129, 1198
1736:
1737: \bibitem[\protect\astroncite{{Hickox} et~al.}{2006}]{hjf+06}
1738: {Hickox}, R.~C., et~al. 2006, ArXiv Astrophysics e-prints,
1739: \eprint{astro-ph/0603678}
1740:
1741: \bibitem[\protect\astroncite{{Hickox} et~al.}{2007}]{hjf+07}
1742: --- 2007, \apj, 671, 1365, \eprint{arXiv:0708.3678}
1743:
1744: \bibitem[\protect\astroncite{{Hopkins} et~al.}{2006}]{hhc+06}
1745: {Hopkins}, P.~F., {Hernquist}, L., {Cox}, T.~J., {Di Matteo}, T., {Robertson},
1746: B., \& {Springel}, V. 2006, \apjs, 163, 1, \eprint{arXiv:astro-ph/0506398}
1747:
1748: \bibitem[\protect\astroncite{{Hopkins} et~al.}{2008}]{hhc+08}
1749: {Hopkins}, P.~F., {Hernquist}, L., {Cox}, T.~J., \& {Kere{\v s}}, D. 2008,
1750: \apjs, 175, 356, \eprint{0706.1243}
1751:
1752: \bibitem[\protect\astroncite{{Hopkins} et~al.}{2007}]{hrh07}
1753: {Hopkins}, P.~F., {Richards}, G.~T., \& {Hernquist}, L. 2007, \apj, 654, 731,
1754: \eprint{arXiv:astro-ph/0605678}
1755:
1756: \bibitem[\protect\astroncite{{Hora} et~al.}{2008}]{hcs+08}
1757: {Hora}, J.~L., et~al. 2008, ArXiv e-prints, 809, \eprint{0809.3411}
1758:
1759: \bibitem[\protect\astroncite{{Ivezic} et~al.}{2008}]{lsst08}
1760: {Ivezic}, Z., {Tyson}, J.~A., {Allsman}, R., {Andrew}, J., {Angel}, R., \& {for
1761: the LSST Collaboration} 2008, ArXiv e-prints, 805, \eprint{0805.2366}
1762:
1763: \bibitem[\protect\astroncite{{Jiang} et~al.}{2006}]{jfh+06}
1764: {Jiang}, L., et~al. 2006, \aj, 132, 2127, \eprint{arXiv:astro-ph/0608006}
1765:
1766: \bibitem[\protect\astroncite{{Kaiser} et~al.}{2002}]{kab+02}
1767: {Kaiser}, N., et~al. 2002, in Survey and Other Telescope Technologies and
1768: Discoveries. Edited by Tyson, J. Anthony; Wolff, Sidney. Proceedings of the
1769: SPIE, Volume 4836, pp. 154-164 (2002)., eds. J.~A. {Tyson} \& S.~{Wolff},
1770: vol. 4836 of {\em Presented at the Society of Photo-Optical Instrumentation
1771: Engineers (SPIE) Conference\/}, 154--164
1772:
1773: \bibitem[\protect\astroncite{{Kauffmann} \& {Haehnelt}}{2000}]{kh00}
1774: {Kauffmann}, G. \& {Haehnelt}, M. 2000, \mnras, 311, 576,
1775: \eprint{arXiv:astro-ph/9906493}
1776:
1777: \bibitem[\protect\astroncite{{Lacy} et~al.}{2007{\natexlab{a}}}]{lps+07}
1778: {Lacy}, M., {Petric}, A.~O., {Sajina}, A., {Canalizo}, G., {Storrie-Lombardi},
1779: L.~J., {Armus}, L., {Fadda}, D., \& {Marleau}, F.~R. 2007{\natexlab{a}}, \aj,
1780: 133, 186, \eprint{arXiv:astro-ph/0609594}
1781:
1782: \bibitem[\protect\astroncite{{Lacy} et~al.}{2007{\natexlab{b}}}]{lps+06}
1783: --- 2007{\natexlab{b}}, \aj, 133, 186, \eprint{arXiv:astro-ph/0609594}
1784:
1785: \bibitem[\protect\astroncite{{Lacy} et~al.}{2007{\natexlab{c}}}]{lsp+07}
1786: {Lacy}, M., {Sajina}, A., {Petric}, A.~O., {Seymour}, N., {Canalizo}, G.,
1787: {Ridgway}, S.~E., {Armus}, L., \& {Storrie-Lombardi}, L.~J.
1788: 2007{\natexlab{c}}, \apjl, 669, L61, \eprint{arXiv:0709.4069}
1789:
1790: \bibitem[\protect\astroncite{{Lacy} et~al.}{2004a}]{lss+04}
1791: {Lacy}, M., et~al. 2004a, \apjs, 154, 166
1792:
1793: \bibitem[\protect\astroncite{{Lacy} et~al.}{2005}]{lacy05}
1794: --- 2005, \apjs, 161, 41
1795:
1796: \bibitem[\protect\astroncite{{Lidz} et~al.}{2006}]{lhc+06}
1797: {Lidz}, A., {Hopkins}, P.~F., {Cox}, T.~J., {Hernquist}, L., \& {Robertson}, B.
1798: 2006, \apj, 641, 41, \eprint{arXiv:astro-ph/0507361}
1799:
1800: \bibitem[\protect\astroncite{{Lonsdale} et~al.}{2003}]{lsr+03}
1801: {Lonsdale}, C.~J., et~al. 2003, \pasp, 115, 897
1802:
1803: \bibitem[\protect\astroncite{{Lupton} et~al.}{2001}]{lgi+01}
1804: {Lupton}, R.~H., {Gunn}, J.~E., {Ivezi{\' c}}, Z., {Knapp}, G.~R., {Kent}, S.,
1805: \& {Yasuda}, N. 2001, in ASP Conf. Ser. 238: Astronomical Data Analysis
1806: Software and Systems X, vol.~10, 269
1807:
1808: \bibitem[\protect\astroncite{{Maddox} et~al.}{2008}]{mhw+08}
1809: {Maddox}, N., {Hewett}, P.~C., {Warren}, S.~J., \& {Croom}, S.~M. 2008, \mnras,
1810: 386, 1605, \eprint{arXiv:0802.3650}
1811:
1812: \bibitem[\protect\astroncite{{Mainieri} et~al.}{2002}]{mbh+02}
1813: {Mainieri}, V., {Bergeron}, J., {Hasinger}, G., {Lehmann}, I., {Rosati}, P.,
1814: {Schmidt}, M., {Szokoly}, G., \& {Della Ceca}, R. 2002, \aap, 393, 425,
1815: \eprint{arXiv:astro-ph/0207166}
1816:
1817: \bibitem[\protect\astroncite{{Mainzer} et~al.}{2006}]{WISE}
1818: {Mainzer}, A.~K., {Eisenhardt}, P., {Wright}, E.~L., {Liu}, F.-C., {Irace}, W.,
1819: {Heinrichsen}, I., {Cutri}, R., \& {Duval}, V. 2006, in Space Telescopes and
1820: Instrumentation I: Optical, Infrared, and Millimeter. Edited by Mather, John
1821: C.; MacEwen, Howard A.; de Graauw, Mattheus W. M.. Proceedings of the SPIE,
1822: Volume 6265, pp. 626521 (2006)., vol. 6265 of {\em Presented at the Society
1823: of Photo-Optical Instrumentation Engineers (SPIE) Conference\/}
1824:
1825: \bibitem[\protect\astroncite{{Malkan} et~al.}{1998}]{mgt98}
1826: {Malkan}, M.~A., {Gorjian}, V., \& {Tam}, R. 1998, \apjs, 117, 25,
1827: \eprint{arXiv:astro-ph/9803123}
1828:
1829: \bibitem[\protect\astroncite{{Myers} et~al.}{2007}]{mbn+07}
1830: {Myers}, A.~D., {Brunner}, R.~J., {Nichol}, R.~C., {Richards}, G.~T.,
1831: {Schneider}, D.~P., \& {Bahcall}, N.~A. 2007, \apj, 658, 85,
1832: \eprint{arXiv:astro-ph/0612190}
1833:
1834: \bibitem[\protect\astroncite{{Norman} et~al.}{2002}]{nhg+02}
1835: {Norman}, C., et~al. 2002, \apj, 571, 218, \eprint{arXiv:astro-ph/0103198}
1836:
1837: \bibitem[\protect\astroncite{{Papovich} et~al.}{2006}]{pap+05}
1838: {Papovich}, C., et~al. 2006, \aj, 132, 231, \eprint{arXiv:astro-ph/0512623}
1839:
1840: \bibitem[\protect\astroncite{{Patten} et~al.}{2006}]{psb+06}
1841: {Patten}, B.~M., et~al. 2006, \apj, 651, 502, \eprint{arXiv:astro-ph/0606432}
1842:
1843: \bibitem[\protect\astroncite{{Polletta} et~al.}{2008}]{pwh+08}
1844: {Polletta}, M., {Weedman}, D., {H{\"o}nig}, S., {Lonsdale}, C.~J., {Smith},
1845: H.~E., \& {Houck}, J. 2008, \apj, 675, 960, \eprint{arXiv:0709.4458}
1846:
1847: \bibitem[\protect\astroncite{{Polletta} et~al.}{2007}]{ptm+07}
1848: {Polletta}, M., et~al. 2007, \apj, 663, 81
1849:
1850: \bibitem[\protect\astroncite{{Polletta} et~al.}{2006}]{pws+06}
1851: {Polletta}, M.~d.~C., et~al. 2006, \apj, 642, 673,
1852: \eprint{arXiv:astro-ph/0602228}
1853:
1854: \bibitem[\protect\astroncite{{Porciani} \& {Norberg}}{2006}]{pn06}
1855: {Porciani}, C. \& {Norberg}, P. 2006, \mnras, 371, 1824,
1856: \eprint{arXiv:astro-ph/0607348}
1857:
1858: \bibitem[\protect\astroncite{{Ptak} \& {Griffiths}}{2003}]{pg03}
1859: {Ptak}, A. \& {Griffiths}, R. 2003, in Astronomical Data Analysis Software and
1860: Systems XII, eds. H.~E. {Payne}, R.~I. {Jedrzejewski}, \& R.~N. {Hook}, vol.
1861: 295 of {\em Astronomical Society of the Pacific Conference Series\/}, 465--+
1862:
1863: \bibitem[\protect\astroncite{{Ptak} et~al.}{2007}]{pmh+07}
1864: {Ptak}, A., {Mobasher}, B., {Hornschemeier}, A., {Bauer}, F., \& {Norman}, C.
1865: 2007, \apj, 667, 826
1866:
1867: \bibitem[\protect\astroncite{{Reyes} et~al.}{2008}]{rzs+08}
1868: {Reyes}, R., et~al. 2008, ArXiv e-prints, 801, \eprint{0801.1115}
1869:
1870: \bibitem[\protect\astroncite{{Richards} et~al.}{2001}]{rws+01}
1871: {Richards}, G.~T., et~al. 2001, \aj, 122, 1151, \eprint{arXiv:astro-ph/0106038}
1872:
1873: \bibitem[\protect\astroncite{{Richards} et~al.}{2002}]{rfn+02}
1874: --- 2002, \aj, 123, 2945
1875:
1876: \bibitem[\protect\astroncite{{Richards} et~al.}{2003}]{rhv+03}
1877: --- 2003, \aj, 126, 1131
1878:
1879: \bibitem[\protect\astroncite{{Richards} et~al.}{2004}]{rng+04}
1880: --- 2004, \apjs, 155, 257
1881:
1882: \bibitem[\protect\astroncite{Richards et~al.}{2006}]{rsf+06}
1883: Richards, G.~T., et~al. 2006, AJ, 131, 2766
1884:
1885: \bibitem[\protect\astroncite{{Richards} et~al.}{2006}]{rls+06}
1886: {Richards}, G.~T., et~al. 2006, \apjs, 166, 470,
1887: \eprint{arXiv:astro-ph/0601558}
1888:
1889: \bibitem[\protect\astroncite{{Richards} et~al.}{2008}]{rmg+09}
1890: --- 2008, ArXiv e-prints, \eprint{0809.3952}
1891:
1892: \bibitem[\protect\astroncite{{Rowan-Robinson} et~al.}{2008}]{rbo+08}
1893: {Rowan-Robinson}, M., et~al. 2008, \mnras, 386, 697, \eprint{arXiv:0802.1890}
1894:
1895: \bibitem[\protect\astroncite{{Salvato} et~al.}{2008}]{shi+08}
1896: {Salvato}, M., et~al. 2008, ArXiv e-prints, 809, \eprint{0809.2098}
1897:
1898: \bibitem[\protect\astroncite{{Sanders} et~al.}{2007}]{scosmos07}
1899: {Sanders}, D.~B., et~al. 2007, \apjs, 172, 86, \eprint{arXiv:astro-ph/0701318}
1900:
1901: \bibitem[\protect\astroncite{{Sawicki}}{2002}]{saw02}
1902: {Sawicki}, M. 2002, \aj, 124, 3050, \eprint{arXiv:astro-ph/0209437}
1903:
1904: \bibitem[\protect\astroncite{{Schlegel} et~al.}{1998}]{sfd98}
1905: {Schlegel}, D.~J., {Finkbeiner}, D.~P., \& {Davis}, M. 1998, \apj, 500, 525
1906:
1907: \bibitem[\protect\astroncite{{Schneider} et~al.}{2007}]{shr+07}
1908: {Schneider}, D.~P., et~al. 2007, \aj, 134, 102, \eprint{arXiv:0704.0806}
1909:
1910: \bibitem[\protect\astroncite{{Scranton} et~al.}{2002}]{sjd+02}
1911: {Scranton}, R., et~al. 2002, \apj, 579, 48, \eprint{arXiv:astro-ph/0107416}
1912:
1913: \bibitem[\protect\astroncite{{Silk} \& {Rees}}{1998}]{sr98}
1914: {Silk}, J. \& {Rees}, M.~J. 1998, \aap, 331, L1,
1915: \eprint{arXiv:astro-ph/9801013}
1916:
1917: \bibitem[\protect\astroncite{{Spoon} et~al.}{2007}]{smh+07}
1918: {Spoon}, H.~W.~W., {Marshall}, J.~A., {Houck}, J.~R., {Elitzur}, M., {Hao}, L.,
1919: {Armus}, L., {Brandl}, B.~R., \& {Charmandaris}, V. 2007, \apjl, 654, L49,
1920: \eprint{arXiv:astro-ph/0611918}
1921:
1922: \bibitem[\protect\astroncite{{Stern} et~al.}{2002}]{smc+02}
1923: {Stern}, D., et~al. 2002, \apj, 568, 71, \eprint{arXiv:astro-ph/0111513}
1924:
1925: \bibitem[\protect\astroncite{{Stern} et~al.}{2005}]{seg+04}
1926: --- 2005, \apj, 631, 163
1927:
1928: \bibitem[\protect\astroncite{{Stern} et~al.}{2007}]{ska+07}
1929: --- 2007, \apj, 663, 677, \eprint{arXiv:astro-ph/0608603}
1930:
1931: \bibitem[\protect\astroncite{{Storrie-Lombardi} et~al.}{2007}]{warm}
1932: {Storrie-Lombardi}, L.~J., et~al. 2007, in The Science Opportunities of the
1933: Warm Spitzer Mission Workshop, eds. L.~J. {Storrie-Lombardi} \& N.~A.
1934: {Silbermann}, vol. 943 of {\em American Institute of Physics Conference
1935: Series\/}, 67--85
1936:
1937: \bibitem[\protect\astroncite{{Stoughton} et~al.}{2002}]{slb+02}
1938: {Stoughton}, C., et~al. 2002, \aj, 123, 485
1939:
1940: \bibitem[\protect\astroncite{{Sturm} et~al.}{2006}]{shl+06}
1941: {Sturm}, E., {Hasinger}, G., {Lehmann}, I., {Mainieri}, V., {Genzel}, R.,
1942: {Lehnert}, M.~D., {Lutz}, D., \& {Tacconi}, L.~J. 2006, \apj, 642, 81,
1943: \eprint{arXiv:astro-ph/0601204}
1944:
1945: \bibitem[\protect\astroncite{{Tajer} et~al.}{2007}]{tpc+07}
1946: {Tajer}, M., et~al. 2007, \aap, 467, 73, \eprint{arXiv:astro-ph/0703263}
1947:
1948: \bibitem[\protect\astroncite{{The Dark Energy Survey
1949: Collaboration}}{2005}]{des05}
1950: {The Dark Energy Survey Collaboration} 2005, ArXiv Astrophysics e-prints,
1951: \eprint{astro-ph/0510346}
1952:
1953: \bibitem[\protect\astroncite{{Tremaine} et~al.}{2002}]{tgb+02}
1954: {Tremaine}, S., et~al. 2002, \apj, 574, 740
1955:
1956: \bibitem[\protect\astroncite{{Trump} et~al.}{2007}]{tim+07}
1957: {Trump}, J.~R., et~al. 2007, \apjs, 172, 383, \eprint{arXiv:astro-ph/0606016}
1958:
1959: \bibitem[\protect\astroncite{{Weinstein} et~al.}{2004}]{wrs+04}
1960: {Weinstein}, M.~A., et~al. 2004, \apjs, 155, 243
1961:
1962: \bibitem[\protect\astroncite{{Werner} et~al.}{2004}]{wrl+04}
1963: {Werner}, M.~W., et~al. 2004, \apjs, 154, 1
1964:
1965: \bibitem[\protect\astroncite{{White} et~al.}{2003}]{whb+03}
1966: {White}, R.~L., {Helfand}, D.~J., {Becker}, R.~H., {Gregg}, M.~D., {Postman},
1967: M., {Lauer}, T.~R., \& {Oegerle}, W. 2003, \aj, 126, 706,
1968: \eprint{arXiv:astro-ph/0304028}
1969:
1970: \bibitem[\protect\astroncite{{York} et~al.}{2000}]{yaa+00}
1971: {York}, D.~G., et~al. 2000, \aj, 120, 1579
1972:
1973: \bibitem[\protect\astroncite{{Zakamska} et~al.}{2003}]{zsk+03}
1974: {Zakamska}, N.~L., et~al. 2003, \aj, 126, 2125, \eprint{arXiv:astro-ph/0309551}
1975:
1976: \bibitem[\protect\astroncite{{Zheng} et~al.}{2004}]{zmm+04}
1977: {Zheng}, W., et~al. 2004, \apjs, 155, 73, \eprint{arXiv:astro-ph/0406482}
1978:
1979: \end{thebibliography}
1980:
1981: \clearpage
1982:
1983: \begin{figure}
1984: %\plotone{figures/iracstargal.ps}
1985: \plotone{f1.eps}
1986: \caption{Comparison of MIR colors of point ({\em blue} contours/dots)
1987: and extended ({\em red} contours/dots) for various MIR color and
1988: flux combinations. Green points depicts known type 1 quasars, while
1989: open grey squares are type 2 quasars. Open magenta triangles
1990: indicate brown dwarfs \citep{psb+06}. Comparison of these panels
1991: with Table~1 demonstrates that point sources with red MIR colors are
1992: robust AGNs candidates.
1993: %For relatively faint MIR sources, however,
1994: %photometric errors begin to blur the line. Similarly, for faint
1995: %optical sources, star-galaxy separation begins to break down and
1996: %galaxies become significant ``stellar'' contaminants. At $i\sim20.8
1997: %\;(S_{8\mu \rm m}\sim115\mu {\rm Jy}$ at $z=1.5$ for a type 1
1998: %quasars), roughly 10\% of SDSS point sources are likely to actually
1999: %be galaxies.
2000: The dashed lines depict the Lacy wedge region (upper left),
2001: Stern wedge region (lower left) and a simple $[3.6]-[8.0]$ color+morphology
2002: cut (right panels); see \S~\ref{sec:overview}. Statistical errors are generally less than 1\%,
2003: but systematic errors can be $\sim10$\%.
2004: \label{fig:iracstargal}}
2005: \end{figure}
2006:
2007: \begin{figure}
2008: %\plotone{figures/lacywedgemany.ps}
2009: \plotone{f2.eps}
2010: \caption{Illustration of the power of combining {\em Spitzer} data
2011: with morphology information. Top left: All 4-band data. Top-right:
2012: Removed very faint and saturated objects. Bottom-left: Limited to
2013: SDSS point sources. Bottom-right: Limited to SDSS point sources with
2014: $g<21.5$ (grey contours indicate extended sources). Dashed lines in
2015: each panel show the Lacy wedge. In short, MIR-only quasar selection is
2016: robust only for relatively bright sources or when coupled with
2017: accurate morphologies. Similar results would be seen for the Stern
2018: wedge (albeit with somewhat less contamination from galaxies at the
2019: price of reduced completeness to high-$z$ quasars).
2020: \label{fig:lacywedgemany}}
2021: \end{figure}
2022:
2023: \begin{figure}
2024: \epsscale{0.9}
2025: %\plotone{figures/czplotnew.ps}
2026: \plotone{f3.eps}
2027: \caption{MIR colors of known broad-line, type 1 quasars (black x's).
2028: The solid light gray and dashed dark gray curves show theoretical
2029: color redshift relation for a broad-band SED \citep{rls+06} and for
2030: the higher-resolution \citep{ghw06} IR spectral template,
2031: respectively, the latter more accurately representing the effects of
2032: the H$\alpha$ emission line. In the upper two panels the dotted
2033: lines show the blue limits of the selection criteria for the Stern
2034: wedge (see Fig.~\ref{fig:iracstargal}, objects redder than these
2035: lines being inside the Stern wedge; the $[3.6]-[4.5]$ blue limit is
2036: a function of $[5.8]-[8.0]$, here only the extrema are plotted). In
2037: the middle two panels the dashed lines shows the same for the color
2038: combination used in the Lacy wedge (objects redder than these lines
2039: being inside the Lacy wedge). The bottom left-hand panel shows
2040: that, when coupled with morphology (see Fig~\ref{fig:iracstargal}),
2041: relatively complete and efficient selection is possible using only
2042: the $[3.6]-[8.0]$ color (dash-dot line). The bottom right-hand
2043: panels shows the 1$\mu$m inflection induced color change in
2044: $z-[3.6]$, which is extremely beneficial for photo-z's.
2045: \label{fig:czplotnew}}
2046: \end{figure}
2047:
2048: \clearpage
2049:
2050: \begin{figure}
2051: %\plotone{figures/iraccolorstrainlib.ps}
2052: \plotone{f4.eps}
2053: \caption{MIR color-color and color-mag distributions for the quasar (blue
2054: contours/dots) and non-quasar (red contours/dots) training sets. Green
2055: points indicate known quasars; black crosses mark $z>3.5$ quasars.
2056: The long dashed line in the bottom right-hand panel shows our adopted
2057: $8.0\mu$m flux limit and how it is affected by the $3.6\mu$m and
2058: $5.8\mu$m limits.
2059: \label{fig:iraccolorstrainlib}}
2060: \end{figure}
2061:
2062: \clearpage
2063:
2064: \begin{figure}
2065: %\plotone{figures/sdsscolorstrainlib.ps}
2066: \plotone{f5.eps}
2067: \caption{SDSS color-color and color-mag distribution for the quasar (blue
2068: contours/dots) and non-quasar (red contours/dots) training sets. Green
2069: points indicate known quasars.
2070: \label{fig:sdsscolorstrainlib}}
2071: \end{figure}
2072:
2073:
2074: \clearpage
2075:
2076: \begin{figure}
2077: \epsscale{0.9}
2078: %\plotone{figures/wedgecomp.ps}
2079: \plotone{f6.eps}
2080: \caption{Comparison of the two standard MIR ``wedge'' color selection
2081: algorithms. The top left panel shows that of \citet{lss+04} (dashed
2082: lines). The bottom left panel that of \citet{seg+04}, again as
2083: dashed lines. Black contours and dots are all sources detected in
2084: all 4 IRAC bands that have matches to SDSS sources. Green points
2085: indicate known quasars. Red contours/dots show objects inside the
2086: Lacy wedge, but outside the Stern wedge. Purple crosses indicate
2087: objects in the Stern wedge, but outside the Lacy wedge. The right
2088: panels show additional regions of color-color and flux-color space.
2089: The dashed line in the right-hand panels indicates a logical
2090: dividing line between stars and AGNs among point sources. The
2091: region enclosed by yellow dashed lines in the lower left-hand panel
2092: indicates our modified Stern wedge, which is more conservative on
2093: the galaxy/AGN boundary in $[3.6]-[4.5]$, but is more inclusive to
2094: potential AGNs that have blue $[5.8]-[8.0]$ colors. Open grey
2095: squares indicate the colors of 3 confirmed and one likely AGN
2096: resulting from spectroscopic follow-up of 9 random sources with
2097: $[5.8]-[8.0]<-0.07$.
2098: \label{fig:wedgecomp}}
2099: \end{figure}
2100:
2101:
2102: \clearpage
2103:
2104: \begin{figure}
2105: %\plotone{figures/iraccolorstest79Dclass.ps}
2106: \plotone{f7.eps}
2107: \caption{MIR colors/flux densities of Bayesian selected quasars.
2108: Black contours/dots show the full sample. Blue contours/dots show
2109: objects selected by both the 8-D and 6-D Bayesian algorithms. Gold
2110: (red) points are objects selected only by the 8-D (6-D) algorithm.
2111: Known quasars are shown in green with $z>3.5$ quasars shown by black
2112: crosses. Theoretical power-law colors ($-2<\alpha_{\nu}<2$) are
2113: given by the grey line (where we assume a delta function filter
2114: curve at the nominal effective wavelength). Deviations from
2115: power-laws are seen in the Stern wedge due to the use of adjacent
2116: bandpasses where the small-wavelength structure can dominate the
2117: overall shape of the MIR SED.
2118: \label{fig:iraccolorstest79Dclass}}
2119: \end{figure}
2120:
2121: \begin{figure}
2122: %\plotone{figures/sdsscolorstest79Dclass.ps}
2123: \plotone{f8.eps}
2124: \caption{Optical colors/mags of Bayesian selected quasars. Black
2125: contours/dots show the full sample. Blue contours/dots show objects
2126: selected by both the 8-D and 6-D Bayesian algorithms. Gold (red)
2127: points are objects selected only by the 8-D (6-D) algorithm. Again,
2128: known quasars are shown in green.
2129: %with $z>3.5$ quasars shown by black crosses.
2130: \label{fig:sdsscolorstest79Dclass}}
2131: \end{figure}
2132:
2133: \clearpage
2134:
2135: \begin{figure}
2136: %\plotone{figures/zhisttrainall.ps}
2137: \plotone{f9.eps}
2138: \caption{Fraction of known (type 1) quasars recovered. {\em Left:} Our 8-D
2139: (blue) and 6-D(red) Bayesian selection algorithms. {\em Right:} The
2140: Lacy wedge (gray), the Stern wedge (purple) and a $[3.6]-[4.5]$
2141: color cut (green). The dashed black line in each panel indicates
2142: the number of known objects in each redshift bin. The peak at
2143: $z=1.2$ has 89 objects. $z=5.4$ has 0 objects and $z=4.2$ and
2144: $z=4.8$ have only one each, thus statistics are poor in these bins.
2145: \label{fig:zhisttrainall}}
2146: \end{figure}
2147:
2148: \clearpage
2149:
2150: \begin{figure}
2151: %\plotone{figures/s34compeff.ps}
2152: \plotone{f10.eps}
2153: \caption{Completeness vs.\ contamination for quasar selection based on
2154: a cut in $[3.6]-[4.5]$ alone. Crosses indicate the fraction of
2155: known (type 1) quasars recovered as a function of $[3.6]-[4.5]$
2156: color. For $z\lesssim2$, type 2 quasars have similar $[3.6]-[4.5]$
2157: colors and should be equally included/excluded. The squares show
2158: the contamination fraction for the same color cut.
2159: \label{fig:s34compeff}}
2160: \end{figure}
2161:
2162: \begin{figure}
2163: %\plottwo{figures/zzplot.ps}{figures/dzhist2.ps}
2164: \plottwo{f11a.eps}{f11b.eps}
2165: \caption{{\em Left:} Photometric vs.\ spectroscopic redshift for 399 known
2166: quasars. Open black squares show photometric redshifts determined
2167: from 8 colors. Grey crosses show photometric
2168: redshifts determined from 6 colors. {\em Right:} Histogram of the
2169: fractional error in photometric redshift. 8-color photo-z results
2170: are shown by the solid line; 6-color by the dashed line. Little
2171: accuracy is lost by going from 8 to 6 colors. The bin size is 0.05 in redshift.
2172: \label{fig:zzplot}}
2173: \end{figure}
2174:
2175:
2176: \begin{figure}
2177: %\plotone{figures/nmi.ps}
2178: \plotone{f12.eps}
2179: \caption{Number counts ($i$-band) of {\em Spitzer}+SDSS selected quasars as
2180: compared to SDSS and 2QZ number counts. Two redshift bins are
2181: shown: $0.3<z<2.2$ as circles/squares and $3<z<5$ as triangles. Our
2182: new results are shown for two extreme cases: objects selected by
2183: either the 6-D or 8-D algorithms and with photometric redshift
2184: probability larger than 0.5 (closed points) and for objects selected
2185: by both the 6-D and 8-D algorithms and with photometric redshift
2186: probability larger than 0.8 (open points). For objects brighter
2187: than the nominal SDSS flux limit of $i=21.3$, these extremes should
2188: bracket the true values. Points are truncated at the bright end
2189: where the errors become large due to the lack of area.
2190: \label{fig:nmi}}
2191: \end{figure}
2192:
2193:
2194: \begin{figure}
2195: \epsscale{0.9}
2196: %\plotone{figures/zMplot.ps}
2197: \plotone{f13.eps}
2198: \caption{Absolute $i$-band magnitude vs.\ redshift. Known SDSS
2199: quasars are shown as black triangles. Robust MIR+optical quasar
2200: candidates are shown by grey crosses. These are objects detected by
2201: both the 8-D and 6-D Bayesian methods, that have point-like SDSS
2202: morphologies, $i<21.3$ (the SDSS flux limit), and photometric
2203: redshift probabilities greater than 90\%. The remaining quasar
2204: candidates are plotted as red dots and more likely to include
2205: contaminants and erroneous photometric redshifts. The dotted black
2206: line shows the approximate division between the bright and faint
2207: ends of the quasar luminosity function (i.e., $L^*_Q$) as derived
2208: from \citet{hrh07}. This demonstrates that SDSS quasars only probe
2209: the bright end of the QLF, while adding MIR information enables us
2210: to probe the faint end of the QLF. The dashed black line indicates
2211: the depth that can be reached if optical+MIR selection could be
2212: performed to $i=23$ (e.g., SDSS Stripe 82). At that depth, the
2213: faint end of the QLF can be probed to nearly $z=4$, and there is
2214: sufficient dynamic range in luminosity at $z<3$ to determine the
2215: luminosity dependence of quasar clustering as a function of
2216: redshift.
2217: \label{fig:zmplot}} \end{figure}
2218:
2219: \clearpage
2220:
2221: \begin{figure}
2222: %\plotone{figures/hickoxall.ps}
2223: \plotone{f14.eps}
2224: \caption{Optical/MIR color-mag relationship showing the separation of
2225: type 1 and type 2 quasars in this plane, similar to \citet{hjf+07}.
2226: Known type 1 quasars are shown by green crosses. Known type 2
2227: quasars are open grey squares; filled gray triangles indicate those
2228: recovered by our algorithm. Our point source quasar candidates are
2229: shown in blue. Extended quasar candidates are shown in red. The
2230: separation between these populations is similar to that seen by
2231: \citet{hjf+07} for type 1 and type 2 AGNs using a dividing line of
2232: $r-[4.5]=2.5$. Note however, that optical magnitude and redshift
2233: must be considered to some extent. The dotted lines show $i=19.1,
2234: 20.2$, and $22$ from bottom to top and demonstrates that the
2235: apparent diagonal locus of type 1 sources is artificial.
2236: \label{fig:hickoxall}}
2237: \end{figure}
2238:
2239: \iffalse
2240: \begin{figure}
2241: \plotone{figures/limitssedwarm.eps}
2242: \caption{[GTR: For the ADIOS proposal, rather than for the paper but
2243: included here for people to see.] Comparison of {\em Spitzer} warm
2244: mission (120s depth) flux density limits with those of UKIDSS, SDSS
2245: Main, SDSS Stripe 82, and GALEX. The mean SED for a $z=1$, $z=2.5$,
2246: and $z=4$ type 1 quasar are shown. The depth of Stripe 82 is more
2247: than sufficient for finding all such objects for an IRAC depth of
2248: 4.25$\mu$Jy at $3.6\mu$m.
2249: \label{fig:limitssedwarm}}
2250: \end{figure}
2251: \fi
2252:
2253: \begin{deluxetable}{lcccccccc}
2254: %\tabletypesize{\scriptsize}
2255: %\rotate
2256: \tablewidth{0pt}
2257: \tablecaption{Wide-area MIR Field Parameters\label{tab:tab0}}
2258: \tablehead{
2259: \colhead{Field} &
2260: \colhead{RA} &
2261: \colhead{Dec} &
2262: \colhead{Area} &
2263: \colhead{Exp} &
2264: \colhead{$3.6$/$8.0\mu$m $5\sigma$ Depth} &
2265: \colhead{$3.6$/$8.0\mu$m $95\%$ Comp. Depth} \\
2266: \colhead{} &
2267: \colhead{(deg)} &
2268: \colhead{(deg)} &
2269: \colhead{(deg$^2$)} &
2270: \colhead{(s)} &
2271: \colhead{($\mu$Jy)} &
2272: \colhead{($\mu$Jy)}
2273: }
2274: \startdata
2275: XFLS & 259.5 & 59.5 & 4 & 60 & $\cdots$/$\cdots$ & 20(77\%)/100(94\%) \\
2276: Bo\"{o}tes & 218.02 & 34.28 & 8.5 & 90 & 6.4/56 & $\cdots$/$\cdots$ \\
2277: ELAIS-N1 & 242.75 & 55.0 & 9.3 & 120 & 3.7/37.8 & 14/56 \\
2278: ELAIS-N2 & 249.2 & 41.029 & 4.2 & 120 & 3.7/37.8 & 14/56 \\
2279: Lockman & 161.25 & 58.0 & 11.1 & 120 & 3.7/37.8 & 14/56 \\
2280: COSMOS & 150.62 & 2.21 & 2.0 & 1200 & 0.9/14.6 & $\cdots$/$\cdots$ \\
2281: \enddata
2282: \end{deluxetable}
2283:
2284:
2285: \begin{deluxetable}{llllll}
2286: %\tabletypesize{\scriptsize}
2287: %\rotate
2288: \tablewidth{0pt}
2289: \tablecaption{Mean Observed and Theoretical Star and Quasar MIR Colors\label{tab:tab1}}
2290: \tablehead{
2291: \colhead{Color} &
2292: \colhead{Star($\alpha=2$)} &
2293: \colhead{Star(obs.)} &
2294: \colhead{QSO($\alpha=-1$)} &
2295: \colhead{QSO(obs.)}
2296: }
2297: \startdata
2298: $[3.6]-[4.5]$ & $-$0.485 & $-$0.497 & 0.242 & 0.287 \\
2299: $[5.8]-[8.0]$ & $-$0.698 & $-$0.604 & 0.349 & 0.454 \\
2300: $[3.6]-[8.0]$ & $-$1.734 & $-$1.534 & 0.867 & 1.162 \\
2301: $[3.6]-[5.8]$ & $-$1.036 & $-$0.926 & 0.518 & 0.703 \\
2302: $[4.5]-[8.0]$ & $-$1.249 & $-$1.031 & 0.625 & 0.853 \\
2303: \enddata
2304: \tablecomments{For $\alpha$ we adopt the nomenclature: $f_{\nu}\propto\nu^{\alpha}$.}
2305: \end{deluxetable}
2306:
2307:
2308: %combination of wedgestats and wedgestatscat (for 8/6-D)
2309:
2310: \begin{deluxetable}{lrrrr}
2311: %\tabletypesize{\scriptsize}
2312: %\rotate
2313: \tablewidth{0pt}
2314: \tablecaption{MIR Color Selection Comparison\label{tab:tab2}}
2315: \tablehead{
2316: \colhead{} &
2317: \colhead{All} &
2318: \colhead{Point} &
2319: \colhead{Extended} &
2320: \colhead{Bright}
2321: }
2322: \startdata
2323: N Objects & 52659 & 22473 & 30186 & 2225 \\
2324: N 8-D & 5468 & 3426 & 2042 & 273 \\
2325: N 6-D & 5222 & 3426 & 1796 & 271 \\
2326: N Lacy & 15776 & 3424 & 12352 & 299 \\
2327: N Stern & 5659 & 2981 & 2678 & 268 \\
2328: N Ch1/2 Cut & 12360 & 3207 & 9153 & 474 \\
2329: \enddata
2330: \end{deluxetable}
2331:
2332: %% CURRENT
2333: \iffalse
2334: N objects 52659 22473 30186 2225
2335: N 9-D 5497 3451 2046 273
2336: N 7-D 5250 3451 1799 271
2337: N Lacy 15776 3424 12352 299
2338: N Stern 5659 2981 2678 268
2339: N Cut 12360 3207 9153 474
2340: N 9-D and Lacy 5449
2341: N 9-D and Stern 4939
2342: N 9-D and cut 5128
2343: N 7-D and Lacy 5201
2344: N 7-D and Stern 4693
2345: N 7-D and cut 4917
2346: \fi
2347:
2348: \iffalse
2349: %% OLD
2350: N Objects & 53889 & 23688 & 30201 & 2747 \\
2351: %N 3-D & 4323 & 2822 & 1501 & 260 \\
2352: N 8-D & 3828 & 3060 & 768 & 220 \\
2353: N 6-D & 3766 & 3051 & 715 & 224 \\
2354: N Lacy & 15776 & 3424 & 12352 & 299 \\
2355: N Stern & 5659 & 2981 & 2678 & 268 \\
2356: N Ch1/2 Cut & 12360 & 3207 & 9153 & 474 \\
2357: \fi
2358:
2359: \iffalse
2360: \begin{deluxetable}{lrrrr}
2361: %\tabletypesize{\scriptsize}
2362: %\rotate
2363: \tablewidth{0pt}
2364: \tablecaption{Deep MIR Color Selection Comparison\label{tab:tab3}}
2365: \tablehead{
2366: \colhead{} &
2367: \colhead{All} &
2368: \colhead{Bright}
2369: }
2370: \startdata
2371: N objects & 109953 & 4781 \\
2372: N Bayes & 16254 & 405 \\
2373: N Lacy & 42702 & 463 \\
2374: N Stern & 16850 & 408 \\
2375: N Ch1/2 & 30632 & 720 \\
2376: \enddata
2377: \end{deluxetable}
2378: \fi
2379:
2380: %n ra dec name objid u g r i z s3 s4 s5 s8 uerr gerr rerr ierr zerr s3err s4err s5err s8err ured type lacy stern class9cons class9lib class7cons class7lib ckass4cons class4lib zphot9 zphotlo9 zphothi9 zphotprob9 zphot7 zphotlo7 zphothi7 zphotprob7 zem
2381:
2382: \begin{deluxetable}{lcl}
2383: \tabletypesize{\small}
2384: \tablewidth{0pt}
2385: \tablecaption{MIR/Optical Quasar Candidate Catalog Format\label{tab:cat}}
2386: \tablehead{
2387: \colhead{Column} &
2388: \colhead{Format} &
2389: \colhead{Description}
2390: }
2391: \startdata
2392: 1 & I7 & Unique catalog number \\
2393: 2 & F10.6 & Right ascension in decimal degrees (J2000.0) \\
2394: 3 & F10.6 & Declination in decimal degrees (J2000.0) \\
2395: 4 & A18 & Name: SDSS J$hhmmss.ss+ddmmss.s$ (J2000.0) \\
2396: 5 & A19 & SDSS Object ID \\
2397: 6 & F6.3 & $u$ PSF asinh magnitude (corrected for Galactic extinction) \\
2398: 7 & F6.3 & $g$ PSF asinh magnitude (corrected for Galactic extinction) \\
2399: 8 & F6.3 & $r$ PSF asinh magnitude (corrected for Galactic extinction) \\
2400: 9 & F6.3 & $i$ PSF asinh magnitude (corrected for Galactic extinction) \\
2401: 10 & F6.3 & $z$ PSF asinh magnitude (corrected for Galactic extinction) \\
2402: 11 & F8.3 & {\em Spitzer}-IRAC Channel 1 $3.6\mu$m flux density ($\mu$Jy) \\
2403: 12 & F8.3 & {\em Spitzer}-IRAC Channel 2 $4.5\mu$m flux density ($\mu$Jy) \\
2404: 13 & F8.3 & {\em Spitzer}-IRAC Channel 3 $5.8\mu$m flux density ($\mu$Jy) \\
2405: 14 & F8.3 & {\em Spitzer}-IRAC Channel 4 $8.0\mu$m flux density ($\mu$Jy) \\
2406: 15 & F6.3 & Error in PSF $u$ asinh magnitude \\
2407: 16 & F5.3 & Error in PSF $g$ asinh magnitude \\
2408: 17 & F5.3 & Error in PSF $r$ asinh magnitude \\
2409: 18 & F5.3 & Error in PSF $i$ asinh magnitude \\
2410: 19 & F5.3 & Error in PSF $z$ asinh magnitude \\
2411: 20 & F6.3 & Error in $3.6\mu$m flux density ($\mu$Jy) \\
2412: 21 & F6.3 & Error in $4.5\mu$m flux density ($\mu$Jy) \\
2413: 22 & F6.3 & Error in $5.8\mu$m flux density ($\mu$Jy) \\
2414: 23 & F6.3 & Error in $8.0\mu$m flux density ($\mu$Jy) \\
2415: 24 & F6.3 & $u$-band Galactic extinction, $A_u$ (mag); $A_u/A_g/A_r/A_i/A_z=5.155/3.793/2.751/2.086/1.479 \times E(B-V)$ \\
2416: 25 & I1 & SDSS Morphology (point$=6$, extended$=3$) \\
2417: 26 & I1 & Lacy wedge flag (in$=1$, out$=0$) \\
2418: 27 & I1 & Stern wedge flag (in$=1$, out$=0$) \\
2419: 28 & I1 & Modified Stern wedge flag (in$=1$, out$=0$) \\
2420: 29 & I1 & 8-D Bayesian classification (in$=1$, out$=0$) \\
2421: 30 & I1 & 6-D Bayesian classification (in$=1$, out$=0$) \\
2422: 31 & F6.3 & 8-D Photometric redshift (see \citealt{wrs+04})\\
2423: 32 & F6.3 & Lower limit of 8-D photometric redshift range \\
2424: 33 & F6.3 & Upper limit of 8-D photometric redshift range \\
2425: 34 & F6.3 & 8-D Photometric redshift range probability \\
2426: 35 & F6.3 & 6-D Photometric redshift (see \citealt{wrs+04})\\
2427: 36 & F6.3 & Lower limit of 6-D photometric redshift range \\
2428: 37 & F6.3 & Upper limit of 6-D photometric redshift range \\
2429: 38 & F6.3 & 6-D Photometric redshift range probability \\
2430: 39 & F5.3 & Optical selection flag (from Richards et al. 2008) \\
2431: 40 & F5.3 & Previous catalog identification \\
2432: 41 & F5.3 & Previous catalog object redshift
2433: \enddata
2434: \end{deluxetable}
2435:
2436: %awk 'NR>=1 && NR<=5 {print $1"\\ldots & "$2" & "$3" & "$4" & "$5" & "$6" & "$7" & "$8" & "$9" & "$10" & "$11" & "$12"\\\\"}' sdss_irac_qsos_79D_bayes.zphot.match.dat > tab3.stub
2437:
2438: \begin{deluxetable}{lllrlllllllll}
2439: \tabletypesize{\scriptsize}
2440: \rotate
2441: \tablewidth{0pt}
2442: \tablecaption{Bayesian Quasar Candidates\label{tab:tab3}}
2443: \tablehead{
2444: \colhead{} &
2445: \colhead{R.A.} &
2446: \colhead{Decl.} &
2447: \colhead{Name} &
2448: \colhead{} &
2449: \colhead{} &
2450: \colhead{} &
2451: \colhead{} &
2452: \colhead{} &
2453: \colhead{} &
2454: \colhead{$f_{3.6}$} &
2455: \colhead{$f_{4.5}$} \\
2456: \colhead{Number} &
2457: \colhead{(deg)} &
2458: \colhead{(deg)} &
2459: \colhead{(SDSS J)} &
2460: \colhead{ObjID} &
2461: \colhead{$u$} &
2462: \colhead{$g$} &
2463: \colhead{$r$} &
2464: \colhead{$i$} &
2465: \colhead{$z$} &
2466: \colhead{($\mu$Jy)} &
2467: \colhead{($\mu$Jy)} \\
2468: \colhead{(1)} &
2469: \colhead{(2)} &
2470: \colhead{(3)} &
2471: \colhead{(4)} &
2472: \colhead{(5)} &
2473: \colhead{(6)} &
2474: \colhead{(7)} &
2475: \colhead{(8)} &
2476: \colhead{(9)} &
2477: \colhead{(10)} &
2478: \colhead{(11)} &
2479: \colhead{(12)}
2480: }
2481: \startdata
2482: 1\ldots & 149.326148 & 2.713954 & 095718.27+024250.2 & 587727944570503405 & 20.378 & 20.151
2483: & 19.805 & 19.581 & 19.705 & 126.810 & 185.392\\
2484: 2\ldots & 149.335600 & 2.488276 & 095720.54+022917.7 & 587726033308877340 & 22.961 & 22.082
2485: & 21.033 & 20.779 & 20.784 & 37.673 & 35.500\\
2486: 3\ldots & 149.349168 & 1.937556 & 095723.80+015615.2 & 587727943496761583 & 19.998 & 20.076
2487: & 19.893 & 19.914 & 19.977 & 61.660 & 103.770\\
2488: 5\ldots & 149.358575 & 2.750292 & 095726.05+024501.0 & 587727944570569013 & 20.661 & 20.946
2489: & 20.572 & 20.453 & 20.477 & 58.928 & 81.784\\
2490: \enddata
2491: \end{deluxetable}
2492:
2493:
2494: %awk 'NR>=1 && NR<=5 {print $1"\\ldots & "$2" & "$3" & "$4" & "$5" & "$6" & "$7" & "$8" & "$9" & "$10" & "$11" & "$12"\\\\"}' sdss_irac_qsos_79D_wedge.zphot.match.dat > tab4.stub
2495:
2496: \begin{deluxetable}{lllrlllllllll}
2497: \tabletypesize{\scriptsize}
2498: \rotate
2499: \tablewidth{0pt}
2500: \tablecaption{Wedge Quasar Candidates\label{tab:tab4}}
2501: \tablehead{
2502: \colhead{} &
2503: \colhead{R.A.} &
2504: \colhead{Decl.} &
2505: \colhead{Name} &
2506: \colhead{} &
2507: \colhead{} &
2508: \colhead{} &
2509: \colhead{} &
2510: \colhead{} &
2511: \colhead{} &
2512: \colhead{$f_{3.6}$} &
2513: \colhead{$f_{4.5}$} \\
2514: \colhead{Number} &
2515: \colhead{(deg)} &
2516: \colhead{(deg)} &
2517: \colhead{(SDSS J)} &
2518: \colhead{ObjID} &
2519: \colhead{$u$} &
2520: \colhead{$g$} &
2521: \colhead{$r$} &
2522: \colhead{$i$} &
2523: \colhead{$z$} &
2524: \colhead{$\mu$Jy} &
2525: \colhead{$\mu$Jy} \\
2526: \colhead{(1)} &
2527: \colhead{(2)} &
2528: \colhead{(3)} &
2529: \colhead{(4)} &
2530: \colhead{(5)} &
2531: \colhead{(6)} &
2532: \colhead{(7)} &
2533: \colhead{(8)} &
2534: \colhead{(9)} &
2535: \colhead{(10)} &
2536: \colhead{(11)} &
2537: \colhead{(12)}
2538: }
2539: \startdata
2540: 4\ldots & 149.351001 & 1.955768 & 095724.24+015720.7 & 587727943496762081 & 24.292 & 21.987
2541: & 20.286 & 19.699 & 19.323 & 146.039 & 131.027\\
2542: 17\ldots & 149.390413 & 2.919368 & 095733.69+025509.7 & 587726033845813857 & 23.405 & 23.41
2543: 5 & 22.363 & 21.831 & 22.037 & 137.255 & 125.932\\
2544: 26\ldots & 149.404266 & 1.737216 & 095737.02+014413.9 & 587726032235135314 & 22.585 & 21.36
2545: 8 & 20.113 & 19.658 & 19.406 & 110.340 & 100.392\\
2546: 30\ldots & 149.413157 & 1.668670 & 095739.15+014007.2 & 587726032235135709 & 24.339 & 22.69
2547: 9 & 21.685 & 21.067 & 20.674 & 42.732 & 42.716\\
2548: \enddata
2549: \end{deluxetable}
2550:
2551:
2552:
2553: \begin{deluxetable}{lllrlrrlllll}
2554: \tabletypesize{\scriptsize}
2555: \rotate
2556: \tablewidth{0pt}
2557: \tablecaption{{\em XMM-Newton} Objects in the XFLS Area\label{tab:xmm}}
2558: \tablehead{
2559: \colhead{Name} &
2560: \colhead{RA} &
2561: \colhead{Dec} &
2562: \colhead{Exp.} &
2563: \colhead{Total Flux} &
2564: \colhead{Soft Cnts} &
2565: \colhead{Hard Cnts} &
2566: \colhead{HR} &
2567: \colhead{$\sigma_{\rm HR}$} &
2568: \colhead{CATID} &
2569: \colhead{Ref/ID} &
2570: \colhead{z}
2571: }
2572: \startdata
2573: X171007.10+591127.7 & 257.529588 & 59.191014 & 12.8 & 6.98477e-14 & 23.1 & 8.8 & -0.45 & 0.34 & & Lacy05 & \\
2574: X171029.30+590833.7 & 257.622073 & 59.142688 & 12.8 & 2.53155e-13 & 86.2 & 27.0 & -0.52 & 0.12 & & DR5QSO & 0.864 \\
2575: X171049.65+590802.9 & 257.706889 & 59.134134 & 12.8 & 8.59703e-14 & 43.6 & 11.0 & -0.60 & 0.21 & 5606 & P06QSO & 1.234 \\
2576: X171126.68+585541.8 & 257.861152 & 58.928266 & 12.8 & 2.86859e-13 & 123.3 & 24.9 & -0.66 & 0.11 & 5629 & DR5QSO & 0.537 \\
2577: X171136.73+590115.7 & 257.903044 & 59.021019 & 12.8 & 4.43253e-14 & 21.1 & 15.7 & -0.15 & 0.23 & & & \\
2578: X171156.98+591220.0 & 257.987412 & 59.205546 & 12.8 & 6.17736e-14 & 21.3 & 9.2 & -0.40 & 0.34 & 5648 & P06QSO & 2.043 \\
2579: X171159.26+590433.1 & 257.996904 & 59.075852 & 12.8 & 4.83567e-14 & 18.8 & 11.4 & -0.24 & 0.25 & 5651 & & \\
2580: X171231.71+591217.6 & 258.132137 & 59.204879 & 12.8 & 1.13168e-13 & 22.0 & 10.2 & -0.37 & 0.29 & & & \\
2581: %X171629.97+594145.4 & 259.124862 & 59.695954 & 11.5 & 0 & -786.0 & -1356.0 & 0.00 & 0.00 & 0 & ? & -1 \\
2582: X171634.82+594310.9 & 259.145078 & 59.719695 & 9.3 & 1.15027e-13 & 15.6 & 21.1 & 0.15 & 0.57 & & & \\
2583: X171638.06+594514.6 & 259.158591 & 59.754067 & 11.5 & 5.98434e-14 & -0.7 & 23.4 & 1.00 & 1.77 & & & \\
2584: X171641.88+593758.7 & 259.174511 & 59.632980 & 9.3 & 2.5577e-13 & 39.4 & 20.2 & -0.32 & 0.37 & & & \\
2585: X171652.39+593543.6 & 259.218298 & 59.595450 & 9.3 & 3.81194e-14 & -5.1 & 13.1 & 1.00 & 2.36 & & & \\
2586: X171712.89+593828.7 & 259.303714 & 59.641318 & 11.5 & 2.6894e-13 & & & & & 5845 & P06QSO & 0.233 \\
2587: X171717.25+594640.4 & 259.321867 & 59.777894 & 9.3 & 2.13858e-13 & 51.9 & 27.6 & -0.31 & 0.23 & & & \\
2588: X171736.53+593010.9 & 259.402214 & 59.503022 & 11.5 & & & & & & & P96QSO & 0.599 \\
2589: X171737.02+593011.1 & 259.404248 & 59.503075 & 8.4 & 1.73472e-13 & 35.0 & 8.6 & -0.61 & 0.35 & 5854 & DR5QSO & 0.599 \\
2590: X171746.28+594123.7 & 259.442816 & 59.689926 & 1.1 & 3.24386e-13 & 40.0 & 18.4 & -0.37 & 0.25 & & Fadda06 & \\
2591: X171747.39+593258.9 & 259.447471 & 59.549688 & 9.3 & 4.64486e-13 & 97.3 & 51.9 & -0.30 & 0.16 & 5858 & P06Sy1 & 0.248 \\
2592: X171802.80+594001.0 & 259.511675 & 59.666935 & 11.5 & 1.59579e-13 & & & & & & & \\
2593: X171806.56+593312.6 & 259.527341 & 59.553487 & 9.3 & 8.9476e-13 & 227.2 & 51.9 & -0.63 & 0.10 & 5878 & P06QSO & 0.273 \\
2594: X171839.19+593402.0 & 259.663301 & 59.567217 & 9.3 & 3.81391e-13 & 101.4 & 49.4 & -0.34 & 0.14 & 5899 & P06QSO & 0.383 \\
2595: X171902.17+593715.7 & 259.759039 & 59.621019 & 9.3 & 9.07265e-13 & 306.9 & 92.8 & -0.54 & 0.08 & 5914 & DR5QSO & 0.179 \\
2596: X171943.91+594100.0 & 259.932958 & 59.683335 & 9.3 & 5.13025e-13 & 27.1 & -6.2 & -1.00 & 1.50 & & P06QSO & 0.129
2597: \enddata
2598: \end{deluxetable}
2599:
2600: \end{document}
2601:
2602: %Old X-ray table
2603: X171007.10+591127.7 & 257.529588 & 59.191014 & 12.7 & 1.11461e-13 & 0 & 0 & 0.00 & 0.00 & \\
2604: X171029.29+590833.8 & 257.622055 & 59.142711 & 12.7 & 2.42707e-13 & 0 & 0 & 0.00 & 0.00 & \\
2605: X171049.81+590802.2 & 257.707522 & 59.133946 & 6.5 & 1.01565e-13 & 0 & 0 & 0.00 & 0.00 & 5697 \\
2606: X171126.71+585542.0 & 257.861309 & 58.928322 & 12.7 & 2.65779e-13 & 0 & 0 & 0.00 & 0.00 & 5722 \\
2607: X171136.72+590117.3 & 257.902993 & 59.021465 & 12.7 & 4.7381e-14 & 0 & 0 & 0.00 & 0.00 & \\
2608: X171156.98+591220.0 & 257.987412 & 59.205546 & 12.7 & 6.52957e-14 & 0 & 0 & 0.00 & 0.00 & 5741 \\
2609: X171158.92+590433.0 & 257.995519 & 59.075844 & 12.8 & 8.64732e-14 & 0 & 0 & 0.00 & 0.00 & 5744 \\
2610: X171205.89+584754.9 & 258.024554 & 58.798590 & 6.5 & 1.27334e-13 & 0 & 0 & 0.00 & 0.00 & \\
2611: X171231.71+591217.6 & 258.132137 & 59.204879 & 12.7 & 9.93246e-14 & 0 & 0 & 0.00 & 0.00 & \\
2612: X171712.89+593828.7 & 259.303714 & 59.641318 & 11.4 & 2.40212e-13 & 0 & 0 & 0.00 & 0.00 & 5943 \\
2613: X171716.83+594644.9 & 259.320123 & 59.779144 & 11.4 & 1.59868e-13 & 0 & 0 & 0.00 & 0.00 & \\
2614: X171717.25+594640.4 & 259.321867 & 59.777894 & 9.2 & 2.1457e-13 & 51 & 27 & $-0.31$ & 0.23 & \\
2615: X171747.39+593258.9 & 259.447471 & 59.549688 & 9.2 & 4.36407e-13 & 97 & 51 & $-0.30$ & 0.16 & 5956 \\
2616: X171806.56+593312.6 & 259.527341 & 59.553487 & 9.2 & 8.98261e-13 & 227 & 51 & $-0.63$ & 0.10 & 5976 \\
2617: X171839.19+593402.0 & 259.663301 & 59.567217 & 9.2 & 3.83021e-13 & 101 & 49 & $-0.34$ & 0.14 & 5997 \\
2618: X171902.17+593715.7 & 259.759039 & 59.621019 & 9.2 & 1.16189e-12 & 306 & 92 & $-0.54$ & 0.08 & 6012 \\
2619: X171943.91+594100.0 & 259.932958 & 59.683335 & 9.2 & 3.19777e-13 & 27 & 0 & $-1.00$ & 1.50 & \\
2620:
2621: % LocalWords: Szalay Brunner Deo Reigel Parejko Storrie Zakamska Ptak Drexel
2622: % LocalWords: LSST hlh blackbody wrl lss seg drp fha yaa Lockman lsr Surace QZ
2623: % LocalWords: rls jfh IRAC's rng KDE pdf sjd hjf Mayall Kitt blueward QSO rws
2624: % LocalWords: wrs hpp rss Kron PIs XAssist ptak Spitzer's UKIDSS PAHs Caltech
2625: % LocalWords: ESA dps llllll lrrrr lcl lllrlrrlll Cnts CATID bandpasses Jy fm
2626: % LocalWords: unobscured lifecycle gbb tgb sr hhc bbm csw cdd lhc pn mbn dsc
2627: % LocalWords: lsst STARRS kab des VST anp bba rbo Akari Mpc dkn fww hck Cowie
2628: % LocalWords: Lidz baum ccs bcs SMBHs kht lps apa mrl esb scosmos tbl IRSA sfd
2629: % LocalWords: SDSS's sdssdr deblended starburst akb Ellipticals Im shr lsp nhg
2630: % LocalWords: mbh smc shl pws DS tpc ptm CDFS zmm htf gzz pdfs grn ewm rsf tim
2631: % LocalWords: deblending priori versa rmg aaa hcs z's photoz's shi csb alm hrh
2632: % LocalWords: qso NL rhv whb ghw mhw mgt smh dck pwh Sansigre NIR starbursts
2633: % LocalWords: AJ NNX AE Alamos KPNO sdsstech flsqsos richards shortnames rpd
2634: % LocalWords: psb x's lllrlllllllll Decl ObjID lllrlrrlllll Fadda Sy
2635: