1: %% LyX 1.4.1 created this file.  For more info, see http://www.lyx.org/.

2: %% Do not edit unless you really know what you are doing.

3: \documentclass[english,twocolumn]{emulateapj}

4: \usepackage{helvet}

5: \usepackage[latin1]{inputenc}

6: \setcounter{tocdepth}{3}

7: \usepackage{graphicx}

8: 

9: \makeatletter

10: 

11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LyX specific LaTeX commands.

12: %% Because html converters don't know tabularnewline

13: \providecommand{\tabularnewline}{\\}

14: 

15: 

16: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% User specified LaTeX commands.

17: \usepackage{natbib}

18: \citestyle{aa}

19: 

20: \usepackage{babel}

21: \makeatother

22: \begin{document}

23: 

24: \title{Source Matching in the SDSS and RASS: Which Galaxies are Really X-ray

25: Sources?}

26: 

27: 

28: \author{John K. Parejko, Anca Constantin, Michael S. Vogeley}

29: 

30: 

31: \affil{Department of Physics, Drexel University, Philadelphia, PA 19104}

32: 

33: 

34: \email{parejkoj@drexel.edu}

35: 

36: 

37: \and{}

38: 

39: 

40: \author{Fiona Hoyle}

41: 

42: 

43: \affil{Department of Physics and Astronomy, Widener University, Chester,

44: PA 19013}

45: 

46: 

47: \keywords{galaxies: active --- X-rays: general --- X-rays: galaxies --- quasars:

48: general}

49: 

50: \begin{abstract}

51: The current view of galaxy formation holds that all massive galaxies

52: harbor a massive black hole at their center, but that these black

53: holes are not always in an actively accreting phase. X-ray emission

54: is often used to identify accreting sources, but for galaxies that

55: are not harboring quasars (low-luminosity active galaxies), the X-ray

56: flux may be weak, or obscured by dust. To aid in the understanding

57: of weakly accreting black holes in the local universe, a large sample

58: of galaxies with X-ray detections is needed. We cross-match the ROSAT

59: All Sky Survey (RASS) with galaxies from the Sloan Digital Sky Survey

60: Data Release 4 (SDSS DR4) to create such a sample. Because of the

61: high SDSS source density and large RASS positional errors, the cross-matched

62: catalog is highly contaminated by random associations. We investigate

63: the overlap of these surveys and provide a statistical test of the

64: validity of RASS-SDSS galaxy cross-matches. SDSS quasars provide a

65: test of our cross-match validation scheme, as they have a very high

66: fraction of true RASS matches. We find that the number of true matches

67: between the SDSS main galaxy sample and the RASS is highly dependent

68: on the optical spectral classification of the galaxy; essentially

69: no star-forming galaxies are detected, while more than $0.6\%$ of

70: narrow-line Seyferts are detected in the RASS. Also, galaxies with

71: ambiguous optical classification have a surprisingly high RASS detection

72: fraction. This allows us to further constrain the SEDs of low-luminosity

73: active galaxies. Our technique is quite general, and can be applied

74: to any cross-matching between surveys with well-understood positional

75: errors.

76: \end{abstract}

77: 

78: \section{Introduction}

79: 

80: Distinguishing the processes that contribute to the emission from

81: the centers of galaxies is vital to understanding the co-evolution

82: of galaxies and their central black holes. Among nearby galaxies,

83: a large fraction of central emission sources are of ambiguous nature

84: \citep{1997ApJS..112..315H}; emission-lines in optical spectra of

85: many galaxies seem to reflect a mix of behavior between bona-fide

86: accretion (Seyfert-like) and active star formation (H II-like). In

87: order to discriminate between the various possible ionization mechanisms

88: and penetrate the obscuring dust layers that encircle these sources,

89: we need observations at multiple wavelengths. In particular, X-rays

90: are less prone to dust absorption and thus can be used to distinguish

91: between accretion sources and emission from young, hot stars. This

92: can clarify the observed optical emission spectra and allow us to

93: better describe the central accretion sources in low-luminosity active

94: galactic nuclei (AGN).

95: 

96: For an accurate census of the local galactic population, one must

97: study a statistically significant number of sources. The Sloan Digital

98: Sky Survey (SDSS, \citealp{2000AJ....120.1579Y}) provides the largest

99: sample of galaxies with spectra which allow emission-line classification

100: of central sources. The ROSAT All-Sky Survey (RASS, \citealp{1999A&A...349..389V,2000IAUC.7432....3V})

101: is the widest and deepest survey of the X-ray sky. The SDSS and RASS

102: are well matched in terms of depth, but have quite different astrometry

103: and spatial resolution. Previous studies matching a variety of SDSS

104: and RASS sources include analyzing the X-ray properties of spectroscopically

105: confirmed quasars \citep{2003AJ....126.2209A,2007AJ....133..313A},

106: generating an X-ray detected galaxy cluster catalog \citep{2004A&A...423..449P},

107: searching for optically unidentified neutron stars \citep{2006AJ....131.1740A},

108: and surveying the multi-wavelength properties of SDSS galaxies \citep{2006MNRAS.370.1677O}.

109: 

110: Because broad-line quasars are expected to be strong X-ray sources,

111: one would expect a large number of matches between RASS and SDSS for

112: these objects. \citet{2003AJ....126.2209A} characterized the RASS

113: properties of spectroscopically identified broad-line quasars from

114: the SDSS as well as some narrow-line sources. They qualitatively discuss

115: the likelihood that a given RASS-SDSS match is a true match and include

116: {}``normal'' (non or weakly emitting) galaxies as a comparison of

117: what a weak correlation would look like. They study more than $1000$

118: RASS-SDSS quasar/AGN and briefly discuss a few properties of the sample.

119: Their sample reproduces the expected non-linear optical/X-ray (2500\AA/2keV)

120: relationship among broad-line sources. The follow-up study, \citet{2007AJ....133..313A},

121: examines $\sim7000$ sources with similar results.

122: 

123: A different investigation involves identifying RASS sources with no

124: obvious optical counterpart. For example, this is useful for finding

125: optically dim neutron stars. \citet{2006AJ....131.1740A} identified

126: all SDSS sources within 4 times the positional error of each RASS

127: source. They then removed from their catalog any RASS source with

128: an SDSS match which could have produced the X-ray flux. After removing

129: objects with NED identifications, visually-identified bad fields,

130: and known galaxy clusters, 11 RASS sources with no plausible SDSS

131: optical counterpart remained. They claim this number is consistent

132: with the number of isolated neutron stars expected in the SDSS field.

133: Studying poorly understood matching samples in this way can clarify

134: whether the sample includes primarily true matches or primarily false

135: matches.

136: 

137: A recent comparison of RASS and SDSS in a multi-wavelength study \citep{2006MNRAS.370.1677O}

138: identified 267 RASS matches within $30\arcsec$ of SDSS DR1 main sample

139: galaxies \citep{2002AJ....124.1810S,2003AJ....126.2081A}. They list

140: a false association fraction of $\sim9\%$ (computed statistically

141: based on the RASS source density) and also show the positions of their

142: galaxies on an optical emission-line classification diagram \citep[the BPT diagram: ][]{1981PASP...93....5B}.

143: They did not investigate known-bad matches (as in \citealt{2006AJ....131.1740A}),

144: nor did they elaborate on the positions of the RASS detected galaxies

145: on their BPT diagram.

146: 

147: The ROSAT All Sky Survey was produced from data acquired in ROSAT's

148: scanning mode, but ROSAT also performed many individual targeted observations,

149: resulting in several pointed catalogs. These catalogs were generated

150: from serendipitous source discoveries made during individual targeted

151: observations. Because of this, they contain a large number of sources

152: in very small fields scattered over the sky with highly varying exposure

153: durations, making source upper limits difficult to compute. Previous

154: studies \citep[e.g.][]{2006AJ....132.1475S} have examined the properties

155: of SDSS quasars found in these catalogs.

156: 

157: \citet{2005mmgf.conf..320S} looked at star forming galaxies in the

158: SDSS DR1 and matched them to several different ROSAT catalogs, including

159: the RASS. Their final results involve 14 star forming galaxies which

160: they claim to be X-ray sources (four of which were previously studied).

161: We were not able to determine exactly which catalog they used in their

162: published results. Therefore, we cannot check whether the results

163: represent true matches between RASS and SDSS. Some star-forming galaxies

164: are expected to be X-ray emitters, but whether these galaxies are

165: actually detected in RASS remains to be seen.

166: 

167: The XMM-Newton and Chandra X-ray satellites both provide much improved

168: pointing, resolution and depth over ROSAT, but their fields of view

169: are quite small. Both have produced serendipitous source catalogs

170: similar to the ROSAT pointed catalogs mentioned above. The initial

171: XMM serendipitous source catalog was compared with the USNO A2.0 optical

172: catalog \citep{2006MNRAS.367.1017G} to find 46 optically identified

173: non-AGN galaxies with substantial X-ray flux. \citet{2005AJ....129...86H}

174: matched serendipitous source detections in Chandra with SDSS DR2 \citep{2004AJ....128..502A}

175: to find 42 X-ray emitting galaxies of a variety of types. The XMM-slew

176: survey \citep{freyberg-2005-} aims to solve the field of view and

177: uniformity problems by taking data during spacecraft slews between

178: targets. It will produce an all-sky map of equivalent depth to RASS,

179: with more than six times better resolution and pointing accuracy,

180: in roughly 6 years.

181: 

182: In this paper, we investigate the accuracy of matching RASS sources

183: with SDSS galaxies. In Section \ref{sec:Data} we describe the data

184: sets used in this study, including the systematics of selecting an

185: appropriate galaxy sample from SDSS. The details of the cross-matching

186: procedure and the statistical methods are described in Section \ref{sec:Cross-matching}

187: and the final matched data sets, separated by galaxy spectroscopic

188: class are detailed in Section \ref{sec:RASS-Detections-by}. We find

189: that a RASS/SDSS galaxy match cannot be trusted to represent the galaxy's

190: true X-ray flux without first identifying the galaxy's spectral type.

191: Section \ref{sec:Future-directions:-XMM-slew} provides a preliminary

192: analysis of the new XMM-slew catalog and shows its utility in clarifying

193: the presence of X-ray sources in galaxies.

194: 

195: 

196: \section{Data}

197: 

198: \label{sec:Data}

199: 

200: 

201: \subsection{SDSS}

202: 

203: This study employs data from the Sloan Digital Sky Survey Data Release

204: 4 (SDSS DR4), an optical imaging and spectroscopic survey with spectroscopic

205: coverage of $\sim16\%$ of the sky as described in \citet{2000AJ....120.1579Y}

206: and \citet{2006ApJS..162...38A}. Technical details of the photometric

207: camera, telescope, analysis pipeline, monitor and related systems

208: can be found in \citet{1998AJ....116.3040G}, \citet{2006AJ....131.2332G},

209: \citet{2001ASPC..238..269L}, \citet{2001AJ....122.2129H} and \citet{2002AJ....123.2121S},

210: while \citet{2003AJ....125.1559P} describe the astrometric calibration

211: and \citealp{1996AJ....111.1748F} describe the u'g'r'i'z' photometric

212: system. The SDSS has very good photometric resolution ($\sim1\arcsec.4$

213: PSF) and astrometric precision ($<0\arcsec.1$ rms per coordinate)

214: \citep{2002AJ....123..485S}. The spectroscopic survey is constructed

215: from tilings of the photometric data \citep{2003AJ....125.2276B}

216: and includes the main galaxy sample, quasar sample and luminous red

217: galaxy sample which are described in \citet{2002AJ....124.1810S},

218: \citet{2002AJ....123.2945R}, and \citet{2001AJ....122.2267E} respectively.

219: This spectroscopic survey includes uniform, high quality spectra of

220: more than half a million galaxies and nearly 100,000 quasars, via

221: plates containing $3\arcsec$ diameter optical fibers.

222: 

223: 

224: \subsubsection{SDSS Galaxies}

225: 

226: Our focus is on the main galaxy spectroscopic sample which includes

227: all galaxies with Petrosian \emph{r} magnitudes brighter than 17.77

228: with the exception of those not observed due to fiber collision. Because

229: of the size of the fiber-plugs, spectroscopic targets for a single

230: plate must be separated by at least $55\arcsec$. This was more of

231: a problem in DR1 and DR2, before overlapping plates and follow-up

232: observations filled in many of the missing objects. The complete SDSS

233: spectroscopic catalog includes more galaxies with spectra than just

234: the main galaxy sample. We restrict ourselves to the main galaxy sample

235: to avoid sample bias; some SDSS objects were selected for spectroscopy

236: due to their proximity to FIRST radio sources \citep{1995ApJ...450..559B}

237: and/or RASS X-ray sources. See the appendix for details on our SDSS

238: source selection process, and the importance of using the main galaxy

239: sample in cross-matching studies.

240: 

241: SDSS studies at MPA/JHU produced a catalog%

242: \footnote{ Data catalogues from SDSS studies at MPA/JHU \url{http://www.mpa-garching.mpg.de/SDSS/}%

243: } of secondary source products generated from the SDSS spectroscopic

244: data \citep[see][for the DR2 catalog paper]{2004astro.ph..6220B}.

245: This catalog includes simultaneous measurements of the emission and

246: absorption line profiles. The complete catalog includes all objects

247: in SDSS (regardless of magnitude) that are spectroscopically identified

248: as galaxies; sources with emission-line widths greater than 1000km/s

249: are not included (thus all objects identified as Seyferts and LINERs

250: in this paper are type 2 objects). We restrict ourselves to the intersection

251: of the MPA/JHU catalog and the SDSS DR4 main galaxy sample described

252: above.

253: 

254: 

255: \subsubsection{\label{sub:Spectral-Classification-of}Spectral Classification of

256: Main Sample Galaxies}

257: 

258: %

259: \begin{figure}

260: \includegraphics[width=1\columnwidth]{f1}

261: 

262: 

263: \caption{\label{fig:emission-line classification diagram}Emission-line galaxy

264: classification diagram used to separate H IIs, Transitions, LINERs

265: and Seyferts. {}``Unclassified emission'' galaxies are those which

266: lie in a different region in each diagram.}

267: \end{figure}

268: 

269: We classify galaxies based on their optical emission-line properties.

270: Galaxies showing at least a $2\sigma$ detection of  flux in the emission-features

271: H$\alpha$, H$\beta$, {[}O III], {[}N II], {[}S II] and {[}O I] are

272: classified as emission-line galaxies, while those that show some but

273: not all of these lines are called {}``unclassifiable'' galaxies.

274: The strong line emitters are further separated into sources dominated

275: by accretion and those dominated by light from hot, young stars. We

276: classify H IIs, Seyferts, LINERs and Transition objects based on their

277: positions in a 4-dimensional space defined by the line-flux ratios

278: {[}O III]$\lambda5007$/H$\beta$, {[}N II]$\lambda6583$/H$\alpha$,

279: {[}S II]$\lambda\lambda6716,6731$/H$\alpha$, and {[}O I]$\lambda6300$/H$\alpha$.

280: We use the classification criteria from \citet{2006MNRAS.372..961K}.

281: Fig. \ref{fig:emission-line classification diagram} shows the regions

282: defining each galaxy subclass. We call those galaxies that do not

283: lie in the same classification region in each diagram, {}``unclassified

284: emission'' galaxies. Finally, galaxies showing no signs of emission

285: in H$\alpha$, H$\beta$ and {[}O III] are classified as {}``Passive''

286: galaxies. More details on this classification scheme can be found

287: in \citet{2007-Anca-inpress}.

288: 

289: 

290: \subsection{ROSAT All Sky Survey}

291: 

292: %

293: \begin{figure}

294: \includegraphics[angle=-90,width=1\columnwidth]{f2}

295: 

296: 

297: \caption{\label{fig:ROSAT-positional-errors}RASS positional errors (rms)

298: in the Faint Source Catalog (FSC) and Bright Source Catalog (BSC).

299: The numbers after the hash (\#) in this, and all subsequent histograms,

300: give the total number of points included in that histogram.}

301: \end{figure}

302: 

303: Over the course of its eight year mission, the R\"{o}ntgensatellit

304: (ROSAT) produced a variety of distinct source catalogs from its two

305: X-ray detectors, the Position Sensitive Proportional Counter (PSPC)

306: and High Resolution Imager (HRI). The WGACAT \citep{2000yCat.9031....0W}

307: and 2RXP are serendipitous catalogs from pointed ROSAT observations

308: covering $\sim15\%$ and $\sim17\%$ of the sky, respectively. The

309: High Resolution Imager catalog (1RXH, \citealp{2000yCat.9028....0R})

310: covers $\sim2\%$ of the sky with much greater precision.

311: 

312: The PSPC scanning-mode data are the primary focus of this study: the

313: RASS Faint Source Catalog (FSC, \citealp{2000IAUC.7432....3V}) and

314: RASS Bright Source Catalog (BSC, \citealp{1999A&A...349..389V}) together

315: covering 92\% of the sky. We restrict ourselves to the RASS because

316: we would eventually like to compute source upper-limits. The average

317: integration time per target in the RASS varies between $<100$ seconds

318: for sources near the equator to $>5000$ seconds for sources near

319: the ecliptic poles, with $>97\%$ receiving more than 100 seconds.%

320: \footnote{RASS exposure map and ancillary data \url{http://www.xray.mpe.mpg.de/rosat/survey/rass-bsc/sup/}%

321: } 

322: 

323: The ROSAT PSPC operated between 0.2 and 2.4 keV, with the highest

324: sensitivity and resolution at roughly 1 keV. The PSPC optics were

325: focused for 1keV X-rays resulting in a $1\sigma$ PSF of roughly $25\arcsec$

326: at that energy. The resolution is worse for both higher energy (poor

327: focus) and lower energy (diffraction limit) X-rays. The scan-mode

328: observations that produced the RASS resulted in an astrometric positional

329: error ($1\sigma$ statistical error plus a $6\arcsec$ systematic

330: error) of 10-20'' (Fig. \ref{fig:ROSAT-positional-errors}). We show

331: in Section \ref{sub:Confirming-Quasars} that the $6\arcsec$ systematic

332: error is likely overestimated; $3\arcsec$ is likely more correct.

333: 

334: 

335: \subsection{XMM-Newton Slew Survey}

336: 

337: The RASS catalog is the current best compromise between width and

338: depth for X-ray data, but it has limitations, as noted above. To produce

339: an improved catalog, the X-ray Multi Mirror satellite (XMM-Newton)

340: is collecting X-ray counts during slews between targeted observations.

341: The first release of the XMM-Newton Slew Survey (XMM-slew, \citealp{freyberg-2005-})

342: covers 6240 square degrees of sky, in narrow north-south slews, using

343: the EPIC-pn CCD because of its large detector area, fast read-out

344: rate and high sensitivity to hard X-rays. Although average exposure

345: time is only $\sim10$s for any given source, the large mirror area

346: and sensitive detector make it nearly as deep as the RASS in the soft

347: band (0.2-2keV), and deeper and wider than any previous survey in

348: the hard band (2-12keV). The quoted $8\arcsec$ positional error along

349: the slew direction is dominated by the accuracy of the attitude reconstruction.

350: The EPIC-pn resolution of $4\arcsec$ is roughly a factor of 6 better

351: than the RASS resolution, thus XMM-slew can resolve many of the confused

352: RASS sources.

353: 

354: Two XMM-slew catalogs were released, a {}``total'' catalog containing

355: all detected sources, and a {}``clean'' catalog with known bad sources

356: removed and a higher detection threshold. We examine the clean sample

357: in this study; it contains 2713 sources with detections in at least

358: one band.

359: 

360: 

361: \section{Cross-matching}

362: 

363: \label{sec:Cross-matching}Cross-matching two surveys is simple enough:

364: count all objects separated by less than some threshold distance (in

365: our case, $60\arcsec$) as possible matches. But the validity of such

366: a match depends on the differing sky coverage, sensitivity, positional

367: accuracy and spatial resolution of the two matched surveys. These

368: differences lead to matches due to purely random associations, multiple

369: cross-matches for single sources, and erroneous flux measurements

370: due to contributions from multiple sources. For example, the ROSAT

371: PSPC is more than an order of magnitude worse than the SDSS in both

372: resolution and astrometry, and the SDSS source density is much higher.

373: Understanding the RASS-SDSS galaxy sample is particularly difficult

374: for sources that are not necessarily expected to be strong X-ray emitters,

375: such as spectroscopically identified low-luminosity narrow-line AGN,

376: passive or starburst galaxies. In this section, we attempt to quantify

377: the true and random components of RASS-SDSS cross-matches.

378: 

379: %

380: \begin{figure}

381: \includegraphics[width=1\columnwidth]{f3}

382: 

383: 

384: \caption{\label{fig:2 Quasars and 1 H II}An example of RASS source confusion

385: (North is up). SDSS g-band is shown in green (spectroscopic classifications

386: are labeled), RASS pixels in blue (white X marks the source center)

387: and FIRST sources in red (both quasars are FIRST sources). Notice

388: that the RASS source covers two SDSS spectroscopic galaxies but is

389: centered on neither. There are many other cases where there is no

390: obvious source for the X-ray emission besides a single SDSS galaxy,

391: because of lack of SDSS spectroscopic information about all sources

392: in the field. The resolved quasar is SDSS J101643.87+421027.5 for

393: reference.}

394: \end{figure}

395: 

396: Fig. \ref{fig:2 Quasars and 1 H II} illustrates an example of the

397: issues faced in matching RASS and SDSS. Here a RASS source overlaps

398: two spectroscopically identified SDSS galaxies and is not centered

399: on either of them. One of the galaxies hosts a quasar and thus is

400: the likely source of the X-ray flux, while the other is identified

401: as a star-forming galaxy (H II-type optical spectrum) and thus is

402: expected to contribute little to the X-ray flux. If the quasar were

403: unidentified---because it had no spectrum taken---the star-forming

404: galaxy could have been considered the X-ray source. Another problem

405: is that the center of the X-ray source does not coincide with any

406: of the optical sources. This could be simply due to the astrometric

407: errors in the RASS catalog (Fig. \ref{fig:ROSAT-positional-errors}),

408: or to contributions to the total X-ray emission from the other quasar

409: at the top of the image. This example is not singular: there are many

410: such confusing matches in the RASS-SDSS galaxy sample because of the

411: high SDSS source density. Also, this RASS source is relatively bright,

412: and thus has better centroiding (positional error given as $8\arcsec$)

413: than most RASS sources and was particularly easy to catch.

414: 

415: 

416: \subsection{Obvious X-ray emitters}

417: 

418: %

419: \begin{figure}

420: \includegraphics[angle=-90,width=1\columnwidth]{f4}

421: 

422: 

423: \caption{\label{fig:SDSS-other-sources}Source-separation histogram for SDSS

424: sources that are expected to have a RASS identification. {}``Quasar''

425: includes spectroscopically identified quasars, {}``blue'' includes

426: point sources with $u-g<0.6$, {}``blue2'' is a subset of {}``blue''

427: restricted to $15.5<u<21$ and {}``random'' is a random match between

428: SDSS galaxies and RASS (see Section \ref{sub:Purely-random-matches}).

429: The lower plot shows the {}``blue'' sample, with the magnitude cuts

430: included. Note the vastly different tails for the three quasar-like

431: samples, implying a significantly different contamination fraction

432: in each.}

433: \end{figure}

434: When an SDSS object is the actual source of the RASS X-rays (a true

435: match), the distance between the X-ray and optical source positions

436: should be small. Some obvious choices for true matches are quasars

437: and quasar candidates. To qualitatively assess whether these {}``obvious''

438: choices are correct, we plot the distribution of distances between

439: the center of the RASS and SDSS sources---the source-separation histogram---for

440: these particular systems in Fig. \ref{fig:SDSS-other-sources}. The

441: upper panel includes the following RASS-matched SDSS sources: spectroscopically

442: identified quasars, sources with $u-g<0.6$ (quasar candidates), a

443: subset of the quasar candidates restricted to $15.5<u<21$ and {}``random

444: match'' between galaxies and the RASS, as described in the following

445: section. The lower panel shows the u-magnitude vs. source-separation

446: distribution for blue sources. Note the clustering of points at small

447: source-separations for $15.5<u<21$, suggesting that these are true

448: RASS-SDSS matches.

449: 

450: From the upper panel, spectroscopically identified quasars show an

451: obvious peak at small source-separations. {}``Blue'' objects (all

452: SDSS sources with $u-g<0.6$), which include some objects in the {}``quasar''

453: sample, have a peak at small separations as well as a prominent tail.

454: The {}``blue2'' sample (subset of {}``blue'' with $15.5<u<21$)

455: has a much smaller tail, suggesting a smaller fraction of incorrect

456: matches. Out of these samples, spectroscopically identified quasars

457: appear to represent the most reliable RASS-SDSS cross-match, with

458: the fewest points with large separations.

459: 

460: 

461: \subsection{Purely random matches}

462: 

463: \label{sub:Purely-random-matches}

464: 

465: Incorrect cross-matches between catalogs are due to random associations

466: between optical and X-ray sources. Previous work estimated the random

467: contamination by comparing the source density of the two catalogs,

468: which works well for samples with a small random contamination fraction.

469: We model these incorrect matches by generating {}``offset'' SDSS

470: object catalogs and matching them to the RASS. We produced 10 such

471: offset catalogs each from the SDSS galaxy and quasar catalogs by offsetting

472: all objects (either galaxies or quasars, respectively) from their

473: true RA and Dec by a fixed amount in a fixed direction, with a different

474: offset and direction for each offset catalog to reduce systematic

475: effects. The maximum offset was $1\arcdeg$ in RA and Dec. This procedure

476: preserves the on-sky source distribution of the SDSS, while moving

477: sources far away from their original RASS associations. When these

478: catalogs are matched to the RASS, the result is a linearly increasing

479: source-separation histogram, $\frac{dN}{dr}\propto r$; as the radius

480: increases, more sources fall within the matching circle. We compare

481: these random catalogs with our galaxy or quasar RASS matches to determine

482: the fractional contamination by purely random associations.

483: 

484: 

485: \subsection{\label{sub:Confirming-Quasars}Confirming Quasars}

486: 

487: %

488: \begin{figure}

489: \includegraphics[angle=-90,width=1\columnwidth]{f5}

490: 

491: 

492: \caption{\label{fig:Simulated-quasar-source}Simulated quasar source-separation

493: histogram for cross-matching between RASS and SDSS quasars. Note the

494: differing tails: the real distribution does not fall off as quickly,

495: as there is a small fraction of random matches at large radii. The

496: thin upper curve gives the true matching fraction at that radius (percent,

497: right axis).}

498: \end{figure}

499: X-ray source positional measurements have independent, normally distributed

500: errors in both planar components. This is analogous to darts thrown

501: at a small target. The precision of each throw is known, but individual

502: throws may have different precisions. The distribution of dart-target

503: distances is given by a Rayleigh distribution having a probability

504: density function (PDF), \[

505: P(r)\propto\frac{re^{-r^{2}/2\sigma^{2}}}{\sigma^{2}}\]

506: with scale parameter $\sigma$ and separation distance $r$. In the

507: case of X-ray measurements, the positional precision, $\sigma$ is

508: affected by the X-ray flux (reliability of centroiding depends on

509: the number of X-rays) and the pointing accuracy and resolution of

510: the measuring apparatus. The precision of each RASS source measurement

511: is listed in the catalog as the positional error (Fig. \ref{fig:ROSAT-positional-errors}).

512: 

513: We reproduce the source-separation histogram for RASS-SDSS quasar

514: matches by simulating X-ray source measurements using the corresponding

515: RASS positional errors plus a small random component. Because the

516: RASS positional errors are dependent on the X-ray flux, we use the

517: positional errors from the RASS-SDSS quasar matched catalog. For each

518: such RASS source, we generate a Rayleigh distribution with the positional

519: error of that source as the scale parameter $\sigma$. The sum of

520: the probability distribution function from each source gives our {}``simulated

521: true match'' curve. This PDF is the parent distribution for the true

522: matches between RASS and SDSS quasars. Random associations between

523: RASS and SDSS quasars have a linearly increasing source-separation

524: histogram, as shown above. A linear combination of these two distributions

525: (simulation PDF and random straight-line) should reproduce the observed

526: RASS-SDSS quasar source-separation histogram.

527: 

528: We show the quasar source-separation histogram, simulated true match

529: curve, and random component in Fig. \ref{fig:Simulated-quasar-source}.

530: The simulation curve, which does not include the random component,

531: matches the actual quasar source-separation histogram very well except

532: at the tail end. Combining the simulation and random components via

533: a $\chi^{2}$-minimization on the amplitude of each component yields

534: an excellent fit. The total fit is not shown in Fig. \ref{fig:Simulated-quasar-source}

535: because it would be completely masked by the data. This fit has a

536: $\chi^{2}$ per degree of freedom of $0.68$. However, the distributions

537: match only if the RASS positional errors are all reduced by $3\arcsec$,

538: implying that the quoted $6\arcsec$ systematic offset was overestimated.

539: 

540: The thin upper curve in Fig. \ref{fig:Simulated-quasar-source} gives

541: the {}``true matching fraction'' for RASS-SDSS quasar matches (percent,

542: right axis). This is the number of true matches (simulation curve)

543: divided by the total fit (simulation+random) at that radius. Note

544: that at $30\arcsec$, about 90\% of the RASS-SDSS quasars matches

545: are legimate. We also find that at $60\arcsec$ there is $\sim6\%$

546: total contamination to the RASS-SDSS quasar catalog. This agrees with

547: the estimate from \citet{2007AJ....133..313A} of $\sim5\%$ contamination

548: for their sample.

549: 

550: 

551: \subsection{Galaxies}

552: 

553: %

554: \begin{figure}

555: \includegraphics[angle=-90,width=1\columnwidth]{f6}

556: 

557: 

558: \caption{\label{fig:Reconstructed-ROSAT/SDSS-galaxy}Reconstructed ROSAT/SDSS

559: galaxy source-separation with $\sim36\%$ true matches + $\sim64\%$

560: random matches out to $60\arcsec$. The thin curve from the upper

561: left to the lower right, cutting through the histogram, is the true

562: matching fraction (percent, right axis). There is very good agreement

563: between the total simulation curve and the actual distribution.}

564: \end{figure}

565: Matching SDSS main sample galaxies to the RASS results in 3169 total

566: matches. In contrast to quasars, the RASS-SDSS galaxy source-separation

567: histogram rises quickly, but is then relatively flat out to $60\arcsec$,

568: as seen in Fig. \ref{fig:Reconstructed-ROSAT/SDSS-galaxy}. This suggests

569: that while some galaxies are detected as X-ray sources, a large fraction

570: are simply random associations. We model the RASS-SDSS galaxy source-separation

571: histogram following the procedure outlined for quasars above. In this

572: case, the positional errors are those of the RASS-SDSS galaxy matched

573: catalog. Fig. \ref{fig:Reconstructed-ROSAT/SDSS-galaxy} compares

574: this model with the actual histogram. Note that the simulated true

575: match distribution is somewhat wider than the equivalent quasar curve,

576: as RASS sources associated with galaxies have a lower mean flux and

577: thus have larger positional errors. The $\chi^{2}$ per degree of

578: freedom of the total fit (simulated+random) is $1.18$ for galaxies.

579: 

580: To reduce the effect of source confusion in our RASS-SDSS galaxy sample,

581: we remove from our matched galaxy catalog RASS sources that are also

582: positionally matched with likely X-ray emitters. Our method is similar

583: to that employed by \citet{2006AJ....131.1740A} who removed RASS

584: sources that overlapped with spectroscopically identified quasars,

585: blue point sources (potential quasars), bright objects (ROSAT contaminant)

586: and sources with a quasar-like X-ray/optical spectral slope. Our requirements

587: are more relaxed, as our aim is not to eliminate all obvious x-ray

588: sources, but rather to identify X-ray counterparts of galaxies. Thus,

589: we only remove RASS sources from our matched galaxy catalog that are

590: close to the most reliable RASS cross-matches: within $40\arcsec$

591: of an SDSS quasar or within $30\arcsec$ of an object in the {}``blue2''

592: list described above. Also, if two SDSS galaxies match to one RASS

593: source, we take only the nearest match. This reduces the sample to

594: 1970 galaxies, with many obviously incorrect matches removed, such

595: as the {}``match'' shown in Fig. \ref{fig:2 Quasars and 1 H II}.

596: This {}``cleaned'' catalog improves the $\chi^{2}$ of the simulation+random

597: fit to 0.96 and is the catalog employed in the analysis that follows.

598: 

599: 

600: \section{RASS Detections by Galaxy Spectroscopic Class}

601: 

602: \label{sec:RASS-Detections-by}

603: 

604: %

605: \begin{figure*}

606: %\columnwidth=.45\columnwidth

607: \includegraphics[angle=-90,width=1\columnwidth]{f7a}\hfil

608: \includegraphics[angle=-90,width=1\columnwidth]{f7b}

609: 

610: 

611: \caption{\label{fig:ROSAT/SDSS_galaxy_subclass}ROSAT/SDSS galaxy source separation

612: by galaxy sub-type. Each sub-plot follows the structure of the individual

613: quasar and galaxy plots shown previously. The original matched distribution

614: (black, thick with $1\sigma$ Poisson errors), simulation (blue, short-dash)

615: and random (purple, dot) distributions produce the total simulated

616: distribution (green, long-dash). The thin solid curve (red, right

617: axis) gives the {}``true matching fraction'' described in Section

618: \ref{sub:Confirming-Quasars}.}

619: \end{figure*}

620: %

621: 

622: \begin{table*}

623: \tabletypesize{\scriptsize}

624: \begin{center}

625: \caption{\label{tab:chi-squared}$\chi^{2}$ per degree of freedom}

626: 

627: \begin{tabular}{cccccccccc}

628: \hline 

629: \hline&

630: all&

631: passive&

632: unclassifiable&

633: emission&

634: unc. emission&

635: H II&

636: transition&

637: LINER&

638: Seyfert\tabularnewline

639: \hline

640: $\chi^{2}$&

641: 1.01&

642: 1.24&

643: 1.14&

644: 0.47&

645: 0.90&

646: 1.53&

647: 1.47&

648: 0.47&

649: 1.00\tabularnewline

650: \hline

651: \end{tabular}

652: \end{center}

653: \end{table*}

654: One would expect galaxies with different optical spectroscopic classes

655: to produce different X-ray fluxes and thus to have different matching

656: fractions. \citet{2006MNRAS.370.1677O} list the RASS matching fractions

657: for SDSS galaxies showing no emission as well as AGN, star-forming

658: and unknown emission-line galaxies. They also plot their RASS matches

659: on an emission-line classification diagram analogous to the left-most

660: plot in Fig. \ref{fig:emission-line classification diagram}. However,

661: they do not discuss random matches, nor do they remove known invalid

662: matches (e.g. quasars). Thus, their sample includes many SDSS galaxies

663: which are unlikely to be true matches to RASS sources. To investigate

664: the connection between RASS detection likelihood and optical spectroscopic

665: class, we separate the cleaned RASS-SDSS galaxy catalog into subclasses

666: as described in section \ref{sub:Spectral-Classification-of}. For

667: each of these subclasses, we simulate the source-separation histogram

668: via their corresponding RASS positional errors and linear random components

669: as before, and list the $\chi^{2}$ of the fits in Table \ref{tab:chi-squared}.

670: 

671: Fig. \ref{fig:ROSAT/SDSS_galaxy_subclass} compares the actual and

672: simulated distributions for the different galaxy classes. The left

673: plot shows the four different types of classified emission-line galaxies,

674: while the right plot shows the unclassified and passive galaxies.

675: The thin red curves show the true matching fraction at a given radius.

676: Note the high true matching fraction for galaxies with potentially

677: significant optical emission from a central accretion source: the

678: Seyfert, LINER and transition objects. Also note the relatively high

679: true matching fraction for unclassified emission and passive galaxies.

680: Galaxies with their optical emission dominated by star formation have

681: a very small true matching fraction; though there are a large number

682: of RASS-SDSS matches for H II and unclassifiable galaxies, most of

683: those matches are purely random associations.

684: 

685: %

686: 

687: \begin{table*}

688: \tabletypesize{\scriptsize}

689: \begin{center}

690: \caption{\label{tab:X-ray-detection-fractions}X-ray detection fractions (percent)}

691: 

692: \begin{tabular}{cccccccccc}

693: \hline 

694: \hline quasar&

695: all&

696: passive&

697: unclassifiable&

698: emission&

699: unc. emission&

700: H II&

701: transition&

702: LINER&

703: Seyfert\tabularnewline

704: \hline

705: 8.3&

706: .12&

707: .41&

708: .11&

709: .11&

710: .28&

711: .004&

712: .19&

713: .41&

714: .66\tabularnewline

715: \hline

716: \end{tabular}

717: \end{center}

718: \end{table*}

719: We list the detection fractions for the various spectral classes in

720: Table \ref{tab:X-ray-detection-fractions}, including quasars for

721: comparison. This detection fraction is the integrated simulation curve

722: divided by the total number of galaxies in that class. Note the relatively

723: high detection fraction for galaxies with AGN-dominated optical emission,

724: including the tansition objects. The large X-ray detection fraction

725: for unclassified emission sources (defined in Section \ref{sub:Spectral-Classification-of})

726: suggests that many of these objects harbor obscured accretion.

727: 

728: The number of passive galaxy, unclassifiable galaxy, and LINER matches

729: to RASS are slightly under-predicted by the model at moderate radii

730: ($20-40\arcsec$). Visual inspection of these galaxies confirms that

731: some of them are in or near clusters, which would produce an X-ray

732: source near to, but not coincident with, the galaxy. We do not have

733: a cluster catalog to remove these {}``contaminants'' but a visual

734: tally shows that between half and two-thirds of the RASS-matched passive

735: galaxies may be contaminated by the presence of a galaxy cluster.

736: However, some of these galaxies appear to be field galaxies, and thus

737: we may be finding X-ray bright, Optically Normal Galaxies \citep[XBONGs, see  ][]{2005MNRAS.358..131G}.

738: We plan to examine these objects in more detail in future work.

739: 

740: %

741: 

742: \begin{table*}

743: \tabletypesize{\scriptsize}

744: \begin{center}

745: \caption{\label{tab:Radius-within-which}Source-separation distance at fixed

746: true matching fraction}

747: 

748: \begin{tabular}{ccccccccccc}

749: \hline 

750: \hline fraction&

751: quasar&

752: all&

753: passive&

754: unclassifiable&

755: emission&

756: unc. emission&

757: H II&

758: transition &

759: LINER&

760: Seyfert\tabularnewline

761: \hline

762: 85\%&

763: 33&

764: 6&

765: 14&

766: \nodata&

767: 11&

768: 21&

769: \nodata&

770: 16&

771: 17&

772: 23\tabularnewline

773: 70\%&

774: 40&

775: 15&

776: 24&

777: 13&

778: 25&

779: 29&

780: \nodata&

781: 24&

782: 25&

783: 30\tabularnewline

784: 50\%&

785: 47&

786: 23&

787: 32&

788: 21&

789: 32&

790: 37&

791: \nodata&

792: 32&

793: 32&

794: 38\tabularnewline

795: \hline

796: \end{tabular}

797: \tablecomments{ Listed here are the maximum

798: radii (expressed in arcseconds) for a given matching fraction, based

799: on the true match fraction shown in Fig. \ref{fig:ROSAT/SDSS_galaxy_subclass},

800: with quasars for comparison.}

801: \end{center}

802: \end{table*}

803: We list the RASS-SDSS source-separation radii at various fixed matching

804: fractions in Table \ref{tab:Radius-within-which} for all the objects

805: discussed in this paper. Notice that at no radius do H II galaxies

806: show even a $50\%$ true matching fraction. The true matching fraction

807: for star forming galaxies is extremely low because such galaxies do

808: not produce X-rays at a level detectable by the RASS and/or because

809: the X-rays they produce are completely obscured by dust. Because nearly

810: all RASS-SDSS star-forming galaxy matches are due to random associations,

811: no claims can be made about X-ray emitting star-forming galaxies from

812: these data alone. The XMM-Slew survey, XMM-Newton serendipitous source

813: catalog and the Swift BAT catalog all observe at higher X-ray energies

814: (less attenuated by dust), and so could help clarify the X-ray emission

815: properties of these galaxies.

816: 

817: %

818: 

819: \begin{table*}

820: \tabletypesize{\scriptsize}

821: \begin{center}

822: \caption{\label{tab:Percent-at-radius}Cumulative true matching fraction at

823: fixed radius}

824: 

825: \begin{tabular}{ccccccccccc}

826: \hline 

827: \hline radius&

828: quasar&

829: all&

830: passive&

831: unclassifiable&

832: emission&

833: unc. emission&

834: H II&

835: transition &

836: LINER&

837: Seyfert\tabularnewline

838: \hline

839: $40\arcsec$&

840: 96.9&

841: 53&

842: 71&

843: 48&

844: 64&

845: 84&

846: 7&

847: 75&

848: 77&

849: 88\tabularnewline

850: $30\arcsec$&

851: 98.1&

852: 64&

853: 79&

854: 59&

855: 74&

856: 89&

857: 10&

858: 83&

859: 85&

860: 92\tabularnewline

861: $20\arcsec$&

862: 99.0&

863: 75&

864: 86&

865: 70&

866: 83&

867: 94&

868: 16&

869: 89&

870: 91&

871: 95\tabularnewline

872: \hline

873: \end{tabular}

874: 

875: \tablecomments{ Listed here are the fractions

876: of each sample that are true matches, at the given radius, computed

877: by integrating the curves shown in Fig. \ref{fig:ROSAT/SDSS_galaxy_subclass},

878: with quasars for comparison.}

879: \end{center}

880: \end{table*}

881: For comparison with previous studies, we give the cumulative true

882: matching fraction at fixed radii in Table \ref{tab:Percent-at-radius}.

883: These values are computed from the ratio of the integrals of the simulated

884: and total curves in Fig. \ref{fig:ROSAT/SDSS_galaxy_subclass}, in

885: contrast with the previous table, derived from the point-wise ratios.

886: Again, note that H II galaxies have a very small cumulative true-match

887: fraction, even at small radii. All other matched sub-samples, except

888: for the unclassifiable galaxies, contain more than $85\%$ true RASS

889: matches below $20\arcsec$.

890: 

891: 

892: \section{Future directions: XMM-slew}

893: 

894: \label{sec:Future-directions:-XMM-slew}

895: 

896: %

897: \begin{figure}

898: \includegraphics[angle=-90,width=1\columnwidth]{f8}

899: 

900: 

901: \caption{\label{fig:XMM-slew}XMM-slew vs. SDSS cross matches for galaxies

902: and quasars. The $\sim8\arcsec$ XMM-slew positional errors are readily

903: visible in the quasar matches.}

904: \end{figure}

905: For comparison with RASS, we have matched the XMM-slew clean catalog

906: (first release) to both SDSS galaxies and quasars. Fig. \ref{fig:XMM-slew}

907: plots the source-separation histogram for these sources. The total

908: number of matches is quite small, due to the small number of XMM-slew

909: sources and the small overlap area between the surveys. Because of

910: the nature of the XMM-slew survey, we cannot perform the same analysis

911: as above; the narrow width of the slew strips is too small for a reliable

912: random fraction to be determined, yet. From the source-separation

913: histogram, $20\arcsec$ appears to be a reliable cut-off for true

914: matches. Accepting only those matches within this radius results in

915: $38$ galaxy matches and $115$ quasar matches to XMM-slew.

916: 

917: %

918: \begin{figure}

919: \includegraphics[width=1\columnwidth]{f9}

920: 

921: 

922: \caption{\label{fig:SDSS-HII-in-XMM-slew}An interacting galaxy pair not detected

923: in soft X-rays. The central-source optical classifications are marked.

924: The SDSS optically-classified transition galaxy is found in the XMM-slew

925: survey at only hard (2-12keV) energies. The stripe on the left of

926: the GALEX image is due to detector edge effects. The pair is UGC 08327.}

927: \end{figure}

928: A coverage map for the XMM-slew data is not yet available, so it is

929: not possible to determine the percentage of ROSAT detections that

930: are non-detections in XMM-slew. However, amongst the $38$ {}``reliable''

931: matches are member(s) of each galaxy class described above. Most of

932: these XMM-slew detections are in the soft band ($0.2-2$ keV), but

933: there are a few galaxies with a detected hard X-ray flux. An example

934: is shown in Fig. \ref{fig:SDSS-HII-in-XMM-slew}: an interacting pair

935: of galaxies optically classified as a Transition and a Seyfert. The

936: Transition galaxy shows a hard X-ray flux and substantial radio point

937: source, while the Seyfert is unidentified in hard and soft X-rays

938: and shows a $\sim2\sigma$ detection in the FIRST catalog. There is

939: no RASS source at this location. We plan to followup on this intriguing

940: pair to better understand their emission properties and spectral shape.

941: 

942: 

943: \section{Conclusions}

944: 

945: We have examined the matching statistics between the ROSAT All Sky

946: Survey and the SDSS main galaxy sample. Our technique---simulating

947: the RASS-SDSS source-separation via the RASS positional errors plus

948: a linear random component---can reproduce the measured source-separations

949: for RASS-SDSS quasar matches as well as RASS-SDSS galaxies and subclassifications

950: of galaxies. We find that the likelihood of a given cross-match match

951: being a true match depends strongly on the optical spectral classification

952: of a given galaxy. We find that essentially no optically classified

953: star-forming galaxy has a true RASS counterpart, while LINERs, Seyfert

954: 2s and Transition and unclassified emission galaxies do have reliable

955: X-ray detections. We also find a surprising number of galaxies lacking

956: optical emission lines which appear to be detected in the RASS. A

957: complete, low-redshift SDSS galaxy cluster catalog could be used to

958: clarify these XBONG candidates.

959: 

960: Our technique can be applied to any cross-matching between two surveys.

961: The only requirement is that the positional errors of each measurement

962: be known; no arbitrary fitting parameters are needed. By comparing

963: the observed source-separation histogram with a linear combination

964: of the probability distribution functions computed from the positional

965: errors and a random matched catalog, a {}``true matching fraction''

966: can be determined for any two matched catalogs. This is not limited

967: to X-rays: as a test, we were also able to reproduce the source-separation

968: histogram for a matched catalog of SDSS spectroscopic stars and GALEX

969: UV sources. The technique works best for catalogs containing mostly

970: point sources, as centroiding extended sources can be difficult and

971: the centers of sources may be wavelength-dependent.

972: 

973: 

974: \acknowledgements{We are grateful to Jenny Greene, Alina Badus and Danny Pan for providing

975: useful comments on earlier drafts of this paper. We acknowledge support

976: from NASA grant NAG5-12243 and NSF grant AST-0507647.\\

977: Funding for the Sloan Digital Sky Survey (SDSS) and SDSS-II has been

978: provided by the Alfred P. Sloan Foundation, the Participating Institutions,

979: the National Science Foundation, the U.S. Department of Energy, the

980: National Aeronautics and Space Administration, the Japanese Monbukagakusho,

981: and the Max Planck Society, and the Higher Education Funding Council

982: for England. The SDSS Web site is http://www.sdss.org/.\\

983: The SDSS is managed by the Astrophysical Research Consortium (ARC)

984: for the Participating Institutions. The Participating Institutions

985: are the American Museum of Natural History, Astrophysical Institute

986: Potsdam, University of Basel, University of Cambridge, Case Western

987: Reserve University, The University of Chicago, Drexel University,

988: Fermilab, the Institute for Advanced Study, the Japan Participation

989: Group, The Johns Hopkins University, the Joint Institute for Nuclear

990: Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology,

991: the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST),

992: Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy

993: (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico

994: State University, Ohio State University, University of Pittsburgh,

995: University of Portsmouth, Princeton University, the United States

996: Naval Observatory, and the University of Washington\\

997: We have made use of the ROSAT Data Archive of the Max-Planck-Institut

998: für extraterrestrische Physik (MPE) at Garching, Germany.}

999: 

1000: 

1001: \appendix{\label{app:SDSS-galaxy-selection}}

1002: 

1003: 

1004: \section{SDSS galaxy selection}

1005: 

1006: The main galaxy sample does not contain all the galaxies with spectra:

1007: a galaxy could also have a spectrum taken if it is within $2\arcsec$

1008: of a FIRST radio source or within the error-circle ($10-30\arcsec$)

1009: of a RASS source. Luminous red galaxies are selected for follow-up

1010: spectra based on their position in the (g-r, r-i, i) color-color-magnitude

1011: cube. Spectra are also taken for a variety of serendipitous sources

1012: including low surface-brightness galaxies. These other sources are

1013: all dimmer than 17.77 in the r-band, and biased toward AGN and star-forming

1014: galaxies. The systematics of these serendipitous sources are poorly

1015: understood.

1016: 

1017: %

1018: \begin{figure}

1019: \includegraphics[angle=-90,width=1\columnwidth]{f10}

1020: 

1021: 

1022: \caption{\label{fig:RASS_photo_vs_spectro}SDSS Galaxies with RASS matches

1023: within 1'. Note the sharp drop at $30\arcsec$ within the sample of

1024: objects spectroscopically classified as galaxies compared to the main

1025: galaxy sample (all galaxies with $r_{petro}<17.77$). Some galaxies

1026: which are not in the main sample were targeted specifically because

1027: they were within $30\arcsec$ of a RASS source.}

1028: \end{figure}

1029: The primary method for downloading large data sets from SDSS is CasJobs%

1030: \footnote{SDSS CasJobs website \url{http://casjobs.sdss.org/CasJobs/}%

1031: }. To extract the main galaxy sample from SDSS CasJobs, use the SpecObj

1032: parameter ObjType and select those objects classified as {}``Galaxy''.

1033: This includes all objects that were targeted for spectroscopy because

1034: they met the main galaxy sample criterion. This classification is

1035: \emph{before} the spectra were taken, and is thus a uniform sample.

1036: A more na\"{\i}ve selection might be to take all objects spectroscopically

1037: classified as galaxies: those with SpecObj parameter SpecClass listed

1038: as {}``Galaxy''. However, this sample includes all objects with

1039: a galaxy-like spectrum, which includes objects targeted for the above

1040: reasons in addition to the main galaxy sample.

1041: 

1042: In Fig. \ref{fig:RASS_photo_vs_spectro} we show the source-separation

1043: histogram for these two different samples. The {}``photometric''

1044: sample is the main galaxy sample used in this study. The spectroscopic

1045: sample, with a peak at $30\arcsec$, includes objects specifically

1046: targeted because they were near a RASS source. The fiber-selection

1047: process allocates spare spectroscopic fibers to sources within $30\arcsec$$ $

1048: of a RASS source. These objects, having SpecObj parameter ObjType

1049: classifications {}``ROSAT\_A'', {}``ROSAT\_B'', {}``ROSAT\_C''

1050: or {}``ROSAT\_D'' account for roughly $2\%$ of all objects with

1051: spectra in SDSS. \citet{2002AJ....123..485S} claim over half of these

1052: ROSAT-based targets turn out to be quasars or AGN. This results in

1053: a factor of two increase in potential matches at matching radii below

1054: 30''. This is why a statistical analysis of RASS matches to SDSS

1055: must stick with the main galaxy sample; the other sources were selected

1056: non-uniformly, and though they may result in odd and interesting spectra,

1057: they produce a strong bias in X-ray matching properties.

1058: 

1059: \bibliographystyle{apj}

1060: \bibliography{bib/xray,bib/sdss,bib/sdss-tech,bib/radio,bib/voids,bib/agn-general}

1061: 

1062: \end{document}

1063: