1: \documentclass[12pt, preprint]{emulateapj}% {aastex}%{emulateapj}% 

2: \usepackage{natbib} 

3:  

4: 

5: %%%%% AUTHORS - PLACE YOUR OWN MACROS HERE %%%%% 

6: \def\grtsim{\mathrel{\hbox{\rlap{\hbox{\lower2pt\hbox{$\sim$}}}\raise2pt\hbox{$>$}}}} 

7: \def\lesssim{\mathrel{\hbox{\rlap{\hbox{\lower2pt\hbox{$\sim$}}}\raise2pt\hbox{$<$}}}} 

8:  

9:  

10: \def\degree{\nobreak\ifmmode{^\circ}\else{$^\circ$}\fi} 

11:  

12:  

13: \newcommand{\mc}{\multicolumn} 

14: \newcommand{\mb}{\mbox} 

15: \newcommand{\f}{\phantom{2}} 

16:  

17: %units

18: %\newcommand{\muJy}{$\mu$Jy}

19: \newcommand{\mujy}{$\mu$Jy}

20: \newcommand{\mume}{$\mu$}

21: \newcommand{\wm}{W~m$^{-2}$}

22: \newcommand{\whz}{W~Hz$^{-1}$}

23: \newcommand{\whzsr}{W~Hz$^{-1}$~sr$^{-1}$}

24: \newcommand{\wsr}{W~sr$^{-1}$}

25: \newcommand{\kms}{km~s$^{-1}$}

26: 

27: 

28: %emission lines

29: \newcommand{\vdag}{(v)^\dagger}

30: \newcommand{\civ}{C\,{\sc iv}}

31: \newcommand{\lya}{Ly\,$\alpha$}

32: \newcommand{\nv}{N\,{\sc v}}

33: %obscuration

34: \newcommand{\nh}{$N_{\rm H}$}

35: \newcommand{\av}{$A_{\rm V}$}

36: %luminosity

37: \newcommand{\lx}{$L_{\rm X}$}

38: \newcommand{\lrada}{$L_{1.4 \rm GHz}$}

39: \newcommand{\lradb}{$L_{4.9 \rm GHz}$}

40: 

41: 

42: \newcommand{\lbol}{$L_{\rm bol}$}

43: \newcommand{\ltot}{$L_{\rm tot}$}

44: \newcommand{\lfir}{$L_{\rm FIR}$}

45: \newcommand{\lfiru}{$L_{\rm FIR,u}$}

46: \newcommand{\lfiro}{$L_{\rm FIR,o}$}

47: \newcommand{\llim}{$L_{\rm lim}$}

48: \newcommand{\lnu}{$L_{\nu}$}

49: \newcommand{\nulnu}{$\nu L_{\nu}$}

50: 

51: \newcommand{\rtt}{$\langle R \rangle$}

52: \newcommand{\fe}{$\langle f_{\rm E} \rangle $}

53: \newcommand{\logfe}{log$_{10}$[${F_{\rm I} \over \langle f_{\rm E} \rangle}$]}

54: \newcommand{\ffi}{$F_{\rm I}$}

55: \newcommand{\logfi}{log$_{10}$[$F_{\rm I}$]}

56: \newcommand{\eff}{$\langle \epsilon \rangle $}

57: \newcommand{\logeff}{log$_{10}$[$\langle \epsilon \rangle $]}

58: 

59: \newcommand{\cbol}{$C_{\nu}$}

60: \newcommand{\ctot}{$C_{\rm tot,u}$}

61: \newcommand{\ffiro}{$F_{\rm FIR,o}$}

62: \newcommand{\ffiru}{$F_{\rm FIR,u}$}

63: 

64: \newcommand{\snu}{$S_{\nu 1.2}$}

65: \newcommand{\slim}{$S_{\rm lim}$}

66: 

67: \def\lsol{L$_{\odot}$}

68: \def\lsolsr{L$_{\odot}$~sr$^{-1}$}

69: \def\msol{M$_{\odot}$}

70: \def\msolyr{M$_{\odot}$~yr$^{-1}$}

71: \def\msolmpc{M$_{\odot}$~Mpc$^{-2}$}

72: 

73: \newcommand{\mbh}{$M_{\bullet}$}

74: \newcommand{\densloc}{$\rho_{\bullet ,\rm loc}$}

75: \newcommand{\densacc}{$\rho_{\bullet ,\rm acc}$}

76: 

77: \newcommand{\mstar}{$M^{\star}_{\bullet}$}

78: \newcommand{\lstar}{$L^{\star}_{\nu}$}

79: 

80: 

81: \newcommand{\ratio}{$\lambda/\epsilon$}

82: 

83: 

84:  \newcommand{\myemail}{martinez@mpia-hd.mpg.de}

85: 

86: \shorttitle{Obscured AGNs and the radiation efficiency}

87: \shortauthors{A. Mart\'\i nez-Sansigre \& A.~M. Taylor }

88: 

89: 

90: 

91: 

92: \begin{document} 

93:  

94: 

95: \title{The cosmological consequence of an obscured AGN population on the radiation efficiency} 

96: 

97: \author{Alejo Mart\'\i nez-Sansigre}

98: 

99:  \affil{Max-Planck-Institut f\"ur Astronomie, K\"onigstuhl 17,

100:     D-69117 Heidelberg, Germany; ams@mpia.de}

101: 

102: \author{Andrew M. Taylor}

103: 

104: \affil{Max-Planck-Institut f\"ur Kernphysik, Postfach 103980,

105:   D-69029 Heidelberg, Germany; andrew.taylor@mpi-hd.mpg.de}

106: 

107: 

108: 

109:  

110:  

111: \begin{abstract}  

112:   In light of recent indications for a large population of obscured

113:   active galactic nuclei (AGNs), we revisit the mean radiation

114:   efficiency from accretion onto supermassive black holes (SMBHs),

115:   \eff, applying a bayesian approach.  We use the integrated comoving

116:   energy density emitted by AGNs and compare it to the mass density of

117:   local SMBHs. When considering only optically-selected unobscured

118:   AGNs, we derive log$_{10}$[\eff]$=$-1.77$^{+0.16}_{-0.11}$ or

119:   \eff$=$1.7$^{+0.8}_{-0.4}$\%.  Using the AGNs selected using hard

120:   X-rays, which include unabsorbed and Compton-thin absorbed AGNs, we

121:   find log$_{10}$[\eff]$=$-1.20$^{+0.15}_{-0.10}$ or

122:   \eff$=$6.4$^{+2.6}_{-1.3}$\%.  Using optically-selected AGNs, and

123:   correcting for the obscured population, we inferr

124:   log$_{10}$[\eff]$=$-1.17$^{+0.11}_{-0.08}$ or

125:   \eff$=$6.7$^{+1.9}_{-1.1}$\%, which we consider our best

126:   estimate. We also repeat this calculation for intrinsically luminous

127:   AGNs ($M_{\rm B}<-23$, quasars), comparing to the SMBH mass density

128:   in local elliptical galaxies, and find

129:   log$_{10}$[\eff]$=$-1.27$^{+0.15}_{-0.11}$ or

130:   \eff$=$5.4$^{+2.2}_{-1.2}$\%.  We discuss how our results can be

131:   used to estimate the mean spin of accreting SMBHs, both assuming the

132:   standard thin-disk model of accretion onto black holes and also

133:   comparing to more recent simulations. Our results can rule out

134:   maximally rotating SMBHs ($\langle a \rangle=0.998 {G m_{\bullet}

135:     \over c^{2}}$) at the $\geq$98\% confidence level, as well as high

136:   rotation values ($\langle a \rangle\geq0.9 {G m_{\bullet} \over

137:     c^{2}}$) with $\geq$92\% confidence. Our preferred values of

138:   $\langle a \rangle$ are $\sim$0.25-0.60$ {G m_{\bullet} \over

139:     c^{2}}$, although even these might be overestimated. Hence, we

140:   find that on average, SMBHs are not rapidly spinning during

141:   accretion.  Finally, using an independent measurement of Eddington

142:   ratios, we estimate the SMBH $e$-folding time for the brightest AGNs

143:   (quasars) to be $\langle \tau \rangle=100^{+151}_{-60}$~Myr.

144:  \end{abstract} 

145:  

146: \keywords{galaxies : active --galaxies: nuclei  --  quasars: general

147:    --black hole physics -- cosmology: miscellaneous --}

148: 

149:  

150: \section{Introduction} 

151: 

152: 

153: 

154: Active galactic nuclei (AGNs) are believed to be powered by accretion

155: onto supermassive black holes (SMBHs), and are expected to leave

156: behind dormant SMBHs

157: \citep[][]{1964ApJ...140..796S,1969Natur.223..690L,1984ARA&A..22..471R}.

158: The observed energy per comoving volume radiated by AGNs, the density

159: of local SMBHs and the radiation efficiency of accreting SMBHs can all

160: be related, using conservation of energy, in an elegant argument

161: proposed by \citet{1982MNRAS.200..115S}.

162: 

163: The original form of the ``So{\l}tan argument'' estimated the density

164: of relic SMBHs from the distribution of observed quasar fluxes

165: (magnitudes) and making an educated guess for the

166: efficiency. Nowadays, SMBHs are believed to reside at the centres of

167: galaxies, and estimates of the SMBHs masses are possible

168: \citep[e.g.][]{1995ARA&A..33..581K,1996Natur.383..415E,1998AJ....115.2285M}.

169: More accurate measurements of the AGN luminosity functions (LFs)

170: exist, meaning that the least constrained parameter is the radiation

171: efficiency.  Using variations of the So{\l}tan argument, the

172: efficiency can be estimated from the comparison between the estimated

173: total radiated energy from AGNs with the local SMBH

174: \citep[e.g.][]{2002ApJ...565L..75E,2002MNRAS.335..965Y,2004MNRAS.351..169M,2004MNRAS.354.1020S}.

175: 

176: 

177: 

178: 

179: The discovery of a significant population of obscured AGNs at a range

180: of redshifts (see Section~\ref{sec:demo}) has important consequences

181: for the global energetics of quasars and accretion onto SMBHs, given

182: that the total energy output from the obscured population is

183: comparable to that of the unobscured population (if not greater).

184: 

185: 

186: In this article, we address the issue of how this population can have

187: a significant effect on the mean radiation efficiency from accretion

188: onto SMBHs.  We apply a bayesian inference technique to derive

189: posterior probability distribution functions (PDFs) for the mean

190: radiation efficiency of SMBHs, \eff.

191: 

192: In Section~\ref{sec:soltan} we outline how the total radiated energy

193: density and the local SMBHs density can be used to estimate \eff. In

194: Section~\ref{sec:lf} we discuss the relative contributions of AGNs at

195: different redshifts to the total radiated energy density.

196: Section~\ref{sec:demo} summarises the current knowledge on the

197: obscured AGN population. In Section~\ref{sec:infer} we explain in

198: detail how we inferr \eff, using a bayesian approach, and how we

199: combine different estimates of \eff. Section~\ref{sec:spin} discusses

200: how the mean SMBH spin during periods of accretion can be estimated

201: from \eff.  Finally, in Section~\ref{sec:efold} we estimate the

202: $e$-folding time for quasars, and in Section~\ref{sec:conclusion} we

203: summarise our results.

204: 

205: 

206: Throughout this article we refer to powerful AGNs as quasars where,

207: for unobscured AGNs, the dividing line is generally taken at $M_{\rm

208:   B}=-23$ ($L_{\nu \rm B}=$$7\times10^{22}$~\whz).  We assume a

209: $\Lambda$CDM cosmology with the following parameters: $h = H_{0} /

210: (100 ~ \rm km ~ s^{-1} ~ Mpc^{-1}) = 0.7$; $\Omega_{\rm m} = 0.3$;

211: $\Omega_{\Lambda} = 0.7$.

212:  

213: 

214: \section{The efficiency of accretion onto SMBHs} \label{sec:soltan}

215: 

216: 

217: We outline here the main steps of the So{\l}tan argument, in the form that we

218: will use to inferr the mean radiation efficiency onto SMBHs, \eff.

219: 

220: The total (comoving) energy density radiated by AGNs can be related to

221: the integrated local SMBH density, \densloc, using three assumptions:

222: i) that SMBHs grew all their mass by accretion, thus neglecting

223: mergers between black holes (or rather taking the simplifying

224: assumption that mergers conserve SMBH mass) ii) that the mean

225: efficiency \eff\, is constant with luminosity and over the Hubble time

226: and iii) that AGNs are the manifestation of these periods of growth by

227: accretion.

228: 

229: For an individual SMBH, accretion with radiation efficiency $\epsilon$

230: will lead to a bolometric energy:

231: 

232: 

233: \noindent \begin{equation}

234: E_{\rm bol} = \int  \epsilon { {\rm d}m \over {\rm d}t}c^{2}  {\rm d}t ,

235: \end{equation}

236: 

237: \noindent while a relic SMBH will be left behind, with mass, 

238: 

239: \noindent \begin{equation}

240:  m_{\bullet , \rm acc}  = \int (1-\epsilon){ {\rm d}m \over {\rm d}t}{\rm

241:    d}t .

242: \end{equation}

243: 

244: \noindent The radiated energy no longer contributes to the accreted

245: mass energy, which the distant observer will see as the SMBH mass,

246: hence the $(1-\epsilon)$ term.  Rather than tracking individual

247: sources, we are interested in the relic SMBH mass density

248: ($\rho_{\bullet ,\rm acc}$) inferred from the summed comoving energy

249: density emitted by AGNs over the age of the Universe, which can be

250: estimated from the AGN LF.  Converting to observable quantities, we

251: obtain:

252: 

253: \noindent\begin{equation} \label{eq:integral_no_r}

254:   \rho_{\bullet , \rm acc}  =  { (1- \langle \epsilon \rangle) \over \langle \epsilon \rangle~c^{2}} \int {{\rm d}t \over {\rm d}z} {\rm d}z \int \it  \,

255:   C_{\nu} \nu L_{\nu} ~\phi (\it L_{\nu}, \it z){\rm d}L_{\nu} , 

256: \end{equation} 

257: 

258: \noindent where $\phi (\it L_{\nu}, \it z)$ is the LF for all AGNs,

259: $C_{\nu}$ is the bolometric correction and \eff\, is the mean

260: efficiency. The bolometric correction (defined so $L_{\rm bol}=C_{\nu}

261: \nu L_{\nu}$) is necessary to convert from the monochromatic LF to the

262: bolometric luminosity density.

263: 

264: If only an unobscured AGN LF is available, one can instead correct for

265: the missing obscured AGNs:

266: 

267: \noindent\begin{equation} \label{eq:integral}

268:   \rho_{\bullet , \rm acc}  =  { (1- \langle \epsilon \rangle) \over \langle \epsilon \rangle~c^{2}}(1+\langle R \rangle) \int {{\rm d}t \over {\rm d}z} {\rm d}z \int \it  \,

269:   C_{\nu} \nu L_{\nu} ~\phi_{\rm u} (\it L_{\nu}, \it z){\rm d}L_{\nu} . 

270: \end{equation} 

271: 

272: \noindent Here $\phi_{\rm u} (\it L_{\nu}, \it z)$ is the unobscured

273: AGN LF.  \rtt\, is the mean ratio of obscured to unobscured AGNs so

274: the term (1+$\langle R \rangle$) is a parametrised way of accounting

275: for the obscured population.  Estimates exist for the parameters

276: \cbol\, and \rtt. As mentioned earlier, the least constrained

277: parameter is \eff, so we treat it here as a free parameter.

278: 

279: \section{Radiated energy density}\label{sec:lf}

280: 

281: A crucial quantity for the estimation of \eff\, is the total energy

282: density radiated by AGNs, $U$ (the time-integrated luminosity

283: density).  The AGN population shows a strong evolution so that the

284: contribution of the low-redshift AGNs ($z\lesssim$0.3) to the energy

285: density is very modest.

286:  

287: Amongst optically-selected AGNs, the luminosity density of AGNs

288: strongly increases with redshift, with a peak around $z\sim2$-2.5

289: \citep[e.g.][]{2003A&A...408..499W}, and for the entire AGN population

290: over cosmic time, $U$ is dominated by the AGNs at $z=$1-3

291: (particularly those around $L^{\star}$) with the lower-redshift AGNs

292: contributing a smaller fraction.

293: 

294: This is illustrated in the left panel of Figure~\ref{fig:ut}, which

295: shows the cumulative energy density $U(z)$ [relative to $U(z$$=$0)]

296: for several optical LFs and one hard X-ray LF.  For the optical

297: surveys, all defined in the B-band (centred on 4400~\AA, or 2.8~eV),

298: $U(z)$ is defined as:

299: 

300: \noindent\begin{equation} \label{eq:uto}

301:   U_{\rm B}(z)  =    \int_{z'=5}^{z'=z} {{\rm d}t \over {\rm d}z'} {\rm d}z' \int \it  \,

302:    C_{\rm B} \nu_{\rm B} L_{\nu \rm B} ~\phi_{\rm B} (\it L_{\nu \rm B}, \it z'){\rm d}L_{\nu \rm B}.

303: \end{equation} 

304: 

305: \noindent where $h_{\rm p}\nu_{\rm B}=$2.8~eV ($h_{\rm p}$ is Planck's

306: constant).  $L_{\nu \rm B}$, $\phi_{\rm B}$ and $C_{\rm B}$ are,

307: respectively, the luminosity density, LF and bolometric correction at

308: the B-band.  Plotting $U(z)/U(0)$ removes the dependence on the

309: bolometric correction and shows the relative contribution at different

310: epochs.

311: 

312: As can be seen in the left panel of Figure~\ref{fig:ut}, for

313: optically-selected AGNs, approximately 70\% of $U(z$$=$0) is radiated

314: between $z=1$ and 3, with 20\% radiated at $z<1$ and 10\% at $z>3$.

315: 

316: However, similar studies undertaken at hard X-ray energies (in the

317: 2-10~keV band) have shown that the lower-luminosity AGNs evolve

318: differently to high-luminosity AGNs, with their activity peaking at

319: lower redshifts \citep[e.g.][]{2003ApJ...598..886U}.  Again, we

320: illustrate the effect by plotting $U(z)$ (Figure~\ref{fig:ut}, right

321: panel), with $U(z)$ defined as:

322: 

323: \noindent\begin{equation} \label{eq:utx}

324:   U_{\rm X}(z)  =    \int_{z'=5}^{z'=z} {{\rm d}t \over {\rm d}z'} {\rm d}z' \int \it  \,

325:    C_{\rm X}  L_{\rm X} ~\phi_{\rm X} (\it L_{\rm X}, \it z'){\rm d}L_{\rm X}.

326: \end{equation} 

327: 

328: \noindent where $L_{\rm X}$, $\phi_{\rm X}$ and $C_{\rm X}$ are the

329: hard X-ray luminosity, LF and bolometric correction, but note that

330: here $L_{\rm bol}=C_{\rm X} L_{\rm X}$.

331: 

332: From the right panel of Figure~\ref{fig:ut} we can see that only about

333: 50\% of $U$ was radiated between $z=1$ and 3, with $\sim$45\% radiated

334: at $z<1$ and $\sim$5\% at $z>3$.

335: 

336: Thus, from optically-selected LFs one expects the AGNs at $1 \leq z

337: \leq 3$ to dominate, although the hard X-ray LF indicates that the

338: contribution of $z < 1$ AGNs is similarly important and should not be

339: overlooked.

340: 

341:  

342: 

343: 

344: \section{Demography of obscured AGNs}\label{sec:demo}

345: 

346: Obscured AGNs are believed to be accreting SMBHs where, between X-ray

347: and near-infrared wavelengths, gas and dust block the line of sight to

348: the central region. The strong continuum from the accretion disk and

349: the broad emission lines suffer from heavy extinction by dust, while

350: the narrow emission lines may sometimes be visible, depending on the

351: obscuring geometry.  Photoelectric absorption by the intervening gas

352: and dust will supress the soft X-ray emission ($\lesssim$5~keV). If

353: the column density is large enough to be optically-thick to Compton

354: scattering\footnote{A source is Compton-thick if the absorbing column

355:   density, $N_{\rm H}$, is $\geq {1/\sigma_{T}}=$$1.5\times10^{28}$

356:   m$^{-2}$, where $\sigma_{T}$ is the Thomson electron scattering

357:   cross-section. X-ray absorbed, Compton-thin sources have $10^{26}$

358:   m$^{-2} \lesssim N_{\rm H} \lesssim 10^{28}$ m$^{-2}$, and X-ray

359:   unabsorbed sources have $N_{\rm H} \lesssim 10^{26}$ m$^{-2}$}, then

360: even the hard X-rays ($\grtsim$5~keV) will be heavily affected.

361: 

362: 

363: The obscured AGNs are indistinguishable from galaxies in optical and

364: near-infrared imaging surveys, which has made their identification

365: significantly harder than that of their unobscured counterparts. There

366: are mainly four methods for looking for these objects, which we

367: summarise in chronological order, and we concentrate here on the most

368: recent results.

369: 

370: Radio selection (wavelength $\grtsim$6~cm, energy

371: $\lesssim$20~$\mu$eV). Dust and gas are transparent to moderately

372: high-frequency radio waves, so radio-loud obscured AGNs can be readily

373: identified, in the form of narrow-line radio galaxies.  Amongst the

374: radio-loud population, the obscured to unobscured ratio is found to be

375: $\sim$1:1 \citep{2000MNRAS.316..449W}. However, the fraction of

376: unobscured AGNs that are radio loud is small \citep[$\sim$10\%,

377: e.g. ][]{2002AJ....124.2364I}, so that the narrow-line radio galaxies

378: are expected to represent a similarly small fraction of the obscured

379: AGN population \citep[e.g.][]{2004AJ....128.1002Z}.

380:   

381: Selection at hard X-ray energies ($\sim$2-10~keV). This method is

382: sensitive to the Compton-thin absorbed AGN population, and it has

383: identified large numbers of sources in deep X-ray surveys

384: \citep[see][for a review]{2005ARA&A..43..827B}. \citet

385: {2007A&A...463...79G} find the ratio of Compton-thin absorbed to

386: unabsorbed AGNs to vary from $\sim$1:1 at the high luminosity end

387: (\lx$\sim10^{12}$~\lsol) to $\sim$4:1 at the low luminosity end

388: (\lx$\sim10^{8}$~\lsol). Using the hard X-ray background (at energies

389: $\sim$3-100~keV) as an extra constraint, these authors find that the

390: Compton-thick and Compton-thin absorbed populations are likely to have

391: similar numbers (i.e $\sim$2:1 at high luminosities and up to

392: $\sim$8:1 at low luminosities).

393: 

394: 

395: 

396: Selection using optical spectroscopy (wavelengths $\sim$3500-7500~\AA,

397: energies $\sim$1.5-3.5~eV). If the geometry of the dust is such that

398: the narrow emission lines are not obscured, these are detectable via

399: optical spectroscopy. Using the spectroscopic database from the Sloan

400: Digital Sky Survey, \citet {2003AJ....126.2125Z} were able to select

401: large numbers of obscured luminous AGNs (obscured quasars). Using the

402: [O~III] 5007~\AA\, line strength as a isotropic proxy for the

403: intrinsic AGN luminosity, the ratio of spectroscopically-selected

404: obscured quasars to unobscured quasars is $\sim$1.2-1.5:1 \citep[][

405: for $L_{\rm [O~III]}\geq10^{9}$~\lsol]{2008arXiv0801.1115R}. At lower

406: [O~III] luminosities ($L_{\rm [O~III]}\sim 10^{7}$~\lsol) \citet

407: {2005MNRAS.360..565S} estimate the obscured to unobscured ratio to be

408: $\sim$4:1.

409: 

410: Mid-infrared selection (wavelengths $\sim$3-30~$\mu$m, energies

411: 0.04-0.4~eV).  The extinction due to dust becomes very small at these

412: wavelengths, so mid-infrared selection should be sensitive to the

413: radio-quiet population, including obscured AGNs showing no narrow

414: lines.  Recent work has suggested it is also sensitive to the

415: Compton-thick population

416: \citep[e.g.][]{2006ApJ...642..673P,2007MNRAS.379L...6M}. However, care

417: must be taken to separate starburst contaminants from real AGNs, since

418: these star-forming galaxies also show bright mid-infrared emission.

419: Combining mid-infrared and radio criteria, \citet{2008ApJ...674..676M}

420: found a $\sim$2:1 ratio at the high-luminosity end ($\nu L_{\nu

421:   \rm~8~\mu m}\grtsim 10^{12}$~\lsol), while

422: \citet{2005ApJ...634..169D} found a $\sim$4:1 ratio amongst lower

423: luminosity AGNs ($\nu L_{\nu \rm~24~\mu m}\sim 10^{10}$~\lsol).

424: 

425: 

426: Based on mid-infrared selection with no radio criteria,

427: \citet{2007AJ....133..186L} found a $\sim$2:1 ratio amongst bright

428: AGNs (with a median value $\nu L_{\nu \rm~5~\mu m}\sim

429: 5\times10^{11}$~\lsol). Using a power-law criterion, \citet

430: {2006ApJ...640..167A} also found a $\sim$2:1 ratio amongst powerful

431: AGNs (70\% of their sources having $L_{\rm IR}\geq 10^{12}$~\lsol)

432: while \citet{2007ApJ...660..167D} find a 4:1 ratio for less luminous

433: sources (with typically $L_{\rm X}\lesssim 5\times 10^{9}$~\lsol\, or

434: expected total infrared luminosities $L_{\rm IR}\lesssim

435: 10^{11}$~\lsol) but only 2:1 if the sources are required to be X-ray

436: detected.

437: 

438: A possible picture fitting these observations is the following:

439: amongst luminous AGNs (quasars), the obscured to unobscured ratio

440: seems to be $\sim$2:1, with about half of the obscured quasars being

441: Compton-thick \citep{2007A&A...463...79G,2007MNRAS.379L...6M}, and

442: about half showing no narrow emission lines \citep[][which explains

443: the $\sim$1:1 ratio found from spectroscopically-selected samples by

444: Reyes et al., 2008]{2008ApJ...674..676M}. Amongst radio-loud quasars,

445: the obscured to unobscured ratio is $\sim$1:1 with almost all obscured

446: quasars (narrow-line radio galaxies) showing narrow emission lines.

447: An explanation for this is that the lack of narrow lines is due to

448: kpc-scale dust \citep[e.g.][]{2006MNRAS.370.1479M,2006ApJ...645..115R}

449: and radio jets are efficient at clearing dust

450: \citep{2002ApJ...568..592B,2002MNRAS.331..435W}, so the fraction of

451: radio galaxies obscured by kpc-scale dust is small.

452: 

453: Amongst the low-luminosity AGNs, from X-ray, optical spectroscopy and

454: mid-infrared studies the obscured to unobscured ratio seems to be

455: $\sim$4:1, consistent with the ratio inferred in the local Universe

456: \citep[e.g.][]{1999ApJ...522..157R}.  However, modelling of the hard

457: X-ray background allows this ratio to be as high as 8:1 provided that

458: half of the absorbed AGNs are Compton-thick

459: \citep{2007A&A...463...79G}.

460: 

461: All the evidence points towards an obscured fraction that increases

462: with decreasing AGN luminosity \citep[which was noted a long time ago,

463: see ][]{1982ApJ...256..410L}. Whether the obscured fraction evolves or

464: not with redshift is still unclear \citep[see

465: e.g.][]{2003ApJ...598..886U,2006ApJ...652L..79T,2007A&A...463...79G}. The

466: studies cited above range from $z\leq0.3$ to $z\sim2$, and luminous

467: AGNs are present in the entire range

468: \citep{2003AJ....126.2125Z,2007AJ....133..186L,2008ApJ...674..676M}.

469: Lower-luminosity AGNs, however, are typically only found at low and

470: intermediate redshifts ($z\lesssim$1), as expected from flux-limited

471: samples.  Thus, it is difficult at this stage to judge whether the

472: obscured fraction varies with redshift.

473: 

474: 

475: 

476: \section{Inferring  \eff}\label{sec:infer}

477: 

478: Having discussed the radiated energy density from AGNs and the

479: inferred ratio of obscured to unobscured sources, we proceed to use

480: these to estimate the mean radiation efficiency of AGNs. We will

481: inferr different values of \eff, from the optical LF, with and without

482: adding obscured AGNs, as well as the hard X-ray LF. This will

483: illustrate the effect on \eff\, of the obscured AGN population.

484: 

485: 

486: Under the assumption that the local SMBHs

487: grew by accretion:

488: 

489: 

490: \begin{equation}

491: \rho_{\bullet ,\rm loc}=\rho_{\bullet ,\rm acc}

492: \end{equation}

493: 

494: 

495: \noindent With $\rho_{\bullet ,\rm acc}$ given by

496: Equations~\ref{eq:integral_no_r} or \ref{eq:integral}, we can estimate

497: the value of \eff\, which yields the closest agreement between

498: $\rho_{\bullet ,\rm loc}$ and $\rho_{\bullet ,\rm acc}$.  For

499: convenience we define $u\equiv U/C_{\nu}$, where $U$ is defined by

500: Equation~\ref{eq:uto} or \ref{eq:utx}.  This is done to keep separate

501: and explicit the uncertainties in the LF (represented by $u$) and on

502: the bolometric correction (\cbol).

503: 

504: 

505: 

506: We now apply Bayes' theorem, to obtain the posterior PDF of \eff,

507: given the measurement of \densloc\, in the local Universe and

508: marginalising over the parameters $u$, \rtt\, and \cbol:

509: 

510: 

511: \begin{eqnarray}\label{eq:post}

512: {\rm P(\langle \epsilon \rangle|~\rho_{\bullet,\rm loc})} = 

513:  {{\rm P(\langle \epsilon \rangle)} \over {\rm P(\rho_{\bullet,\rm

514:        loc})}}~{\rm P(\rho_{\bullet,\rm loc}|~\langle \epsilon \rangle)} =

515:  \nonumber \\

516:  {{\rm P(\langle \epsilon \rangle)}\over {\rm P(\rho_{\bullet,\rm

517:        loc})}} \int   {\rm P({\it u})}  {\rm

518:    d}u \times \nonumber\\\int   {\rm P({\it \langle R \rangle})}  {\rm

519:    d}\langle R \rangle  \int   {\rm

520:    P({\it C_{\nu}})} {\rm P(\rho_{\bullet,\rm loc}|~\langle \epsilon \rangle,

521:    {\it u},  {\it \langle R \rangle},  {\it C_{\nu}})}{\rm d}C_{\nu}.\,\,\,\,\,\,\, 

522: \end{eqnarray}

523: 

524: \noindent For our purpose, \densloc\, is our data, so that P(\densloc)

525: is the evidence, P($u$), P(\rtt) and P(\cbol) are the prior PDFs for

526: $u$, \rtt\, and \cbol\, respectively. We have made use of the fact

527: that our parameters are independent, so that P(\rtt $|$

528: \eff)$=$P(\rtt), P($u$$|$ \eff)$=$P($u$) and P(\cbol $|$

529: \eff)$=$P(\cbol). P(\densloc$|$ \eff, $u$, \rtt, \cbol) is the

530: likelihood of \densloc\, given \eff, \rtt\, and \cbol:

531: 

532: 

533: \begin{eqnarray}\label{eq:like} 

534: {\rm P(\rho_{\bullet,\rm loc}|~\langle \epsilon \rangle, {\it u},  {\it \langle R

535:     \rangle},  {\it C_{\nu}})} = %\nonumber \\ 

536: {1 \over (2\pi)^{1/2} \sigma_{\rho_{\bullet ,\rm loc}} }e^{-{1\over2}\left(

537:       { \rho_{\bullet ,\rm loc}-\rho_{\bullet ,\rm acc} \over \sigma_{\rho_{\bullet ,\rm loc}}}\right)^{2}}.

538: \end{eqnarray}

539: 

540: \noindent  From

541: Equation~\ref{eq:integral} and the definition of $u$,  $\rho_{\bullet ,\rm acc}$ is:

542: 

543: \begin{equation}\label{eq:densacc_no_r}

544: \rho_{\bullet ,\rm acc} = { (1- \langle \epsilon \rangle) \over \langle

545:   \epsilon \rangle~c^{2}} C_{\nu} u,

546: \end{equation}

547: 

548: \noindent or: 

549: 

550: \begin{equation}\label{eq:densacc}

551: \rho_{\bullet ,\rm acc} = { (1- \langle \epsilon \rangle) \over \langle

552:   \epsilon \rangle~c^{2}}(1+\langle R \rangle)C_{\nu} u,

553: \end{equation}

554: 

555: 

556: \noindent depending on whether obscured AGNs are to be added on via

557: the (1$+$\rtt) term. Note that we have made the simplifying assumption

558: that \rtt\, is independent of luminosity, this will be addressed in

559: Section~\ref{sec:best}.

560: 

561: \subsection{Priors}\label{sec:priors}

562: 

563: We now assign a prior PDF to each of the parameters \cbol, \rtt, $u$,

564: and \eff, considering three scenarios: a) considering only unobscured

565: AGNs b) considering the X-ray absorbed and unabsorbed AGNs, but not

566: the Compton-thick ones and c) attempting to include all obscured and

567: unobscured AGNs d) same as c) but considering only the bright AGNs

568: (quasars).

569: 

570: 

571: 

572: \subsubsection{{\rm P(\cbol)}}

573: 

574: In this work we are interested in the bolometric output from accretion

575: onto the SMBH, which we assume produces X-ray, optical and ultraviolet

576: light predominantly and emits isotropically.  We also assume that the

577: dust responsible for the observed infrared emission (and the

578: orientation-dependent SED) is on a larger scale, and not directly

579: relevant to the accretion disk.  Thus, the bolometric corrections for

580: accretion onto the SMBH must convert from B-band luminosity to

581: bolometric luminosity integrated over the optical, ultraviolet and

582: X-ray range only.

583: 

584: We adopt the luminosity-dependent bolometric corrections of

585: \citet{2004MNRAS.351..169M} and \citet{2007ApJ...654..731H}, both of

586: which follow these assumptions.  We assume gaussian priors around

587: these values, with the former having typical uncertainties

588: ${\sigma_{\rm C_{\rm bol}} \over C_{\rm bol}}=$0.05 and 0.10 for the

589: B-band and hard X-rays, respectively, and the latter ${\sigma_{\rm

590:     C_{\rm bol}} \over C_{\rm bol}}=$0.08 and 0.11 for the same bands.

591: 

592: 

593: 

594: 

595: \subsubsection{{\rm P($u$)}}\label{sec:pu}

596: 

597: We estimate the total radiated energy density from unobscured AGNs by

598: using the B-band LF, and the total radiated energy density from

599: unabsorbed and absorbed (but Compton-thin) AGNs using the hard X-ray

600: LF. The evolution of the LF is such that the luminosity density at $z>

601: 5$ is similar to that at $z < 1$, but the lookback time is much

602: shorter, meaning the energy emitted at $z\geq 5$ is negligible. In

603: other words, the integral in Equation~\ref{eq:integral} has converged

604: by $z=5$. However, the LFs are not as well constrained in the range $3

605: \leq z \leq 5$, and the functional form chosen to represent the

606: evolution (e.g. a 2nd or 3rd order polynomial) can have unphysical

607: effects.  We therefore prefer to restrict the range over which the LFs

608: are integrated to $0 \leq z \leq 3$. The missing radiated energy is

609: expected to be $\lesssim$10\% (see Section~\ref{sec:lf}), and

610: therefore negligible compared to our uncertainties.

611: 

612: Due to the fact that the bolometric corrections are

613: luminosity-dependent and need to be included inside the integral when

614: calculating $U$, we quote here the product of $u$ and \cbol\, assuming

615: the mean value of \cbol. The uncertainty quoted is only due to $u$,

616: estimated by integrating different LFs, while the uncertainty in

617: \cbol\, is encoded in the prior P(\cbol).

618: 

619: For the B-band, due to the different survey areas and sensitivities,

620: we use the LFs of \citet {2000MNRAS.317.1014B} and \citet

621: {2004MNRAS.349.1397C} at $z< 1$ and of \citet{2003A&A...408..499W} in

622: the range $1 \leq z \leq 3$ (using both pure density and pure

623: luminosity evolution). The LF is integrated for all magnitudes

624: brighter than $M_{\rm B}<-18$ ($L_{\nu \rm B}$$\geq$8$\times$$10^{20}$

625: \whz). The estimated total energy from the B-band LFs is found to be

626: $u$\cbol$=$(1.1$\pm$0.2)$\times 10^{51}$~J~Mpc$^{-3}$, using the

627: bolometric corrections of \citet{2004MNRAS.351..169M}, and

628: $u$\cbol$=$(1.6$\pm$0.4)$\times 10^{51}$~J~Mpc$^{-3}$ when using

629: \cbol\, from \citet{2007ApJ...654..731H}.

630: 

631: 

632: If only the most powerful AGNs are considered (quasars, brighter than $M_{\rm

633:   B}$$=$-23 or $L_{\nu \rm B}$$=$7$\times$$10^{22}$ \whz), we find

634: $u$\cbol$=$(0.9$\pm$0.2)$\times10^{51}$ J~Mpc$^{-3}$ \citep[\cbol\,

635: from][]{2004MNRAS.351..169M} and $u$\cbol$=$(1.4$\pm$0.3)$\times

636: 10^{51}$ J~Mpc$^{-3}$ \citep[\cbol\, from][]{2007ApJ...654..731H}.

637: 

638: We note all of the B-band LFs are likely to be missing the population

639: of reddened AGNs. These sources suffer from moderate amounts of

640: extinction in the visual band, $A_{\rm V}\sim1-5$, which suppresses

641: the AGN signatures at optical but not at near-infrared wavelengths

642: ($\sim$1-2~$\mu$m, energies $\sim$0.5-1~eV).

643: \citet{2006ApJ...638...88B} have estimated these to comprise

644: $\sim$20\% of the unobscured population, so the above estimates of the

645: energy density include this uncertainty.

646: 

647: 

648: For the hard X-ray LF, we use the LFs of \citet{2003ApJ...598..886U}

649: and \citet{2008ApJ...679..118S}. The LF is then integrated over

650: $L_{\rm X}\geq$$3\times10^{34}$~W. The estimated radiated energy

651: density, which includes Compton-thin absorbed AGNs, is then

652: $u$\cbol$=$(4.5$\pm$0.1)$\times 10^{51}$~J~Mpc$^{-3}$ \citep[using

653: \cbol\, from][]{2004MNRAS.351..169M} and

654: $u$\cbol$=$(6.4$\pm$0.1)$\times 10^{51}$~J~Mpc$^{-3}$ \citep[\cbol\,

655: from][]{2007ApJ...654..731H}.

656: 

657: 

658: \subsubsection{{\rm P(\rtt)}}

659: 

660: As we have discussed in Section~\ref{sec:demo}, the value of \rtt\, inferred

661: amongst high-luminosity AGNs is $\sim$2, while amongst low-luminosity AGNs it

662: is $\sim$4. In Equations~\ref{eq:integral} and \ref{eq:densacc}, we have made

663: the simplification of keeping \rtt\, independent of luminosity and redshift

664: (and hence outside the integral).  The question of whether \rtt\, evolves or

665: not with $z$ is still unclear but it seems likely that it does vary with

666: luminosity (see Section~\ref{sec:demo}).

667: 

668: We will repeat our analysis using priors centred on \rtt$\sim$2 and 4, to show

669: the different values inferred for \eff. In addition we will also carry out the

670: analysis without including obscured AGNs, using

671: Equations~\ref{eq:integral_no_r} and \ref{eq:densacc_no_r}, which is

672: equivalent to using a prior with \rtt$=$0.

673: 

674: 

675: In the case of \rtt$\sim$2, inferred from the bright AGNs, we assign a

676: prior based on the estimate of \citet{2008ApJ...674..676M}.  This can be

677: approximated to a gaussian PDF with $\mu_{\langle R \rangle}=$1.7 and

678: $\sigma_{R}=$1.2 (their large uncertainty arising from small number

679: statistics).

680: 

681: 

682: Amongst low-luminosity AGNs, several studies suggest \rtt$\sim$4, but there is

683: no explicit estimate that we can use as a prior. For this case, we use a

684: $\delta$-function prior centred on 4. Due to the lack of scatter in \rtt, the

685: posterior PDF for \rtt$=$4 obtained from the $\delta$-function prior will be

686: artificially narrower than the posterior PDF assuming $\mu_{\langle R

687:   \rangle}=$1.7 and $\sigma_{R}=$1.2. This will be dealt with in

688: Section~\ref{sec:best}.

689: 

690: The hard X-ray LF includes unabsorbed and Compton-thin absorbed AGN, so that

691: only Compton-thick AGNs should be missing.  The luminosity-dependent absorbed

692: AGN fraction is also encoded in the LF.  We do not attempt to correct for the

693: missing Compton-thin population, so we are again using

694: Equation~\ref{eq:integral_no_r} and \ref{eq:densacc_no_r} instead of

695: ~\ref{eq:integral} and \ref{eq:densacc}.

696: 

697: 

698: \subsubsection{{\rm P(\eff)}}

699: 

700: 

701: For the efficiency, we reflect our  ignorance by setting a prior in log

702: space, P(log$_{10}$[\eff]), which is flat in the range $-3.0

703: \leq$log$_{10}$[\eff]$\leq 0.0$.

704: 

705: 

706: 

707: 

708: \subsection{The local density of SMBHs}

709: 

710: 

711: For \densloc, we assume a value of $4.6^{+1.9}_{-1.4}$$\times 10^{5}$ \msol\,

712: Mpc$^{-3}$, following \citet{2004MNRAS.351..169M}.  The likelihood is then

713: assumed to be a gaussian distribution around this value

714: (Equation~\ref{eq:like}). We compare the total \densloc\, to the energy

715: density computed using all AGNs brighter than $M_{\rm B}<-18$ (for the B-band

716: LF) or $L_{\rm X}\geq$$3\times10^{34}$~W (hard X-ray LF).

717: 

718: For the case of \rtt$\sim$2, given that this value is motivated by luminous

719: obscured AGNs (quasars), we also consider only the bright end of the optical

720: LF (AGNs with $M_{\rm B}<-23$). We make the assumption that they are hosted

721: by the progenitors of present-day elliptical galaxies, so we compare their integrated

722: energy to the SMBH density in elliptical galaxies only.

723: \citet{2004MNRAS.351..169M} estimated the SMBH density in elliptical

724: galaxies to be 70\% of the total SMBH density, or $3.2^{+1.3}_{-1.0}$$\times

725: 10^{5}$ \msol\, Mpc$^{-3}$.

726: 

727: \subsection{The posterior PDFs for \eff}\label{sec:poste}

728: 

729: Figure~\ref{fig:posteriors} (panels a, b, c and d) shows the resulting

730: posterior PDFs for log$_{10}$[\eff], given our parametrisation, given our

731: priors and given our assumed local density of SMBHs, \densloc. Panel (a) shows

732: the results from the B-band LFs without any correction for obscured AGNs

733: (\rtt$=$0), using our two sets of  values for \cbol. 

734: 

735: Panel (b) shows the result of using the hard X-ray LF. The value of $u$

736: derived from the hard X-ray LF (Section~\ref{sec:pu}) includes both X-ray

737: unabsorbed AGNs and X-ray absorbed Compton-thin AGNs, but not absorbed

738: Compton-thick AGNs.  

739: 

740: Panel (c) shows the results from the B-band LFs and the two different

741: bolometric corrections, with two different values of \rtt\, ($\sim$2 and 4)

742: used to correct for the obscured population (which includes the Compton-thick

743: AGNs as well as the Compton-thin ones).  Note that the posteriors for \eff\,

744: derived from \rtt$=$0 and \rtt$=$4 are artificially narrower than that for

745: \rtt$\sim$2, due to the $\delta$-function priors for \rtt. This is dealt with

746: in Section~\ref{sec:best}.

747: 

748: Panel d) shows the

749: posterior PDFs for the quasars. It is derived from the B-band LFs, integrating

750: only over AGNs brighter than $M_{\rm B}=-23$, considering only \rtt$\sim$2,

751: and comparing to the SMBH density in local elliptical galaxies.

752: 

753: 

754: 

755: 

756: \subsection{Combining  different estimates for \eff}\label{sec:best}

757: 

758: 

759: 

760: The estimates for \eff\, shown in Figure~2 have been derived using different

761: values for \cbol, and in the case of pannel~c, different values for \rtt. We

762: now wish to combine the estimates in each panel, to obtain our four estimates

763: for \eff\, for each of the three scenarios: a) considering only unobscured

764: quasars from the B-band LF b) considering the absorbed and unabsorbed AGNs

765: from the X-ray LF, but not the Compton-thick ones and c) attempting to

766: include all obscured and unobscured AGNs by using the B-band LF and correcting

767: by (1$+$\rtt) finally d) is the same as c) but only for the quasars.

768: 

769: In all four scenarios, we have used two different sets of bolometric

770: corrections. We have no strong reason to believe one set is more appropriate

771: than the other, but they do yield different values of \eff. 

772: Additionally, in scenario c, have derived estimates for \eff\, in the presence

773: of obscured AGNs based on two different priors for \rtt, one where

774: \rtt$\sim$2, the other with \rtt$=$4. Both of these values for \rtt\, are

775: believed to be accurate, but only appropriate at a certain luminosity.

776: Instead, we have used a fixed value of \rtt\, for the entire LF. We now wish

777: to improve our estimate, by considering how appropriate each of our 

778: estimates is, and combining them.

779: 

780: 

781: We follow the arguments of \citet{1997upa..conf...49P}, and consider our

782: different estimates of \eff\, from different combinations of the parameters

783: \cbol\, and \rtt\, as indepedent measurements. Although these inferred values

784: of \eff\, have been derived using the same method, and are therefore

785: apparently not independent, the reason they differ is due to the different

786: assumptions for \cbol\, and \rtt, which are themselves independent.  What we

787: are doing here is to estimate the most likely value of \eff, given the

788: independent parameters assumed.

789: 

790:  We must now assess the appropriateness of each estimate for \eff\, resulting

791:  from using each combination of parameters.  Each resulting estimate of \eff,

792:  \eff$_{i}$, will be considered a good estimate or a bad estimate, depending

793:  on whether the values of \cbol\, and \rtt\, assumed were appropriate or not.

794:  Bad estimates of \eff\, are simply those where the standard deviation has

795:  been underestimated.

796: 

797: 

798: 

799:  To each measurement \eff$_{i}$, we assign the probability $p_{i}$ that the

800:  choice of parameters was the most appropriate and probability (1-$p_{i}$)

801:  that it was not the most appropriate.  We here consider all $p_{i}$'s equal,

802:  simplifying to $p$.

803: 

804: 

805: The posterior probability for the efficiency \eff, given data $D$,  can then

806: be written as

807: 

808: \begin{eqnarray}\label{eq:press}

809: {\rm P({\rm log}_{10}[\langle \epsilon \rangle] | D)} = \nonumber \\{\rm P({\rm log}_{10}[\langle \epsilon \rangle]) \over

810: P(D)}\int  {\rm P({\it p})}\prod_{i}[p{\rm P}_{Gi} + (1-p){\rm P}_{Bi}]{\rm d}p

811: \end{eqnarray}

812: 

813: \noindent where P$_{Gi}$ and P$_{Bi}$ are the probability distribution

814: functions for good and bad measurements, derived from appropriate and

815: inappropriate parameter values \citep[see] [ for a derivation of this

816: result]{1997upa..conf...49P}.

817: 

818: 

819: We  marginalise over $p$. Thus, even if

820: we do not have any information on the probability of a term being correct or

821: not, and have no objective way of assigning values of $p$, we can still

822: proceed. We simply integrate over the entire range of values for $p$, weighted

823: by the prior P($p$).

824: 

825: 

826: To estimate ${\rm P({\rm log}_{10}[\langle \epsilon \rangle] | D)}$, the

827: posterior PDF for \eff\, given our different measurements, we assume a flat prior for

828: log$_{10}$[\eff].  ${\rm P({\rm log}_{10}[\langle \epsilon \rangle]) /

829:   P(D)}$ is in practice simply a normalisation term. For $p$, we assign a

830: prior flat in the range $0 \leq p \leq 1$, reflecting complete ignorance.

831: 

832: 

833: The good probability distribution functions P$_{Gi}$ are those where the

834: uncertainty is correctly estimated, so we use our individual posteriors for

835: each combination of parameters (the dashed, dotted or dash-dotted curves in

836: Figure~2 panel a, b, c or d). For the bad PDFs, P$_{Bi}$, we choose to assume they

837: would have been correct if they had quoted an uncertainty twice as large. We

838: rederive the curves, using exactly the same procedure and priors of

839: Section~\ref{sec:infer}, but with a final uncertainty which is made

840: artificially to be twice as large.

841: 

842:  

843: When combining the different estimates, we obtain an improved one that takes

844: into account not only the different mean parameter values, but also the

845: different uncertainties. For example, the potentially incorrect

846: (underestimated) width of the estimates for \eff\, derived from a

847: $\delta$-function prior at \rtt$=$4 has been dealt with.

848: 

849: The resulting combined posterior PDFs for \eff\, can be seen as the solid

850: curves in Figure~\ref{fig:posteriors} a, b, c and d. 

851: These are our best estimate for

852: the mean radiation efficiency, \eff, given each scenario.

853: 

854: 

855: 

856: \subsection{Results}\label{sec:results}

857: 

858: Our results for the different scenarios illustrate how the efficiency

859: estimated from the So{\l}tan argument varies as the population of obscured

860: AGNs is taken into account.  In all panels of Figure~2, the solid lines show

861: the result from combining the individual posterior PDFs assuming different

862: parameters. The uncertainties quoted below correspond to the gaussian

863: approximation of the solid lines in each figure. This gaussian approximation

864: is extremely accurate for the low-efficiency tail of the PDFs, but

865: underestimates the high-efficiency tail typically above the full-width

866: half-maximum.

867: 

868: When we consider only the unobscured AGNs, we find

869: log$_{10}$[\eff]$=$-1.77$^{+0.16}_{-0.11}$ (Figure~2a) or a relatively low

870: efficiency, \eff$=$1.7$^{+0.8}_{-0.4}$\%.  This is the efficiency derived when

871: no obscured AGNs are taken into account.

872: 

873: Using the hard X-ray LF

874: (Figure~2b), we find log$_{10}$[\eff]$=$-1.20$^{+0.15}_{-0.10}$ or

875: \eff$=$6.4$^{+2.6}_{-1.3}$\%. This estimate for \eff\, includes the absorbed

876: Compton-thin population, but is still missing the Compton-thick AGNs.

877: 

878: 

879: Figure~2c shows the result if we integrate over the optical LF and consider

880: the values of \rtt\, for high-luminosity and low-luminosity AGNs,

881: log$_{10}$[\eff]$=$-1.17$^{+0.11}_{-0.08}$ or \eff$=$6.7$^{+1.9}_{-1.1}$\%. We

882: consider this our overall best estimate for \eff, because it is expected to

883: include the Compton-thick AGNs and therefore to be the least incomplete in

884: terms of obscured AGNs.

885: 

886: 

887: Finally, when only obscured and unobscured quasars are considered we obtain

888: the solid curve in Figure~2d. Here we are using the optical LF integrated

889: above $M_{\rm B}=-23$ and assuming \rtt$\sim$2. We are also only comparing the

890: radiated energy density to the mass density of local SMBHs in elliptical

891: galaxies. The results can then be approximated as

892: log$_{10}$[\eff]$=$-1.27$^{+0.15}_{-0.11}$ or \eff$=$5.4$^{+2.2}_{-1.2}$\%.

893: 

894: 

895: The difference in efficiency between the B-band LF with \rtt$=$0 and the other

896: results shows the cosmological importance of obscured (or absorbed) AGNs.  In

897: Section~\ref{sec:spin}, we discuss how this affects the estimate of SMBH

898: spins, under a set of assumptions.  However, from the different results

899: obtained using different values of \cbol, visible in the different panels of

900: Figure~\ref{fig:posteriors}, we can also infer that the bolometric correction

901: is still a critical source of uncertainty. 

902: 

903: The value of \eff\, has been a matter of some debate, particularly whether its

904: value is above or below $\sim$10\%. In Section~\ref{sec:comp} we compare to

905: results obtained by other studies. We note that, from our best-estimate PDFs

906: from Figure~\ref{fig:posteriors}b, c and d, we find P(\eff$\leq$10\%) ranges

907: between 79 and 88\%. Hence, our results strongly favour values of \eff\, below

908: 10\%.

909: 

910: 

911: 

912: 

913: \subsection{Comparison to previous estimates}\label{sec:comp}

914: 

915: Given the large uncertainties in \rtt, in \cbol\, and in \densloc, our

916: integrated approach is appropriate. Using this method, we obtain

917: \eff$\sim$7\%, and find our results consistent with those of \citet

918: {2004MNRAS.351..169M}, \citet{2004MNRAS.354.1020S} and

919: \citet{2008MNRAS.388.1011M} who used differential LFs (typically hard X-ray

920: LFs) and local SMBH mass functions, and applied extra corrections to account

921: for the Compton-thick population. 

922: 

923: Our results are partially in agreement with those of

924: \citet{2008arXiv0808.0759C}, who find \eff$\sim$8\% when considering

925: the AGNs leading to SMBHs with masses $<10^{8}$~M$_{\odot}$. However,

926: they also find \eff$\grtsim$18\% for $>10^{9}$~M$_{\odot}$, whereas we

927: have found that \eff$\sim$5\% when comparing quasars to the SMBH mass

928: density in elliptical galaxies. The SMBHs in ellipticals do not

929: necessarily map one-to-one with the SMBHs with $>10^{9}$~M$_{\odot}$

930: \citep[see e.g.][]{2004MNRAS.351..169M}, but our results suggest that

931: the mean radiation efficiency remains approximately constant.

932: 

933: We do not, however, find a good agreement with the results of

934: \citet{2008arXiv0808.3777Y}, who derive a value of \eff$\sim$16\%, or

935: those of \citet{2006ApJ...642L.111W} who used a completely independent

936: approach to estimate \eff$=$20-35\%.

937: 

938: Surprisingly, our result is also consistent with that of \citet

939: {2002MNRAS.335..965Y}, despite the fact that they did not include the

940: obscured AGN population. These authors used a lower value for

941: \densloc\, ($2.5\pm0.4$$\times 10^{5}$ \msol\, Mpc$^{-3}$) and also used

942: a higher value of \cbol, since they did not

943: remove the infrared component of the quasar SED. These factors

944: approximately cancel the effect of including the (1$+$\rtt) term,

945: explaining why our estimate of \eff\, is consistent with that of

946: \citet {2002MNRAS.335..965Y}.  

947: 

948: Using the intensity of the X-ray background, and assuming a typical redshift

949: of 2 for the sources dominating it, \citet {2002ApJ...565L..75E} estimated

950: \eff$\grtsim15$\%.  X-ray selected AGNs, however, have been observed to lie

951: predominantly at lower redshifts

952: \citep[e.g.][]{2003AJ....126..632B,2003ApJ...598..886U}. If a typical value of

953: $z=0.7-1$ is assumed instead, then the value of \eff\, derived following the

954: method of \citet {2002ApJ...565L..75E} is reduced to $\sim$8-10\%, in reasonable

955: agreement with our best estimate.

956: 

957: 

958: 

959: \section{Estimating the mean SMBH spin}\label{sec:spin}

960: 

961: The rotation of a black hole determines the radius of the innermost

962: stable circular orbit (ISCO) and therefore its binding energy. The

963: rotation is measured in terms of the parameter $a\equiv{J\over c

964:   m_{\bullet}}$ (in units of ${G m_{\bullet} \over c^{2}}$), where $J$

965: is the angular momentum, $m_{\bullet}$ is the mass of the SMBH, $G$ is

966: the gravitational constant and $c$ is the speed of light.  The

967: accretion efficiency ($\eta$, equal to 1 minus the binding energy and

968: shown in Figure~\ref{fig:spin}) is therefore determined by $a$.

969: 

970: The standard model for a thin accretion disk assumes no stresses are

971: present between the ISCO and the Schwartzschild radius and neglects

972: the effects of radiation from matter falling through this section, as

973: well as assuming that any heat dissipated in the disk is immediately

974: radiated away \citep{1973blho.conf..343N}. These assumptions then lead

975: to $\epsilon = \eta(a)$, so that the mean radiation efficiency can be

976: used to estimate the mean SMBH spin.

977: 

978: Relaxing these assumptions has two important consequences, $\epsilon$

979: is altered and, through accretion alone the SMBH might reach an

980: equilibrium spin value lower than $a={G m_{\bullet} \over c^{2}}$.

981: 

982: When radiation inside the ISCO is considered, the black hole captures

983: preferentially photons with opposite angular momentum, which limits

984: the final spin to $a=$0.998${G m_{\bullet} \over c^{2}}$

985: \citep{1970PhRvD...1.2721G,1974ApJ...191..507T}. Magnetic fields in

986: the region inside the ISCO can transport angular momentum outwards

987: \citep{1977MNRAS.179..433B}, possibly further limiting the equilibrium value

988: of $a$ \citep[][ estimate a limit of $\sim$0.9${G m_{\bullet} \over

989:   c^{2}}$]{2004ApJ...602..312G,2005ApJ...622.1008K}.

990: 

991: 

992: The assumption that $\epsilon = \eta$ requires that no torques are present

993: within the ISCO, and that the matter ``plunges'' into the black hole too

994: quickly for any radiation to be significant. In this case, no energy can be

995: extracted once matter has crossed the ISCO. However, the infalling matter can

996: exhert a torque through magnetic fields, so that work is still done on the

997: disk even after crossing the ISCO. The dissipation of this work can then lead

998: to a higher radiative efficiency

999: \citep[e.g.][]{2008MNRAS.tmp.1003B,2008arXiv0808.3140N}.

1000: 

1001: 

1002: 

1003: In the following section, we discuss our estimates of $\langle a

1004: \rangle$ from \eff\, assuming $\epsilon = \eta$, but from the

1005: discussion above these values should be considered as upper limits,

1006: since $\epsilon$ is really expected to be $>\eta(a)$. For the case of

1007: SMBH and disk co-rotating with $a\sim0.9 {G m_{\bullet} \over c^{2}}$,

1008: the ISCO is very close to the Schwartzschild radius, so that there is

1009: little time to radiate in the region inside of the ISCO. In addition,

1010: the gravitational redshift diminishes the energy of the emergent

1011: radiation. For this case, \citet{2008arXiv0808.3140N} have found that

1012: $\epsilon\sim 1.06 \eta$, so the difference is negligible compared to

1013: our uncertainties. However, for lower values of $a$ the difference

1014: between $\epsilon$ and $\eta$ might be larger, even of order unity

1015: \citep{2008MNRAS.tmp.1003B}.

1016: 

1017: 

1018: Our value of \eff\, derived in the absence of obscured AGNs

1019: (\eff$=$1.7$^{+0.8}_{-0.4}$\% for \rtt$=$0) is consistent with SMBHs having no

1020: rotation during periods of accretion (\eff$<$5.7\% so that $a=0$). An

1021: important point is therefore that when obscured AGNs are not included, the

1022: derived efficiency suggests that SMBHs were not spinning during periods of

1023: accretion.

1024: 

1025: 

1026: When we do include the obscured AGNs, we find that values of \eff$>$5.7\% are

1027: likely, so that the possibility that the SMBHs are rotating must be

1028: considered.

1029: 

1030: Maximally rotating black holes have $a={G m_{\bullet} \over c^{2}}$

1031: with $\eta=$42\% \citep{1970Natur.226...64B}, although in the standard

1032: accretion disk model they are only expected to reach $a=0.998{G

1033:   m_{\bullet} \over c^{2}}$ at most, corresponding to $\eta=$30\%

1034: \citep{1974ApJ...191..507T}.  From our posterior PDFs,

1035: P(\eff$\geq$30\%) ranges from 0.3\%\footnote{These probabilities have

1036:   been calculated from the actual PDFs, not the gaussian approximation

1037:   which underestimates the probability in the tails of the PDFs.} to

1038: 2\%$^{2}$.  Thus, under our parametrisation, we can rule out that

1039: SMBHs are, on average, maximally rotating with $\geq$98\%

1040: confidence. More recent work suggests the upper limit on $a$ might be

1041: lower, $\sim$0.9${G m_{\bullet} \over c^{2}}$, and with a radiative

1042: efficiency $\sim$15\% \citep{2008arXiv0808.3140N}. Again, from our

1043: posterior PDFs, P(\eff$\geq$15\%) ranges from 2\% to 8\%, so we rule

1044: out $\langle a \rangle \geq 0.9 {G m_{\bullet} \over c^{2}}$ at the

1045: $\geq$92\% confidence.

1046: 

1047: 

1048: 

1049: 

1050: For spinning black holes, $\langle \eta \rangle$ will vary depending on whether the ISCO is

1051: co-rotating or counter-rotating with respect to the black hole spin (see

1052: Figure~\ref{fig:spin}). To estimate the mean value of $a$  two possible limiting scenarios can be envisaged:

1053: 

1054: {\it A) The SMBHs grow from the seed in one long episode of accretion}. In

1055: this case one single accretion disk is expected to be able to provide enough

1056: mass to turn the seed into a SMBH. The

1057: angular momentum of the seed is negligible compared to that of the accretion

1058: disk, so the disk will inevitably force the black hole to co-align

1059: \citep{1996MNRAS.282..291S,2005MNRAS.363...49K}. It is therefore safe to

1060: estimate the efficiency assuming the black hole and ISCO are

1061: always co-rotating. In this case, our best estimate of \eff$=$6.7$^{+1.9}_{-1.1}$\%

1062: corresponds to SMBHs rotating with $\langle a \rangle=0.25^{+0.30}_{-0.25}{G m_{\bullet} \over

1063:   c^{2}}$ during periods of accretion (Figure~\ref{fig:spin}, solid line).

1064: 

1065: 

1066: 

1067: 

1068: {\it B) The SMBHs grow in many shorter periods of accretion}. The implication

1069: here is that the angular momenta of the accretion disk and black hole are

1070: comparable and are also initially randomly oriented with respect to each

1071: other. Depending on the initial configuration and ratio of angular momenta,

1072: the black hole and disk will either co-align or counter-align

1073: \citep{2005MNRAS.363...49K,2006MNRAS.368.1196L}. It is expected that co- and

1074: counter-aligned accretion is equally probable, so when estimating $a$, a mean $\langle \eta(a) \rangle$ from co- and counter-rotating

1075: orbits should be used \citep[the symmetrised curve suggested by][shown in

1076: Figure~\ref{fig:spin} by the dashed line]{2008MNRAS.385.1621K}. In this case,

1077: for our best estimate of \eff\, we inferr 

1078: $\langle a \rangle=0.59^{+0.25}_{-0.59}{G m_{\bullet} \over c^{2}}$.

1079: 

1080: 

1081: 

1082: 

1083: Mergers involving SMBHs are expected to prevent $a$ from reaching the

1084: highest values \citep[see the work of ][ who find half of their

1085: parameter space will lead to $a \leq 0.4{G m_{\bullet} \over

1086:   c^{2}}$]{2003ApJ...585L.101H}. For both accretion scenarios, and for

1087: $a\lesssim0.8{G m_{\bullet} \over c^{2}}$, the dependence of

1088: $\eta$ on $a$ is weak so that it is difficult to constrain

1089: $\langle a \rangle$ from our estimate of \eff.  However, we have found

1090: that maximally rotating SMBHs are not favoured since

1091: P(\eff$\geq$30\%)$\leq$2\%, and our best estimates for $a$ are in

1092: reasonable agreement with the results of \citet{2003ApJ...585L.101H}.

1093: 

1094: 

1095: 

1096: 

1097: Our estimates for $\langle a \rangle$ are nevertheless also consistent with

1098: the value of $a\geq0.5{G m_{\bullet} \over c^{2}}$ estimated for the SMBH in

1099: the Milky Way, Sagittarius A$^{\star}$ \citep[][ who explained the short

1100: near-infrared flare period of 17 minutes as being close the ISCO for

1101: co-rotating disk and SMBH]{2003Natur.425..934G}. A SMBH with $a\geq0.5{G

1102:   m_{\bullet} \over c^{2}}$ and a prograde orbit has an accretion efficiency

1103: $\eta\geq$8\%. Thus, the limit on $a$ inferred for the individual SMBH

1104: Sagittarius A$^{\star}$ is in good agreement with our best estimate of the

1105: mean radiation efficiency, \eff, assuming the standard thin disk model and

1106: ignoring the magnetic fields.

1107: 

1108: Indeed, we remind the reader that we have estimated $\langle a \rangle$

1109: assuming $\epsilon=\eta$, but at the beginning of this Section we have argued

1110: for $\epsilon>\eta$. Hence, our estimates of $\langle a \rangle$ can be

1111: considered upper limits, since for a given true value of $a$,

1112: $\epsilon>\eta(a)$

1113: \citep{2005ApJ...622.1008K,2008arXiv0808.3140N,2008MNRAS.tmp.1003B}. For

1114: $a\sim0.9 {G m_{\bullet} \over

1115:     c^{2}}$, it seems that $\epsilon$ is only slightly larger than $\eta$, but

1116: for lower values of $a$ the difference is expected to be larger, since the

1117: time available for radiation is longer and the gravitational redshift is less

1118: severe.  This only strengthens our main conclusion that our inferred value of

1119: \eff\, does not require accreting SMBHs to be rotating particularly rapidly,

1120: on average, during periods of accretion.

1121: 

1122: 

1123: 

1124: 

1125: \section{The mean $e$-folding time}\label{sec:efold}

1126: 

1127: Given our assumption that AGNs grow by accretion, their mass will grow

1128: exponentially provided that the luminosity is proportional to the mass. Thus,

1129: if AGNs radiate at a fraction $\lambda$ of their Eddington luminosity, then

1130: they will grow exponentially with an $e$-folding time:

1131: 

1132: 

1133: 

1134: \begin{equation}

1135: \tau = {\epsilon \over \lambda (1-\epsilon)}{c \sigma_{\rm T} \over  4

1136:     \pi G m_{\rm p} }

1137: \end{equation}

1138: 

1139: \noindent where $m_{\rm p}$ is the mass of the proton and $\sigma_{T}$ is the

1140:   Thomson electron scattering cross-section. Our estimate of \eff\, is

1141:   completely independent of $\lambda$, and it is only to estimate $\tau$ that

1142:   we require an estimate of the Eddington ratio $\lambda$.

1143: 

1144:   Estimates of $\lambda$ exist for unobscured quasars, derived from their

1145:   bolometric luminosities  and virial-mass estimates for the SMBH

1146:   mass.  \citet{2004MNRAS.352.1390M} found $\lambda\sim$0.15-0.5 in the range

1147:   $0.1\leq z \leq 2.1$, with significant scatter.  We approximate the 

1148:  distribution of

1149:   ${\rm log}_{10}[\lambda]$ to a gaussian with $\mu_{{\rm

1150:       log}_{10}[\lambda]}$$=$-0.6 and $\sigma_{{\rm

1151:       log}_{10}(\lambda)}$$=$0.4.

1152: 

1153: 

1154:   When combined with our estimate of \eff\, for quasars (5.4$^{+2.2}_{-1.2}$\%),

1155:   this leads to an estimate of log$_{10}$[$\langle \tau

1156:   \rangle$/Myr]$=2.0\pm0.4$, or $\langle \tau \rangle=100^{+151}_{-60}$~Myr.

1157:   This value is consistent with (although at the high end of) other current

1158:   constraints from quasar clustering measurements, from arguments based on

1159:   coevolution of quasars, SMBHs and host galaxies, from arguments of

1160:   simultaneous triggering of quasars and starbursts and from the lengths of

1161:   radio jets. These estimates all suggest $\tau$ is in the range 10-100~Myr

1162:   \citep[see][ for a review]{2004cbhg.symp..169M}.

1163: 

1164: 

1165: 

1166:   Such a long value of $\tau$ suggests SMBHs require a long time to build up

1167:   their mass. At $z=6.4$, SMBH masses of 2-6$\times10^{9}$~\msol\, have been

1168:   estimated \citep{2003ApJ...587L..15W,2003ApJ...594L..95B}, when the Universe

1169:   was only 840~Myrs old. This suggests that the seeds that formed the $z=6$

1170:   quasars might have been as large as $\sim10^{5}-10^{6}$~\msol.  

1171: 

1172:   The estimates of the SMBH mass and particularly our estimate of the

1173:   $e$-folding time have very large uncertainties and the quasars

1174:   at $z=6.4$ could have  shorter $e$-folding times. They might

1175:   be accreting closer to the Eddington limit than $\lambda$$\sim$0.25, the

1176:   mean of the $0.1 \leq z \leq 2.1$ quasars \citep[][]{2004MNRAS.352.1390M},

1177:   or radiating less efficiently than $\sim$7\% . This would then imply

1178:   significantly smaller seeds. For example, if $\tau\sim$60~Myr (40~Myr), the

1179:   seeds could be $\sim$$10^{3}$~\msol\, ($\sim$1~\msol).

1180: 

1181: \section {Summary}\label{sec:conclusion} 

1182: 

1183: We have revisited the mean radiation efficiency from accretion onto SMBHs,

1184: \eff, quantifying the effect introduced by taking into account a large

1185: population of obscured AGNs.  We estimate the cumulative energy density of the

1186: AGN population from the unobscured AGN LF by applying appropriate bolometric

1187: corrections and correcting for the obscured population. The efficiency \eff\,

1188: is estimated by comparing this energy density to the SMBH mass density in the

1189: local Universe.

1190: 

1191: When no obscured AGNs are included, we find \eff$=$1.7$^{+0.8}_{-0.4}$\%  which

1192: does not require SMBHs to rotate at all during periods of accretion ($\langle a\rangle$$=0$).

1193: 

1194: Using the hard X-ray LF, which includes absorbed Compton-thin AGNs, we find

1195: \eff$=$6.4$^{+2.6}_{-1.3}$\%. Using the B-band LFs, correcting for the

1196: obscured population (which includes the Compton-thick AGNs), we find

1197: \eff$=$6.7$^{+1.9}_{-1.1}$\%, which we consider to be our best estimate.

1198: 

1199: We have also derived \eff\, for the powerful AGNs only (the quasars), and find  \eff$=$5.4$^{+2.2}_{-1.2}$\%.

1200: 

1201: 

1202: 

1203: Depending on whether the SMBHs and accretion disks are assumed to be always

1204: co-rotating, or instead to counter-rotate in half of the cases, we estimate

1205: the rotation parameter during periods of accretion to be $\langle a

1206: \rangle=0.25^{+0.30}_{-0.25}{G m_{\bullet} \over c^{2}}$ or $\langle a

1207: \rangle=0.59^{+0.25}_{-0.59}{G m_{\bullet} \over c^{2}}$, respectively

1208: 

1209: 

1210: Efficiencies $\geq$30\% ($\geq$15\%) are ruled out at $\geq$98\% ($\geq$92\%)

1211: confidence, so the SMBHs are not, on average, maximally rotating during the

1212: periods of accretion.  Efficiencies $\leq$10\% are favoured, with 79\%-88\% of

1213: the probability density in this region.

1214: 

1215: 

1216: Finally, combining our best estimate for \eff\, with estimates of the

1217: Eddington ratio for unobscured quasars yields an estimate of the $e$-folding

1218: time of $\langle \tau \rangle=100^{+151}_{-60}$~Myr. This 

1219: suggests that the seeds of $z=6$ quasars might have been very large,

1220:  potentially as large as $\sim10^{6}$~\msol.  However, the large

1221: uncertainties in $\tau$ translate into order-of-magnitude uncertainties for

1222: the masses of the seeds, which could be as small as $\sim1$~\msol.

1223: 

1224: 

1225: 

1226: 

1227:  \acknowledgements 

1228: 

1229:  We are grateful to John Miller, Andrew King and Giuseppe Lodato for useful

1230:  discussions, to Ross McLure for access to the distribution of Eddington

1231:  ratios, to Fabio Fontanot and Mark Sargent for comments on the manuscript and

1232:  to the anonymous referee for suggestions that improved the article

1233:  significantly.

1234: 

1235: \bibliographystyle{apj} 

1236: 

1237: 

1238:  

1239: \begin{thebibliography}{55}

1240: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

1241: 

1242: \bibitem[{{Alonso-Herrero} {et~al.}(2006){Alonso-Herrero},

1243:   {P{\'e}rez-Gonz{\'a}lez}, {Alexander}, {Rieke}, {Rigopoulou}, {Le Floc'h},

1244:   {Barmby}, {Papovich}, {Rigby}, {Bauer}, {Brandt}, {Egami}, {Willner}, {Dole},

1245:   \& {Huang}}]{2006ApJ...640..167A}

1246: {Alonso-Herrero}, A., {et~al.} 2006, \apj, 640, 167

1247: 

1248: \bibitem[{{Baker} {et~al.}(2002){Baker}, {Hunstead}, {Athreya}, {Barthel}, {de

1249:   Silva}, {Lehnert}, \& {Saunders}}]{2002ApJ...568..592B}

1250: {Baker}, J.~C., {Hunstead}, R.~W., {Athreya}, R.~M., {Barthel}, P.~D., {de

1251:   Silva}, E., {Lehnert}, M.~D., \& {Saunders}, R.~D.~E. 2002, \apj, 568, 592

1252: 

1253: \bibitem[{{Bardeen}(1970)}]{1970Natur.226...64B}

1254: {Bardeen}, J.~M. 1970, \nat, 226, 64

1255: 

1256: \bibitem[{{Barger} {et~al.}(2003){Barger}, {Cowie}, {Capak}, {Alexander},

1257:   {Bauer}, {Fernandez}, {Brandt}, {Garmire}, \&

1258:   {Hornschemeier}}]{2003AJ....126..632B}

1259: {Barger}, A.~J., {et~al.} 2003, \aj, 126, 632

1260: 

1261: \bibitem[{{Barth} {et~al.}(2003){Barth}, {Martini}, {Nelson}, \&

1262:   {Ho}}]{2003ApJ...594L..95B}

1263: {Barth}, A.~J., {Martini}, P., {Nelson}, C.~H., \& {Ho}, L.~C. 2003, \apjl,

1264:   594, L95

1265: 

1266: \bibitem[{{Beckwith} {et~al.}(2008){Beckwith}, {Hawley}, \&

1267:   {Krolik}}]{2008MNRAS.tmp.1003B}

1268: {Beckwith}, K., {Hawley}, J.~F., \& {Krolik}, J.~H. 2008, \mnras, 1003

1269: 

1270: \bibitem[{{Blandford} \& {Znajek}(1977)}]{1977MNRAS.179..433B}

1271: {Blandford}, R.~D. \& {Znajek}, R.~L. 1977, \mnras, 179, 433

1272: 

1273: \bibitem[{{Boyle} {et~al.}(2000){Boyle}, {Shanks}, {Croom}, {Smith}, {Miller},

1274:   {Loaring}, \& {Heymans}}]{2000MNRAS.317.1014B}

1275: {Boyle}, B.~J., {Shanks}, T., {Croom}, S.~M., {Smith}, R.~J., {Miller}, L.,

1276:   {Loaring}, N., \& {Heymans}, C. 2000, \mnras, 317, 1014

1277: 

1278: \bibitem[{{Brandt} \& {Hasinger}(2005)}]{2005ARA&A..43..827B}

1279: {Brandt}, W.~N. \& {Hasinger}, G. 2005, \araa, 43, 827

1280: 

1281: \bibitem[{{Brown} {et~al.}(2006)}]{2006ApJ...638...88B}

1282: {Brown}, M.~J.~I. {et~al.} 2006, \apj, 638, 88

1283: 

1284: \bibitem[{{Cao} \& {Li}(2008)}]{2008arXiv0808.0759C}

1285: {Cao}, X. \& {Li}, F. 2008, ArXiv e-prints, 808

1286: 

1287: \bibitem[{{Croom} {et~al.}(2004){Croom}, {Smith}, {Boyle}, {Shanks}, {Miller},

1288:   {Outram}, \& {Loaring}}]{2004MNRAS.349.1397C}

1289: {Croom}, S.~M., {Smith}, R.~J., {Boyle}, B.~J., {Shanks}, T., {Miller}, L.,

1290:   {Outram}, P.~J., \& {Loaring}, N.~S. 2004, \mnras, 349, 1397

1291: 

1292: \bibitem[{{Donley} {et~al.}(2007){Donley}, {Rieke}, {P{\'e}rez-Gonz{\'a}lez},

1293:   {Rigby}, \& {Alonso-Herrero}}]{2007ApJ...660..167D}

1294: {Donley}, J.~L., {Rieke}, G.~H., {P{\'e}rez-Gonz{\'a}lez}, P.~G., {Rigby},

1295:   J.~R., \& {Alonso-Herrero}, A. 2007, \apj, 660, 167

1296: 

1297: \bibitem[{{Donley} {et~al.}(2005){Donley}, {Rieke}, {Rigby}, \&

1298:   {P{\'e}rez-Gonz{\'a}lez}}]{2005ApJ...634..169D}

1299: {Donley}, J.~L., {Rieke}, G.~H., {Rigby}, J.~R., \& {P{\'e}rez-Gonz{\'a}lez},

1300:   P.~G. 2005, \apj, 634, 169

1301: 

1302: \bibitem[{{Eckart} \& {Genzel}(1996)}]{1996Natur.383..415E}

1303: {Eckart}, A. \& {Genzel}, R. 1996, \nat, 383, 415

1304: 

1305: \bibitem[{{Elvis} {et~al.}(2002){Elvis}, {Risaliti}, \&

1306:   {Zamorani}}]{2002ApJ...565L..75E}

1307: {Elvis}, M., {Risaliti}, G., \& {Zamorani}, G. 2002, \apjl, 565, L75

1308: 

1309: \bibitem[{{Gammie} {et~al.}(2004){Gammie}, {Shapiro}, \&

1310:   {McKinney}}]{2004ApJ...602..312G}

1311: {Gammie}, C.~F., {Shapiro}, S.~L., \& {McKinney}, J.~C. 2004, \apj, 602, 312

1312: 

1313: \bibitem[{{Genzel} {et~al.}(2003){Genzel}, {Sch{\"o}del}, {Ott}, {Eckart},

1314:   {Alexander}, {Lacombe}, {Rouan}, \& {Aschenbach}}]{2003Natur.425..934G}

1315: {Genzel}, R., {Sch{\"o}del}, R., {Ott}, T., {Eckart}, A., {Alexander}, T.,

1316:   {Lacombe}, F., {Rouan}, D., \& {Aschenbach}, B. 2003, \nat, 425, 934

1317: 

1318: \bibitem[{{Gilli} {et~al.}(2007){Gilli}, {Comastri}, \&

1319:   {Hasinger}}]{2007A&A...463...79G}

1320: {Gilli}, R., {Comastri}, A., \& {Hasinger}, G. 2007, \aap, 463, 79

1321: 

1322: \bibitem[{{Godfrey}(1970)}]{1970PhRvD...1.2721G}

1323: {Godfrey}, B.~B. 1970, \prd, 1, 2721

1324: 

1325: \bibitem[{{Hopkins} {et~al.}(2007){Hopkins}, {Richards}, \&

1326:   {Hernquist}}]{2007ApJ...654..731H}

1327: {Hopkins}, P.~F., {Richards}, G.~T., \& {Hernquist}, L. 2007, \apj, 654, 731

1328: 

1329: \bibitem[{{Hughes} \& {Blandford}(2003)}]{2003ApJ...585L.101H}

1330: {Hughes}, S.~A. \& {Blandford}, R.~D. 2003, \apjl, 585, L101

1331: 

1332: \bibitem[{{Ivezi{\'c}} {et~al.}(2002){Ivezi{\'c}}, {Menou}, {Knapp}, {Strauss},

1333:   {Lupton}, {Vanden Berk}, {Richards}, {Tremonti}, {Weinstein}, {Anderson},

1334:   {Bahcall}, {Becker}, {Bernardi}, {Blanton}, {Eisenstein}, {Fan},

1335:   {Finkbeiner}, {Finlator}, {Frieman}, {Gunn}, {Hall}, {Kim}, {Kinkhabwala},

1336:   {Narayanan}, {Rockosi}, {Schlegel}, {Schneider}, {Strateva}, {SubbaRao},

1337:   {Thakar}, {Voges}, {White}, {Yanny}, {Brinkmann}, {Doi}, {Fukugita},

1338:   {Hennessy}, {Munn}, {Nichol}, \& {York}}]{2002AJ....124.2364I}

1339: {Ivezi{\'c}}, {\v Z}., {et~al.} 2002, \aj, 124, 2364

1340: 

1341: \bibitem[{{King} {et~al.}(2005){King}, {Lubow}, {Ogilvie}, \&

1342:   {Pringle}}]{2005MNRAS.363...49K}

1343: {King}, A.~R., {Lubow}, S.~H., {Ogilvie}, G.~I., \& {Pringle}, J.~E. 2005,

1344:   \mnras, 363, 49

1345: 

1346: \bibitem[{{King} {et~al.}(2008){King}, {Pringle}, \&

1347:   {Hofmann}}]{2008MNRAS.385.1621K}

1348: {King}, A.~R., {Pringle}, J.~E., \& {Hofmann}, J.~A. 2008, \mnras, 385, 1621

1349: 

1350: \bibitem[{{Kormendy} \& {Richstone}(1995)}]{1995ARA&A..33..581K}

1351: {Kormendy}, J. \& {Richstone}, D. 1995, \araa, 33, 581

1352: 

1353: \bibitem[{{Krolik} {et~al.}(2005){Krolik}, {Hawley}, \&

1354:   {Hirose}}]{2005ApJ...622.1008K}

1355: {Krolik}, J.~H., {Hawley}, J.~F., \& {Hirose}, S. 2005, \apj, 622, 1008

1356: 

1357: \bibitem[{{Lacy} {et~al.}(2007){Lacy}, {Petric}, {Sajina}, {Canalizo},

1358:   {Storrie-Lombardi}, {Armus}, {Fadda}, \& {Marleau}}]{2007AJ....133..186L}

1359: {Lacy}, M., {Petric}, A.~O., {Sajina}, A., {Canalizo}, G., {Storrie-Lombardi},

1360:   L.~J., {Armus}, L., {Fadda}, D., \& {Marleau}, F.~R. 2007, \aj, 133, 186

1361: 

1362: \bibitem[{{Lawrence} \& {Elvis}(1982)}]{1982ApJ...256..410L}

1363: {Lawrence}, A. \& {Elvis}, M. 1982, \apj, 256, 410

1364: 

1365: \bibitem[{{Lodato} \& {Pringle}(2006)}]{2006MNRAS.368.1196L}

1366: {Lodato}, G. \& {Pringle}, J.~E. 2006, \mnras, 368, 1196

1367: 

1368: \bibitem[{{Lynden-Bell}(1969)}]{1969Natur.223..690L}

1369: {Lynden-Bell}, D. 1969, \nat, 223, 690

1370: 

1371: \bibitem[{{Magorrian} {et~al.}(1998){Magorrian}, {Tremaine}, {Richstone},

1372:   {Bender}, {Bower}, {Dressler}, {Faber}, {Gebhardt}, {Green}, {Grillmair},

1373:   {Kormendy}, \& {Lauer}}]{1998AJ....115.2285M}

1374: {Magorrian}, J., {et~al.} 1998, \aj, 115, 2285

1375: 

1376: \bibitem[{{Marconi} {et~al.}(2004){Marconi}, {Risaliti}, {Gilli}, {Hunt},

1377:   {Maiolino}, \& {Salvati}}]{2004MNRAS.351..169M}

1378: {Marconi}, A., {Risaliti}, G., {Gilli}, R., {Hunt}, L.~K., {Maiolino}, R., \&

1379:   {Salvati}, M. 2004, \mnras, 351, 169

1380: 

1381: \bibitem[{{Mart{\'{\i}}nez-Sansigre} {et~al.}(2008){Mart{\'{\i}}nez-Sansigre},

1382:   {Lacy}, {Sajina}, \& {Rawlings}}]{2008ApJ...674..676M}

1383: {Mart{\'{\i}}nez-Sansigre}, A., {Lacy}, M., {Sajina}, A., \& {Rawlings}, S.

1384:   2008, \apj, 674, 676

1385: 

1386: \bibitem[{{Mart{\'{\i}}nez-Sansigre} {et~al.}(2007){Mart{\'{\i}}nez-Sansigre},

1387:   {Rawlings}, {Bonfield}, {Mateos}, {Simpson}, {Watson}, {Almaini}, {Foucaud},

1388:   {Sekiguchi}, \& {Ueda}}]{2007MNRAS.379L...6M}

1389: {Mart{\'{\i}}nez-Sansigre}, A., {et~al.} 2007, \mnras, 379, L6

1390: 

1391: \bibitem[{{Mart{\'{\i}}nez-Sansigre} {et~al.}(2006){Mart{\'{\i}}nez-Sansigre},

1392:   {Rawlings}, {Lacy}, {Fadda}, {Jarvis}, {Marleau}, {Simpson}, \&

1393:   {Willott}}]{2006MNRAS.370.1479M}

1394: {Mart{\'{\i}}nez-Sansigre}, A., {Rawlings}, S., {Lacy}, M., {Fadda}, D.,

1395:   {Jarvis}, M.~J., {Marleau}, F.~R., {Simpson}, C., \& {Willott}, C.~J. 2006,

1396:   \mnras, 370, 1479

1397: 

1398: \bibitem[{{Martini}(2004)}]{2004cbhg.symp..169M}

1399: {Martini}, P. 2004, in Coevolution of Black Holes and Galaxies, ed. L.~C. {Ho},

1400:   169--+

1401: 

1402: \bibitem[{{McLure} \& {Dunlop}(2004)}]{2004MNRAS.352.1390M}

1403: {McLure}, R.~J. \& {Dunlop}, J.~S. 2004, \mnras, 352, 1390

1404: 

1405: \bibitem[{{Merloni} \& {Heinz}(2008)}]{2008MNRAS.388.1011M}

1406: {Merloni}, A. \& {Heinz}, S. 2008, \mnras, 388, 1011

1407: 

1408: \bibitem[{{Noble} {et~al.}(2008){Noble}, {Krolik}, \&

1409:   {Hawley}}]{2008arXiv0808.3140N}

1410: {Noble}, S.~C., {Krolik}, J.~H., \& {Hawley}, J.~F. 2008, ArXiv e-prints, 808

1411: 

1412: \bibitem[{{Novikov} \& {Thorne}(1973)}]{1973blho.conf..343N}

1413: {Novikov}, I.~D. \& {Thorne}, K.~S. 1973, in Black Holes (Les Astres Occlus),

1414:   343--450

1415: 

1416: \bibitem[{{Polletta} {et~al.}(2006)}]{2006ApJ...642..673P}

1417: {Polletta}, M.~d.~C. {et~al.} 2006, \apj, 642, 673

1418: 

1419: \bibitem[{{Press}(1997)}]{1997upa..conf...49P}

1420: {Press}, W.~H. 1997, in Unsolved Problems in Astrophysics, 49--60

1421: 

1422: \bibitem[{{Rees}(1984)}]{1984ARA&A..22..471R}

1423: {Rees}, M.~J. 1984, \araa, 22, 471

1424: 

1425: \bibitem[{{Reyes} {et~al.}(2008){Reyes}, {Zakamska}, {Strauss}, {Green},

1426:   {Krolik}, {Shen}, {Richards}, {Anderson}, \&

1427:   {Schneider}}]{2008arXiv0801.1115R}

1428: {Reyes}, R., {et~al.} 2008, ArXiv

1429:   e-prints, 801

1430: 

1431: \bibitem[{{Rigby} {et~al.}(2006){Rigby}, {Rieke}, {Donley}, {Alonso-Herrero},

1432:   \& {P{\'e}rez-Gonz{\'a}lez}}]{2006ApJ...645..115R}

1433: {Rigby}, J.~R., {Rieke}, G.~H., {Donley}, J.~L., {Alonso-Herrero}, A., \&

1434:   {P{\'e}rez-Gonz{\'a}lez}, P.~G. 2006, \apj, 645, 115

1435: 

1436: \bibitem[{{Risaliti} {et~al.}(1999){Risaliti}, {Maiolino}, \&

1437:   {Salvati}}]{1999ApJ...522..157R}

1438: {Risaliti}, G., {Maiolino}, R., \& {Salvati}, M. 1999, \apj, 522, 157

1439: 

1440: \bibitem[{{Salpeter}(1964)}]{1964ApJ...140..796S}

1441: {Salpeter}, E.~E. 1964, \apj, 140, 796

1442: 

1443: \bibitem[{{Scheuer} \& {Feiler}(1996)}]{1996MNRAS.282..291S}

1444: {Scheuer}, P.~A.~G. \& {Feiler}, R. 1996, \mnras, 282, 291

1445: 

1446: \bibitem[{{Shankar} {et~al.}(2004){Shankar}, {Salucci}, {Granato}, {De Zotti},

1447:   \& {Danese}}]{2004MNRAS.354.1020S}

1448: {Shankar}, F., {Salucci}, P., {Granato}, G.~L., {De Zotti}, G., \& {Danese}, L.

1449:   2004, \mnras, 354, 1020

1450: 

1451: \bibitem[{{Silverman} {et~al.}(2008){Silverman}, {Green}, {Barkhouse}, {Kim},

1452:   {Kim}, {Wilkes}, {Cameron}, {Hasinger}, {Jannuzi}, {Smith}, {Smith}, \&

1453:   {Tananbaum}}]{2008ApJ...679..118S}

1454: {Silverman}, J.~D., {et~al.} 2008, \apj, 679, 118

1455: 

1456: \bibitem[{{Simpson}(2005)}]{2005MNRAS.360..565S}

1457: {Simpson}, C. 2005, \mnras, 360, 565

1458: 

1459: \bibitem[{{So{\l}tan}(1982)}]{1982MNRAS.200..115S}

1460: {So{\l}tan}, A. 1982, \mnras, 200, 115

1461: 

1462: \bibitem[{{Thorne}(1974)}]{1974ApJ...191..507T}

1463: {Thorne}, K.~S. 1974, \apj, 191, 507

1464: 

1465: \bibitem[{{Treister} \& {Urry}(2006)}]{2006ApJ...652L..79T}

1466: {Treister}, E. \& {Urry}, C.~M. 2006, \apjl, 652, L79

1467: 

1468: \bibitem[{{Ueda} {et~al.}(2003){Ueda}, {Akiyama}, {Ohta}, \&

1469:   {Miyaji}}]{2003ApJ...598..886U}

1470: {Ueda}, Y., {Akiyama}, M., {Ohta}, K., \& {Miyaji}, T. 2003, \apj, 598, 886

1471: 

1472: 

1473: \bibitem[{{Wang} {et~al.}(2006){Wang}, {Chen}, {Ho}, \&

1474:   {McLure}}]{2006ApJ...642L.111W}

1475: {Wang}, J.-M., {Chen}, Y.-M., {Ho}, L.~C., \& {McLure}, R.~J. 2006, \apjl, 642,

1476:   L111

1477: 

1478: \bibitem[{{Willott} {et~al.}(2003){Willott}, {McLure}, \&

1479:   {Jarvis}}]{2003ApJ...587L..15W}

1480: {Willott}, C.~J., {McLure}, R.~J., \& {Jarvis}, M.~J. 2003, \apjl, 587, L15

1481: 

1482: \bibitem[{{Willott} {et~al.}(2002){Willott}, {Rawlings}, {Archibald}, \&

1483:   {Dunlop}}]{2002MNRAS.331..435W}

1484: {Willott}, C.~J., {Rawlings}, S., {Archibald}, E.~N., \& {Dunlop}, J.~S. 2002,

1485:   \mnras, 331, 435

1486: 

1487: \bibitem[{{Willott} {et~al.}(2000){Willott}, {Rawlings}, {Blundell}, \&

1488:   {Lacy}}]{2000MNRAS.316..449W}

1489: {Willott}, C.~J., {Rawlings}, S., {Blundell}, K.~M., \& {Lacy}, M. 2000,

1490:   \mnras, 316, 449

1491: 

1492: \bibitem[{{Wolf} {et~al.}(2003){Wolf}, {Wisotzki}, {Borch}, {Dye},

1493:   {Kleinheinrich}, \& {Meisenheimer}}]{2003A&A...408..499W}

1494: {Wolf}, C., {Wisotzki}, L., {Borch}, A., {Dye}, S., {Kleinheinrich}, M., \&

1495:   {Meisenheimer}, K. 2003, \aap, 408, 499

1496: 

1497: \bibitem[{{Yu} \& {Lu}(2008)}]{2008arXiv0808.3777Y}

1498: {Yu}, Q. \& {Lu}, Y. 2008, ArXiv e-prints, 808

1499: 

1500: \bibitem[{{Yu} \& {Tremaine}(2002)}]{2002MNRAS.335..965Y}

1501: {Yu}, Q. \& {Tremaine}, S. 2002, \mnras, 335, 965

1502: 

1503: \bibitem[{{Zakamska} {et~al.}(2004){Zakamska}, {Strauss}, {Heckman},

1504:   {Ivezi{\'c}}, \& {Krolik}}]{2004AJ....128.1002Z}

1505: {Zakamska}, N.~L., {Strauss}, M.~A., {Heckman}, T.~M., {Ivezi{\'c}}, {\v Z}.,

1506:   \& {Krolik}, J.~H. 2004, \aj, 128, 1002

1507: 

1508: \bibitem[{{Zakamska} {et~al.}(2003){Zakamska}, {Strauss}, {Krolik}, {Collinge},

1509:   {Hall}, {Hao}, {Heckman}, {Ivezi{\'c}}, {Richards}, {Schlegel}, {Schneider},

1510:   {Strateva}, {Vanden Berk}, {Anderson}, \& {Brinkmann}}]{2003AJ....126.2125Z}

1511: {Zakamska}, N.~L., {et~al.} 2003, \aj, 126,

1512:   2125

1513: 

1514: \end{thebibliography}

1515: 

1516: 

1517: 

1518: \clearpage 

1519: 

1520:  \begin{figure} 

1521: \plottwo{f1a.eps}{f1b.eps}

1522: \caption{\noindent Cumulative energy densities for the optical LFs

1523:   (left) and hard X-ray LF (right). We remove any dependence on

1524:   \cbol\, by dividing them by the maximum value (at $z=0$). The

1525:   maximum values are taken from the \citet{2004MNRAS.349.1397C} LF

1526:   (left) and from the \citet{2003ApJ...598..886U} LF (right). The

1527:   \citet{2003A&A...408..499W} LF, defined only for $z\geq1$, agrees

1528:   well with the other two LFs assuming pure density evolution (PDE). 

1529:   However, pure luminosity evolution (PLE) predicts a somewhat larger

1530:   value for $U_{\rm B}$. The optical LFs include only unobscured AGNs, while

1531:   the X-ray LF also includes Compton-thin absorbed AGNs. }

1532: \label{fig:ut} 

1533: \end{figure} 

1534: 

1535: \begin{figure} 

1536: \plottwo{f2a.eps}{f2b.eps}

1537: \caption{\noindent Posterior PDF for \eff\, derived from the optical

1538:   LFs for unobscured AGNs only (a) and using the hard X-ray LF, with

1539:   no correction for Compton-thick absorbed AGNs (b). The posteriors

1540:   derived from the bolometric corrections of

1541:   \citet{2004MNRAS.351..169M} are marked as 'M04', whereas those from

1542:   \citet{2007ApJ...654..731H} are marked as 'H07'. The combined

1543:   posteriors, derived in Section~\ref{sec:best} are shown as the solid

1544:   curve. For panel (a), the optical LF is integrated over all

1545:   magnitudes brighter than $M_{\rm B}=-18$ ($L_{\nu \rm

1546:     B}$$\geq$8$\times$$10^{20}$ \whz), whereas for panel (b) the LF

1547:   has been integrated over all $L_{\rm X}\geq3\times10^{34}$~W.}

1548: \label{fig:posteriors} 

1549: \end{figure} 

1550: 

1551: \addtocounter{figure}{-1}

1552: \begin{figure} 

1553: \plottwo{f2c.eps}{f2d.eps}

1554: \caption{ Continued. Posterior PDFs for \eff\, derived from the optical LFs

1555:   correcting for the obscured AGNs using \rtt. In panel (c), the optical LF is

1556:   integrated over all magnitudes brighter than $M_{\rm B}=-18$, and two

1557:   different values of \rtt\, are used (\rtt$\sim$2 and $=$4). We consider this

1558:   our best estimate for \eff. In panel (d), the LF has been

1559:   integrated over all magnitudes brighter than $M_{\rm B}=-23$, and only

1560:   \rtt$\sim$2 is used, to estimate the value of \eff\, from the bright AGN

1561:   population (the quasars). }

1562: \end{figure} 

1563: 

1564: 

1565: 

1566: \begin{figure} 

1567: \plotone{f3.eps}

1568: \caption{\noindent Accretion efficiency $\eta$ as a function of rotation

1569:   parameter $a$. The solid curve represents the case of the SMBH co-rotating

1570:   with the inner part of the accretion disk, so the innermost stable circular

1571:   stable orbit (ISCO) is prograde. The dotted line represents the case of

1572:   counter-rotating SMBH and inner accretion disk (the ISCO is retrograde). The dashed curve represents the mean of both curves

1573:   \citep [the ``symmetrised'' curve, see ][]{2008MNRAS.385.1621K}. 

1574: }

1575: \label{fig:spin} 

1576: \end{figure} 

1577: 

1578: 

1579: 

1580: 

1581:  

1582: \label{lastpage} 

1583:  

1584: \end{document} 

1585: