1: % mn2esample.tex

2: %    

3: % v2.1 released 22nd May 2002 (G. Hutton)

4: %

5: % The mnsample.tex file has been amended to highlight

6: % the proper use of LaTeX2e code with the class file

7: % and using natbib cross-referencing. These changes

8: % do not reflect the original paper by A. V. Raveendran.

9: %

10: % Previous versions of this sample document were

11: % compatible with the LaTeX 2.09 style file mn.sty

12: % v1.2 released 5th September 1994 (M. Reed)

13: % v1.1 released 18th July 1994

14: % v1.0 released 28th January 1994

15: 

16: \documentclass[useAMS,usenatbib]{mn2e}

17: %\usepackage{psfig}  

18: 

19: % If your system does not have the AMS fonts version 2.0 installed, then

20: % remove the useAMS option.

21: %

22: % useAMS allows you to obtain upright Greek characters.

23: % e.g. \umu, \upi etc.  See the section on "Upright Greek characters" in

24: % this guide for further information.

25: %

26: % If you are using AMS 2.0 fonts, bold math letters/symbols are available

27: % at a larger range of sizes for NFSS release 1 and 2 (using \boldmath or

28: % preferably \bmath).

29: %

30: % The usenatbib command allows the use of Patrick Daly's natbib.sty for

31: % cross-referencing.

32: %

33: % If you wish to typeset the paper in Times font (if you do not have the

34: % PostScript Type 1 Computer Modern fonts you will need to do this to get

35: % smoother fonts in a PDF file) then uncomment the next line

36: % \usepackage{Times}

37: 

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

39: 

40: %\usepackage{amstext}

41: %\usepackage{amsmath, float}

42: \usepackage{graphicx}

43: %\usepackage{txfonts}

44: 

45: \newcommand       \be           {\begin{equation}}

46: \newcommand       \ee           {\end{equation}}

47: \newcommand       \bea          {\begin{eqnarray}}

48: \newcommand       \eea          {\end{eqnarray}}

49: \newcommand       \apj          {ApJ}

50: \newcommand       \apjl         {ApJL}

51: \newcommand       \aap          {A\&A}

52: \newcommand       \nat          {Nature}

53: \newcommand       \mnras        {MNRAS}

54: \def\simlt{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}}

55: \def\simgt{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}}

56: \def\lesssim{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}}

57: 

58: \def\gtrsim{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}}

59: 

60: 

61: 

62: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

63: \title[Time-dependent Models of Accretion Disks Formed from Compact Object Mergers]{Time-dependent models of accretion disks formed from compact object mergers} \author[B.~D. Metzger, A.~L. Piro, E. Quataert]{B.~D. Metzger\thanks{E-mail:

64: bmetzger@astro.berkeley.edu}, A.~L. Piro, and E. Quataert \\

65: Astronomy Department and Theoretical Astrophysics Center,

66: University of California, Berkeley, 601 Campbell Hall, Berkeley CA,

67: 94720\\}

68: \begin{document}

69: \date{Accepted . Received ; in original form }

70: \pagerange{\pageref{firstpage}--\pageref{lastpage}} \pubyear{????}

71: \maketitle

72: \label{firstpage}

73: 

74: \begin{abstract}

75: We present time-dependent models of the remnant accretion disks

76: created during compact object mergers, focusing on the energy

77: available from accretion at late times and the composition of the disk

78: and its outflows.  We calculate the dynamics near the outer edge of

79: the disk, which contains the majority of the disk's mass and

80: determines the accretion rate onto the central black hole. This

81: treatment allows us to follow the evolution over much longer

82: timescales ($100\ {\rm s}$ or longer) than current hydrodynamic

83: simulations.  At late times the disk becomes advective and its

84: properties asymptote to self-similar solutions with an accretion rate $\dot{M}_{d} \propto

85: t^{-4/3}$ (neglecting outflows).  This late-time accretion can in

86: principle provide sufficient energy to power the late-time activity

87: observed by {\it Swift} from some short-duration gamma-ray bursts.  However, because outflows during the advective phase unbind the majority of the

88: remaining mass, it is difficult for the remnant disk alone to produce

89: significant accretion power well beyond the onset of the advective

90: phase.  Unless the viscosity is quite low ($\alpha\lesssim10^{-3}$),

91: this occurs before the start of observed flaring at $\sim 30$ s;

92: continued mass inflow at late times thus appears required to explain

93: the late-time activity from short-duration gamma-ray bursts.  We show that the composition of the disk freezes-out when the disk is

94: relatively neutron rich (electron fraction $Y_e\simeq 0.3$). Roughly

95: $10^{-2} M_\odot$ of this neutron-rich material is ejected by winds at

96: late times.  During earlier, neutrino-cooled phases of accretion, neutrino

97: irradiation of the disk produces a wind with $Y_e\simeq0.5$, which synthesizes at

98: most $\sim10^{-3}M_\odot$ of $^{56}$Ni.  We highlight what conditions

99: are favorable for $^{56}$Ni production and predict, in the best cases,

100: optical and infrared transients peaking $\sim 0.5-2$ days after the

101: burst, with fluxes a factor of $\sim 10$ below the current

102: observational limits.

103: \end{abstract}

104: 

105: \begin{keywords}

106: {accretion disks ---

107: 	black hole physics ---

108: 	gamma rays: bursts ---

109: 	neutrinos}

110: \end{keywords}

111: 

112: %\vspace{-0.7cm}

113: \section{Introduction}

114: \label{sec:int}

115: \voffset=-2cm

116: 

117: \vspace{0.2 cm}

118: \label{sec:introduction}

119: The most popular model for the creation of short duration gamma-ray

120: bursts (GRBs) is either binary neutron star (NS/NS) or black

121: hole-neutron star (BH/NS) coalescence \citep{pac86,pac91,eic89,nar91}.

122: Support for the merger hypothesis comes from their durations of $\lesssim2\ {\rm s}$, observations of well-localized short GRBs in galaxies

123: without strong star formation

124: \citep{ber05,ger05,2005ApJ...630L.117H}, and the lack of a detectable

125: coincident supernovae \citep{2005ApJ...630L.117H,blo06,sod06,fer07}, as

126: is found in the case of long ($\gtrsim2\ {\rm s}$) GRBs

127: \citep{gal98,hjo03,sta03}.

128: 

129: Previous theoretical studies of the merger process have focused on one

130: of two stages. The first is the {\it dynamical} portion in which the less massive companion is tidally disrupted by the more

131: massive BH \citep{lk95,lk98,lk99,kl98,jan99,ros04} or NS

132: \citep{ruf96,rj99,oj06}. The details of whether a dynamical

133: instability \citep{rs94,lai94} or Roche lobe overflow occurs depends

134: on the mass ratio and the nuclear equation of state

135: \citep{bc92,ue99}.  

136: %Detailed studies of the neutrino emission during the merger stage \citep{ruf96,ruf97,rj99,rj01} indicate that the resulting $\nu-\bar{\nu}$ annihilation along the rotation axis is insufficient to power the GRB.

137: 

138:    Nevertheless, generally $\sim~0.01-0.1M_\odot$ of material remains in a remnant disk following the dynamical stage.  The accretion of this material onto the central object gives rise to the second, {\it

139: disk} portion of the merger.  The energetics and timescale of the accretion phase are reasonably consistent with observations of short GRBs, as

140: was shown by models of steady-state, azimuthally symmetric, vertically

141: averaged disks \citep{pop99,nar01,km02,dim02,cb07}.  More recently,

142: these disks have been modeled with time-dependent calculations in 1D

143: \citep{jan04}, 2D \citep{lee04,lee05b}, and 3D \citep{set04,set06}.  The typical time interval that present multi-dimensional calculations can simulate is on the order of the burst duration or less ($\sim1-2\ {\rm s}$ for 2D and $\sim50\ {\rm ms}$ for 3D).  

144: 

145: Recent observations of short GRBs by {\it Swift}, however, indicate continued activity from the central engine on much longer timescales.  X-ray flares with durations of $\sim100\ {\rm s}$ after a delay of $\sim30\ {\rm s}$

146: have been seen from several bursts

147: \citep{bar05,vil05,cam06,lap06}. Stacked lightcurves of many bursts

148: indicate continued activity on a similar timescale

149: \citep{laz01,mon05}. In one extreme case, GRB 050724 displayed an

150: X-ray flare $12$ hours post-burst. This flaring activity has been

151: attributed to a number of different sources, including fragmentation

152: of a rapidly rotating core \citep{kin05}, magnetic regulation of the

153: accretion flow \citep{pz06}, fragmentation of the accretion disk

154: (Perna et al. 2005; although this explanation may have difficulty

155: reproducing the observed timescales, Piro \& Pfahl 2007), differential

156: rotation in a post-merger millisecond pulsar \citep{dai06}, and an

157: infalling tidal tail of material stripped from the disrupted NS (Lee $\&$ Ramirez-Ruiz 2007; Rosswog 2007).

158: 

159: In order to determine whether the late-time activity from short GRBs is consistent with a compact merger origin, the disk evolution should be followed for timescales much longer than the initial viscous time. With this aim, we perform time-dependent calculations modeling

160: the disk as an annulus that contains the majority of the mass.  This

161: simplification allows us to study the disk evolution for arbitrarily

162: long timescales, and to readily determine important properties such as

163: the disk's composition and when it becomes advective. We are

164: also able to survey much of the parameter space of initial disk mass

165: and angular momentum. In \S \ref{sec:initial} we discuss the initial

166: conditions for disks formed from compact object mergers. This is

167: followed by \S \ref{sec:physics}, in which we summarize the main

168: assumptions of our ring model.  In \S \ref{sec:results} we present the

169: results of our calculations and summarize the main properties of the

170: models.  We then calculate outflows from our disk solutions in \S

171: \ref{sec:winds}. We investigate the composition of the outflows and

172: argue that they generally consist of neutron-rich isotopes, but can

173: produce $^{56}$Ni in some circumstances.  The presence or lack of an

174: optical transient from short GRBs therefore provides an important

175: constraint on progenitor models.  We conclude in \S \ref{sec:theend}

176: with a discussion of our results. In Appendix A we summarize the

177: Green's function solution to the viscous spreading of a ring, which is

178: important for connecting our ring model to the true extended disk

179: geometry. In Appendix B we present analytic self-similar solutions

180: that reproduce many of the features of our numerical solutions.

181: 

182: \section{Initial Conditions}

183: \label{sec:initial}

184: 

185:    The dynamical phase of NS/NS or BH/NS mergers has

186: been studied extensively using a number of different numerical

187: techniques and methods for including general

188: relativity (GR).  Here we summarize some of the most relevant features for our study

189: \citep[for a more detailed review, see][]{lr07}.  

190: 

191:    When the lighter companion NS is first tidally disrupted, a debris disk is

192: formed within only a few dynamical timescales.  The initial disk mass, $M_{d,0}$, is generally larger for more

193: asymmetric mass ratios (i.e., small $q$, where $q$ is the ratio of the

194: lighter to the heavier binary component). For example, \citet{st06}

195: find that for a NS/NS merger with $q=0.7$ that $M_{d,0}=0.03\

196: M_\odot$, but for $q=0.9$ the disk is much less massive with

197: $M_{d,0}=10^{-3}\ M_\odot$.  Another trend is that including strong

198: gravity gives less massive remnant disks.  The BH spin is also important, with larger spin favoring disk formation \citep{ras05}

199: and the production of a tidal tail. These have masses of

200: $\simeq0.01-0.05\ M_\odot$ and may provide prolonged mass inflow

201: \citep{lr07}, but for simplicity this will be ignored here.  Taken 

202: together, these simulations generally find $M_{d,0}\simeq0.01-0.3M_\odot$, with the disk containing a substantial fraction of the angular momentum of the disrupted companion.

203: 

204: In the standard picture of NS-NS mergers, the resulting hypermassive NS collapses to a BH shortly following the merger.  However, simulations show that when (and if) collapse actually occurs depends on the mass of the central NS and its ability to transport angular momentum to the surrounding disk (Shibata et al.~2005; Shibata $\&$ Taniguchi 2006; Shibata et al.~2006).  In fact, if the NS remains supported by differential rotaton for several seconds (Baumgarte et al.~2000; Morrison et al.~2004; Duez et al.~2004, 2006) or loses sufficient mass via a centrifugally-driven outflow (e.g., Thompson et al.~2004; Dessart et al.~2008a), the NS itself may power the GRB (e.g., Price $\&$ Rosswog 2006).  In this paper we assume that the central object promptly collapses to a BH; our model, however, would be reasonably applicable for the case of a central NS as well, the primary difference being that the significant neutrino flux from the newly-formed NS and from the boundary layer between the disk and the NS could modify the composition and thermal properties of the disk.

205: 

206: We present some characteristic numbers to motivate our choice of

207: initial conditions.  Consider a binary with masses $M$ and $m$

208: ($M>m$), where the latter is the NS (with radius $R$) that is tidally

209: disrupted. The disruption radius, $a_t$, is estimated to be (Kopal

210: 1959, adding Fishbone's 1973 10\% strong gravity correction) \be

211: a_t\simeq2.4R \left(\frac{M+m}{m}\right)^{1/3}.  \ee The characteristic

212: orbital period at this radius is \be P_t \simeq

213: 23.4\left(\frac{R^3}{Gm}\right)^{1/2} \simeq

214: 2\times10^{-3}m_{1.4}^{-1/2}R_6^{3/2}\ {\rm s}, \ee where

215: $m_{1.4}=m/1.4\ M_\odot$ and $R_6=R/10^6\ {\rm cm}$, with an orbital

216: angular momentum of \bea J_t &=&  \left( G(M+m)a_t\right)^{1/2}m \nonumber \\

217: &\simeq& 6\times10^{49}(1/q+1)^{2/3}m_{1.4}^{3/2}R_6^{1/2}\ {\rm ergs\

218: s}, \eea where $q=m/M$. The disrupted NS also contains spin angular

219: momentum. This is negligible since the NS is not strongly affected by

220: tidal coupling \citep{bc92}.  Even a rapidly rotating NS ($\simeq5\

221: {\rm ms}$) has an associated angular momentum of merely $\sim10^{48}\

222: {\rm ergs\ s}$.

223: 

224:    Once disrupted, a considerable fraction of the NS is either lost

225: from the system or immediately swallowed by the BH. The remaining material forms a thick torus surrounding the central BH. Its associated viscous

226: timescale can be estimated by assuming that the majority of the torus'

227: mass lies at a single radius, $r_{d,0}$. Taking the angular momentum

228: of the disk to be $J_d\simeq(GMr_{d,0})^{1/2}M_{d,0}$, we estimate \be

229: r_{d,0}\simeq3\times10^7M_3^{-1}M_{0.1}^{-2}\left(\frac{J_{49}}{2}\right)^2\

230: {\rm cm},

231: 	\label{eq:rd0}

232: \ee

233: where $M_3= M/3\ M_\odot$, $M_{0.1}=M_{d,0}/0.1\ M_\odot$, and

234: $J_{49}=J_d/10^{49}\ {\rm ergs\ s}$.

235: For a disk with half-thickness $H$, the viscous timescale is

236: \bea

237: 	t_{\rm visc,0}& = &\alpha^{-1}\left(\frac{r_d}{H}\right)^2 \left(\frac{r_d^3}{GM}\right)^{1/2}

238: 	\nonumber

239: 	\\ &\simeq &6\times10^{-2}\alpha_{0.1}^{-1}

240: 	M_3^{-1/2}r_7^{3/2}\left(\frac{H}{0.5r_d}\right)^{-2}\ {\rm s},

241: 	\label{eq:tvisc0}

242: \eea where $\alpha = 0.1\alpha_{0.1}$ is the standard

243: dimensionless viscosity \citep{ss73}, $r_7=r_{d,0}/10^7\ {\rm cm}$,

244: and we have scaled to an initial ratio of $H/r_d=0.5$, consistent with

245: our numerical solutions.  The initial viscous time $t_{\rm visc,0}$ is

246: roughly the time at which the central BH begins accreting in earnest.

247: The strong dependence of $t_{\rm visc,0}$ on disk mass and radius demonstrates that the initial evolution of the disk is sensitive to the

248: outcome of the dynamical phase of the merger. But as we will show, the

249: late time evolution is much less sensitive to initial conditions and

250: is well described by self-similar solutions.

251: 

252: \section{Physics of the Expanding Ring Model}

253: \label{sec:physics}

254: 

255:    Given these initial conditions, one would like to know how the disk

256: then evolves. Modeling the entire disk requires resolving timescales

257: over $\sim4-6$ orders of magnitude.  This makes it expensive to carry

258: out simulations for long periods of time. We consider instead a

259: simplified model that captures most of the features of interest. At

260: any given time, $t$, the disk can be broken into three regions

261: depending on the local viscous time, $t_{\rm visc}$, which increases

262: with radius, roughly as $t_{\rm visc}\sim r^{3/2}$.  At small radii,

263: $t_{\rm visc}<t$, and the disk comes into steady-state.  This is the

264: region most often modeled in previous studies

265: \citep{pop99,nar01,km02,dim02,cb07}. The radii where $t_{\rm visc}\sim

266: t$ contain the majority of the disk's mass and angular

267: momentum. Therefore, this region determines the viscous evolution of

268: the rest of the disk, including the mass accretion rate that is fed to

269: the interior steady-state region. Motivated by this fact, we focus on

270: this radius and model the disk as a ring. Exterior to this point is a

271: third region where $t_{\rm visc}>t$, but this contains a small amount

272: of mass and is negligible for the viscous evolution.

273: 

274: \subsection{Dynamical Equations}

275: \label{sec:dynamical}

276: 

277: Our ring model treats the disk as a single annulus that is evolved

278: forward in time. In this picture, the properties of the ring, such as

279: its surface density $\Sigma$ and temperature $T$, are representative

280: of the location where $\Sigma r^2$ peaks.  The main drawback of this

281: method is that the material in the disk is in fact distributed

282: spatially in radius.  Thus, although the mass of the disk in the

283: vicinity of $r_d$ is $\simeq \pi \Sigma r_d^2$, the total mass of the

284: disk (integrated over all radii) is $M_d=A \pi \Sigma r_d^2$, where A

285: is a factor of order unity that accounts for the distinction between

286: the total mass of the disk and the mass of the material near $r_d$.

287: Similarly, we write the total angular momentum of the disk as $J_d =

288: B(GMr_d)^{1/2}\pi r_d^2\Sigma$.  At early times the constants $A$ and

289: $B$ depend on the initial conditions of how matter is spatially

290: distributed; however, at times much greater than the initial viscous

291: time (given by eq. [\ref{eq:tvisc0}]), material initially concentrated

292: at a given radius becomes spread out in a manner determined by the

293: viscosity. As described in detail in Appendix A, we choose the

294: constants $A$ and $B$ by setting the solution of our simplified ring

295: model at late times equal to the Green's function solution for a

296: spreading ring with a viscosity $\nu \propto r^{1/2}$ (as is

297: appropriate for the radiatively inefficient disk at late-times).  This

298: fixes $A = 3.62$ and $B = 3.24$.\footnote{In fact, when the total

299: angular momentum is conserved, the viscous evolution is independent of

300: $A/B$ as long as $A/B$ is nearly constant with time.}  Conveniently

301: $A/B\simeq1$, so that it is a good approximation to take

302: $J_d\simeq(GMr_d)^{1/2}M_d$.

303: 

304:    The time evolution of the disk is determined by the conservation equations.

305: Conservation of mass is

306: \be

307: 	\frac{d}{dt}\left(A\pi\Sigma r_d^2\right)=-\dot{M}_d,

308: 	\label{eq:mass}

309: \ee where $\dot{M}_d$ is in general the total mass loss rate, which could include both accretion and a wind (for now we ignore the effects of a

310: wind). Conservation of angular momentum is \be \frac{d}{dt}\left[

311: B(GMr_d)^{1/2}\pi\Sigma r_d^2\right]=-\dot{J},

312: 	\label{eq:ang_mom}

313: \ee

314: where $\dot{J}$ is the angular momentum loss rate.

315: Equations (\ref{eq:mass}) and (\ref{eq:ang_mom}) provide two coupled

316: equations that can be solved for the dependent variables $r_d$ and $\Sigma$.

317: 

318:    The accretion rate must depend on the characteristic mass and

319: viscous timescale of the ring, so we use

320: \be

321: 	\dot{M}_{d} = fM_d/t_{\rm visc},

322: 	\label{eq:mdotacc}

323: \ee where $t_{\rm visc}=r_d^2/\nu$ and $\nu$ is the viscosity.  The

324: factor $f$ is set like $A$ and $B$ to match the exact solution of a

325: spreading ring with $\nu \propto r^{1/2}$ (Appendix A), which gives

326: $f=1.6$.\footnote{Although we set $t_{\rm

327: visc}=r^2/\nu$, any prefactors that could go into this prescription

328: would just be absorbed into a re-definition of $f$.}  Requiring a

329: no-torque boundary condition at a radius $r_*$, we take \be f =

330: 1.6/[1-(r_*/r_d)^{1/2}].  \ee In contrast, a steady-state disk obeys

331: $\dot{M}_{d}=3\pi\nu\Sigma$ (ignoring the no-torque condition), which

332: instead gives $f=3/A\simeq0.83$.  

333: 

334: For the viscosity, we use an $\alpha$-prescription, \be \nu=\alpha

335: c_sH, \ee where $c_s=(P/\rho)^{1/2}$ is the isothermal sound speed.

336: The equation of state includes contributions from radiation pressure,

337: gas pressure, relativistic degeneracy pressure, and neutrino pressure

338: as in \citet{dim02}.

339: 

340: \subsection{Energetics}

341: 

342: For the energy equation, we take \be q_{\rm visc} = q_\nu^- +q_{\rm

343: adv}, \label{eq:energy} \ee where $q_{\rm visc}$ is the viscous

344: heating, $q_\nu^-$ is the neutrino cooling \citep[using the

345: prescriptions given by][which includes neutrino optical-depth

346: effects]{dim02}, $q_{\rm adv}$ is the advective heat flux, and all $q$

347: values correspond to half the disk thickness.

348: 

349: For a disk rotating at the Keplerian frequency $\Omega =

350: (GM/r_d^3)^{1/2}$, \be q_{\rm visc}=\frac{9}{8}\nu\Omega^2\Sigma =

351: \frac{9}{8fA}\frac{GM\dot{M}_{d}}{\pi

352: r_d^3}\left[1-\left(\frac{r_*}{r_d}\right)^{1/2} \right],

353: 	\label{eq:qvisc}

354: \ee where the prefactor $9/(8fA)\simeq0.2$ is different from the

355: steady-state value of $3/8$.  The advective term, $q_{\rm adv}$, is

356: set as in \citet{dim02}, with the only difference being that the

357: radial velocity is the expansion rate of the ring's radius \be V_r =

358: \frac{dr_d}{dt} = \frac{2\dot{M}}{A\pi r_d\Sigma}, \ee where we have

359: taken $\dot{J}=0$. 

360: 

361: Fusion to $\alpha$-particles produces heating in addition to

362: $q_{\rm visc}$, with \be q_{\rm

363: nucl}=6.8\times10^{28}\rho_{10}\frac{dX_{\rm \alpha}}{dt}H, \ee where

364: all quantities are expressed in cgs units, $\rho_{10}=\rho/10^{10}\

365: {\rm g\ cm^{-3}}$ and $X_{\rm \alpha}$ is the mass fraction of

366: $\alpha$-particles. Note that in our case $q_{\rm nucl}>0$ since

367: $\alpha$-particles are synthesized as the disk expands (in contrast to

368: studies that follow cooling from photodisintegration as material moves

369: inward).  In our calculations we do not include $q_{\rm nucl}$ in solving

370: equation (\ref{eq:energy}) because we were not able to find reasonable

371: solutions when doing so (for reasons explained in \S \ref{sec:gross}).

372: 

373: \subsection{Composition}

374: 

375:    An advantage of the ring model is that other properties of the

376: disk, such as its composition, can be cast into differential equations

377: and integrated along with equations (\ref{eq:mass}) and

378: (\ref{eq:ang_mom}).  Since the neutron content of the disk is

379: particularly important for determining the properties of the disk's

380: outflows, we evolve the electron fraction $Y_e$ using \be

381: \frac{dY_e}{dt} = -Y_e r_{e^-p}+(1-Y_e) r_{e^+n},

382: 	\label{eq:ye}

383: \ee

384: where $Y_e=X_p/(X_n+X_p)$, $X_p$ and $X_n$ are the proton and

385: neutron mass fraction, respectively, and $r_{e^-p}$ and $r_{e^+n}$ are

386: the electron and positron capture rates, respectively (Beloborodov 2003a).  We have neglected the effect of neutrino absorptions on the evolution of $Y_{e}$ in equation (\ref{eq:ye}).  Although absorptions are important at early times when the disk is optically thick, we are primarily concerned with the late-time value of $Y_{e}$, which does not depend sensitively on the neutrino irradiation (see $\S\ref{sec:composition}$). 

387: 

388:    As the disk evolves, the protons and neutrons eventually burn to

389: form $\alpha$-particles. At these times the disk is sufficiently cold that

390: the positron and electron capture rates are negligible

391: (i.e, $1/r_{e^-p}\gg t_{\rm visc}$) and $Y_e$ has frozen-out. This fixes the

392: difference between the free neutron and proton mass fractions

393: \be

394: 	X_n-X_p=1-2Y_e.

395: 	\label{eq:frozenratio}

396: \ee

397: Since the rates for reactions that synthesize and destroy $\alpha$-particles

398: are all fast in comparison

399: to the viscous time, we determine the composition using

400: nuclear statistical equilibrium (NSE) between protons, neutron, and $\alpha$-particles.

401: This is expressed by the Saha relation \citep{st83}

402: \be

403: 	X_p^2X_n^2

404: 	= 1.57\times10^4\ X_\alpha \rho_{10}^{-3}

405: 	T_{10}^{9/2}

406: 	\exp\left(-\frac{32.81}{T_{10}} \right).

407: 	\label{eq:saha}

408: \ee

409: NSE is a good assumption because the disk temperature is generally $\gtrsim 0.5$ MeV (see Fig.~\ref{fig:temperature}), except at very late

410: times or for very low disk masses (e.g., the $M_{d,0} = 0.03M_{\sun}$ case,   

411: for which we do not calculate the nuclear composition anyways).  By combining equations (\ref{eq:frozenratio}) and (\ref{eq:saha}) with mass conservation,

412: $X_p+X_n+X_\alpha=1$, we solve for all of the mass fractions at a given

413: $\rho$, $T$, and $Y_e$.

414: 

415: \section{Time-Evolving Solutions}

416: \label{sec:results}

417: 

418:    We next present the results of integrating equations

419: (\ref{eq:mass}), (\ref{eq:ang_mom}), and (\ref{eq:ye}) forward in

420: time.  For simplicity, we typically assume that $\dot{J} = 0$.

421: A convenient property of our formalism is the ease with which these

422: complications can be included (for example, we consider the effects of winds at the end

423: of \S 4.1). The disk

424: properties are determined by the initial conditions $M_{d,0}$, $J_d$,

425: and $Y_{e,0}$, and by the viscosity $\alpha$.  For the majority of our

426: study we set the initial $Y_{e,0}=0.1$, which is characteristic of the

427: inner neutron star crust \citep{hz90a,hz90b,pr95}. An additional

428: important parameter is $r_*$, which is set by the spin of the central

429: BH.  In most of our calculations we take $r_*\simeq2.3r_g\simeq1.02\times10^6\ {\rm cm}$, corresponding to the innermost stable circular orbit of a $3\ M_\odot$ BH with spin $a\simeq0.9$; when calculating the properties of disk outflows in $\S\ref{sec:winds}$, however, we also consider the case of a nonrotating ($a=0$) BH.  We consider the general evolution of the disk in \S

430: \ref{sec:gross}, and then focus on the composition in \S

431: \ref{sec:composition}.

432: 

433: \begin{figure}

434: %%\epsscale{1.2}

435: \resizebox{\hsize}{!}{\includegraphics[ ]{mass1.eps}}

436: \caption{Example disk models showing the evolution of the disk radius,

437: $r_d$, disk mass, $M_d$, and accretion rate, $\dot{M}_{d}$, as a

438: function of time.  We compare $M_{d,0}=0.03$ ({\it solid lines}),

439: $0.1$ ({\it dotted lines}) and $0.3\ M_\odot$ ({\it dashed lines})

440: solutions; all use $J_{49}=2$ and $\alpha=0.1$. The inner radius is

441: $r_*\simeq2.3r_g\simeq1.02\times10^6\ {\rm cm}$ (corresponding to a

442: $3\ M_\odot$ BH with a spin of $a\simeq0.9$).

443: }

444: \label{fig:mass1}

445: %\epsscale{1.0}

446: \end{figure}

447: 

448: \begin{figure}

449: %%\epsscale{1.2}

450: \resizebox{\hsize}{!}{\includegraphics[ ]{temperature.eps}}

451: \caption{Comparison of the midplane temperatures and scaleheight for the three

452: models from Fig. 1. In the lowest mass model, the ring

453: is always advectively-dominated, thus $H/r_d$ is constant.}

454: \label{fig:temperature}

455: %\epsscale{1.0}

456: \end{figure}

457: 

458: \subsection{Disk Evolution and Energetics}

459: \label{sec:gross}

460: 

461:     At any given time, a ring model is in one of three phases: (1)

462: early-time, optically thick to neutrinos and advectively dominated,

463: (2) mid-time, optically thin to neutrinos and geometrically thin, and

464: (3) late-time, radiatively-inefficient accretion flow

465: (RIAF).\footnote{An optically thick, geometrically thin stage occurs

466: between stages (1) and (2); however, this phase is brief and is not

467: dynamically very different from phase (2), so we do not consider it

468: separately in our discussion.}  This is analogous to the different

469: regions of steady-state, hyper-accreting accretion disks \citep[see,

470: e.g.,][]{cb07}, but now the transitions occur with time instead of

471: radius. The phases that a certain ring model samples during the course

472: of its viscous expansion depends on $t_{\rm visc,0}$. A more compact

473: disk (a shorter $t_{\rm visc,0}$) will exhibit all three phases, while

474: larger disks may only exhibit phases (2) and (3), or even just (3).

475: 

476:     We present a number of figures that are helpful in understanding

477: these three phases and how they are affected by changing $M_{d,0}$.

478: Figure \ref{fig:mass1} shows the radius $r_d$, mass $M_d$, and

479: accretion rate $\dot{M}_{d}$ as a function of time, for $M_{d,0} =

480: 0.3, 0.1,$ and $0.03 \, M_\odot$.  Figure \ref{fig:temperature} compares the

481: midplane temperature and scaleheight for these same models.

482: Figures \ref{fig:energy1} and

483: \ref{fig:energy2} show key results describing the energetics of the

484: $M_{d,0}=0.3$ and $0.1\ M_\odot$ solutions, respectively, while Figure

485: \ref{fig:pressure} shows the different contributions to the total

486: pressure in the disk as a function of time. Note that we fix the total

487: angular momentum in these calculations ($J_{49}=2$) and thus a larger

488: $M_{d,0}$ corresponds to a smaller $r_{d,0}$ and a shorter $t_{\rm visc,0}$.

489: 

490: \begin{figure}

491: %\epsscale{1.2}

492: \resizebox{\hsize}{!}{\includegraphics[ ]{energy1.eps}}

493: \caption{The cooling rates and neutrino luminosity

494: for the $M_{d,0}=0.3M_\odot$ model from Fig. \ref{fig:mass1}. For the

495: cooling rates we compare the neutrino ({\it solid line}) and

496: advective ({\it dashed line}) rates, normalized to the viscous

497: heating. The implied heating from the creation of $\alpha$-particles

498: is plotted as a dotted line, but is not accounted for in the disk

499: evolution.  The neutrino luminosities are from the entire disk ({\it

500: solid line}) and the ring ({\it dashed line}). The former luminosity

501: is estimated by integrating over a steady-state disk model at each

502: time given $\dot{M}_{d}(t)$.}

503: \label{fig:energy1}

504: %\epsscale{1.0}

505: \end{figure}

506: 

507:    The first transition the disks make is from an optically thick,

508: advective disk to a thin, neutrino-cooled disk; i.e., from phase (1)

509: to (2). This is only exhibited by the $M_{d,0}=0.3M_\odot$ model and is seen most clearly at early times in Figure \ref{fig:temperature}

510: when $H/r_d\simeq0.5$ and in Figure \ref{fig:energy1} when $q_{\rm adv}\gg q_\nu^-$.  Figure

511: \ref{fig:pressure} shows that this phase is ion pressure (ideal gas)

512: dominated.  A simple estimate determines what initial disk mass is

513: required for phase (1) to occur, i.e., for the initial disk to be

514: optically thick and advective. The disk is advective for radii inside

515: of which the neutrino diffusion time out of the disk exceeds the

516: inflow time.  Setting this radius equal to the initial radius of the

517: disk (eq. [\ref{eq:rd0}]), we find that there is a critical disk mass

518: below which the disk never experiences phase (1), 

519: \be

520: M_{d,\rm crit}\sim0.2\alpha_{0.1}^{-1/10}

521: M_3^{-7/10}\left(\frac{J_{49}}{2}\right)^{9/10}\left(\frac{H}{0.5r_d}\right)^{-3/5}M_\odot,

522: 	\label{eq:mcrit}

523: \ee where we have dropped scalings with $f$ and $A$ since they appear

524: raised to the $1/10$ power. This estimate is consistent with the fact that

525: our $M_{d,1}=0.1M_\odot$ model is not advective at early times, as

526: seen in Figures \ref{fig:temperature} and \ref{fig:energy2}.  In this case only phases (2) and (3) are seen,

527: i.e., the disk is initially thin and neutrino cooled and later

528: transitions to being advective.

529: 

530: \begin{figure}

531: %\epsscale{1.2}

532: \resizebox{\hsize}{!}{\includegraphics[ ]{energy2.eps}}

533: \caption{The same as Fig. \ref{fig:energy1}, but for

534: $M_{d,0}=0.1M_\odot$.}

535: \label{fig:energy2}

536: %\epsscale{1.0}

537: \end{figure}

538: 

539:    Once the models reach the late-time, RIAF phase, or phase (3), they

540: asymptote to self-similar solutions, independent of the initial disk

541: mass.  In this phase, the disk has $q_{\rm adv}>q_\nu^-$ and is

542: radiation pressure dominated.  We derive analytic self-similar

543: solutions in Appendix B2 for this limit and show that $r_d \propto

544: t^{2/3}$, $M_d \propto t^{-1/3}$ and $\dot M_{d} \propto t^{-4/3}$.  The

545: RIAF solution occurs external to an ``ignition radius,'' which we estimate as the location where the pair capture cooling rate balances $\sim 1/2$ of the viscous heating for a thick disk,

546: \bea

547: r_{\rm ign}\simeq \nonumber \eea \bea 3\times10^7\alpha_{0.1}^{-2}M_3^{-3/5}\left(\frac{H/r_{d}}{0.4}\right)^{-14/5} \left(\frac{\dot{M}_{d}}{0.1M_\odot{\rm s^{-1}}}\right)^{6/5} {\rm cm},

548: 	\label{eq:rign}

549: \eea

550: where we have scaled $H/r_{d}$ to $\approx 0.4$, a value appropriate for the transition between the thin and thick disk regimes.  We combine this with the analytic results for $r_d(t)$ and

551: $\dot{M}_{d}(t)$ in the RIAF limit (eqs. [\ref{eq:rd_analytic2}] and

552: [\ref{eq:mdot_analytic2}])\footnote{We use these solutions rather than

553: the thin-disk ones because the numerical results follow these more

554: closely (Fig. \ref{fig:analytic}).}  to estimate the time when the

555: disk transitions to being thick, which yields \be t_{\rm

556: thick}\sim0.1\alpha_{0.1}^{-23/17}M_3^{-13/17}\left(\frac{J_{\rm

557: 49}}{2}\right)^{9/17} {\rm s}.

558: 	\label{eq:tthick}

559: \ee Equation (\ref{eq:tthick}) is only applicable if the disk is thin

560: at early times.  For sufficiently small initial disk masses, less than

561: \be M_{d,\rm

562: thick}\sim0.1\alpha_{0.1}^{2/17}M_3^{-7/17}\left(\frac{J_{\rm

563: 49}}{2}\right)^{14/17} M_\odot, \ee this is no longer true, and the disk

564: is always a RIAF at its outer radius.

565: 

566: \begin{figure}

567: %\epsscale{1.2}

568: \resizebox{\hsize}{!}{\includegraphics[ ]{pressure.eps}}

569: \caption{Pressure contributions for $M_{d,0}=0.3M_\odot$ ({\it top

570: panel}), $0.1M_\odot$ ({\it middle panel}) and $0.03M_\odot$ ({\it

571: bottom panel}). The pressures are all normalized to the total pressure

572: and include the ion pressure ({\it solid lines}), radiation pressure

573: ({\it dotted lines}), degenerate electron pressure ({\it dashed

574: line}), and neutrino pressure ({\it dot-dashed line}).}

575: \label{fig:pressure}

576: %\epsscale{1.0}

577: \end{figure}

578: 

579: Figures \ref{fig:energy1} and \ref{fig:energy2} show that at

580: approximately the same time as the disk transitions from being thin to

581: thick, protons and neutrons are fused to He.  Although the nuclear

582: heating rate $q_{\rm nuc}$ is shown in Figures \ref{fig:energy1} and

583: \ref{fig:energy2}, this heating was not included in our time-dependent

584: calculations so that we could obtain solutions at late times.  The

585: nuclear heating rate is sufficiently large, i.e, $q_{\rm nucl}\gtrsim

586: q_{\rm visc}$, that the disk is not able to accommodate this added

587: energy (it is already thick with $H \simeq r$ due to viscous heating

588: alone). This probably implies that the burning contributes to driving

589: a powerful wind \citep[as described by][]{lr07}.

590: 

591: However, such a wind

592: already begins at this time by virtue of the disk being advective (as

593: discussed in \S \ref{thickdiskwinds}).  In Appendix B3, we

594: present analytic self-similar solutions for advective disks with mass

595: loss and show that this significant mass loss causes $M_d$

596: and $\dot M_{d}$ to decline much more rapidly with time than is shown in

597: Figure \ref{fig:mass1}.  This is shown explicitly in Figure

598: \ref{fig:wind}, where we present disk models calculated using the mass

599: and angular momentum loss prescriptions described in Appendix B3;

600: such losses are assumed to occur only when the disk is thick, between

601: $\sim \max(r_*, r_{\rm ign})$ and $\sim r_d$.  Figure \ref{fig:wind}

602: compares time-dependent solutions with no wind ({\it solid line}), a

603: wind with $p=0.5$ ({\it dotted line}; see eq. [B8]), and a wind with

604: $p=1$ ({\it dashed line}).\footnote{See Appendix $\ref{sec:riaf}$ for the definition of p.}  The loss of angular momentum does not appreciably slow the radial expansion of the disk, but it does substantially accelerate the decline in the disk mass and accretion rate (see also eqs. [\ref 

605: {MdCne1}] and [\ref{mdotinCne1}]).  If the models with winds are  

606: accurate, significant accretion is only likely to last for a few  

607: viscous times once the disk enters the late-time advective phase.   

608: Continued central engine activity at much later times could result  

609: from late-time infall of tidally stripped NS material \citep[e.g.,][] 

610: {lr07}.

611: 

612: \begin{figure}

613: %\epsscale{1.2}

614: \resizebox{\hsize}{!}{\includegraphics[ ]{comparison.eps}}

615: \caption{The radius $r_{d}$, disk mass $M_d$, and mass accretion rate reaching the central BH $\dot{M}_{\rm in}$ for different parameterizations of mass loss during the advective phase.  We initialize a disk with $M_{d,0}=0.1M_\odot$ (and all other parameters fixed as in Fig. \ref{fig:mass1}) and compare solutions with no wind ({\it solid line}), $p=0.5$ ({\it dotted line}; see eq. [B8]), and $p=1$ ({\it dashed line}).}

616: \label{fig:wind}

617: %\epsscale{1.0}

618: \end{figure}

619: 

620: As an additional comparison, we present the effect of varying $J_d$ in

621: Figure \ref{fig:mass2}. The main trend is that a higher $J_d$ has a

622: larger initial radius for a given $M_d$, and therefore a longer

623: viscous time and smaller accretion rate.  The late time behavior is

624: more sensitive to $J_d$ than the initial $M_d$, as predicted by the

625: self-similar solutions, but it still does not affect the late time

626: disk radius (see eq. [\ref{eq:rd_analytic2}]). We do not plot our

627: results for different $\alpha$ since they are generally consistent

628: with the analytic scalings above and in Appendix \ref{sec:analytic}.

629: 

630: \begin{figure}

631: %\epsscale{1.2}

632: \resizebox{\hsize}{!}{\includegraphics[ ]{mass2.eps}}

633: \caption{Similar to Fig. \ref{fig:mass1}, but now taking the angular

634: momentum to be $J_{49}=2$ ({\it solid lines}), $4$ ({\it dotted

635: lines}), and $6$ ({\it dashed lines}). All solutions take

636: $M_{d,0}=0.3M_\odot$ with all other variables the same as in

637: Fig. \ref{fig:mass1}.}

638: \label{fig:mass2}

639: %\epsscale{1.0}

640: \end{figure}

641: 

642: \subsection{Composition}

643: \label{sec:composition}

644:  

645:    The composition of the disk is important for determining the

646: observational effects of any outflows. To this end, we plot the

647: composition of our $M_{d,0}=0.3\ M_\odot$, $J_{49}=2$ disk as a function

648: of time in the upper panel of Figure \ref{fig:composition}.  In the

649: bottom panel we plot the relevant timescales for setting the

650: composition, namely the viscous timescale, $t_{\rm visc}$ ({\it solid

651: line}), the neutronization timescale $t_n=1/r_{e^-p}$ ({\it dotted

652: line}), and the timescale for $\alpha$-particle photodisintegration,

653: $t_{\rm photo}$ ({\it dashed line}).  At early times $t_n\ll t_{\rm

654: visc}$, so that an equilibrium value of $Y_e\simeq0.23$ is reached almost immediately. As the disk leaves the optically thick phase and

655: becomes thinner, degeneracy pressure plays a larger role. This

656: enhances neutron production, with a minimum $Y_e\simeq0.05$. As the

657: neutrino cooling subsides and the disk becomes thick again, $Y_e$

658: increases. Before $Y_e$ can reach $\simeq0.5$, it freezes-out at a

659: value of $Y_e\simeq0.3$ once $t_n>t_{\rm visc}$.

660: 

661: %   Slightly smaller mass disks show both a similar evolution and

662: %final freeze out of $Y_e$. Although disks with initial mass

663: %$M_{d,0}<0.1M_\odot$ are never hot enough have

664: %$t_n\lesssim t_{\rm visc}$. In this case, $Y_e$ is not altered during

665: %the viscous evolution of the ring, and instead reflects the composition

666: %set by the tidally-disrupted progenitor and the subsequent dynamical

667: %stage of the merger.

668: 

669: Besides the neutron abundance, Figure \ref{fig:composition} also

670: highlights the production of $\alpha$-particles. Initially, the reactions needed

671: to convert neutrons and protons to helium as well as

672: photodisintegration of helium all happen on timescales much shorter

673: than the disk evolution timescale (as an example, we plot the helium

674: photodisintegration timescale in the bottom panel of

675: Fig. \ref{fig:composition}), so that we can estimate the

676: $\alpha$-particle mass fraction using chemical balance

677: (eq. [\ref{eq:saha}]). Once the $\alpha$-particle photodisintegration

678: timescale becomes sufficiently long ($t_{\rm visc}<t_{\rm photo}$),

679: chemical equilibrium no longer applies and $X_p=0$,

680: $X_n=1-2Y_e\simeq0.4$, and $X_\alpha=2Y_e\simeq0.6$.

681: \begin{figure}

682: %\epsscale{1.2}

683: \resizebox{\hsize}{!}{\includegraphics[ ]{composition.eps}}

684: \caption{The composition and important reaction

685: timescales as a function of time, for the $M_{d,0}=0.3M_\odot$ model from

686: Fig. \ref{fig:mass1}.  In the top panel we plot the electron fraction,

687: $Y_e$ and the mass fraction of protons, neutrons, and

688: $\alpha$-particles ({\it see inset key}). In the bottom panel we show

689: the viscous time, $t_{\rm visc}$ ({\it thick, solid line}), the

690: neutronization time, $t_n=1/r_{e^-p}$ ({\it dotted line}), and the

691: $\alpha$-particle photodisintegration time, $t_{\rm photo}$ ({\it

692: dashed line}).}

693: \label{fig:composition}

694: %\epsscale{1.0}

695: \end{figure}

696: 

697: Figure \ref{fig:Ye} shows how the late-time, frozen-out value of $Y_e$ in the disk depends

698: on the initial disk mass $M_{d,0}$ and radius $r_{d,0}$, for two

699: different initial electron fractions, $Y_{e,0} = 0.1$ and $Y_{e,0} =

700: 0.5$.  The former is relevant for the disks created from NS-NS or

701: BH-NS mergers (the focus of this paper), while a larger $Y_{e,0}

702: \simeq 0.5$ is appropriate for disks created during the

703: accretion-induced collapse of a white-dwarf to a neutron star (e.g., Woosley $\&$ Baron 1992; Dessart et al.~2006).  Figure \ref{fig:Ye} shows that for sufficiently compact disks, the disk

704: reaches a modestly neutron-rich composition, with $Y_e \simeq

705: 0.3-0.4$, independent of the initial composition.  This is because,

706: as highlighted in Figure \ref{fig:composition}, the timescale to come

707: into $\beta$-equilibrium is shorter than the viscous time.  For disks

708: with a small initial mass and/or a large initial radius (the lower right-hand corner

709: of each panel), $t_n>t_{\rm visc}$

710: and the disk retains its initial composition (set by the tidally-disrupted progenitor

711: and the subsequent dynamical stage of the merger).

712: Finally, neutrino irradiation of the outer disk by the inner disk can increases the freeze-out electron fraction, but we estimate this changes the freeze-out value of $Y_{e}$ by at most $\sim 20\%$.\footnote{Our calculations employ the pair-capture cooling prescription of DiMatteo et al.~(2002), which assume $Y_{e} = 0.5$ and ultra-relativistic electrons; we find, however, that including the effects of degeneracy and arbitrary electron energies on the cooling changes the asymptotic electron fraction by at most a few percent.}

713: 

714: \section{Disk Winds}

715: \label{sec:winds}

716: 

717: Having described the evolution of the accretion disk as a function of

718: time, we now discuss the properties of outflows from these

719: hyper-accreting disks. Winds driven from deep within the BH potential

720: well could produce relativistic jets and power late-time central

721: engine activity.  Outflows driven from larger radii dominate the

722: system's mass loss and may power supernova-like optical transients

723: through the decay of radioactive isotopes that are synthesized in the

724: wind (Li \& Paczy{\'n}ski 1998; Kulkarni 2005). In both cases, the

725: mass loss rate and nuclear composition are critical for determining

726: the observable signature.

727: 

728: \begin{figure}

729: %\epsscale{1.2}

730: %\hskip -2.6cm

731: \resizebox{\hsize}{!}{\includegraphics[ ]{Yecontour.eps}}

732: \caption{Contours of late-time electron fraction in the expanding disk

733: as a function of initial disk mass $M_{d,0}$ and radius $r_{d,0}$, for

734: two different initial compositions.  Relatively compact disks come

735: into $\beta$-equilibrium and reach an electron fraction independent of

736: the initial $Y_e$, while low mass, more extended disks retain their

737: initial composition.  Figure \ref{fig:composition} shows the evolution

738: of $Y_e$ with time for one particular disk solution.}

739: \label{fig:Ye}

740: %\epsscale{1.0}

741: \end{figure}

742: 

743: 

744: 

745: 

746: The type and character of the outflow depends on the disk's

747: thermodynamic state and changes as it passes through the different

748: stages of evolution described in the previous section. In \S

749: \ref{thindiskwinds} we discuss early times when winds are due to

750: neutrino irradiation of the thin, efficiently neutrino-cooled portions

751: of the disk.  We then consider thermally driven winds during thick,

752: radiatively-inefficient accretion in \S \ref{thickdiskwinds}. This

753: dominates the mass loss at late times and blows away most of the

754: remaining disk.  In \S\ref{nuclearcomposition} we summarize the

755: nuclear composition of the outflows during each phase. We predict an

756: ejected $^{56}$Ni mass of at most $\sim 10^{-3} M_{\sun}$ (\S

757: \ref{sec:ni}). Its decay may power transient emission detectable

758: following some short GRBs.

759: 

760: \subsection{Neutrino-Heated Thin-Disk Winds}

761: \label{thindiskwinds}

762: 

763: A wind with a mass loss rate $\dot{M}_w$ driven from a thin disk at

764: radius $r$ must absorb a net power greater than $\dot{E}_{b} =

765: GM\dot{M}_w/2r$ to become unbound from the central BH.  In principle,

766: $\dot{E}_{b}$ may be supplied by dissipation of the turbulence that

767: produces the accretion shear stresses.  ``Viscous'' heating of this

768: kind only efficiently drives an outflow if a substantial fraction of

769: the accretion power is dissipated in the disk's upper atmosphere,

770: where the cooling timescale is long compared to the wind's outward

771: advection timescale.  However, local radiation MHD simulations to date

772: suggest that very little energy dissipation occurs in the corona

773: (e.g., Krolik et al.~2007).  Instead, heating in the atmosphere above

774: a thin, neutrino-cooled disk is likely dominated by neutrino

775: irradiation.  We therefore focus on the neutrino-driven mass loss

776: rate, which sets a \emph{minimum} $\dot{M}_w$, and which can be

777: reliably estimated.  Neutrino-driven outflows from hyper-accreting disks have also been studied by Daigne $\&$ Mochkovitch (2002), Levinson (2006), Metzger et al.~(2008a), and Barzilay $\&$ Levinson (2008); Dessart et al.~(2008b) calculate the neutrino-driven mass loss from the central NS following a NS-NS merger under the assumption that collapse to a BH is not prompt.

778: 

779:    The neutrino-driven mass loss rate is calculated by equating $\dot{E}_b$ to

780: the total neutrino heating rate in the disk's atmosphere. For the

781: radii and entropies that characterize the winds, heating via electron

782: neutrino absorption on baryons ($p+\bar{\nu}_{e}\rightarrow n+e^{+}$

783: and $n+\nu_{e}\rightarrow p+e^{-}$) dominates other forms of neutrino

784: heating (e.g., $\nu-\nu$ annihilation and $\nu-e^{-}$ scattering; see

785: Qian $\&$ Woosley 1996; hereafter QW96).  Since the neutrino

786: absorption cross section,

787: $\sigma_{\nu N}\simeq 5\times 10^{-44}\langle\epsilon_{\nu}^{2}\rangle$ MeV$^{-2}$ cm$^{2}$,

788: increases with neutrino energy, neutrinos radiated from near the inner

789: radius $r_{*}$ dominate. Assuming that the $\nu_{e}$ and $\bar{\nu}_{e}$

790: luminosities and spectra are approximately equal and can be approximated

791: as originating from a point source at small radii, the neutrino heating rate

792: through a surface density $\Sigma$ at radius $r$ is

793: \be

794: 	q_{\nu}^{+} = \frac{L_{\nu}\sigma_{\nu N}\Sigma}{4\pi m_N r^{2}} 

795: 	\simeq 2\times 10^{39}L_{52}\langle\epsilon_{10}^{2}\rangle

796: 	\Sigma_{18}r_{6}^{-2}{\rm \,ergs\,\,s^{-1}\,\,cm^{-2}},

797: 	\nonumber

798: 	\\

799: 	\label{eq:qnuplus}

800: \ee

801: where $r = 10^{6}r_{6}$ cm, $L_{\nu} = 10^{52}L_{52}$ ergs s$^{-1}$,

802: $\langle\epsilon_{\nu}^{2}\rangle$ = 100$\langle\epsilon_{10}^{2}\rangle$ MeV$^{2}$,

803: and $\Sigma = \Sigma_{18}10^{18}$ g cm$^{-2}$.  This expression assumes that the absorbing

804: layer is optically thin, i.e., that

805: $\tau_{\nu} \equiv \Sigma\sigma_{\nu N}/m_{N} \simeq 3\Sigma_{18}\langle\epsilon_{10}^{2}\rangle < 1$.

806: 

807:    First, consider neutrino heating in comparison to viscous heating in the midplane.  This ratio

808: is largest when the disk is marginally optically thick ($\tau_{\nu} \simeq 1$),

809: peaking at a value of

810: \bea

811: 	\left.\frac{q_{\nu}^{+}}{q_{\rm visc}}\right|_{\tau_{\nu} \simeq 1} \simeq \nonumber \eea \bea 0.5\left(\frac{\epsilon}{0.1}\right)\left(\frac{f}{1.6}\right)\left(\frac{A}{3.6}\right)^{3/5}\langle\epsilon_{10}^{2}\rangle^{2/5}J_{49}^{2/5}M_{3}^{-6/5},

812: \label{eq:heating_ratio}

813: \eea where $\epsilon \equiv L_{\nu}/\dot{M}_{d}c^{2}$ is the disk's

814: radiative efficiency.  Thus, although we neglected neutrino heating in

815: \S \ref{sec:results}, it may become somewhat important when $\tau_\nu

816: \sim 1$ and should be included in a more detailed calculation.

817: 

818: We now consider a wind that emerges from the disk in the

819: $z$-direction, parallel to the rotation axis. Away from the disk

820: midplane, neutrino heating dominates over viscous heating, balancing

821: cooling ($q_{\nu}^{+} = q_{\nu}^{-}$) at a slightly lower temperature,

822: $ T_{\nu} \simeq

823: 3.3L_{52}^{1/6}\langle\epsilon_{10}^{2}\rangle^{1/6}r_{6}^{-1/3}{\rm

824: MeV}$.  Moving further out in the hydrostatic atmosphere, the

825: temperature slowly decreases below $T_{\nu}$.  Due to the strong

826: temperature dependence of the pair capture cooling rate ($q_{\nu}^{-}

827: \propto T^{6}$), a ``gain region'' of net neutrino heating (i.e.,

828: $q_{\nu}^{+} > q_{\nu}^{-}$) develops above a height $z_{\rm

829: gain}$. This net heating drives an outflow.

830: 

831: The thermal power deposited in the upper disk atmosphere

832: $\dot{E}_{\nu}$ is the specific heating rate $q^{+}_{\nu}/\Sigma$

833: (eq.~[\ref{eq:qnuplus}]) multiplied by the mass of the atmosphere in

834: the gain region $M_{\rm gain} \simeq 2\pi H(z_{\rm

835: gain})r^{2}\rho(z_{\rm gain})$, where $H(z_{\rm gain})$ is the scale

836: height near the base of the gain region.  Although the midplane of a

837: neutrino-cooled disk is generally dominated by nonrelativistic gas

838: pressure (see Fig. \ref{fig:pressure}), the gain region has a

839: sufficiently low density that it is instead dominated by radiation

840: pressure $P_{\rm rad} = (11/12)a T^{4}$. Its scale height is $H(z_{\rm

841: gain}) \simeq (P_{\rm rad}/\rho g_{\rm z})|_{\rm z_{\rm gain}}$, where

842: $g_{\rm z}$ is the gravitational acceleration in the

843: $z$-direction. Since $H(z_{\rm gain})$ is less than the midplane scale

844: height $H$, $z_{\rm gain} \simeq$ H and $g_{\rm z} \simeq

845: GMH/r^{3}$. The atmosphere in the gain region is roughly isothermal so

846: we set $T(z_{\rm gain}) \approx T_{\nu}$.  By combining these

847: estimates and equating $\dot{E}_{\nu}$ with $\dot{E}_{b}$ we find that

848: the neutrino-driven mass loss rate from a thin disk is \be

849: \dot{M}_{\nu}|_{S^{a} \gg S_{\rm N}} \approx 10^{-6}L_{52}^{5/3}

850: \langle\epsilon_{10}^{2}\rangle^{5/3}r_{6}^{5/3}M_{3}^{-2}(H/r)^{-1}M_{\sun}{\rm s^{-1}},

851: 	\label{eq:mdot_nu}

852: \ee

853: analogous to that derived by QW96 for proto-neutron star winds.  The assumption that the atmosphere is

854: radiation dominated is only valid if the asymptotic entropy in relativistic particles $S^{a}$ exceeds that in

855: nonrelativistic nucleons $S_{\rm N} \simeq 6 + {\rm ln}(T_{\rm MeV}^{3/2}/\rho_{10}) k_{B}$ baryon$^{-1}$,

856: where $T = T_{\rm MeV}$ MeV.  By dividing the energy gained by a nucleon in the wind

857: $\simeq GMm_{N}/2r$ by the gain region temperature $T(z_{\rm gain})$, we estimate

858: \be

859: 	S^{a} \simeq 60 L_{52}^{-1/6}\langle\epsilon_{10}^{2}\rangle^{-1/6}r_{6}^{-2/3}M_{3}\,\,k_{\rm B}{\rm\,baryon^{-1}}

860: 	\label{eq:sa}

861: \ee

862: as the asymptotic wind entropy.

863: 

864: Although equation (\ref{eq:mdot_nu}) does not strictly hold when $S^{a}\sim S_{\rm N}$, QW96 show that

865: $\dot{M}_{\nu}$ scales the same way with $L_{\nu}$, $\langle\epsilon_{\nu}^{2}\rangle$, $M$, and $r$,

866: but with a larger normalization of

867: \be

868: 	\dot{M}_{\nu}|_{S^{a} \sim S_{\rm N}} \approx 10^{-5}L_{52}^{5/3}

869: 	\langle\epsilon_{10}^{2}\rangle^{5/3}r_{6}^{5/3}M_{3}^{-2}(H/r)^{-1}M_{\sun}{\rm s^{-1}}.

870: 	\label{eq:mdot_nu2}

871: \ee The mass loss rate is higher for low entropy winds because

872: neutrino heating peaks further off the disk surface, which reduces the

873: binding energy and gravitational acceleration of matter in the gain

874: region.  Using the numerical disk wind calculations described in

875: Metzger et al. (2008b; hereafter M08b) we have verified that equation

876: (\ref{eq:mdot_nu2}) holds to within a factor $\simeq 2$ when $S^{a}

877: \sim S_{\rm N}$.

878: 

879: In deriving equations (\ref{eq:mdot_nu}) and (\ref{eq:mdot_nu2}), we

880: have implicitly assumed that the timescale for neutrinos to heat matter in the gain region

881: $t_{\rm heat} \equiv (U_{\rm th}\Sigma/\rho q_{\nu}^{+})|_{z_{\rm gain}}$, where

882: $U_{\rm th} \simeq 3P_{\rm rad}$ is the thermal energy density, is short compared to $t_{\rm visc}$,

883: the timescale over which the disk properties appreciably change.  Equating $S^{a}$ (eq.~[\ref{eq:sa}])

884: to the entropy in relativistic particles $\propto T^{3}/\rho$, we find that

885: \be \rho(z_{\rm gain}) \simeq 10^{8} r_{6}^{-1/3}L_{52}^{2/3}\langle\epsilon_{10}^{2}\rangle^{2/3}M_{3}^{-1} {\rm g\,\,cm^{-3}}. 

886: \label{eq:rho_gain}

887: \ee

888: Then, using equations (\ref{eq:qnuplus}) and (\ref{eq:rho_gain}), we have

889: that\footnote{Equation (\ref{eq:theat}) is also approximately equal to the outward advection

890: timescale of the wind in the heating region.} 

891: \be

892: 	t_{\rm heat} \simeq \left.\frac{3P_{\rm rad}}{\rho(q^{+}_{\nu}/\Sigma)}\right|_{z_{\rm gain}}

893: 	\simeq 0.1\,{\rm s}\,\,L_{52}^{-1}r_{6}\langle\epsilon_{10}^{2}\rangle^{-1}M_{3}

894: 	\label{eq:theat}

895: \ee  

896: For most of the disk solutions considered in this paper, we find that

897: $t_{\rm heat} \lesssim t_{\rm visc}$ during the thin disk phase; thus, equations (\ref{eq:mdot_nu})

898: and (\ref{eq:mdot_nu2}) are reasonably applicable near $r_{d}$.

899: 

900:    Figure \ref{fig:mdot} compares the accretion rate $\dot{M}_{d}$

901: ({\it solid line}) with the neutrino-driven mass loss rate

902: $\dot{M}_\nu$. In order to determine $L_{\nu}$ and

903: $\langle\epsilon_{\nu}^{2}\rangle$, we calculated steady-state disk

904: models \cite[e.g.,][]{dim02} with the accretion rate set at each time

905: according to our ring model with $J_{49} = 2$ and $M_d=0.3

906: M_{\sun}$.  We plot the neutrino-driven mass loss rate $\dot{M}_{\nu}$

907: (eqs.~[\ref{eq:mdot_nu}] and [\ref{eq:mdot_nu2}]) at small ({\it

908: dotted line}) and large ({\it short-dashed line}) radii. This shows

909: that the mass loss is dominated by large radii where the majority of

910: the mass lies, as expected since $\dot{M}_{\nu} \propto r^{5/3}$.  The vertical dot-dashed line marks where the

911: disk transitions to being thick (eq. [\ref{eq:tthick}]),

912: after which neutrino-heating no longer dominates the wind mass loss.

913: \begin{figure}

914: %\epsscale{1.2}

915: \resizebox{\hsize}{!}{\includegraphics[ ]{windmdot.eps}}

916: \caption{The accretion rate $\dot{M}_{d}$ ({\it solid line}) and

917: neutrino-driven mass loss rates $\dot{M}_\nu$ for our $J_{49} = 2$ and

918: $M_{d,0} = 0.3 M_{\sun}$ model, focusing on the phase of thin,

919: efficiently neutrino-cooled accretion. The neutrino-driven mass loss

920: rate $\dot{M}_{\nu}$ (interpolated between eqs.~[\ref{eq:mdot_nu}] and

921: [\ref{eq:mdot_nu2}]) is shown at the inner disk radius ($r_{*} =

922: 10^{6}$ cm; {\it dotted line}) and at the outer disk radius (near

923: $r_{d}$, {\it short-dashed line}).  The disk is advective to the right

924: of the vertical line (eq. [\ref{eq:tthick}]), at which point the

925: mass-loss will no longer be dominated by neutrino irradiation.}

926: 

927: \label{fig:mdot}

928: %\epsscale{1.0}

929: \end{figure}

930: 

931: Outflows that are launched from small radii, near $r_{*}$, have the

932: greatest potential to produce relativistic jets and to power high

933: energy emission. But as we now argue, these neutrino-driven winds are

934: too massive to become highly relativistic. Our calculation above

935: focused on purely thermal, neutrino-driven winds, which accelerate

936: matter to only a fraction of the escape speed (and thus are mildly relativistic). However, in the presence of a strong, large scale

937: open poloidal magnetic field, a more powerful, magnetically-driven

938: outflow is possible.  Magnetocentrifugal support in the wind's

939: hydrostatic atmosphere may further enhance mass loss (e.g., Levinson 2006), but equation (\ref{eq:mdot_nu}) still represents the {\it minimum} mass loading on

940: field lines which thread a neutrino-cooled disk.  Figure

941: \ref{fig:mdot} shows that $\dot{M}_{\nu}(r_{*}) \sim

942: 10^{-4}-10^{-2}M_{\sun}$ s$^{-1}$ during the thin disk phase.  The

943: luminosities of the prompt emission and late-time X-ray flares from

944: short GRBs, however, do not typically exceed $L_\gamma \sim 10^{50}$

945: erg s$^{-1}$ (and are often much lower; Nakar 2007).  Thus, even

946: assuming a modest radiative efficiency for the outflow of

947: $\epsilon_{w} \sim 0.1$, the Lorentz factor $\Gamma$ of a

948: neutrino-heated disk wind must obey $\Gamma \simeq

949: L_\gamma/[\epsilon_{w}\dot{M}_{\nu}(r_{*})c^{2}] \lesssim 5$, which is

950: inconsistent with existing compactness constraints on short GRBs

951: (Nakar 2007).  A more likely source for the relativistic outflows that

952: power short GRBs and their late-time flares are nearly baryon-free

953: field lines which thread the BH's event horizon (e.g., McKinney 2005).

954: In addition, in \S \ref{thickdiskwinds} we argue that when the disk becomes

955: advection dominated and neutrino irradiation effectively ceases, jet

956: production may be more likely.

957: 

958: \subsection{Radiatively-Inefficient Thick Disk Winds}

959: \label{thickdiskwinds}

960: 

961: At late times ($t \sim t_{\rm thick}$; eq.~[\ref{eq:tthick}]) the disk

962: transitions from thin and neutrino-cooled to being advective.  At this

963: point a neutrino-driven outflow is unlikely to dominate the mass loss,

964: in part because the neutrino luminosity precipitously drops

965: (Fig. \ref{fig:energy1} \& \ref{fig:energy2}).  In addition, because

966: RIAFs possess a positive Bernoulli parameter, a powerful

967: viscously-driven outflow is likely (Blandford $\&$ Begelman 1999;

968: Stone $\&$ Pringle 2001; Proga $\&$ Begelman 2003).

969: 

970: In \S \ref{sec:gross} we showed that the disk becomes radiatively

971: inefficient external to an ``ignition radius'' $r_{\rm ign} \propto

972: \dot{M}_{d}^{6/5}$ (eq.~[\ref{eq:rign}]). The outer disk, near

973: $r_{d}$, thickens first (when $r_{d} \sim r_{\rm ign}$ at $t \sim

974: t_{\rm thick}$) and radiatively inefficient conditions move inwards as

975: $\dot{M}_{d}$ decreases. In the simplest picture, one might expect

976: that the innermost radii become an RIAF only once $\dot{M}_{d}$ drops

977: from its value at $t \sim t_{\rm thick}$ by an additional factor $\sim

978: (r_{*}/r_{d})^{5/6}$. In fact, the {\it entire} disk probably become

979: radiatively inefficient on a timescale similar to $t_{\rm thick}$ if

980: the accretion rate which reaches small radii abruptly decreases once

981: the outer disk thickens (Fig.~\ref{fig:wind}).  Hence, at a time $t_{\rm thick}$, a

982: significant portion of the accreting matter may be redirected into an

983: outflow, with only a fraction $\sim (r_{*}/r_{d})$ reaching small

984: radii and accreting onto the BH (Stone $\&$ Pringle 2001).

985: 

986: X-ray binaries typically produce radio jets upon transitioning from

987: their ``high-soft'' (radiatively efficient) to ``low-hard''

988: (radiatively inefficient) states (e.g., Remillard $\&$ McClintock

989: 2006).  In analogy, once the inner disk becomes an RIAF, conditions

990: seem to favor the production of relativistic jets (see also Lazzati et al.~2008).\footnote{This is in

991: stark contrast to jets powered by neutrino annihilation along the

992: polar axis, which require a {\it high} radiative efficiency.} 

993: 

994: Even if only a fraction $(r_{*}/r_{d})$ of the mass remaining when the

995: disk thickens actually reaches the origin, the total energy supply

996: available would be \bea E_{\rm jet}&\equiv&\epsilon_{\rm

997: jet}M_d(t_{\rm thick})c^{2}\left(\frac{r_{*}}{r_d(t_{\rm

998: thick})}\right) \nonumber \eea \bea \simeq 3\times

999: 10^{50}\left(\frac{\epsilon_{\rm jet}}{0.1}\right)

1000: \left(\frac{r_{*}}{10^{6}{\rm

1001: cm}}\right)\alpha_{0.1}^{6/17}M_{3}^{-4/17}

1002: \left(\frac{J_{49}}{2}\right)^{8/17}{\rm ergs}, \nonumber \\

1003: 	\label{eq:ejet}

1004: \eea where $\epsilon_{\rm jet}$ is the fraction of the accretion energy

1005: used to power a jet and we have estimated $M_d(t_{\rm thick})$ and

1006: $r_{d}(t_{\rm thick})$ using the self-similar thick disk solutions

1007: (eqs.~[\ref{eq:md_analytic2}] and [\ref{eq:rd_analytic2}],

1008: respectively).  Equation (\ref{eq:ejet}) shows that the accretion

1009: energy available from near $r_{*}$ following the RIAF transition is

1010: more than sufficient to power the late-time X-ray flares observed

1011: following some short GRBs.  If this is the case, $t_{\rm thick}$ sets

1012: a characteristic timescale for late-time central engine activity.  If

1013: $\alpha\lesssim 10^{-3}$, $t_{\rm thick}$ may be large enough to

1014: explain the $\sim 30\ {\rm s}$ delay until flaring observed for some

1015: short GRBs (e.g., Berger et al.~2005; Villasenor et

1016: al.~2005). However, very late time energy injection, such as the {\it

1017: Chandra} flare observed two weeks after GRB050709 (Fox et al.~2005),

1018: appears to require an alternative explanation. In addition, given

1019: observational evidence for $\alpha \sim 0.1$ in a number of

1020: environments (King et al. 2007), it may be more natural to associate

1021: $E_{\rm jet}$ and $t_{\rm thick}$ with the energy and duration,

1022: respectively, of the short GRB itself, rather than the late-time

1023: central engine activity (see \S \ref{sec:theend}).

1024: 

1025: \subsection{Outflow Nuclear Composition}

1026: \label{nuclearcomposition}

1027: 

1028: The outflow nuclear composition has important consequences for the

1029: observable signature of compact object mergers. Nonrelativistic

1030: outflows are sufficiently dense to synthesize heavy isotopes (Pruet et al.~2004; Surman et al.~2006), which

1031: may power transient emission via radioactive decay.  The isotopic

1032: yield depends on the speed, thermodynamic properties, and the asymptotic electron fraction $Y_{e}^{a}$ in the

1033: outflow.\footnote{The {\it asymptotic} electron fraction is germane

1034: because heavy nuclei primarily form after freeze-out from

1035: $\beta$-equilibrium.}  Although relativistic winds from the inner disk

1036: are unlikely to synthesize anything heavier than He (Lemoine 2002;

1037: Beloborodov 2003a), $Y_{e}^{a}$ is important in this

1038: case as well. A neutron-rich outflow may alter the jet's dynamics and

1039: the prompt and afterglow emission from that of the standard GRB

1040: fireball model (e.g., Derishev et al.~1999; Beloborodov 2003b; Rossi

1041: et al.~2006).

1042: 

1043: Figure \ref{fig:mdot_radius} delineates different regimes of outflow properties and

1044: composition (as given by $Y_e^a$) as a function of the wind launching

1045: radius $r$ and accretion rate $\dot{M}_{d}$. We fix $\alpha = 0.1$ and

1046: $M = 3M_{\sun}$. The time-dependent evolution of the ring radius $r_{d}$

1047: is shown for a solution with $J_{49} = 2$ and $M_{d,0} = 0.3M_{\sun}$ ({\it solid line}).

1048: At each time a given steady-state disk profile can be read off of this plot as a horizontal line

1049: that extends from the far left and ends on $r_d$.

1050: Therefore, outflows from radii {\it interior} to $r_{d}$ contribute to the disk's total nucleosynthetic yield.

1051: 

1052: \begin{figure}

1053: %\epsscale{1.2}

1054: \resizebox{\hsize}{!}{\includegraphics[ ]{mdot_radius.eps}}

1055: \caption{Asymptotic electron fraction $Y_e^a$ for disk winds as a

1056: function of the wind launching radius $r$ and accretion rate

1057: $\dot{M}_{d}$ (for $\alpha = 0.1$ and $M = 3M_{\sun}$). The solid line

1058: indicates the location of the ring radius $r_{d}$ for our fiducial

1059: solution with $M_{d,0} = 0.3 M_{\sun}$ and $J_{49} = 2$.  The short

1060: dashed line is the ``ignition'' radius $r_{\rm ign}$

1061: (eq.~[\ref{eq:rign}]).  Exterior to this (marked ``Thick Disk'') the

1062: disk is advective with a viscously driven wind of composition

1063: $Y_e^a\simeq Y_e^{\rm mid}<0.5$.  Interior to $r_{\rm ign}$ (marked

1064: ``Thin Disk'') a neutrino-driven wind occurs. The dotted line shows

1065: $r=r_\nu$ with $Q=2$ (eq. [\ref{eq:rnu}]) and determines where the

1066: neutrino absorptions necessary to unbind matter alter the wind

1067: composition, so that $Y_e^a\simeq Y_e^{\rm mid}<0.5$ ($Y_e^a\simeq

1068: Y_e^\nu$) exterior (interior) to $r_\nu$.  The $\dot{M}_{d}$ above

1069: which $\tau(r_{*})>1$ is plotted for BH spins of $a = 0$ and $a =

1070: 0.9$.  Above this line, the $\bar{\nu}_{e}$ and $\nu_{e}$ spectra

1071: differ and $Y_{e}^{\nu} <0.5$, while below this their spectra are

1072: similar and $Y_{e}^{\nu} \simeq 0.5$.  In the region where $r < r_{\rm

1073: ign}$, $\tau_{\nu}(r_{*}) < 1$, and $r<r_\nu$ (i.e., the middle/lower left-hand trapezoid), $Y_{e}^{a} \simeq Y_{e}^{\nu} \sim 0.5$; these

1074: conditions are favorable for $^{56}$Ni production (see

1075: $\S\ref{sec:ni}$).}

1076: \label{fig:mdot_radius}

1077: %\epsscale{1.0}

1078: \end{figure}

1079: 

1080: The ignition radius $r_{\rm ign}$ (eq.~[\ref{eq:rign}]) is shown in

1081: Figure \ref{fig:mdot_radius} with a short dashed line. For $r \gtrsim

1082: r_{\rm ign}$ the disk is an RIAF and marked in the figure as ``Thick

1083: Disk.'' In this case, a viscously driven outflow dominates (\S

1084: \ref{thickdiskwinds}). Since outflows from \mbox{RIAFs} escape the

1085: disk in roughly the accretion timescale, these winds retain the

1086: midplane electron fraction (M08b), so that $Y_e^a\simeq Y_{e}^{\rm

1087: mid}\ll 0.5$ (because the disk itself freezes-out neutron-rich, as summarized in \S \ref{sec:composition} and

1088: Fig. \ref{fig:Ye}).

1089: 

1090: For $r \lesssim r_{\rm ign}$, the disk is efficiently neutrino-cooled

1091: and marked in Figure \ref{fig:mdot_radius} as ``Thin Disk.'' The

1092: absorption of neutrinos, which heats the outflow and unbinds it from

1093: the BH may also alter its nucleonic composition. This drives

1094: $Y_e^a$ to a value set by the neutrino radiation field $Y_e^\nu$,

1095: which in general is different from $Y_{e}^{\rm mid}$.  A simple

1096: criterion was discussed by M08b for determining when $Y_e^a\simeq

1097: Y_e^\nu$.  A typical nucleon in the accretion disk at radius $r$ must

1098: absorb an energy $\simeq GMm_{N}/2r$ to become unbound from the BH, so that $N_{\nu} \simeq GMm_{N}/2r\langle\epsilon_{\nu}\rangle$

1099: neutrinos must be absorbed per nucleon. If we take $N_{\nu} > Q \sim

1100: 2-3$, then a typical nucleon has changed its identity ($p\rightarrow

1101: n$ or $n\rightarrow p$) at least several times.

1102: 

1103: This implies that all purely neutrino-driven outflows from radii

1104: smaller than \be r_{\nu} \equiv

1105: \frac{GMm_{p}}{2Q\langle\epsilon_{\nu}\rangle} \simeq

1106: 10^{7}M_{3}\langle\epsilon_{10}\rangle^{-1}(Q/2)^{-1}{\rm\,cm},

1107: 	\label{eq:rnu}

1108: \ee where $\langle\epsilon_{\nu}\rangle \equiv

1109: 10\langle\epsilon_{10}\rangle\ {\rm MeV}$, achieve $Y_{e}^{a} \simeq

1110: Y_{e}^{\nu}$, independent of the disk's midplane composition.

1111: 

1112: We plot $r_\nu$ with $Q = 2$ as a dotted line in Figure \ref{fig:mdot_radius}, where $\langle\epsilon_{\nu}\rangle$ is calculated from $\dot{M}_{d}$ using our steady-state disk solutions (see $\S \ref{thindiskwinds}$).  For $r \lesssim r_{\nu}$, any neutrino-driven outflow enters equilibrium with the neutrino radiation field (i.e., $Y_{e}^{a} \simeq Y_{e}^{\nu}$).  For $r \gtrsim r_{\nu}$ the outflow approximately retains the midplane electron fraction (i.e., $Y_{e}^{a} \simeq Y_{e}^{\rm mid}$).

1113: 

1114: Although we have established the conditions under which $Y_{e}^{a}$ is

1115: determined by neutrino absorptions, we must now address what sets

1116: $Y_{e}^{\nu}$ itself.  If the rate of neutrino absorptions exceeds the

1117: rate of degenerate pair captures before the wind falls out of

1118: $\beta$-equilibrium, $Y_e^\nu$ is \be Y_e^\nu \equiv \left(1+\frac{L_{\bar{\nu}_e}}{L_{\nu_e}}

1119: \frac{\langle\epsilon_{\bar{\nu}_e}\rangle-2\Delta +

1120: 1.2\Delta^2/\langle\epsilon_{\bar{\nu}_e}\rangle}

1121: {\langle\epsilon_{\nu_e}\rangle+2\Delta+1.2\Delta^2/\langle\epsilon_{\nu_e}\rangle}

1122: \right)^{-1},

1123: 	\label{eq:yeanu}

1124: \ee where $\Delta = 1.293$ MeV is the neutron-proton mass difference,

1125: and $L_{\nu_{e}}/L_{\bar{\nu}_{e}}$ and

1126: $\langle\epsilon_{\nu_{e}}\rangle$/$\langle\epsilon_{\bar{\nu}_{e}}\rangle$

1127: are the mean $\nu_{e}$/$\bar{\nu}_{e}$ luminosities and energies,

1128: respectively, from a centrally-concentrated source (Qian et al.~1993;

1129: QW96). Equation (\ref{eq:yeanu}) demonstrates that the $\nu_e$ and

1130: $\bar{\nu}_e$ spectra are crucial for setting $Y_e^\nu$.

1131:  

1132: Since the disk's luminosity and temperature peak at just a few

1133: $r_{g}$, $Y_{e}^{\nu}$ is primarily determined by conditions at small

1134: radii. At early times, the accretion disk may be optically thick near

1135: $r_{*}$ and so the $\nu_{e}$ and $\bar{\nu}_{e}$ spectra depend on the

1136: temperatures at $\nu_{e}$ and $\bar{\nu}_{e}$ neutrinospheres,

1137: respectively.  Since there are more neutrons than protons in the disk,

1138: the optical depth to $\nu_{e}$ through the disk is higher than to

1139: $\bar{\nu}_{e}$; thus, the temperature at the $\bar{\nu}_{e}$

1140: neutrinosphere is higher than at the $\nu_{e}$ neutrinosphere.  This

1141: implies $L_{\bar{\nu}_{e}} \gg L_{\nu_{e}}$,

1142: $\langle\epsilon_{\bar{\nu}_{e}}\rangle \gg

1143: \langle\epsilon_{\nu_{e}}\rangle$, and thus $Y_{e}^{\nu} \ll 0.5$.

1144: Using 3-dimensional calculations of the merger of NSs with zero spin, Rosswog $\&$ Liebend{\"o}rfer (2003) find that at $\sim 15$ ms following merger,

1145: $L_{\bar{\nu}_{e}} \simeq 3.5L_{\nu_{e}}$,

1146: $\langle\epsilon_{\nu_{e}}\rangle \simeq 9$ MeV, and

1147: $\langle\epsilon_{\bar{\nu}_{e}}\rangle \simeq$ 15 MeV, which implies

1148: $Y_{e}^{\nu} \simeq 0.21$, consistent with our arguments (see also

1149: Surman et al. 2008).  We conclude that when the disk is optically-thick near $r_{*}$, a

1150: neutron-rich outflow is again the most likely outcome.  The critical accretion rate at which $\tau_{\nu}(r_{*})=1$ is shown in Figure \ref{fig:mdot_radius} with a long dashed line for both $a = 0$ and $a=0.9$.

1151: 

1152: Once the disk becomes optically thin near $r_{*}$, the

1153: difference between the $\nu_{e}$ and $\bar{\nu}_{e}$ spectra is much

1154: less pronounced.  This occurs because (1) the neutrinos and

1155: antineutrinos originate from regions with the same temperature; (2)

1156: any net lepton flux out of the disk must remain modest (i.e.,

1157: $L_{\nu_{e}}/\langle\epsilon_{\nu_{e}}\rangle \simeq

1158: L_{\bar{\nu}_{e}}/\langle\epsilon_{\bar{\nu}_{e}}\rangle$); and (3)

1159: the difference between the $e^{-}$ and $e^{+}$ capture cross sections

1160: for $kT \gg \Delta-m_{e}c^{2}$ is small. Taking

1161: $\langle\epsilon_{\nu_{e}}\rangle\sim\langle\epsilon_{\bar{\nu}_{e}}\rangle

1162: \gg \Delta$, equation (\ref{eq:yeanu}) shows that $Y_{e}^{\nu}\simeq

1163: 0.5$, a value in the range required to produce $^{56}$Ni (which we

1164: discuss further in \S \ref{sec:ni}).  Indeed, M08b used the

1165: steady-state, optically-thin $\alpha$-disk calculations of Chen \&

1166: Beloborodov (2007; hereafter CB07) to calculate the neutrino radiation

1167: fields carefully, and showed that $Y_{e}^{\nu} \gtrsim 0.5$ over the

1168: majority of the disk (see their Fig.~1).  Although the precise spectra

1169: extracted from an $\alpha$-disk calculation should be taken with

1170: caution, the conclusion that the $\nu_{e}$ and $\bar{\nu}_{e}$ spectra

1171: are similar for optically thin accretion (and $Y_{e}^{\nu} \simeq

1172: 0.5$) is probably robust.

1173: 

1174: Figure \ref{fig:mdot_radius} illustrates that under most conditions

1175: the outflows from hyper-accreting disks are neutron-rich.

1176: Neutron-rich material ejected during the initial dynamical phase of

1177: compact object mergers has long been considered a promising source for

1178: producing Galactic $r$-process elements, whose precise astrophysical

1179: origin remains uncertain (Lattimer $\&$ Schramm 1974; see, however,

1180: Qian 2000).  In addition, Surman et al.~(2008) find that winds driven

1181: from the remnant accretion disk at early times (when it is optically

1182: thick; upper left quadrant of Fig. \ref{fig:mdot_radius}) are

1183: sufficiently neutron-rich to produce successful $r$-process.  The

1184: outflows driven from the advective disk at late times, however, are

1185: unlikely to produce r-process elements, given their modest entropies

1186: and electron fractions of $Y_e \gtrsim 0.3$

1187: (Figs. \ref{fig:composition} and \ref{fig:Ye}).  Instead, this modest

1188: $Y_e$ material will be synthesized to form intermediate mass neutron

1189: rich isotopes (Hartmann et al. 1985).

1190: 

1191: \subsection{$^{56}$Ni Production and Optical Transients}

1192: \label{sec:ni}

1193: 

1194: As summarized in Figure \ref{fig:mdot_radius}, most of the material in

1195: the outflow driven from a hyper-accreting disk will be neutron-rich.

1196: Nonrelativistic neutron-rich ejecta are difficult to detect because

1197: isotopes synthesized from low $Y_e$ material are themselves very

1198: neutron-rich and typically possess very short half-lives, on the order

1199: of seconds (e.g., Freiburghaus et al.~1999).  Thus, most of the

1200: radioactive energy is released at high optical depths and suffers

1201: severe adiabatic losses before the photons can diffusively escape.  By

1202: contrast, ejecta with $Y_{e}^{a} \simeq 0.5$ are easier to detect

1203: because they can produce a significant quantity of $^{56}$Ni (Hartmann

1204: et al.~1985), an isotope better suited to powering observable emission

1205: because its half-life $\simeq 6$ days is comparable to the timescale

1206: on which the outflow becomes optically thin.  From Figure

1207: \ref{fig:mdot_radius} we see that outflows in a modest range of

1208: parameter space (middle/lower-left trapezoid) are capable of synthesizing

1209: $^{56}$Ni.  One caveat to this conclusion is that it only applies if

1210: the winds are primarily neutrino driven.  If the outflow is instead

1211: magnetocentrifugally driven by a moderately strong open poloidal

1212: magnetic field (e.g., Levinson 2006; Xie et al.~2007), then $Y_{e}^{a} \ll 0.5$ can result, even if

1213: $Y_{e}^{\nu} \simeq 0.5$ (M08b).  In what follows we assume that the

1214: wind's are primarily neutrino driven.

1215: 

1216: Under this assumption, Figure \ref{fig:Nimass} shows the total

1217: $^{56}$Ni mass, $M_{\rm Ni}= (X_{\rm Ni}/0.4)M_{Y_{e}=0.5}$, produced

1218: in outflows from hyper-accreting disks as a function of the disk's

1219: initial mass $M_{d,0}$ and radius $r_{d,0}$, where $M_{Y_{e} = 0.5}$

1220: is the total mass loss with $Y_{e}^{a} \simeq 0.5$ and $X_{\rm Ni}$ is

1221: the average $^{56}$Ni mass fraction synthesized in the wind.  We

1222: calculate $M_{Y_{e}=0.5}$ by integrating the neutrino-driven mass loss

1223: (eqs.~[\ref{eq:mdot_nu}] and [\ref{eq:mdot_nu2}]) across the

1224: $Y_{e}^{a} \simeq 0.5$ region in Figure \ref{fig:mdot_radius}, using

1225: $r_{d}(t)$ and $\dot{M}_{d}(t)$ from the disk evolution calculations

1226: described in $\S$\ref{sec:results}.  

1227: \begin{figure}

1228: %\epsscale{1.2}

1229: \resizebox{\hsize}{!}{\includegraphics[ ]{Nicontour.eps}}

1230: \caption{Contours of total $^{56}$Ni mass $M_{\rm Ni} \equiv (X_{\rm

1231: Ni}/0.4)M_{Y_{e} = 0.5}$ (in units of $M_{\sun}$) produced in the

1232: neutrino-driven outflows as a function of the initial disk mass

1233: $M_{d,0}$ and initial ring radius $r_{d,0}$, where $M_{Y_{e} = 0.5}$

1234: is the total mass loss with $Y_{e}^{a} \simeq 0.5$ (based on the

1235: arguments in Fig. \ref{fig:mdot_radius}) and $X_{\rm Ni}$ is the

1236: average $^{56}$Ni mass fraction synthesized in the wind.  The upper

1237: and lower panels correspond to non-rotating ($a = 0$) and rapidly

1238: spinning ($a = 0.9$) BHs, respectively.}

1239: \label{fig:Nimass}

1240: %\epsscale{1.0}

1241: \end{figure}

1242: 

1243: Pruet et al.~(2004) present calculations of $X_{\rm Ni}$ which are

1244: parameterized in terms of the asymptotic entropy $S^{a}$, mass loss

1245: rate $\dot{M}_{w}$, and asymptotic velocity $v^{a}$ of an outflow with

1246: $Y_{e}^{a} \simeq 0.51$.  $M_{Y_{e} = 0.5}$ is dominated by

1247: outflows from radii $\sim 3\times 10^{6}-10^{7}$ cm when

1248: $\dot{M}_{d}\sim 0.1-1 M_{\sun}$ s$^{-1}$ (corresponding to $L_{52}

1249: \sim$ few); equation (\ref{eq:sa}) thus gives $S^{a}\sim 10-30 k_{B}$

1250: baryon$^{-1}$ for the ejecta with $Y_{e}^{a} \simeq 0.5$.  Purely

1251: neutrino-driven winds achieve asymptotic velocities which are

1252: typically below the escape speed of the central object (e.g., Thompson

1253: et al.~2001); thus, the asymptotic kinetic energy is most likely

1254: dominated by energy released during the formation of heavy elements.

1255: Because $\sim 8$ MeV baryon$^{-1}$ is released in producing Fe-peak

1256: elements, we estimate that $v^{a} \simeq 0.1-0.15$ c.  Applying these

1257: wind parameters to Figure 3 of Pruet et al.~(2004), we estimate that $X_{\rm

1258: Ni} \sim 0.2-0.5$, thereby justifying our scaling for $X_{\rm Ni}$ in

1259: Figure \ref{fig:Nimass}.

1260: 

1261: Figure \ref{fig:Nimass} shows that for large initial disk masses

1262: ($M_{d,0} \gtrsim 0.1M_{\sun}$), the ejected Ni mass, $\sim 3 \times

1263: 10^{-4}-10^{-3}M_{\sun}$, can be appreciable.  Disks with moderate

1264: initial radii $r_{d,0} \sim 10^{7}$ cm are optimal for producing

1265: $^{56}$Ni because they are sufficiently large to contain the radius

1266: $r_{\nu} \sim 10^{7}$ cm and yet are sufficiently compact to have a

1267: large initial accretion rate, which maximizes the neutrino luminosity

1268: and thus the neutrino-driven mass loss.  Conveniently, initial disk

1269: parameters from many compact object merger simulations (see \S

1270: \ref{sec:initial}) are in the range required to produce $\sim

1271: 10^{-4}-10^{-3}M_{\sun}$ of Ni.

1272: 

1273: \begin{figure}

1274: %\epsscale{1.2}

1275: \resizebox{\hsize}{!}{\includegraphics[ ]{lightcurve.eps}}

1276: \caption{Luminosity of Ni decay-powered ``macronovae'' as a function

1277: of time since merger for Ni mass $M_{\rm Ni} = 10^{-3}M_{\sun}$ and

1278: ejecta velocity $v^{a} = 0.1$ c.  Light curves are shown for three

1279: values of the total ejected mass $M_{\rm tot} = 10^{-3}$ ({\it solid

1280: line}), $10^{-2}$ ({\it dotted line}), and $10^{-1}M_{\sun}$ ({\it dashed

1281: line}).  The luminosities in V and J-Band (0.44 and 1.26 $\mu$m,

1282: respectively) are shown with thick and thin lines, respectively.  The V-band upper limit on emission following GRB050509B from Hjorth et al.~(2005) is shown with a filled triangle.}

1283: \label{fig:lightcurve}

1284: %\epsscale{1.0}

1285: \end{figure}

1286: 

1287: 

1288: The decay of $M_{\rm Ni} \sim 10^{-4}-10^{-3}M_{\sun}$ can reheat the

1289: (adiabatically cooled) ejecta sufficiently to produce detectable

1290: transient emission.  In order to explore this possibility, we

1291: calculate the light curves of ejecta heated by Ni decay

1292: (``macronovae'') using the method of Kulkarni (2005).  This simplified

1293: one-zone model accounts for the fraction of the gamma-rays produced by

1294: the Ni decay which are absorbed by the expanding material (Colgate et

1295: al.~1980) and assumes blackbody emission at the photosphere,

1296: neglecting Comptonization.

1297: 

1298: 

1299: Figure \ref{fig:lightcurve} shows the V and J-band luminosities as a

1300: function of time since the merger for an outflow with Ni mass $M_{\rm

1301: Ni} = 10^{-3}M_{\sun}$ which is expanding at $v^{a} = 0.1$ c.  The V-band light curve peaks earlier because the temperature at the photosphere decreases as the material expands.  Somewhat after the peak in the light curves, recombination will decrease the opacity well below that considered here; thus our calculations are not quantitatively reliable at these times.  The {\it total} mass $M_{\rm tot}$ ejected during the merger event, most of it

1302: neutron rich, is likely to be significantly larger than $M_{\rm Ni}$;

1303: this provides additional opacity for the Ni-rich material.  To explore

1304: the effect of this additional material on the detectability of the Ni

1305: decay, the light curves in Figure \ref{fig:lightcurve} are shown for

1306: three values of $M_{\rm tot}$: $10^{-3} M_{\sun}$ ({\it solid line}),

1307: $10^{-2}M_{\sun}$ ({\it dotted line}), and $10^{-1}M_{\sun}$ ({\it dashed

1308: line}).  As Figure \ref{fig:lightcurve} shows, larger $M_{\rm tot}$:

1309: (1) delays the time to peak emission ($t_{\rm peak}$ is roughly

1310: $\propto M_{\rm tot}^{1/2}$); (2) increases the total fluence of the

1311: event by trapping a higher fraction of the gamma-ray emission; and (3)

1312: increases the peak wavelength of the emission, pushing it into the

1313: near-IR for large $M_{\rm tot}$.  We conclude that long wavelength

1314: ($\lambda \gtrsim \mu$m) observations at $t \sim 1$ day are the most

1315: promising for the detection of a Ni decay-powered macronova.

1316: 

1317: Hjorth et al.~(2005) place an upper limit of $M_{V} > 27.5$ at $t

1318: =3.9$ days on any emission associated with the short GRB 050509B

1319: (redshift $z \simeq 0.22$); we mark this constraint in Figure

1320: \ref{fig:lightcurve} with an arrow.  For $M_{\rm tot} = 0.1M_{\sun}$

1321: this constrains the ejected Ni mass to be $M_{\rm Ni} \lesssim

1322: 10^{-2}M_{\sun}$ (see also Kulkarni 2005).  As Figure \ref{fig:Nimass}

1323: illustrates, compact object mergers are very unlikely to produce this

1324: much Ni, so the absence of a detection thus far is unsurprising.

1325: 

1326: \section{Conclusion and Discussion}

1327: \label{sec:theend}

1328: 

1329: We have calculated the time-dependent evolution of accretion disks

1330: formed from compact mergers, and the properties of their outflows.

1331: Since most of the disk mass resides at large

1332: radii, we approximate the disk as a ring at a given radius and

1333: calculate the dynamics and composition of the ring as a function of

1334: time. This ring model is calibrated to correctly reproduces

1335: the Green's function solution for a viscously spreading ring with

1336: viscosity $\nu \propto r^{1/2}$ (appropriate for a thick disk; see

1337: Appendix \ref{appendix:ring}). With this simplified model, we have

1338: studied the full parameter space of remnant accretion disks (different

1339: initial masses, compositions, etc.) and can follow the viscous

1340: evolution for arbitrarily long timescales.

1341: 

1342: The energetics of the ring at a given time can be described by one of

1343: three models: (1) optically thick to neutrinos and advective, (2)

1344: optically thin to neutrinos and geometrically thin, and (3) optically

1345: thin to neutrinos and advective. A massive, compact disk (with a short

1346: initial viscous time $t_{\rm visc,0}$; eq. [\ref{eq:tvisc0}]) will

1347: exhibit all three of these accretion phases, evolving from (1) to (3)

1348: as a function of time (Figs. \ref{fig:mass1}-\ref{fig:energy2}).

1349: Less massive disks, on the other hand, only pass through phases

1350: (2) and (3), or even just (3). Note that these phases refer to the

1351: energetics of the disk near the outer radius.  At a given time, the disk may also undergo similar transitions as a

1352: function of radius; e.g., a disk that is advective at large radii will

1353: be neutrino cooled and geometrically thin inside the ignition radius

1354: $r_{\rm ign}$ (eq. [\ref{eq:rign}]).

1355: 

1356: Neutrino-driven winds during the early-time optically thick and

1357: neutrino-cooled (thin disk) phases unbind so much mass that field

1358: lines connected to the disk cannot produce sufficiently relativistic

1359: material to power short-duration GRBs (\S \ref{thindiskwinds} and

1360: Fig. \ref{fig:mdot}).  An alternative source for the relativistic

1361: material needed to produce short GRBs are nearly baryon-free magnetic

1362: field lines that thread the BH's event horizon (e.g., McKinney 2005).

1363: In addition, when the inner disk becomes advective ($\dot M_{d} \lesssim

1364: 0.07 \, \alpha_{0.1}^{5/3} \, M_\odot \, s^{-1}$ for a = 0),

1365: conditions appear particularly suitable for the formation of

1366: relativistic jets (by analogy to X-ray binaries, which produce jets

1367: when making a similar transition; e.g., Remillard $\&$ McClintock

1368: 2006; see Lazatti et al.~2008 for a similar argument in the context of long-duration GRBs).

1369: 

1370: Once the disk has transitioned to a late-time advective phase (phase 3

1371: above), the properties of the disk become well-described by

1372: self-similar solutions.  Ignoring for the moment outflows from the

1373: disk, these solutions are $r_d \propto t^{2/3}$, $M_d \propto

1374: t^{-1/3}$, and $\dot M_{d} \propto t^{-4/3}$.  Power-law variations in the

1375: disk properties are a generic feature of a viscously evolving disk

1376: that conserves total angular momentum.  These scalings are not,

1377: however, likely to be applicable in practice because outflows during

1378: the advective phase unbind most of the remaining material (\S

1379: \ref{sec:gross} \& \ref{thickdiskwinds}).  Energy produced by fusion

1380: to He and heavier elements also contributes to driving an outflow

1381: (Figs. \ref{fig:energy1} \& \ref{fig:energy2}).  Such outflows remove a significant fraction of the angular momentum of the disk.

1382: This leads to a much more rapid decrease in the disk mass and accretion rate at

1383: late times (Appendix B3 and Fig. \ref{fig:wind}).  Significant

1384: accretion onto the central black hole will thus only last for a few

1385: viscous times after the onset of the advective phase.

1386: 

1387: At the outer edge of the disk, the transition from a neutrino-cooled

1388: thin disk to the late-time advective phase occurs at a time $t_{\rm

1389: thick} \sim 0.1 \, \alpha_{0.1}^{-23/17} \, (J_{\rm 49}/2)^{9/17}$ s

1390: (eq. [\ref{eq:tthick}]).  The rapid decrease in $\dot M_{d}$ after the

1391: onset of the advective phase implies that the inner disk becomes

1392: advective at a similar time (\S \ref{thickdiskwinds} and

1393: Fig. \ref{fig:wind}).  Quantitatively, we find that for powerful winds

1394: with $p = 1$ (see eq. [\ref{outflow}]), the inner disk becomes

1395: advective at $t \sim 0.2$, $5$, and $100$ sec, for $\alpha = 0.1,

1396: 0.01,$ and $0.001$, respectively (for our fiducial model with an

1397: initial mass of $0.1 M_\odot$ and an initial radius of $\simeq 3

1398: \times 10^7$ cm).  Thus, for $\alpha \sim 10^{-3}$, the timescale for

1399: the inner disk to become advective is comparable to the onset of

1400: observed flaring at $\sim 30$ sec in some short GRBs (e.g., Berger et

1401: al. 2005).  Given the slow decline in disk mass with time before

1402: $t_{\rm thick}$, there is ample accretion energy available in the disk

1403: at this point to power the observed flaring.  However, there is

1404: observational evidence for $\alpha \sim 0.1$ in a number of

1405: astrophysical disks (King et al. 2007); we thus doubt that $t_{\rm

1406: thick}$ is large enough to coincide with the onset of observed

1407: flaring.  Instead $t_{\rm thick}$ is likely to be $\sim 0.1-1$ sec,

1408: comparable to the duration of the short GRB itself. In this case, the

1409: rapid decrease in the disk mass and accretion rate in the advective

1410: phase imply that the remnant accretion disk alone does not contain

1411: sufficient mass at $\sim 30$ sec to power the observed late-time

1412: activity from short GRBs, nor is there any physical reason to expect a

1413: sudden change in the disk or jet properties at this time.  

1414: 

1415: A more likely source of late-time flaring in compact object merger models is a continued inflow of mass at late

1416: times, such as is produced by the infalling tidal tail found in Lee \&

1417: Ramirez-Ruiz's (2007) NS-NS merger simulations (see also Rosswog 2007).  Similarly, the BH-NS merger simulations of Faber et al.~(2006a,b) show that $\sim 0.03 M_{\sun}$ of material is ejected into highly eccentric orbits during the merger, which returns to the BH on a timescale $\gtrsim 1$ s.  However, final conclusions regarding the quantity and ubiquity of late-time fall-back from NS-NS and BH-NS mergers must await full-GR simulations which include BH spin and realistic EOSs.

1418: 

1419: The second major focus of this paper has been on the composition of

1420: the accretion disk and its outflows as a function of time.  For

1421: initial disk properties expected in compact object mergers (\S

1422: \ref{sec:initial}), the disk typically comes into $\beta$-equilibrium

1423: given the high temperatures and densities at small radii.  As material

1424: spreads to larger radii, however, the composition of the disk freezes

1425: out before it becomes advective at late times; at freeze-out the disk

1426: is modestly neutron rich, with an electron fraction $Y_e \approx 0.3$ (\S \ref{sec:composition} and Fig. \ref{fig:Ye}).  This

1427: neutron rich material -- $\sim 10^{-2} M_\odot$ for typical initial

1428: disk parameters -- is blown away once the disk enters the advective

1429: phase at $\sim t_{\rm thick}$.  These outflows are particularly

1430: interesting given the low solar system abundance of material produced

1431: in nuclear statistical equilibrium at $Y_e \sim 0.3$ (Hartmann et

1432: al. 1985).  In a separate paper, we will study this nucleosynthesis

1433: and its implications in more detail.

1434: 

1435: Although outflows from compact object merger accretion disks are

1436: neutron rich in most circumstances, neutrino-driven winds from radii

1437: $\simeq 10^6-10^7$ cm at accretion rates $\dot M_{d} \sim 0.03-1 \,

1438: M_\odot \, {\rm s^{-1}}$ have electron fractions $Y_e \simeq 0.5$,

1439: precisely that required to synthesize significant amounts of $^{56}$Ni

1440: (Fig. \ref{fig:mdot_radius}). We have calculated the total Ni mass

1441: ejected by compact object merger disks as a function of their initial

1442: mass and radius (\S \ref{sec:ni} and Fig. \ref{fig:Nimass}). Disks

1443: with initial masses $\gtrsim 0.1 M_\odot$ can produce up to $\sim

1444: 10^{-3} M_\odot$ of $^{56}$Ni.  The radioactive decay of this Ni as

1445: the outflow expands to large radii will produce an optical and

1446: infrared transient peaking $\sim 0.5-2$ days after the merger, with a

1447: peak flux of $\nu L_\nu \simeq 10^{40}$ ergs s$^{-1}$

1448: (Fig. \ref{fig:lightcurve}).  Because the Ni mass is likely to be a

1449: small fraction of the total mass of the ejecta (most of which is

1450: neutron rich), this transient is best detected at $\sim 1 \, \mu$m.

1451: As Figure \ref{fig:lightcurve} shows, current observational limits on

1452: SN-like transients coincident with short GRBs are about a factor of

1453: $\sim 10$ above our predictions.  However, somewhat deeper limits from

1454: a moderately closer burst could start to put interesting constraints

1455: on short GRB progenitors.  It is also possible that the decay of some

1456: neutron-rich isotopes could heat the outflow and contribute to the

1457: late-time thermal emission (although most such isotopes have very

1458: short half-lives).  This possibility should be investigated in future

1459: calculations using a nuclear reaction network.

1460: 

1461: Although we have focused on short GRBs throughout this paper, many of

1462: our results can be applied more broadly.  For example, long duration

1463: GRBs show late-time activity and flaring similar to that seen in short

1464: GRBs (e.g., Falcone et al.~2007).  For the reasons described above, this activity is

1465: probably produced by a continued inflow of mass at late times

1466: (fallback from the stellar progenitor's envelope) rather than solely

1467: by the viscous evolution of the small-scale disk.  As a final

1468: application of our results, we note that the accretion-induced

1469: collapse of a white dwarf to a neutron star (AIC) is expected to

1470: produce a compact disk of $\sim 0.1-0.5 M_{\sun}$ outside the newly

1471: formed neutron star's surface (Dessart et al. 2006).  The calculations

1472: presented here describe the evolution of this remnant disk, with the

1473: one caveat that the composition of the disk in the AIC context may be

1474: strongly affected by neutrino irradiation from the newly-formed

1475: neutron star.

1476: 

1477: \section*{Acknowledgments}

1478: 

1479: We thank Josh Bloom, Davide Lazzati, and Daniel Perley for useful conversations.  A. L. P. is supported by the Theoretical Astrophysics Center at UC

1480: Berkeley.  B. D. M.  and E. Q. are supported in part by the David and

1481: Lucile Packard Foundation, NASA Grant NNG06GI68G, and a NASA GSRP

1482: Fellowship to B.D.M.

1483: 

1484: \begin{appendix}

1485: 

1486: \section{Calibration of the Ring Model}

1487: \label{appendix:ring}

1488: 

1489: The surface density $\Sigma$ of an axisymmetric disk in a Keplerian

1490: potential with constant total angular momentum evolves according to a

1491: diffusion equation (e.g., Frank et al.~2002): \be \frac{\partial

1492: \Sigma}{\partial t} = \frac{3}{r}\frac{\partial}{\partial

1493: r}\left[r^{1/2}\frac{\partial}{\partial r}\left(\nu\Sigma

1494: r^{1/2}\right)\right],

1495: \label{eq:sigma_evo}

1496: \ee where $\nu$ is the kinematic viscosity.  Assuming that $\nu$

1497: depends only on radius as a power law, viz: $\nu =

1498: \nu_{0}(r/R_{0})^{n}$, equation (\ref{eq:sigma_evo}) is linear and,

1499: for an initial surface density distribution $\Sigma(r,t=0) =

1500: (M_{0}/2\pi R_{0})\delta(r-R_{0})$ which is narrowly peaked about the

1501: radius $R_{0}$, the solution (for $n < 2$) is given by \bea \Sigma(r,t) = \nonumber \eea \be

1502: \frac{M_{0}(1-n/2)}{\pi

1503: R_{0}^{2}x^{(n+1/4)}\tau}\exp\left[\frac{-(1+x^{2-n})}{\tau}\right]I_{1/|4-2n|}\left[\frac{2x^{1-n/2}}{\tau}\right], \nonumber \\

1504: \label{eq:sigma}

1505: \ee where $M_{0}$ is the initial disk mass, $x \equiv r/R_{0}$, $\tau

1506: \equiv t[12\nu_{0}(1-n/2)^{2}/R_{0}^{2}]$, and $I_{m}$ is a modified

1507: Bessel function of order $m$.  For small argument $y \ll 1$,

1508: $I_{m}(y)$ takes the asymptotic form $I_{m} \simeq

1509: (y/2)^{m}/\Gamma(m+1)$, where $\Gamma$ is the Gamma function; thus,

1510: for late times or small radii such that $\tau \gg 2x^{1-n/2}$,

1511: equation (\ref{eq:sigma}) reduces to \bea \Sigma(r,t)|_{\tau \gg

1512: 2x^{1-n/2}} = \nonumber \eea \bea \frac{M_{0}}{\pi

1513: R_{0}^{2}}\frac{(1-n/2)}{\Gamma[\frac{5-2n}{4-2n}]}\frac{1}{\tau^{\left(\frac{5-2n}{4-2n}\right)}x^{n}}\exp\left[\frac{-(1+x^{2-n})}{\tau}\right]

1514: \label{eq:sigma_asym}

1515: \eea Most of the mass in the disk is located near the radius where the

1516: local mass $M_{d} \propto \Sigma r^{2}$ peaks; using equation

1517: (\ref{eq:sigma_asym}), at late times this radius is found to be

1518: $r_{\rm peak} = R_{0}\tau^{1/(2-n)}$.  Hence, equation

1519: (\ref{eq:sigma_asym}) becomes valid near $r_{\rm peak}$ for $\tau \gg

1520: 1$.

1521: 

1522: The constant $A$, which relates the total disk mass at late times from

1523: the exact solution of equation (\ref{eq:sigma_evo}) to the mass

1524: defined by $\pi\Sigma(r_{\rm peak})r_{\rm peak}^{2}$, can

1525: be calculated from equation (\ref{eq:sigma_asym}) to be \be A(\tau \gg

1526: 1) \equiv \left.\frac{\int_{0}^{\infty}2\pi \Sigma r

1527: dr}{\pi\Sigma(r_{\rm peak})r_{\rm peak}^{2}}\right|_{\tau \gg 1} =

1528: \frac{2e}{2-n}

1529: \label{eq:a_const}

1530: \ee Similarly, the constant $B$, which relates the total disk angular

1531: momentum at late times from the exact solution to that estimated by

1532: $\pi\Sigma r_{\rm peak}^{2}(GMr_{\rm peak})^{1/2}$, is

1533: given by \be B(\tau \gg 1) \equiv \left.\frac{\int_{0}^{\infty}2\pi

1534: \Sigma r^{3/2} dr}{\pi\Sigma(r_{\rm peak})r_{\rm

1535: peak}^{5/2}}\right|_{\tau \gg 1} =

1536: \frac{2e}{2-n}\Gamma\left[\frac{5-2n}{4-2n}\right]

1537: \label{eq:b_const}

1538: \ee 

1539: 

1540: From mass continuity, the radial velocity is given by

1541: \be

1542: v_{r} = \frac{-3}{\Sigma r^{1/2}}\frac{\partial}{\partial r}\left[\nu\Sigma r^{1/2}\right] = \frac{-3\nu_{0}}{R_{0}}\frac{1}{\Sigma x^{1/2}}\frac{\partial}{\partial x}\left[\Sigma x^{n+1/2}\right],

1543: \label{eq:velocity}

1544: \ee

1545: which, using equation (\ref{eq:sigma_asym}), gives the accretion rate at small radii

1546: \bea \dot{M}_{\rm in} &=& -2\pi\Sigma r v_{r}|_{\tau \gg 2x^{1-n/2}} \nonumber \\ &=& \frac{M_{0}}{R_{0}^{2}/\nu_{0}}\frac{3(1-n/2)}{\Gamma[(5-2n)/(4-2n)]}\exp[-1/\tau]\tau^{-\left(\frac{5-2n}{4-2n}\right)} \nonumber \\

1547: \label{eq:mdot_analytic}

1548: \eea

1549: Equation (\ref{eq:mdot_analytic}) is easily checked by noting that $\int_{0}^{\infty}\dot{M}_{\rm in}dt = M_{0}$, which shows that the entire initial disk eventually accretes onto the central object.  In $\S\ref{sec:dynamical}$ we introduced the following prescription for evolving the disk mass:

1550: \be

1551: \dot{M}_{d} = \frac{fM_{d}}{t_{\rm visc}},

1552: \label{eq:mdot_f}

1553: \ee where, in terms of the viscosity prescription adopted above,

1554: $t_{\rm visc} = r_{d}^{2}/\nu = t_{\rm visc,0}(r_{d}/R_{0})^{2-n}$ and

1555: $t_{\rm visc,0} \equiv R_{0}^{2}/\nu_{0}$ is the initial viscous time.

1556: Assuming that the total disk angular momentum remains constant, $J

1557: \propto M_{d}r_{d}^{1/2} = M_{0}R_{0}^{1/2}$, the solution to equation

1558: (\ref{eq:mdot_f}) is given by \be M_{d}(t) = M_{0}[1 +

1559: (4-2n)f(t/t_{\rm visc,0})]^{-1/(4-2n)}

1560: \label{eq:mdisk}

1561: \ee

1562: In our evolutionary calculations we set $f$ so that the accretion rate from the exact solution to equation (\ref{eq:sigma_evo}) ($\dot{M}_{\rm in}$; eq.~[\ref{eq:mdot_analytic}]) matches the solution to equation (\ref{eq:mdot_f}) at late times (i.e., in the self-similar limit).  This requires

1563: \be

1564: f = 3(1-n/2)\Gamma[(5-2n)/(4-2n)]^{4-2n}

1565: \label{eq:fspecial}

1566: \ee

1567: 

1568: For an advection-dominated disk, $\nu = \alpha c_{s}H \propto \Omega R^{2} \propto r^{1/2}$; thus, $n=1/2$, $f \simeq 1.602$, $A \simeq 3.62$,

1569: and $B \simeq 3.23$.  For a neutrino-cooled, optically-thin disk which

1570: is dominated by gas pressure, $T \propto r^{-3/10}$ and $\nu \propto

1571: r^{6/5}$; thus, $n=6/5$, $f \simeq 1.01$, $A \simeq 6.80$, and $B

1572: \simeq 6.09$.

1573: 

1574: In Figure \ref{fig:comparison} we show $\dot{M}_{\rm in}/\dot{M}_{d}$

1575: as a function of $t/t_{\rm visc,0}$ for $n=1/2$ in order to compare

1576: the disk evolution derived from the exact solution of equation

1577: (\ref{eq:sigma_evo}) to that calculated from our simplified model.

1578: Figure \ref{fig:comparison} also shows the ratio of the total disk

1579: mass $M_{\rm tot} \equiv \int_{0}^{\infty}2\pi \Sigma r dr$ calculated

1580: from equation (\ref{eq:sigma}) to the disk mass $M_{d}$

1581: (eq.~[\ref{eq:mdisk}]) of the simplified model, as well as the ratio

1582: of $r_{\rm peak}$ (the radius where $\Sigma r^{2}$ peaks, using

1583: eq.~[\ref{eq:sigma}] for $\Sigma$) to the radius determined by angular

1584: momentum conservation: $r_{d} = R_{0}(M_{d}/M_{0})^{2}$.  Figure

1585: \ref{fig:comparison} shows that, although the accretion rate in the

1586: two models differ at very early times (the initially

1587: narrowly-concentrated ring takes a short period of time to spread to

1588: small radii), they approach one another to $\lesssim 20\%$ by $t

1589: \gtrsim 0.1t_{\rm visc,0}$.  Likewise, the disk mass and radii from

1590: the exact solution and simplified model are quite similar at all

1591: times.

1592: 

1593: The numerical values for $A$ and $B$ given in equations

1594: (\ref{eq:a_const}) and (\ref{eq:b_const}) and employed in our

1595: calculations apply only to the mass and angular momentum distribution

1596: in the disk at late times ($\tau \gg 1$).  Initially, the disk is

1597: entirely concentrated at a single radius and $A(t = 0) = B(t = 0) =

1598: 1$; thus, $A(t)$ and $B(t)$ evolve significantly from early times

1599: until the disk enters the self-similar limit and so one might worry

1600: that the early-time description of the disk's evolution depends

1601: sensitively on the initial mass distribution.  Our model only assumes,

1602: however, that the $ratio$ $A(t)/B(t)$ remains constant, which is a

1603: good approximation.  To illustrate this, Figure \ref{fig:comparison}

1604: shows $A(t)/B(t)$ calculated from the exact solution

1605: (eq.~[\ref{eq:sigma}]) for $n = 1/2$.  Note that $A(t)/B(t)$ increases

1606: from unity to its asymptotic value $A/B = \Gamma[(5-2n)/(4-2n)]$,

1607: which is $\simeq 1.12$ for $n = 1/2$.

1608: 

1609: \begin{figure}

1610: %\epsscale{0.6}

1611: \resizebox{\hsize}{!}{\includegraphics[]{appendixA.eps}}

1612: \caption{Comparison of the accretion rate ($solid$), disk mass

1613: ($short$ $dashed$), and disk radius (where the local disk mass peaks;

1614: $dotted$) as calculated from our simplified ring model to that derived

1615: from the exact solution of the diffusion equation for a

1616: $\delta-$function initial mass distribution (eq.~[\ref{eq:sigma}]); we

1617: assume $\nu \propto r^{1/2}$, as applies for a thick disk.  The

1618: parameter $f \simeq 1.6$ (eq.~[\ref{eq:fspecial}]) adopted in our

1619: model is chosen to ensure that the accretion rates match at late times

1620: (i.e., $\dot{M}_{\rm in}/\dot{M}_{d} \rightarrow 1$).  Also shown is

1621: the ratio $A(t)/B(t)$ (eqs.~[\ref{eq:a_const}] and

1622: [\ref{eq:b_const}]), a measure of the relative distribution of mass

1623: and angular momentum, which asymptotes to $\Gamma[(5-2n)/(4-2n)]

1624: \simeq 1.12$ at late times.}

1625: \label{fig:comparison}

1626: %\epsscale{1.0}

1627: \end{figure}

1628: 

1629: \section{Analytic Self-Similar Solutions}

1630: \label{sec:analytic}

1631: 

1632: The late-time evolution of our disk calculations asymptote to power

1633: laws that are well approximated by analytic self-similar solutions.

1634: We derive these here to aid in interpreting our numerical

1635: results. Presentation is divided between neutrino-cooled, thin-disk

1636: solutions and late-time advective solutions. One could just as well

1637: derive analogous results for disks that are optically thick to

1638: neutrinos. We forgo this here since the initial viscous time is always

1639: sufficiently long that these solutions are never applicable to our

1640: numerical results.  We conclude by presenting self-similar solutions

1641: for advective disks with substantial mass loss, since these differ

1642: significantly from the solutions without mass loss.

1643: 

1644: \subsection{Neutrino-cooled, Thin-disk Solutions}

1645: 

1646: In the neutrino-cooled, thin-disk limit, the cooling is dominated by

1647: Urca, and the pressure is given by ideal gas. Combining local energy

1648: balance and continuity, $\dot{M}_{d}=fA\pi\nu\Sigma$, allows us to

1649: solve for the temperature and column density as functions of radius.

1650: We substitute these into the angular momentum equation,

1651: $B(GMr_d)^{1/2}\pi r_d^2\Sigma=J_d$, to solve for $M_d$ as a function

1652: of $\dot{M}_{d}$ and $J_d$. We then assume the solutions have a

1653: self-similar form of $M_d\propto t^{-\beta}$, so that

1654: $\dot{M}_{d}=-dM_d/dt=\beta M_d/t$. In this way we solve for

1655: $\beta=5/8$, $\dot{M}_{d}(t)$, and subsequently any other variable of

1656: interest. The results are \bea M_d= \nonumber \eea \bea 1.3\times10^{-2}

1657: f_{1.6}^{-5/8}\left(\frac{A_{3.6}}{B_{3.2}}\right)\alpha_{0.1}^{-3/4}M_3^{-1/4}

1658: \left(\frac{J_{49}}{2}\right) t^{-5/8} M_\odot,

1659: 	\label{eq:md_analytic1}

1660: \eea

1661: \bea

1662: 	\dot{M}_{d}= \nonumber

1663: \eea \bea

1664: 2.7\times10^{-2}f_{1.6}^{-5/8}\left(\frac{A_{3.6}}{B_{3.2}}\right)\alpha_{0.1}^{-3/4}M_3^{-1/4}

1665: 		\left(\frac{J_{49}}{2}\right) t^{-13/8}M_\odot{\rm s^{-1}}

1666: 	\label{eq:mdot_analytic1}

1667: \eea

1668: and

1669: \be

1670: 	r_d=4.1\times10^8f_{1.6}^{5/4}\alpha_{0.1}^{3/2}M_3^{-1/2}

1671: 	t^{5/4}{\rm cm}.

1672: 	\label{eq:rd_analytic1}

1673: \ee where $f_{1.6}=f/1.6$, $A_{3.6}=A/3.6$, $B_{3.2}=B/3.2$, and $t$

1674: is measured in seconds, and the prefactors have been scaled to match

1675: our numerical results.  The first thing to notice is that both $M_d$

1676: and $\dot{M}_{d}$ are rather insensitive to the choice of $f$ as long

1677: as it is near unity, and $A$ and $B$ only appear as a ratio, which is

1678: also nearly unity. This provides confidence in using this

1679: parameterization, and these specific values for the corresponding

1680: parameters, when the disk is not well-described by $n=1/2$. This

1681: analysis also demonstrates the relative dependence on $\alpha$.  In

1682: Figure \ref{fig:analytic} we compare these scaling ({\it dotted

1683: lines}) with the numerical calculations. This shows that these

1684: solutions are only applicable for a short time.  At times when

1685: $t<t_{\rm visc}$ the evolution is much flatter and is dominated by

1686: initial conditions. At later times the disk becomes advective and the

1687: solutions of the next section apply.

1688: 

1689: \begin{figure}

1690: %\epsscale{0.6}

1691: \resizebox{\hsize}{!}{\includegraphics[ ]{analytic.eps}}

1692: \caption{Comparison of the numerical disk solutions ({\it solid lines}) with

1693: the analytic solutions for the thin, neutrino-cooled ({\it dotted lines})

1694: and thick, advective limits ({\it dashed lines}). The numerical solution is the

1695: $0.3M_\odot$ disk from Fig. \ref{fig:mass1}.}

1696: \label{fig:analytic}

1697: %\epsscale{1.0}

1698: \end{figure}

1699: 

1700: \subsection{Late-time Advective Solutions}

1701: 

1702: \label{sec:riaf}

1703: 

1704: In this limit, self-similar solutions can be found in an analogous way. The viscous

1705: energy release is carried by advection with the internal energy dominated

1706: by relativistic particles, so that

1707: \be

1708: 	\frac{9}{8fA\pi}\Omega^2\dot{M}=V_r\frac{H}{r}\frac{11}{6}aT^4.

1709: 	\label{eq:appendix_energy}

1710: \ee Combining this with mass continuity, gives the column depth as a

1711: function of radius, $\Sigma(r)=(16/9A\pi\alpha)(\dot{M}/r^2\Omega)$.

1712: We then use this relation with $B(GMr_d)^{1/2}M_d=J_d$ and

1713: $\dot{M}_{d}=\beta M_d/t$, to find $\beta=1/3$ and the self-similar

1714: solutions \be M_d =

1715: 3.7\times10^{-2}\left(\frac{A_{3.6}}{B_{3.2}}\right)\alpha_{0.1}^{-1/3}M_3^{-2/3}

1716: \left(\frac{J_{49}}{2}\right) t^{-1/3}M_\odot,

1717: 	\label{eq:md_analytic2}

1718: \ee

1719: \be

1720: 	\dot{M}_{d}=1.2\times10^{-2}\left(\frac{A_{3.6}}{B_{3.2}}\right)\alpha_{0.1}^{-1/3}M_3^{-2/3}

1721: 		\left(\frac{J_{49}}{2}\right) t^{-4/3}M_\odot{\rm s^{-1}},

1722: 	\label{eq:mdot_analytic2}

1723: \ee

1724: and

1725: \be

1726: 	r_d=2.3\times10^8\alpha_{0.1}^{2/3}M_3^{1/3}

1727: 	t^{2/3}{\rm cm}.

1728: 	\label{eq:rd_analytic2}

1729: \ee

1730: These advective results are even more insensitive to $A$, $B$, and $f$ than the

1731: thin-disk results. Equation (\ref{eq:md_analytic2})-(\ref{eq:rd_analytic2})

1732: are plotted in Figure \ref{fig:analytic} as dashed lines. The numerical calculations follow

1733: these solutions very closely for times later than $t_{\rm thick}$

1734: (given by eq. [\ref{eq:tthick}]).

1735: 

1736:    Equations (\ref{eq:md_analytic2})-(\ref{eq:rd_analytic2}) can also

1737: be derived ignoring equation (\ref{eq:appendix_energy}), but assuming

1738: that the scaleheight is fixed at $H/r\simeq0.6$. This introduces the

1739: additional dependencies $M_d\propto(H/r)^{-2/3}$,

1740: $\dot{M}_{d}\propto(H/r)^{-2/3}$, and $r_d\propto(H/r)^{4/3}$, but

1741: gives nearly identical prefactors.

1742: 

1743: \subsection{Advective Solutions with Mass Loss}

1744: 

1745: \label{sec:riaf}

1746: 

1747: In \S \ref{thickdiskwinds} we described how advective disks are likely

1748: to lose a substantial fraction of their mass to viscously driven

1749: outflows.  Because the outflow removes angular momentum as well -- at

1750: least the specific angular momentum of the mass that is lost -- the

1751: disk need not expand as rapidly to large radii.  In addition, the disk

1752: mass and accretion rate decrease much more rapidly at late times than

1753: in the self-similar solutions described in the previous subsection.

1754: To quantify this effect, we follow Blandford \& Begelman (1999) and

1755: assume that only a fraction $\sim (r_*/r_d)^p$ of the available

1756: material is accreted onto the central BH.  The remainder is lost to an

1757: outflow.  Thus the outflow rate at any time is given by \be \dot

1758: M_{\rm out} = \left( 1 - \left[r_* \over r_d\right]^p\right) \, {f M_d \over

1759: t_{\rm visc}} \label{outflow} \ee We further assume that the angular

1760: momentum loss rate from the disk is given by \be \dot J = - C \dot

1761: M_{\rm out} \left( G M r_d\right)^{1/2}

1762: \label{ang}.  \ee where $C$ is a constant that depends on the torque

1763: exerted by the outflowing mass on the remaining disk.  If the outflow

1764: produces no net torque, an assumption that appears at least

1765: qualitatively consistent with the relatively small-scale magnetic

1766: fields seen in global MHD disk simulations (e.g., Stone \& Pringle

1767: 2001), then the angular momentum loss is only that due to the specific

1768: angular momentum of the outflow, and (Kumar, Narayan, \& Johnson 2008)

1769: \be C = {2p \over 2p + 1} \label{C}.  \ee We solve equations

1770: (\ref{eq:mass}), (\ref{eq:ang_mom}), (\ref{outflow}), and (\ref{ang}),

1771: assuming $A/B = 1$ and $\nu \propto r^{1/2}$ (as appropriate for a

1772: thick disk).  The solution depends on the relative magnitude of $1-C$

1773: and $C(r_*/r_d)^p$.  For $C(r_*/r_d)^p \ll 1-C$, which is true at

1774: nearly all times if equation (\ref{C}) is applicable, then \be r_d

1775: \simeq r_{d,0} \left[1 + 3 f (1-C)\left({t \over t_{\rm

1776: visc,0}}\right)\right]^{2/3}, \label{rdCne1} \ee \be M_d \simeq M_{d,0}

1777: \left[1 + 3 f (1-C) \left({t \over t_{\rm visc,0}}\right)\right]^{-1/ 

1778: [3(1-C)]},

1779: \label{MdCne1}\ee and \bea \dot M_{\rm in} \simeq {f {M_{d,0} \over t_{\rm visc,0}}}\left(r_* \over r_{d,0}\right)^p\times \nonumber \eea

1780: \bea 

1781: \left[1 + 3 f (1-C) \left({t \over t_{\rm visc,0}}\right)\right]^{-[1

1782: + 3(1+2p/3)(1-C)]/[3(1-C)]}

1783: \label{mdotinCne1} \eea Note that if $p = C = 0$ (i.e., no mass or

1784: angular momentum loss), then these self-similar solutions reduce to

1785: those of the previous subsection.  However, for the case $p = 1$

1786: consistent with a number of global advective disk simulations (e.g.,

1787: Hawley \& Balbus 2002), and in the absence of a net torque on the

1788: disk, $C = 2/3$ and these solutions correspond to $r_d \propto

1789: t^{2/3}$, $M_d \propto t^{-1}$, and $\dot M_{\rm in} \propto t^{-8/3}$

1790: (see also Fig. \ref{fig:wind}).  This shows that the disk mass and

1791: accretion rate decrease subsantially more rapidly in time than in the

1792: absence of an outflow, while the disk expands outward at roughly the

1793: same rate.  If there is a net torque on the disk such that $C \simeq

1794: 1$, then equations (\ref{rdCne1})-(\ref{mdotinCne1}) are not

1795: applicable.  Instead, for $C(r_*/r_d)^p \gg 1-C$, the solution is

1796: given by (for $p \ne 0$ and $t \gg t_{\rm visc,0}$) \be r_d(t) \simeq

1797: \left[(3 + 2p) f r_*^p r_{d,0}^{1.5}\right]^{1/(1.5 + p)} \, \left( t

1798: \over t_{\rm visc,0}\right)^{1/(1.5 + p)} \label{rd} \ee and \be M_d(t)

1799: \simeq M_{d,0} \exp[-D (t/t_{\rm visc,0})^{p/(1.5 + p)}]

1800: \label{Md} \ee

1801: where \be D = \left( {1.5 + p \over p \, (3 + 2 p)^{1.5/(1.5 + p)}} \right)

1802: \left( {f r_{d,0}^{1.5} \over [f r^p_{*} r_{d,0}^{1.5}]^{1.5/(1.5 +

1803: p)}}\right).  \ee For $p = 1$ and for $r_{d,0} \sim r_*$, these solutions

1804: become $r_d(t) \sim r_{d,0} (t/t_{\rm visc,0})^{2/5}$ and $M_d(t) \sim

1805: M_{d,0} \exp[-1.15 (t/t_{\rm visc,0})^{2/5}]$.  The radius of the disk

1806: thus increases significantly more slowly, and the mass of the disk

1807: decreases much more rapidly, than in the self-similar solutions

1808: without mass-loss.

1809: 

1810: The numerical solutions including mass-loss during the advective phase

1811: shown in \S 4.1 (Fig. \ref{fig:wind}) assume that equation (\ref{C})

1812: is applicable and are indeed well-described by the self-similar

1813: solutions given in equations (\ref{rdCne1})-(\ref{mdotinCne1}) at late

1814: times.

1815: 

1816: 

1817: \end{appendix}

1818: 

1819: 

1820: \begin{thebibliography}{99}

1821: 

1822: \bibitem[Barthelmy et al.(2005)]{bar05}

1823: Barthelmy, S. D., et al. 2005, Nature, 438, 994 

1824: 

1825: \bibitem[Barzilay 

1826: \& Levinson(2008)]{2008NewA...13..386B} Barzilay, Y., \& Levinson, A.\ 2008, New Astronomy, 13, 386 

1827: 

1828: \bibitem[Baumgarte et al.(2000)]{2000ApJ...528L..29B} Baumgarte, T.~W., 

1829: Shapiro, S.~L., \& Shibata, M.\ 2000, \apjl, 528, L29 

1830: 

1831: \bibitem[Beloborodov(2003a)]{2003ApJ...588..931B} Beloborodov, A.~M.\ 2003, 

1832: \apj, 588, 931 

1833: 

1834: \bibitem[Beloborodov(2003b)]{2003ApJ...585L..19B} Beloborodov, A.~M.\ 2003b, 

1835: \apjl, 585, L19 

1836: 

1837: \bibitem[Berger et al.(2005)]{ber05}

1838: Berger, E., et al, 2005, Nature 438, 988

1839: 

1840: \bibitem[Bildsten \& Cutler(1992)]{bc92}

1841: Bildsten, L. \& Cutler, C. 1992, ApJ, 400, 175 

1842: 

1843: \bibitem[Blandford \& Begelman(1999)]{1999MNRAS.303L...1B} Blandford, R.~D., \& Begelman, M.~C.\ 1999, \mnras, 303, L1 

1844: 

1845: \bibitem[Bloom et al.(2006)]{blo06}

1846: Bloom, J.~S., et al.\ 2006, \apj, 638, 354

1847: 

1848: \bibitem[Campana et al.(2006)]{cam06}

1849: Campana, S., et al. 2006, \aap, 454, 113

1850: 

1851: \bibitem[Chen \& Beloborodov(2007)]{cb07}

1852: Chen, W. X. \& Beloborodov, A. M. 2007, \apj, 657, 383 $(CB07)$

1853: 

1854: \bibitem[Colgate et al.(1980)]{1980ApJ...237L..81C} Colgate, S.~A., 

1855: Petschek, A.~G., \& Kriese, J.~T.\ 1980, \apjl, 237, L81 

1856: 

1857: \bibitem[Dai et al.(2006)]{dai06}

1858: Dai, Z. G., et al. 2006, Science, 311, 1127

1859: 

1860: \bibitem[Daigne 

1861: \& Mochkovitch(2002)]{2002A&A...388..189D} Daigne, F., \& Mochkovitch, R.\ 2002, \aap, 388, 189 

1862: 

1863: \bibitem[Derishev et al.(1999)]{1999ApJ...521..640D} Derishev, E.~V., 

1864: Kocharovsky, V.~V., \& Kocharovsky, V.~V.\ 1999, \apj, 521, 640 

1865: 

1866: \bibitem[Dessart et al.(2006)]{2006ApJ...644.1063D} Dessart, L., Burrows, 

1867: A., Ott, C.~D., Livne, E., Yoon, S.-C., 

1868: \& Langer, N.\ 2006, \apj, 644, 1063 

1869: 

1870: \bibitem[Dessart et al.(2008)]{2008ApJ...673L..43D} Dessart, L., Burrows, 

1871: A., Livne, E., \& Ott, C.~D.\ 2008a, \apjl, 673, L43 

1872: 

1873: \bibitem[Dessart et al.(2008)]{2008arXiv0806.4380D} Dessart, L., Ott, C., 

1874: Burrows, A., Rosswog, S., 

1875: \& Livne, E.\ 2008b, ArXiv e-prints, 806, arXiv:0806.4380 

1876: 

1877: \bibitem[DiMatteo et al.(2002)]{dim02}

1878: DiMatteo, T., Perna, R., \& Narayan, R. 2002, \apj, 579, 706 

1879: 

1880: \bibitem[Duez et al.(2006)]{2006PhRvD..73j4015D} Duez, M.~D., Liu, Y.~T., 

1881: Shapiro, S.~L., Shibata, M., \& Stephens, B.~C.\ 2006, Phys. Rev. D, 73, 104015 

1882: 

1883: \bibitem[Duez et al.(2004)]{2004PhRvD..69j4030D} Duez, M.~D., Liu, Y.~T., 

1884: Shapiro, S.~L., \& Stephens, B.~C.\ 2004, Phys. Rev. D, 69, 104030 

1885: 

1886: \bibitem[Eichler et al.(1989)]{eic89}

1887: Eichler, D., Livio, M., Piran, T., \& Schramm, D. N. 1989, Nature, 340, 126

1888: 

1889: \bibitem[Faber et al.(2006a)]{fab06a}

1890: Faber, J. A., et al. 2006a, Phys. Rev. D, 73, 4012 

1891: 

1892: \bibitem[Faber et al.(2006b)]{fab06b}

1893: Faber, J. A., et al. 2006b, \apj, 641, L93 

1894: 

1895: \bibitem[Falcone et al.(2007)]{2007ApJ...671.1921F} Falcone, A.~D., et al.\ 

1896: 2007, \apj, 671, 1921 

1897: 

1898: \bibitem[Ferrero et al.(2007)]{fer07}

1899: Ferrero, P., et al.\ 2007, AJ, 134, 2118

1900: 

1901: \bibitem[Fishbone(1973)]{fis73}

1902: Fishbone, L. G. 1973, \apj, 185, 43

1903: 

1904: \bibitem[Fox et al.(2005)]{2005Natur.437..845F} Fox, D.~B., et al.\ 2005, 

1905: \nat, 437, 845 

1906: 

1907: \bibitem[Frank et al.(2002)]{fra02}

1908: Frank, J., King, A. R., \& Raine, D. J. 2002, Accretion Power in Astrophysics 

1909: (3d ed.; Cambridge: Cambridge Univ. Press)

1910: 

1911: \bibitem[Freiburghaus et al.(1999)]{1999ApJ...525L.121F} Freiburghaus, C., 

1912: Rosswog, S., \& Thielemann, F.-K.\ 1999, \apjl, 525, L121 

1913: 

1914: \bibitem[Galama et al.(1998)]{gal98}

1915: Galama, T. J., et al. 1998, Nature, 385, 670

1916: 

1917: \bibitem[Gehrels et al.~(2005)]{ger05}

1918: Gehrels, N. et al, 2005, Nature, 437, 851

1919: 

1920: \bibitem[Haensel \& Zdunik(1990a)]{hz90a}

1921: Haensel, P., \& Zdunik, J.~L.\ 1990a, \aap, 227, 431

1922: 

1923: \bibitem[Haensel \& Zdunik(1990b)]{hz90b}

1924: Haensel, P., \& Zdunik, J.~L.\ 1990b, \aap, 229, 117

1925: 

1926: \bibitem[Hartmann et al.(1985)]{1985ApJ...297..837H} Hartmann, D., Woosley, 

1927: S.~E., \& El Eid, M.~F.\ 1985, \apj, 297, 837

1928: 

1929: \bibitem[Hawley \& Balbus(2002)]{hb02} Hawley, J.~F., \& Balbus,

1930: S.~A.\ 2002, \apj, 573, 738

1931: 

1932: \bibitem[Hjorth et al.(2003)]{hjo03}

1933: Hjorth, J., et al. 2003, \nat, 423, 847

1934: 

1935: \bibitem[Hjorth et al.(2005)]{2005ApJ...630L.117H} Hjorth, J., et al.\ 

1936: 2005, \apjl, 630, L117 

1937: 

1938: \bibitem[Janiuk et al.(2004)]{jan04}

1939: Janiuk, A., Perna, R., DiMatteo, T., \& Czerny, B. 2004, \mnras, 355, 950 

1940: 

1941: \bibitem[Janka et al.(1999)]{jan99}

1942: Janka, H.-Th., Eberl, T., Ruffert, M., \& Fryer, C. L. 1999, ApJ, 527, L39 

1943: 

1944: \bibitem[King et al.(2005)]{kin05}

1945: King, A. R., et al. 2005, \apj, 630, L113 

1946: 

1947: \bibitem[King et al.(2007)]{kin07} King, A.~R., Pringle, J.~E., \& Livio, M.\ 2007, \mnras, 376, 1740

1948: 

1949: \bibitem[Klu\'{z}niak \& Lee(1998)]{kl98}

1950: Klu\'{z}niak, W. \& Lee, W. H. 1998, ApJ, 494, L53 

1951: 

1952: \bibitem[Kohri \& Mineshige(2002)]{km02}

1953: Kohri, K. \& Mineshige, S. 2002,\apj, 577, 311 

1954: 

1955: \bibitem[Kopal(1959)]{kop59}

1956: Kopal, Z. 1959, Close Binary Systems (London: Chapman \& Hall)

1957: 

1958: \bibitem[Krolik et al.(2007)]{2007ApJ...664.1045K} Krolik, J.~H., Hirose, 

1959: S., \& Blaes, O.\ 2007, \apj, 664, 1045 

1960: 

1961: \bibitem[Kulkarni(2005)]{2005astro.ph.10256K} Kulkarni, S.~R.\ 2005, ArXiv 

1962: Astrophysics e-prints, arXiv:astro-ph/0510256 

1963: 

1964: \bibitem[Kumar et al.(2008)]{2008arXiv0807.0441K} Kumar, P., Narayan,  

1965: R.,  \& Johnson, J.~L.\ 2008, ArXiv e-prints, 807, arXiv:0807.0441

1966: 

1967: \bibitem[Lai et al.(1994)]{lai94}

1968: Lai, D., Rasio, F. A., \& Shapiro, S. L. 1994, ApJ, 423, 344

1969: 

1970: \bibitem[La Parola et al.(2006)]{lap06}

1971: La Parola, V., et al. 2006, \aap, 454, 753

1972: 

1973: \bibitem[Lattimer \& Schramm(1974)]{1974ApJ...192L.145L}

1974: Lattimer, J.~M., \& Schramm, D.~N.\ 1974, \apjl, 192, L145 

1975: 

1976: \bibitem[Lazzati et al.(2001)]{laz01}

1977: Lazzati, D., Ramirez-Ruiz, E., \& Ghisellini, G. 2001, \aap, 379, L39

1978: 

1979: \bibitem[Lazzati et al.(2008)]{2008arXiv0805.0138L} Lazzati, D., Perna, R., 

1980: \& Begelman, M.~C.\ 2008, ArXiv e-prints, 805, arXiv:0805.0138 

1981: 

1982: \bibitem[Lee \& Klu\'{z}niak(1995)]{lk95}

1983: Lee, W. H. \& Klu\'{z}niak, W. 1995, Acta Astron., 45, 705

1984: 

1985: \bibitem[Lee \& Klu\'{z}niak(1998)]{lk98}

1986: Lee, W. H. \& Klu\'{z}niak, W. 1998, ApJ, 526, 178

1987: 

1988: \bibitem[Lee \& Klu\'{z}niak(1999)]{lk99}

1989: Lee, W. H. \& Klu\'{z}niak, W. 1999, MNRAS, 308, 780 

1990: 

1991: \bibitem[Lee et al.(2004)]{2004ApJ...608L...5L} Lee, W.~H., Ramirez-Ruiz, 

1992: E., \& Page, D.\ 2004, \apjl, 608, L5 

1993: 

1994: \bibitem[Lee \& Ramirez-Ruiz(2007)]{lr07}

1995: Lee, W. \& Ramirez-Ruiz, E. 2007, New J. Phys., 9, 17 

1996: 

1997: \bibitem[Lee et al.(2004)]{lee04}

1998: Lee, W. H., Ramirez-Ruiz, E., \& Page, D. 2004, \apj, 608, L5 

1999: 

2000: \bibitem[Lee et al.(2005b)]{lee05b}

2001: Lee, W. H., Ramirez-Ruiz, E., \& Page, D. 2005b, \apj, 632, 421

2002: 

2003: \bibitem[Lemoine(2002)]{2002A&A...390L..31L} Lemoine, M.\ 2002, \aap, 390, L31 

2004: 

2005: \bibitem[Levinson(2006)]{2006ApJ...648..510L} Levinson, A.\ 2006, \apj, 

2006: 648, 510 

2007: 

2008: \bibitem[Li \& Paczy{\'n}ski(1998)]{1998ApJ...507L..59L} Li, L.-X., \& Paczy{\'n}ski, B.\ 1998, \apjl, 507, L59 

2009: 

2010: \bibitem[McKinney(2005)]{2005astro.ph..6368M} McKinney, J.~C.\ 2005, ArXiv 

2011: Astrophysics e-prints, arXiv:astro-ph/0506368 

2012: 

2013: \bibitem[Metzger et al.(2008)]{2008ApJ...676.1130M} Metzger, B.~D., 

2014: Thompson, T.~A., \& Quataert, E.\ 2008b, \apj, 676, 1130 $(M08B)$

2015: 

2016: \bibitem[Montanari et al.(2005)]{mon05}

2017: Montanari, E., et al. 2005, \apj, 625, L17

2018: 

2019: \bibitem[Morrison et al.(2004)]{2004ApJ...610..941M} Morrison, I.~A., 

2020: Baumgarte, T.~W., \& Shapiro, S.~L.\ 2004, \apj, 610, 941 

2021: 

2022: \bibitem[Nakar(2007)]{2007PhR...442..166N} Nakar, E.\ 2007, Phys.~Rep., 442, 

2023: 166 

2024: 

2025: \bibitem[Narayan et al.(2001)]{nar01}

2026: Narayan, R., Piran, T., \& Kumar, P. 2001, \apj, 557, 949 

2027: 

2028: \bibitem[Narayan et al.(1991)]{nar91}

2029: Narayan, R., Piran, T., \& Shemi, A. 1991, ApJ, 379, L17

2030: 

2031: \bibitem[Oechslin \& Janka(2006)]{oj06}

2032: Oechslin, R., \& Janka, H.-Th. 2006, MNRAS, 368, 1489

2033: 

2034: \bibitem[Paczy\'{n}ski(1986)]{pac86}

2035: Paczy\'{n}ski, B. 1986, ApJ, 308, L43

2036: 

2037: \bibitem[Paczy\'{n}ski(1991)]{pac91}

2038: Paczy\'{n}ski, B. 1991, Acta Astron., 41, 257

2039: 

2040: \bibitem[Perna et al.(2005)]{per05}

2041: Perna, R., Armitage, P. J., \& Zhang, B. 2005, \apj, 636, L29

2042: 

2043: \bibitem[Pethick \& Ravenhall (1995)]{pr95}

2044: Pethick, C. J., \& Ravenhall, D. G. 1995, Annu. Rev. Nucl. Part. Sci., 45, 429

2045: 

2046: \bibitem[Piro \& Pfahl(2007)]{pp07}

2047: Piro, A. L. \& Pfahl, E. 2007, \apj, 658, 1173

2048: 

2049: \bibitem[Popham et al.(1999)]{pop99}

2050: Popham, R., Woosley, S. E., \& Fryer, C. L. 1999, \apj, 518, 356

2051: 

2052: \bibitem[Price 

2053: \& Rosswog(2006)]{2006Sci...312..719P} Price, D.~J., \& Rosswog, S.\ 2006, Science, 312, 719 

2054: 

2055: \bibitem[Proga \& Begelman(2003)]{2003ApJ...592..767P} Proga, D., \& Begelman, M.~C.\ 2003, \apj, 592, 767

2056: 

2057: \bibitem[Proga \& Zhang(2006)]{pz06}

2058: Proga, D. \& Zhang, B. 2006, \mnras, 370, L61

2059: 

2060: \bibitem[Pruet et al.(2004)]{2004ApJ...606.1006P} Pruet, J., Thompson, 

2061: T.~A., \& Hoffman, R.~D.\ 2004, \apj, 606, 1006 

2062: 

2063: \bibitem[Qian(2000)]{2000ApJ...534L..67Q} Qian, Y.-Z.\ 2000, \apjl, 534, 

2064: L67

2065: 

2066: \bibitem[Qian et al.(1993)]{1993PhRvL..71.1965Q} Qian, Y.-Z., Fuller, 

2067: G.~M., Mathews, G.~J., Mayle, R.~W., Wilson, J.~R., 

2068: \& Woosley, S.~E.\ 1993, Physical Review Letters, 71, 1965 

2069: 

2070: \bibitem[Qian 

2071: \& Woosley(1996)]{1996ApJ...471..331Q} Qian, Y.-Z., \& Woosley, S.~E.\ 1996, \apj, 471, 331 $(QW96)$

2072: 

2073: \bibitem[Rasio et al.(2005)]{ras05}

2074: Rasio, F. A., et al. 2005, Proc. JGRG14 Yukawa Institute for Theoretical Physics

2075: (preprint astro-ph/0503007)

2076: 

2077: \bibitem[Rasio \& Shapiro(1994)]{rs94}

2078: Rasio, F. A., \& Shapiro, S. L. 1994, ApJ, 432, 242

2079: 

2080: \bibitem[Remillard 

2081: \& McClintock(2006)]{2006ARA&A..44...49R} Remillard, R.~A., \& McClintock, J.~E.\ 2006, ARA\&A, 44, 49 

2082: 

2083: \bibitem[Rossi et al.(2006)]{2006MNRAS.369.1797R} Rossi, E.~M., 

2084: Beloborodov, A.~M., \& Rees, M.~J.\ 2006, \mnras, 369, 1797

2085: 

2086: \bibitem[Rosswog(2007)]{2007MNRAS.376L..48R} Rosswog, S.\ 2007, \mnras, 

2087: 376, L48 

2088: 

2089: \bibitem[Rosswog \& Liebend{\"o}rfer(2003)]{2003MNRAS.342..673R}

2090: Rosswog, S., \& Liebend{\"o}rfer, M.\ 2003, \mnras, 342, 673 

2091: 

2092: \bibitem[Rosswog et al.(2004)]{ros04}

2093: Rosswog, S., Speith, R., \& Wynn, G. A., 2004, MNRAS, 351, 1121

2094: 

2095: \bibitem[Ruffert \& Janka(1999)]{rj99}

2096: Ruffert, M., \& Janka, H.-Th. 1999, \aap, 344, 573

2097: 

2098: \bibitem[Ruffert \& Janka(2001)]{rj01}

2099: Ruffert, M., \& Janka, H.-Th. 2001, \aap, 380, 544 

2100: 

2101: \bibitem[Ruffert et al.(1996)]{ruf96}

2102: Ruffert, M., Janka, H.-Th., \& Sch\"{a}fer, G. 1996, \aap, 311, 532

2103: 

2104: \bibitem[Ruffert et al.(1997)]{ruf97}

2105: Ruffert, M., Janka, H.-Th., Takahashi, K., \& Sch\"{a}fer, G. 1997, \aap, 319, 122

2106: 

2107: \bibitem[Setiawan et al.(2004)]{set04}

2108: Setiawan, S., Ruffert, M., \& Janka, H.-Th. 2004, \mnras, 352, 753

2109: 

2110: \bibitem[Setiawan et al.(2006)]{set06}

2111: Setiawan, S., Ruffert, M., \& Janka, H.-Th. 2006, \aap, 458, 553 

2112: 

2113: \bibitem[Shakura \& Sunyaev(1973)]{ss73}

2114: Shakura, N. I., \& Sunyaev, R. A. 1973, A\&A, 24, 337 

2115: 

2116: \bibitem[Shapiro \& Teukolsky(1983)]{st83}

2117: Shapiro, S. L., \& Teukolsky, S. A. 1983, Black Holes, White Dwarfs, and 

2118: Neutron Stars: The Physics of Compact Objects (New York: Wiley) 

2119: 

2120: \bibitem[Shibata et al.(2005)]{2005PhRvD..71h4021S} Shibata, M., Taniguchi, 

2121: K., \& Ury{\= u}, K.\ 2005, Phys. Rev. D, 71, 084021 

2122: 

2123: \bibitem[Shibata \& Taniguchi(2006)]{st06}

2124: Shibata, M. \& Taniguchi, K. 2006, Phys. Rev. D, 73, 064027

2125: 

2126: \bibitem[Soderberg et al.(2006)]{sod06}

2127: Soderberg, A.~M., et  al.\ 2006, \apj, 650, 261

2128: 

2129: \bibitem[Stanek et al.(2003)]{sta03}

2130: Stanek, K. Z., et al. 2003, \apj, 591, L17

2131: 

2132: \bibitem[Stone \& Pringle(2001)]{2001MNRAS.322..461S} Stone, J.~M., \& Pringle, J.~E.\ 2001, \mnras, 322, 461 

2133: 

2134: \bibitem[Surman et al.(2006)]{2006ApJ...643.1057S} Surman, R., McLaughlin, 

2135: G.~C., \& Hix, W.~R.\ 2006, \apj, 643, 1057 

2136: 

2137: \bibitem[Surman et al.(2008)]{2008arXiv0803.1785S} Surman, R., McLaughlin, 

2138: G.~C., Ruffert, M., Janka, H.~-., 

2139: \& Hix, W.~R.\ 2008, ArXiv e-prints, 803, arXiv:0803.1785 

2140: 

2141: \bibitem[Thompson et al.(2001)]{2001ApJ...562..887T} Thompson, T.~A., 

2142: Burrows, A., \& Meyer, B.~S.\ 2001, \apj, 562, 887 

2143: 

2144: \bibitem[Thompson et al.(2004)]{2004ApJ...611..380T} Thompson, T.~A., 

2145: Chang, P., \& Quataert, E.\ 2004, \apj, 611, 380 

2146: 

2147: \bibitem[Ury\={u} \& Eriguchi(1999)]{ue99}

2148: Ury\={u}, K., \& Eriguchi, Y. 1999, MNRAS, 303, 329

2149: 

2150: \bibitem[Villasenor et al.(2005)]{vil05}

2151: Villasenor, J. S., et al. 2005, Nature, 437, 855

2152: 

2153: \bibitem[Woosley \& Baron(1992)]{1992ApJ...391..228W}

2154: Woosley, S. E. \& Baron, E. 1992, \apj, 391, 228

2155: 

2156: \bibitem[Xie et al.(2007)]{2007ChJAA...7..685X} Xie, Y., Huang, C.-Y., 

2157: \& Lei, W.-H.\ 2007, Chinese Journal of Astronomy and Astrophysics, 7, 685 

2158: 

2159: \end{thebibliography} 

2160: 

2161: \label{lastpage}

2162: 

2163: \end{document} 

2164: 

2165: