The Astrophysical Journal, 620:59–68, 2005 February 10

# 2005. The American Astronomical Society. All rights reserved. Printed in U.S.A.

SPIN, ACCRETION, AND THE COSMOLOGICAL GROWTH OF SUPERMASSIVE BLACK HOLES

Stuart L. Shapiro1, 2
Receivved 2004 September 23; accepted 2004 October 26

ABSTRACT
If supermassive black holes (SMBHs) are the energy sources that power quasars and active galactic nuclei
(AGNs), then QSO SDSS 1148+5251, the quasar with the highest redshift (zQSO ¼ 6:43), hosts an SMBH formed
within
0.9 Gyr after the big bang. This requirement places constraints on the cosmological formation of SMBHs,
believed to grow from smaller initial seeds by a combination of accretion and mergers. We focus on gas accretion
onto seeds produced by the collapse of Population III stars at high redshift. We incorporate the results of recent
relativistic, MHD simulations of disk accretion onto Kerr black holes to track the coupled evolution of the masses
and spins of the holes. We allow for an additional amplification of
104 in the mass of a typical seed due to mergers,
consistent with recent Monte Carlo simulations of hierarchical mergers of cold dark matter halos containing black
hole seeds. We find that the growth of Population III black hole remnants to
109 M by zQSO k 6:43 favors MHD
accretion disks over standard thin disks. MHD disks tend to drive the holes to a submaximal equilibrium spin rate
a=M
0:95 and radiation efficiency
M
0:2, while standard thin disks drive them to maximal spin (a=M ¼ 1) and
efficiency (
M ¼ 0:42). This small difference in efficiency results in a huge difference in mass amplification by
accretion at the Eddington limit. The MHD equilibrium efficiency is consistent with the observed ratio of the QSO
plus AGN luminosity density to the local SMBH mass density. Our prototype analysis is designed to stimulate the
incorporation of results from relativistic stellar collapse and MHD accretion simulations in future Monte Carlo
simulations of hierarchical structure formation to better determine the cosmological role of SMBHs and their mass
and spin distributions.
Subject headinggs: accretion, accretion disks — black hole physics — cosmology: theory — MHD —
quasars: general

1. INTRODUCTION value
L  1 as a physical upper limit (McLure & Dunlop

2004). Barring an extraordinary coincidence, the range of val-
There is substantial evidence that supermassive black holes
ues inferred from observations for both
M and
L suggest that a
(SMBHs) with masses in the range 106 1010 M exist and are
significant fraction of the mass of SMBHs is acquired by gas
the engines that power active galactic nuclei (AGNs) and qua-
accretion. Together, these two parameters control the rate of
sars (Rees 1984, 1998, 2001; Macchetto 1999). There is also
growth of black holes by accretion and are crucial in determin-
ample evidence that SMBHs reside at the centers of many, and
ing whether initial seed black holes formed at high redshift have
perhaps most, galaxies (Richstone et al. 1998; Ho 1999), in-
sufficient time to grow to SMBHs to explain quasars and AGNs.
cluding the Milky Way (Genzel et al. 1997; Ghez et al. 2000,
The more massive the initial seed, the less time is required for
2003; Schödel et al. 2002).
The highest redshift of a quasar discovered to date is zQSO ¼ it to grow to SMBH scale and the easier it is to have an SMBH
in place by z  6:43. One possible progenitor that readily
6:43, corresponding to QSO SDSS 1148+5251 (Fan et al.
produces an SMBH is a supermassive star (SMS) with M 3
2003). Accordingly, if they are the energy sources in quasars
103 M (see, e.g., Shapiro 2004a for a recent review and ref-
(QSOs), the first SMBHs must have formed prior to zQSO ¼
erences). SMSs can form when gaseous structures build up
6:43 or within t ¼ 0:87 Gyr after the big bang in the concor-
sufficient radiation pressure to inhibit fragmentation and pre-
dance CDM cosmological model. This requirement sets a sig-
vent normal star formation; plausible cosmological scenarios
nificant constraint on black hole seed formation and growth
have been proposed that can lead to this occurrence (Gnedin
mechanisms in the early universe. Once formed, black holes
2001; Bromm & Loeb 2003). SMSs supported by radiation
grow by a combination of mergers and gas accretion. An impor-
pressure will evolve in a quasi-stationary manner to the point of
tant clue to the growth process is provided by the ratio R of the
observed QSO plus AGN luminosity density to the local SMBH onset of dynamical collapse due to a general relativistic radial
instability (Chandrasekhar 1964a, 1964b, 1964c; R. Feynman
mass density, since R is related to the mean radiative efficiency
1964, unpublished, as quoted in Fowler 1964). The collapse of

M of accretion onto black holes (Soltan 1982):
M  R. Recent
measurements suggest 0:1 P R P 0:2 (Yu & Tremaine 2002; a nonrotating, marginally unstable, spherical SMS of mass M
leads directly to the formation of a nonrotating Schwarzschild
Elvis et al. 2002), a range consistent with disk accretion onto
black hole of the same total mass (Shapiro & Teukolsky 1979).
black holes. Another clue is provided by the estimated ratios
L
But like most stars formed in nature, SMSs will be rotating. In
of the bolometric-to-Eddington luminosities of the broad-line
fact, in the event that viscosity and/or turbulent magnetic fields
quasars in a Sloan Digital Sky Survey sample of 12,698 quasars
are present to drive these stars to uniform rotation, they are
in the redshift interval 0:1  z  2:1. This survey supports the
likely to be maximally rotating (i.e., at the mass-shedding limit)
1 by the time they reach the onset of collapse (Baumgarte &
Department of Physics, University of Illinois at Urbana-Champaign,
Urbana, IL 61801-3080. Shapiro 1999). Recent relativistic hydrodynamical simulations
2
Department of Astronomy and NCSA, University of Illinois at Urbana- have shown that unstable, maximally rotating SMSs of arbitrary
Champaign, Urbana, IL 61801-3080. mass M inevitably collapse to SMBHs of mass
0:9M and spin
59
60 SHAPIRO Vol. 620

parameter a=M
0:75; the rest of the mass goes into an am- holes and explore their cosmological implications for SMBH
bient disk about the hole (Shibata & Shapiro 2002; Shapiro & evolution. We show how the evolution and amplification of
Shibata 2002; Shapiro 2004b). black hole mass by accretion is intimately tied to the evolution
But the fact remains that SMSs have yet to be observed, and of black hole spin, probing some of the implications of our ear-
there is no concrete evidence that they actually form in the early lier survey and analysis of black hole spin evolution (Gammie
universe. Moreover, simulations of cosmological structure for- et al. 2004) for cosmology. We demonstrate how, in principle,
mation performed to date indicate that the first generation of the very existence of a quasar at redshift zQSO ¼ 6:43 can help
stars are more likely to be zero-metallicity Population III stars constrain the formation epoch and /or size of black hole seeds
in the range 102 103 M (Bromm et al. 1999, 2002; Abel et al. and select among competing models of accretion.
2000, 2002; but note that Norman [2004] reports preliminary Specifically, recent relativistic MHD simulations predict the
indications that supermassive stellar objects with M k 104 M radiation efficiency
M as a function of the black hole spin pa-
may be forming as second-generation stars at 10 P z P 15 in his rameter a=M ; they also predict the spin-up rate as a function of
latest simulations). So the most conservative hypothesis is that a=M. Monte Carlo simulations that determine cosmological
the seed black holes that later grow to become SMBHs originate SMBH growth must integrate coupled mass and spin evolution
from the collapse of Population III stars (Madau & Rees 2001) prescriptions versus time for each accreting black hole to pro-
and not SMSs, and this is the Ansatz we explore here. Newtonian duce reliable growth histories and final SMBH masses and spins.
simulations suggest that Population III stars with masses in This paper provides prototype integrations of the coupled evo-
the range M
60 140 M and M k 260 M collapse directly lution equations, focusing on the history of the progenitor of
to black holes, while stars with M
140 260 M undergo ex- SDSS 1148+5251 and demonstrating the significant differences
plosive annihilation via pair-creation processes (Heger et al. in the outcomes for standard thin disk accretion models versus
2003). The upper limit to the mass of a Population III star is recent MHD models.
set by accretion over a stellar lifetime, yielding M P 600 M Our discussion is simplified and illustrative at best; much of
(Omukai & Palla 2003; Yoshida et al. 2003). Abel et al. (2002) the input physics involving black hole seed formation, accre-
argue that first-generation stars significantly larger than 100 M tion flows, and mergers is still being developed. Our main goal
are likely to explode before they have time to accrete to larger is to isolate some of the underlying local physical issues and
masses. The Pistol star is an example of an existing star believed parameters pertaining to accretion to better understand their
to have a mass k200 M , although it has high metallicity (Figer role in determining the global outcome of the cosmological
et al. 1998). Very massive stars are dominated by thermal ra- Monte Carlo simulations of SMBH buildup during hierarchical
diation pressure, so the catastrophic collapse of those that do structure formation. Excellent overviews of the input physics
not explode will be hydrodynamically similar to the collapse of have appeared elsewhere (see, e.g., Haiman & Quataert 2005
SMSs, producing black holes with masses comparable to those and references therein). Also, earlier treatments have consid-
of their progenitors. The higher the mass of the black hole seed ered some of the constraints on cosmological growth imposed
and the earlier it forms in the universe, the easier it is for it to by the recent discoveries of luminous quasars at high redshift
grow to an SMBH, and, hence, the more conservative will be (e.g., Haiman & Loeb 2001; Haiman 2004; Yoo & Miralda-
any constraints imposed on the cosmological black hole growth Escudé 2004). But here we specifically want to illustrate in the
rate by the existence of an SMBH by zQSO ¼ 6:43. Accordingly, simplest fashion how the most recent findings related to rela-
we take the highest range of plausible values for the masses of tivistic, MHD accretion flows onto spinning black holes have
black hole seeds to establish the most conservative (robust) con- important implications for evolutionary models of the growth
straints, adopting 100  M =M  600 for the range of black of SMBHs in the early universe. We emphasize by concrete
hole seeds arising from the collapse of Population III stars at example the point made in Gammie et al. (2004) that tracking
z  40. the spin as well as the mass of a black holes is necessary to
The most detailed studies of SMBH formation to date involve determine its growth (see also Volonteri et al. 2005). We also
detailed, Monte Carlo simulations that follow the cosmological show that the value of accretion radiation efficiency,
0.1,
growth of a distribution of black hole seeds by a combination of adopted in many Monte Carlo simulations may not be entirely
discrete, stochastic mergers as well as gas accretion (see, e.g., consistent with the latest MHD accretion disk modeling. De-
Haehnelt & Kauffmann 2000; Volonteri et al. 2003; Bromley termining this parameter, on which the outcome of cosmolog-
et al. 2004; Haehnelt 2004; Haiman 2004; Yoo & Miralda-Escudé ical simulations of SMBH growth depends very sensitively
2004; and references therein). Both processes are assumed to take (exponentially!), is coupled to the spin evolution of the hole;
place in the context of the cold dark matter (CDM) model, where both may now be within our grasp via detailed relativistic MHD
dark matter halos merge hierarchically and black holes are as- simulations of black hole accretion.
sumed to settle, merge, and accrete in their gaseous centers. Typ- The calculations performed here are prototypical only; our
ically, the stellar dynamical processes that lead to mergers, as well main aim is to motivate the incorporation of these parameters in
as the hydrodynamical processes that fuel accretion, are modeled more detailed Monte Carlo studies and thereby sharpen some of
in these analyses by implementing simple, physically plausible, the rules that enter these simulations. We also hope to provide
rules rather than by detailed integrations of the governing dy- those members of the computational relativistic MHD commu-
namical and hydrodynamical equations of motion. Performing nity who may not be SMBH Monte Carlo cognoscenti a simple
such ‘‘first principles’’ integrations would prove prohibitive in means of extracting the essence of the Monte Carlo simulations,
this context. particularly the evolutionary consequences of different accretion
In this paper we focus on SMBH cosmological growth by ac- models.
cretion. We identify and explore the main variables that govern In x 2 we set out the basic equations that describe the coupled
this process and ultimately influence the outcome of detailed evolution of black hole mass and spin by accretion. We also
Monte Carlo simulations that track SMBH growth. We incor- summarize in this section the relations we require from the
porate some of the most recent findings of relativistic magneto- concordance CDM cosmological model. In x 3 we integrate
hydrodynamical (MHD) simulations of gas accretion onto black the coupled evolution equations to track the evolution of black
No. 1, 2005 SUPERMASSIVE BLACK HOLES 61

hole mass and spin as functions of time. In x 4 we apply the re- Inserting equations (4) and (5) into equation (7) yields the evo-
sults to the cosmological problem and identify some constraints lution equation for the black hole spin,
imposed by the existence of an SMBH at z  6:43. In x 5 we
summarize briefly and discuss some caveats and areas for fur- d(a=M )
L s
¼ : ð8Þ
ther study. dt
M 
2. BASIC EQUATIONS In general, equations (5) and (8) must be integrated simulta-
Here we assemble the fundamental black hole accretion neously to determine the mass and spin evolution of the black
evolution equations and review the underlying assumptions on hole.
which they are based. We then specify a background cosmo- Determining
M (a=M ) and s(a=M ), which are needed to in-
logical model in which the growth of the black hole to super- tegrate equations (5) and (8), requires a gasdynamical model for
massive size occurs. black hole accretion. We assume that the gas has sufficient an-
gular momentum to form a disk about the hole, and we consider
2.1. Black Hole Mass and Spin Evvolution two different accretion disk models: (1) a standard, relativistic,
Define
M , the efficiency of conversion of rest-mass energy to Keplerian ‘‘thin disk’’ with ‘‘no-torque boundary conditions’’ at
luminous energy by accretion onto a black hole of mass M, the innermost stable circular orbit (ISCO; Pringle & Rees 1973;
according to Novikov & Thorne 1973; see Shapiro & Teukolsky 1983 for
review and references) and (2) a relativistic, MHD accretion disk

M  L=Ṁ0 c 2 ; ð1Þ that accounts for the presence of a frozen-in magnetic field in a
perfectly conducting plasma (De Villiers & Hawley 2003, De
where Ṁ0 is the rate of rest-mass accretion and L is the luminos- Villiers et al. 2003, Gammie et al. 2004, McKinney & Gammie
ity. Define
L , the efficiency of accretion luminosity, according 2004, and references therein). In the MHD model the magneto-
to rotational instability (MRI; Balbus & Hawley 1991) drives mag-
netic turbulence and provides the necessary torque to remove

L  L=LE ; ð2Þ angular momentum from the gas and drive the inflow. The MHD
model is arguably the most realistic model for disk accretion of
where LE is the Eddington luminosity, given by magnetized plasma onto a black hole. The standard thin disk
model provides a simple, analytic, limiting case that is useful as a
4M e mp c point of comparison.
LE ¼  1:3 ; 1046 e M8 ergs s1 : ð3Þ
T In a standard thin disk corotating with the black hole, the
energy and angular momentum per unit rest mass accreted by
Here we assume that the accretion is dominated by normal, a black hole are the energy and angular momentum of a unit
baryonic matter, and we ignore any contribution of collisionless mass at the ISCO, immediately prior to its rapid plunge and
or self-interacting dark matter (but see, e.g., Ostriker [2000] capture by the hole:
and Balberg & Shapiro [2002] for alternative scenarios). In pffiffiffiffiffiffiffiffiffiffi
particular, we assume that the accreting gas consists of fully r 2ms  2Mrms þ a Mrms
ionized atoms and that the principal opacity source is Thomson ẼISCO ¼ pffiffiffiffiffiffiffiffiffiffi ; ð9Þ
rms (r 2ms  3Mh rms þ 2a Mrms )1=2
scattering. The quantity e is the mean molecular weight per pffiffiffiffiffiffiffiffiffiffi 2 pffiffiffiffiffiffiffiffiffiffi
electron, and mp is the proton mass. The black hole growth rate Mrms (r ms  2a Mrms þ a 2 )
l̃ISCO ¼ pffiffiffiffiffiffiffiffiffiffi ;
must account for the loss of accretion mass energy in the form of rms (r 2ms  3Mh rms þ 2a Mrms )1=2
outgoing radiation according to
where the ISCO radius rms is given by
dM
¼ (1 
M ) Ṁ0 : ð4Þ
dt rms ¼ M f3 þ Z2  ½(3  Z1 )(3 þ Z1 þ 2Z2 )1=2 g; ð10Þ
Combining equations (1)–(4), we obtain the black hole growth
where
rate,
 1=3  
dM
L (1 
M ) M a2 a 1=3  a 1=3
¼ ; ð5Þ Z1  1 þ 1  2 1þ þ 1 ; ð11Þ
dt
M  M M M

where  is the characteristic accretion timescale and

Mc 2  2 1=2
  0:451 ð6Þ a
e Gyr Z2  3 2 þ Z12 ð12Þ
LE M
and is independent of M.
(see, e.g., Shapiro & Teukolsky 1983, eqs. [12.7.17]–[12.7.18]
The mass accretion efficiency
M is typically a function of the
and [12.7.24]). The mass accretion efficiency and spin evolu-
black hole spin parameter a=M ¼ J =M 2. It changes with time
tion parameters corresponding to the thin disk model are then
as the spin evolves. It is convenient to express the spin evolu-
given by
tion in terms of the nondimensional quantity s ¼ s(a=M ), de-
fined by
M ¼ 1  ẼISCO ; ð13Þ
d(a=M ) M a
s : ð7Þ s ¼ l̃ISCO  2 ẼISCO (standard thin disk): ð14Þ
dt Ṁ0 M
62 SHAPIRO Vol. 620

The MHD disk accretion model of Gammie et al. (2004) and Begelman 2003). However, the estimated Eddington ratios of
McKinney & Gammie (2004) is based on a fully relativistic, the bolometric luminosities of the broad-line quasars in a Sloan
axisymmetric simulation of a nonradiative, magnetized plasma Digital Sky Survey sample of 12,698 quasars in the redshift in-
onto a Kerr-Schild black hole within the MHD approxima- terval 0:1  z  2:1 support the value
L  1 as a physical upper
tion. The initial gas configuration is a torus with an inner ra- limit (McLure & Dunlop 2004). (Here the mass of the central
dius at r=M ¼ 6; in the absence of a magnetic field the torus is black hole is estimated from a virial relation, assuming that the
constructed to be in equilibrium about the hole (Fishbone & gas motions in the broad-line region of the quasars, measured
Moncrief 1976). The torus in threaded with a poloidal magnetic by H and Mg ii emission lines, are virialized.) The maximum
field initially and evolves with an adiabatic equation of state allowed value of
L establishes the maximum growth rate of a
(EOS) with an adiabatic index ¼ 4=3 (to model a radiation- black hole from accretion (see eq. [5]) and, as a result, the most
dominated, inner-disk EOS). The source of viscosity is MHD robust (least stringent) constraints on any gas accretion scenario
turbulence driven by the MRI instability. The simulations are for the growth of an SMBH from a smaller initial seed. To es-
performed for various black hole spin parameters a=M , holding tablish these constraints, we therefore set
L ¼ 1 in many of our
the value of the spin parameter fixed during the simulation. The numerical estimates below.
evolution proceeds for many dynamical timescales M, until a As a point of reference, and for later application, it is useful to
crude steady state, with fluctuations, is achieved. integrate equation (5) assuming that both the mass and lumi-
The results of the numerical simulations suggest that in nosity efficiencies remain constant with time, yielding
steady state the radiation efficiency parameter
M as a function  
of a=M is remarkably close to the function characterizing the
L (1 
M ) (t  ti )
M(t)=M (ti ) ¼ exp ; (
M ;
L constant);
standard thin disk (eq. [13]), even though there is no sharp
M 
transition in the surface density at or near the ISCO. However, ð16Þ
the spin evolution parameter s(a=M ) is different and can be
represented reasonably well by the least-squares linear fit where ti is the initial time at which the black hole has a mass
M(ti ). Note that the right-hand side of equation (16) is inde-
a pendent of black hole mass. This fact makes it possible, under
s ¼ 3:14 3:30 (MHD disk) ð15Þ
M certain plausible conditions that we specify, to disentangle and
track separately the amplification of black hole mass by accre-
(see McKinney & Gammie 2004, Table 2). The numerical tion from the amplification by discrete mergers. Let Mn (t) be
simulations demonstrate that the above parameters describing the mass of the black hole at time t following its nth merger
steady-state, MHD accretion-disk behavior are not particularly with another hole at time tn , where tn  t  tnþ1 . Assume that the
sensitive to the initial conditions in the disk (e.g., the initial duration of a merger, as well as the time required for accretion
B-field). As McKinney & Gammie (2004) discuss, the key re- to drive the merged remnant to spin equilibrium (see eq. [22]
sults are also quite comparable to those found by De Villiers below), are both much shorter than the time interval between
et al. (2004), who used a different numerical method and took mergers, and that the hole continues to accrete steadily through-
the adiabatic index of the gas to be ¼ 5=3 instead of ¼ out this interval. Note that black hole mergers can completely
4=3. We therefore model a relativistic MHD accretion disk by eject black holes from halo centers owing to gravitational wave
adopting equations (13) and (15) in our evolution equations. recoil and thereby turn off accretion altogether (see, e.g., dis-
The s versus a=M curve for the MHD disk is roughly parallel cussions of black hole recoil in halos in Hut & Rees 1992,
to, but somewhat below, the curve for the standard disk (see Merritt et al. 2004, and Madau & Quataert 2004). However, in-
Gammie et al. 2004, Fig. 4, for a comparison). In particular, the corporating the most recent recoil calculations (Favata et al.
parameter s never falls to zero for a standard thin disk until 2004) into simple models of dark halo mergers, Yoo & Miralda-
the hole is maximally rotating at a=M ¼ 1, while for an MHD Escudé (2004) conclude that the kick velocities are not suffi-
disk s crosses zero at a=M  0:95. The crucial physical ciently large to impede black hole growth significantly (cf. Haiman
consequence is that steady accretion via a standard thin disk al- 2004). Note also that a major merger between two black holes of
ways causes the central black hole to spin up until it is maxi- comparable mass may change both the magnitude and direction
mally rotating (Bardeen 1970), while accretion via an MHD of the spin of the resulting black hole remnant. Following such a
disk drives the black hole to spin equilibrium at a=M  0:95. merger, the orientation of the black hole spin may not be aligned
While the value of a=M at equilibrium is not determined pre- with the orientation of the asymptotic gaseous disk at radii
cisely and may depend on details of the flow geometry (and the r 3 100 M outside the hole. (Rees [1978] and Natarajan &
presence of radiative cooling), its departure from maximal is Pringle [1998] point out that the black hole exerts a torque on
robust and is typically k0.05 below unity. The difference be- the asymptotic disk, which also exerts a torque back on the hole,
tween a=M ¼ 1 and 0.95 is quite substantial physically; for eventually forcing their mutual alignment but on a timescale
example, the efficiency parameter is
M ¼ 1  1=31=2 ¼ 0:42 that is still uncertain.) Moreover, the orientation of the asymp-
for a=M ¼ 1, while it is considerably smaller,
M ¼ 0:19, for totic disk will likely fluctuate in time because of the redis-
a=M ¼ 0:95. As we see below, this difference will have signif- tribution of gas following mergers of dark halo cores, galaxy
icant consequences for the cosmological growth of a black hole mergers, and the tidal disruptions of passing stars by the central
from an initial seed. For comparison, we note that the radiative hole. However, near the black hole, at radii from r
M to 20M ,
efficiency for disk accretion onto a nonrotating Schwarzschild where the bulk of the disk’s gravitational energy is released and
black hole (a=M ¼ 0) is
M ¼ 1  (8=9)1=2 ¼ 0:057. the hole-disk interactions are strong, the hole’s gravitomagnetic
If we assume that the supply of gas remains sufficiently co- field will exert a force on the disk that, when combined with vis-
pious in the vicinity of the black hole, then the accretion lumi- cous forces and magnetic fields, will drive the disk down into
nosity is likely to be Eddington-limited and nearly constant, with the hole’s equatorial plane (the ‘‘Bardeen-Petterson effect’’;

L  1. Accretion models characterized by super-Eddington lu- Bardeen & Petterson 1975). This effect will maintain the align-
minosities with
L > 1 are possible theoretically (Ruszkowski & ment between the axis of the inner disk and the spin axis of the
No. 1, 2005 SUPERMASSIVE BLACK HOLES 63

hole. Accretion will subsequently drive the hole to spin equi- To evaluate the molecular weight e appearing in equations (3)
librium and restore
M to its equilibrium value. and (6), we assume that the accreting plasma is a zero-metallicity,
For the situation described above, we can take
M to be con- primoridial gas, for which
stant and equal to its value at spin equilibrium throughout most
of the lifetime of the hole. Assume further that
L is also constant, 1
which should be the case if the available gas is sufficiently co- e ¼ : ð21Þ
1  Y =2
pious that the accretion is always Eddington-limited, whereby

L  1. Let fn be the mass amplification of the hole following its We take the primordial helium abundance Y to be Y  0:25
nth merger with another hole: fn ¼ Mn (tn )=Mn1 (tn ) > 1. Then (Cyburt et al. 2003).
we may use equation (16) to calculate the total mass amplifica- For the concordance model, the age of the universe is T ¼
tion from ti to tf according to t(0) ¼ 13:7 Gyr, while the age at redshift z ¼ 6:43, the highest
known quasar redshift (SDSS 1148+525; Fan et al. 2003), is
Mf MN (tf ) M0 (t1 ) M1 (t1 ) M1 (t2 )
¼ ¼ ::: only tQSO ¼ 0:87 Gyr. Hence, tQSO is the upper limit to the time
Mi M0 (ti ) M0 (ti ) M0 (t1 ) M1 (t1 ) available for accretion to occur onto the initial seed black hole
MN 1 (tN ) MN (tN ) MN (tf ) that powers this quasar. In fact, assuming that the black hole
:::
MN1 (tN 1 ) MN 1 (tN ) MN (tN ) seed forms from the collapse of a first-generation, Population III
star at redshift z P 40 and t k t(40) ¼ 0:067 Gyr, the available
¼ exp½C(t1  ti ) f1 exp½C(t2  t1 ) : : :
 time for accretion is reduced to taccrete P 0:80 Gyr. We note that
: : : exp½C(tN  tN 1 ) fN exp C(tf  tN ) the stellar evolution (hydrogen-burning) lifetime of a massive

¼ f1 f2 : : : fN exp C(tf  ti ) ; ð17Þ Population III progenitor, tevo
0:003 Gyr (Wagoner 1969;
Omukai & Palla 2003), is much smaller than t(40), so the delay
where C ¼
L (1 
M )=(
M ) and is essentially constant. Com- between stellar formation and collapse is of little consequence
parison of equations (16) and (17) reveals that the net mass for determining the total time available for accretion growth.
amplification due to accretion can be treated as a single multi- Most important, the exponential accretion growth timescale is
plicative factor that is independent of the net amplification factor given by equations (5), (6), and (21) to be
due to mergers, f ¼ f1 f2 fN . Thus, for the simple scenario en-
visioned here, equation (16) can be applied to determine the
M 1 (
M =0:1) 1
growth ¼  ¼ 0:0394 Gyr; ð22Þ
growth of a black hole by accretion, even when steady growth 1 
M
L 1 
M
L
by accretion is interrupted and augmented by discrete, stochastic
black hole mergers. which is considerably smaller than taccrete . It is the availability
of many exponential growth timescales from the time of black
2.2. Cosmologg
ical Model hole seed formation to the birth of a quasar that makes it pos-
To relate the time parameter t appearing in the evolution sible for the black hole to grow from stellar to supermassive size
equations to an observable timelike quantity such as the redshift by gas accretion (aided by mergers) in the early universe.
z, we adopt the concordance CDM, spatially flat (k ¼ 0), cos-
mological model. All the free parameters in this model that we 3. BLACK HOLE GROWTH AND SPIN-UP
need for our computations have been measured by now, so the
Here we integrate the coupled mass and spin evolution
model is uniquely specified.
equations (5) and (8) to study black hole growth and spin-up
The basic evolution equation for the Friedmann-Robertson-
by gas accretion. In Figure 1 we show the increase in mass by
Walker expansion parameter a(t) is given by
accretion at the Eddington limit (
L ¼ 1) as a function of time.
 2   a 3  In Figure 1a the accretion disk is a standard thin disk with a no-
2 ȧ 2 0 0 0 torque boundary condition at the ISCO (eq. [14]); in Figure 1b
H  ¼ H0 m þ  ; ð18Þ
a a the disk is an MHD disk as calculated by McKinney & Gammie
(2004; eq. [15]). We consider two different initial values for
where H(t) is Hubble’s constant and the normalized mass the black hole spin parameter, a=M . The solid lines show the
density parameter 0m and cosmological constant parameter 0 case in which the initial black hole is nonrotating with
satisfy the relation a=M ¼ 0. The dotted lines show the case in which the black
hole is spinning with a=M ¼ 0:75. The latter is the value cal-
0 0
m þ  ¼ 1: ð19Þ culated for a black hole formed from the catastrophic collapse
of a massive, radiation-dominated star spinning uniformly at
In the above equations the sub(super)script ‘‘0’’ denotes the the mass-shedding limit that has evolved to the onset of relativ-
value of a quantity at the current epoch, z ¼ 0. Recalling that istic radial instability just prior to collapse (Shibata & Shapiro
a=a0 ¼ 1 þ z and substituting equation (19) into equation (18), 2002; Shapiro & Shibata 2002; Shapiro 2004b). For each disk
we can integrate equation (18) to obtain t ¼ t(z), model, the solid dots show the black hole growth that would
occur if the accretion were maintained at the asymptotic black
" ! #1=2 hole spin and efficiency from the beginning. Time is plotted in
0
2 1 1 m 1
t(z) ¼ sinh : ð20Þ units of  given by equations (6) and (21) for a zero-metallicity,
0
3H0 (1  0 1=2
m) m (1 þ z)3 cosmological abundance of H and He. The total duration plotted
corresponds to the age difference between redshift zi ¼ 40 (the
To evaluate t(z) appearing above, we adopt the Wilkinson Mi- earliest plausible birth date of a black hole from the collapse of a
crowave Anisotropy Probe values for the cosmological param- Population III star) and redshift zf ¼ 6:43 (the highest known
eters 0m  0:27 and H0  100h km s1 Mpc1 with h  0:71 quasar redshift, corresponding to SDSS 1148+5251; Fan et al.
(Bennett et al. 2003; Spergel et al. 2003). 2003) in the adopted CDM cosmology.
64 SHAPIRO Vol. 620

Fig. 2.—Same as Fig. 1, but for spin evolution during Eddington-limited

Fig. 1.—Black hole growth by accretion at the Eddington limit. In (a) the accretion. The large figure plots the spin parameter vs. time, while the inset
accretion disk is a standard thin disk with a no-torque boundary condition at plots the spin parameter vs. the black hole mass amplification factor.
the ISCO, ignoring photon recapture; in (b) the disk is an MHD disk as
calculated by McKinney & Gammie (2004). The solid lines show evolution
starting from a=M ¼ 0, while the dotted lines show evolution starting from amplification in lieu of a detailed integration of coupled evolu-
a=M ¼ 0:75. For each disk model, the solid dots show the mass amplification
that would occur if the accretion were maintained at the asymptotic, equilib- tion equations.
rium spin and radiation efficiency from the beginning. Time is plotted in units
of  ¼ Mc 2 =LE ¼ 0:45=e Gyr. The total duration shown corresponds to the 4. COSMOLOGICAL IMPLICATIONS
age difference between redshift zi ¼ 40 (the earliest plausible birth date of a Next we consider some of the cosmological implications of
black hole from the collapse of a Population III star) and redshift zf ¼ 6:43
(the highest known quasar redshift, corresponding to 1148+5251) in a con- the results of the previous section for the growth of SMBHs in
cordance CDM cosmology. the early universe. In Figure 3 we employ equations (16) and
(20) to study the black hole mass amplification factor Mf =Mi as
a function of the redshift zi of the initial seed black hole. Here
The key difference between the two disk models is that the we plot the final amplification factor achieved by redshift zf ¼
standard thin disk drives the black hole to maximal spin and 6:43, the highest known quasar redshift. Each solid curve is
efficiency (a=M ¼ 1 and
M ¼ 0:42; Bardeen 1970), while the labeled by the adopted constant radiation efficiency,
M ; the lu-
MHD disk drives the black hole to spin equilibrium at lower minosity is assumed to be the Eddington value (
L ¼ 1). The
values (a=M  0:95 and
M  0:19). A moderately lower ra- horizontal dashed lines indicate the range of amplification fac-
diation efficiency
M for Eddington-limited accretion results in tors required for accretion alone to grow an SMBH of mass
substantially larger black hole growth. As is evident from the 109 M from an initial seed black hole of mass 100  M =M 
figure, apart from an initial transient lasting
0:1
growth 600 formed from the collapse of a Population III star. A mass
(see Fig. 1, inset), the asymptotic evolution and final black hole of 109 M is the value inferred for typical quasars, including
mass amplification factor is little affected by the initial black 1148+5251 (Fan et al. 2003). The lower the mass of the initial
hole spin. seed, the larger the required mass amplification. The horizontal
In Figure 2 we show the spin evolution during Eddington- dotted lines indicate the required range of accretion-driven mass
limited accretion for the cases shown in Figure 1. The large amplification assuming that mergers assist the growth process by
figure plots the spin parameter versus time, while the inset plots accounting for an amplification of f
104 in black hole mass,
the spin parameter as a function of black hole mass amplifi- with the remaining amplification provided by gas accretion (Yoo
cation. After the initial transient lasting
0.1, during which & Miralda-Escudé 2004).
time the black hole grows by a factor of
2, the black hole spin Some important cosmological implications can be inferred
and efficiency approach their asymptotic values. An analytic from Figure 3. Consider a seed black hole that forms sometime
integration of equation (8) with equation (15), crudely hold- after redshift zi P 40, by which time the earliest stars have formed
ing
M fixed, yields an exponential spin-up timescale spin  and collapsed. In the absence of mergers, steady accretion cannot
(1 
M )growth =3:30, which explains the rapid spin-up rate. by itself achieve the required growth to explain quasars at zf ¼
This result also explains why the asymptotic black hole evo- 6:43 unless the efficiency satisfies
M P 0:13. For disk accretion,
lution and mass amplification are little affected by the initial this constraint requires the steady-state black hole spin parameter
black hole spin and depend only on the equilibrium spin rate to be below a=M  0:83. This value is below (1) the maximal
of the accreting Kerr black hole. Thus, for all but the initial black hole spin a=M ¼ 1, the asymptotic equilibrium value for
transient, it is adequate to adopt the equilibrium spin rate and a standard thin disk; (2) a=M ¼ 0:998 (for which
M  0:32),
use equation (16) with the corresponding constant efficiency the equilibrium value of a standard thin disk accounting for the

M ½(a=M )eq  to calculate the cosmological black hole mass recapture of some of the emitted photons by the black hole
No. 1, 2005 SUPERMASSIVE BLACK HOLES 65

tion to SMBH status is easily achieved for typical MHD accre-

tion disks, marginally possible for a standard thin disk in spin
equilibrium accounting for photon recapture, but not possible
for a standard thin disk that drives the black hole to maximal
rotation. However, if the black hole seed is less than 600 M ,
the standard thin disk appears to be ruled out; the required am-
plification places the lower dotted line in the plot above the curve
for
M ¼ 0:32, corresponding to a=M ¼ 0:998, the value for a
standard thin disk in spin equilibrium, accounting for photon re-
capture. These conclusions also hold even if the black hole seed
forms much earlier than zi  40, while they may be tightened if
the seed forms later. In fact, it may be likely that the seed forms
later, at zi P 40, given that even 4  peaks in the density per-
turbation spectrum for the progenitor halo of SDSS 1148+5251
do not collapse until z
30 in the CDM concordance cosmol-
ogy (see, e.g., Barkana & Loeb 2001, Fig. 5). Moreover, the
potential wells of the earliest halos are quite shallow (
1 km s1)
and may not be able to retain enough gas to form stars. Never-
theless, the effect of altering the date of birth of the black hole
seed is not very great unless zi P 20 25, as is evident from
Figure 3.
It is significant that the range of equilibrium accretion disk
radiation efficiencies required to achieve the necessary growth
Fig. 3.—Black hole accretion mass amplification Mf =Mi vs. redshift zi of of a black hole seed to supermassive size by zf ¼ 6:43 is con-
the initial seed. Here we plot the amplification achieved by redshift zf ¼ 6:43, sistent with the values inferred observationally for R, the ratio
the highest known quasar redshift, corresponding to 1148+5251. Each solid
curve is labeled by the adopted constant radiation efficiency,
M ; the lumi- of the QSO plus AGN luminosity density to the mass density of
nosity is assumed to be the Eddington value (
L ¼ 1). The horizontal dashed SMBHs in nearby galaxies:
M  R
0:1 0:2 (Soltan 1982;
lines bracket the range of amplification required for accretion alone to grow a Yu & Tremaine 2002; Elvis et al. 2002). This consistency sup-
seed black hole of mass 100  M =M  600 formed from the collapse of a ports the notion that accretion plays an important role in SMBH
Population III star to 109 M . The horizontal dotted lines bracket the required
accretion amplification range assuming that mergers account for a growth of
growth and is responsible for the acquisition of the bulk of the
104 in black hole mass, the remainder being by gas accretion. final mass of an SMBH. The range of
M inferred from R favors
accretion disk models that drive the black hole to spin equi-
librium in the range 0:7 P a=M P 0:95, well below maximal
(Thorne 1974); and (3) a=M  0:95, the equilibrium value of a rotation and consistent with the values found in recent simu-
typical MHD disk. Therefore, it is likely that mergers are required lations of relativistic MHD accretion disks.
to assist accretion to achieve black hole growth to supermassive How are our conclusions altered if (when) a quasar is dis-
size by zf ¼ 6:43. As may be inferred from the figure, this con- covered at a higher redshift, zQSO > 6:43? We anticipate this
clusion holds even in the (unlikely) event that the black hole seed possibility in Figure 4, where we solve equation (16) to plot the
forms much earlier than zi  40. mass efficiency
M required to build an SMBH from a seed as a
Monte Carlo simulations by Yoo & Miralda-Escudé (2004) function of the host quasar redshift, zQSO . We again assume that
of hierarchical CDM halo mergers, accompanied by mergers of a seed black hole of mass 100  M =M  600, formed from
their central black holes, suggest that black hole mass ampli- the collapse of a Population III star at zi 3 zQSO , grows by zQSO
fication factors of
104 are achieved via mergers by zf ¼ 6:43. to 109 M . While accreting, the black hole is again assumed to
This result implies that f ¼ f1 f2 : : : fN
104 in equation (17). radiate at the Eddington luminosity, with
L ¼ 1. We set ti ¼ 0,
While other simulations predict lower growth factors due to zi ¼ 1, so that the plotted radiation efficiency represents an
mergers, we adopt 104 to be conservative, noting that our con- upper limit to
M ; increasing ti and lowering zi decreases the
clusions below are strengthened if the smaller factors turn out required efficiency (but the decrease is small if, for example,
to be correct. As equation (17) shows, equation (16) can be used we set zi k 40). The dashed lines indicate the range of efficien-
to estimate the cumulative mass amplication by gas accretion, cies required for gas accretion alone to achieve the necessary
even when steady growth by accretion is interrupted and en- growth. The dotted lines indicate the range required for gas ac-
hanced by stochastic black hole mergers. Minor mergers of mas- cretion, assuming that mergers assist the growth and account for
sive black holes with smaller holes tend to spin down the massive an amplification of f ¼ 104 in black hole mass. The horizontal
hole (Hughes & Blandford 2003; Gammie et al. 2004). Major solid lines bracket the range of efficiencies inferred observa-
mergers of two black holes of comparable mass following bi- tionally for R.
nary inspiral drive the merged remnant to a=m k 0:7. (See, e.g., In the absence of mergers, the upper limit to the radiation
Baumgarte & Shapiro [2003] for a review of binary black hole efficiency required to build an SMBH by zQSO  6:43 is
M 
coalescence and for references; see Gammie et al. [2004] for 0:14, consistent with observational constraints implied by R,
a general discussion of black hole spin-up and spin-down but inconsistent with the much higher values expected for a
mechanisms and for references.) In either case, after a short standard thin disk driven to spin equilibrium, and marginally
transient epoch, accretion will drive the merged remnant to inconsistent for an MHD accretion disk. The inconsistencies
the disk accretion equilibrium spin rate and corresponding become worse as zQSO > 6:43. In fact, the upper limit to
M falls
mass efficiency (see x 2), as assumed in employing equation (16). below 0.1 for zQSO k 8, which would force
M to lie below the
Figure 3 indicates that for black hole seeds that arise from observationally inferred range for R and would therefore be
collapsed Population III stars, merger-assisted mass amplifica- difficult to understand.
66 SHAPIRO Vol. 620

5. SUMMARY
We have explored the evolution of black hole mass and spin
by gas accretion in the early universe. We have illustrated how
for Eddington-limited accretion, the growth of an SMBH de-
pends sensitively on the radiative efficiency,
M . For disk ac-
cretion, the mean efficiency is determined by the equilibrium
black hole spin, which in turn depends on the torques acting on
the gas near the black hole horizon. We have explored the con-
sequences of the assumptions that seed black holes are the rem-
nants of collapsed Population III stars that may form as early as
z P 40 and grow to
109 M by zQSO ¼ 6:43, the highest red-
shift discovered to date, corresponding to QSO SDS 1148+5251.
Allowing for growth by both accretion and mergers, simple the-
ory suggests that the required mass amplification is possible
provided the radiation efficiency satisfies
M P 0:2, with the up-
per limit decreasing should a quasar be discovered at higher red-
shift zQSO > 6:43. The inferred efficiency is consistent with the
observed ratio R of the QSO plus AGN radiation density to the
mass density of SMBHs in local galaxies, which suggests that an
appreciable fraction of the mass of the final black hole is acquired
by disk accretion, rather than by mergers. The inferred efficiency
favors MHD accretion disk models that exert nonzero torques on
Fig. 4.—Radiation efficiency
M required to build an SMBH vs. the redshift the gas at the inner edge of the disk. These disks ultimately drive
zQSO of the host quasar. The seed black hole forms with a mass 100  the black hole to spin equilibrium at a=M
0:95, substantially
M =M  600 from the collapse of a Population III star at zi 3 zQSO and below the maximum spin allowed for a Kerr black hole.
grows by zQSO to 109 M . Accretion occurs at the Eddington luminosity. The
dashed lines bracket the range of efficiencies required for accretion alone to
We have made a number of simplifying assumptions in an
achieve the necessary growth. The dotted lines bracket the range required for effort to provide a concrete calculation that illustrates how the
accretion assuming that mergers account for a growth of 104 in black hole SMBH initial seed, its cosmological growth, and accretion disk
mass. The horizontal solid lines bracket the range of efficiencies inferred from models may all be constrained by the existence of quasars at
R, the observed ratio of the QSO plus AGN luminosity density to the mass high redshift. Our conclusions are tentative, as there exist many
density of local SMBHs.
caveats. For example, if the progenitors of black hole seeds are
SMSs with M 3103 M instead of Population III stars, many of
When mergers are included, the upper limit to the efficiency the constraints on the radiation efficiency
M , and the associated
required to build an SMBH by zQSO  6:43 increases to
M
range of expected black hole spins, may have to be relaxed. On
0:30 (i.e., any black hole seed born at a finite redshift, 1 > the other hand, if the progenitors have masses P100 M , the
zi > zQSO , must accrete at a lower efficiency than 0.30 to reach constraints favoring low efficiencies
M P 0:2 characterizing
109 M ). This upper limit is roughly consistent with the obser- MHD accretion disks in spin equilibrium are strengthened. If
vational constraint on
M  R and with theoretical values for the accretion luminosity is super-Eddington with
L > 1, then the
accretion from an MHD disk in spin equilibrium, but only upper limit for
M increases (see eq. [16]), relaxing the con-
marginally consistent for accretion from a standard thin disk in straints that favor an MHD disk over a standard thin disk. Super-
spin equilibrium accounting for photon recapture, and incon- Eddington accretion is possible theoretically (Ruszkowski &
sistent for accretion from a standard thin disk that drives the Begelman 2003). However, a Sloan Digital Sky Survey of
black hole to maximal rotation. Should a quasar be discovered 12,698 broad-line quasars in the redshift interval 0:1  z  2:1
at zQSO > 6:43, it would appear that accretion from a standard supports the Eddington value as a physical limit (McLure &
thin disk will be ruled out; the upper limit will fall below Dunlop 2004). Moreover, given the observed value for R, a

M ¼ 0:32, the value for a standard thin disk in spin equilibrium higher limit for
M would imply that a large fraction of the
accounting for photon recapture. Should a quasar be discovered emitted radiation lies outside the optical bandwidths included in
at zQSO k 10,
M would fall below 0.19 and the results would be the determination of R, which seems unlikely. Finally, mergers
difficult to reconcile with accretion from a typical MHD disk may lead to black hole mass amplification factors smaller than
as modeled in recent simulations. These critical values of zQSO f
104 , the typical value found in the Monte Carlo calculations
are all smaller if the initial black hole seed is smaller than of Yoo & Miralda-Escudé (2004) and the value assumed here. A
600 M , as the top curve, which sets the limit, is lowered in the smaller merger amplification factor requires larger accretion-
figure. Finally, should a quasar be discovered ( perhaps un- driven amplification in order for a black hole seed to reach
expectedly) at zQSO k 18, the upper limit to
M would drop SMBH size by zQSO ¼ 6:43 (see eq. [17]). This would strengthen
below the observationally inferred value 0.1 for all zi  zQSO . the constraints set by SMBH accretion growth on the upper limit
This drop would not be easily explained by any of these to
M , although the efficiency must still be consistent with R if the
models of accretion and might require super-Eddington ac- bolometric accretion luminosity resides mostly in the optical
cretion with
L > 1 to keep
M k 0:1. Alternatively, the value wavebands included in R and the value of R measured for local
of R, observed for local host galaxies, might be significantly galaxies also applies at higher redshift. Of course, the peak
smaller at high redshift, which would relax the constraint on contribution to the value of R arises from the average behavior of

M . But achieving the inferred lower radiation efficiencies accreting BHs with z
3; this ratio does not necessarily apply to
might then require advection-dominated accretion disks or individual objects like SDSS 1148+5251 nor need it apply to
spherical accretion. SMBHs at z  6.
No. 1, 2005 SUPERMASSIVE BLACK HOLES 67

We have assumed that the mass of SDSS 1148+5251 is 2002), inferred for nearby host galaxies. Only by performing

109 M ; a lower value would relax many of our constraints, detailed simulations that track the formation and growth of
while a higher value would strengthen them. If the flux from this SMBHs in a cosmological setting governed by hierarchical halo
source were amplified by gravitational lensing, or beaming, then mergers, black hole mergers, gas settling, star formation, and
a lower mass estimate would be appropriate. However, no mul- feedback can these correlations be reliably reproduced. We look
tiple images have been seen (Richards et al. 2004), and it has forward to the next generation of simulations that incorporate
been shown that high amplification without at least two images is the recent results of relativistic MHD accretion onto black holes,
very improbable (Keeton et al. 2004). Strong beaming also seems since, as demonstrated here, the outcome of these global simu-
unlikely, since it would reduce the line/continuum ratio (Haiman lations may depend sensitively on the local physics of such ac-
& Cen 2002), which is not observed (Willott et al. 2003). In fact, cretion flows.
assuming that the quasar emits at the Eddington luminosity gives
a mass of 4:6 ; 10 9 M (Fan et al. 2003; Haiman 2004), and this
higher value strengthens our conclusions somewhat. It is a pleasure to thank T. Abel, R. Cyburt, C. Gammie,
Isolating the accretion growth of a seed black hole in the J. McKinney, J. Miralda-Escudé, and Q. Yu for valuable dis-
early universe from the hole’s full dynamical history and envi- cussions. We also thank the referee for a critical reading of
ronment does not allow us to account for other important cor- the manuscript and valuable comments. A portion of this work
relations that provide clues to the formation of SMBHs. Such was performed during the summer of 2004 at the Aspen Center
correlations include the SMBH mass versus bulge luminosity for Physics, whose hospitality is gratefully acknowledged. This
relation, MBH / Lbulge (Kormendy & Richstone 1995), and the work was supported in part by NSF grants PHY-0205155 and
SMBH mass versus velocity dispersion relation, MBH / v 4c PHY-0345151 and NASA grant NNG04GK54G at the Univer-
(Gebhardt et al. 2000; Ferrarese & Merritt 2000; Tremaine et al. sity of Illinois at Urbana-Champaign.

REFERENCES
Abel, T., Bryan, G. L., & Norman, M. L. 2000, ApJ, 540, 39 Heger, A., Fryer, C. L., Woosley, S. E., Langer, N., & Hartmann, D. H. 2003,
———. 2002, Science, 295, 93 ApJ, 591, 288
Balberg, S., & Shapiro, S. L. 2002, Phys. Rev. Lett., 88, 1301 Ho, L. C. 1999, in Observational Evidence for Black Holes in the Universe, ed.
Balbus, S. A., & Hawley, J. F. 1991, ApJ, 376, 214 S. K. Chakrabarti ( Dordrecht: Kluwer), 157
Bardeen, J. M. 1970, Nature, 226, 64 Hughes, S. A., & Blandford, R. D. 2003, ApJ, 585, L101
Bardeen, J. M., & Petterson, J. A. 1975, ApJ, 195, L65 Hut, P., & Rees, M. J. 1992, MNRAS, 259, P27
Barkana, R., & Loeb, A. 2001, Phys. Rep., 349, 125 Keeton, C. R., Kuhlen, M., & Haiman, Z. 2004, ApJ, submitted (astro-ph /
Baumgarte, T. W., & Shapiro, S. L. 1999, ApJ, 527, L5 0405143)
———. 2003, Phys. Rep., 376/2, 41 Kormendy, J., & Richstone, R. D. 1995, ARA&A, 33, 581
Bennett, C. L., et al. 2003, ApJS, 148, 1 Macchetto, F. D. 1999, Ap&SS, 269, 269
Bromley, J. M., Somerville, R. S., & Fabian, A. C. 2004, MNRAS, 350, 456 Madau, P., & Quataert, E. 2004, ApJ, 606, L17
Bromm, V., Coppi, P. S., & Larson, R. B. 1999, ApJ, 527, L5 Madau, P., & Rees, M. 2001, ApJ, 551, L27
———. 2002, ApJ, 564, 23 McKinney, J. C., & Gammie, C. F. 2004, ApJ, 611, 977
Bromm, V., & Loeb, A. 2003, ApJ, 596, 34 McLure, R. J., & Dunlop, J. S. 2004, MNRAS, 352, 1390
Chandrasekhar, S. 1964a, Phys. Rev. Lett., 12, 114 Merritt, D., Milosavljevic, M., Favata, M., Hughes, S. A., & Holz, D. E. 2004,
———. 1964b, Phys. Rev. Lett., 12, 437 ApJ, 607, L9
———. 1964c, ApJ, 140, 417 Natarajan, P., & Pringle, J. E. 1998, ApJ, 506, L97
Cyburt, R. H., Fields, B. D., & Olive, K. A. 2003, Phys. Lett. B, 567, 227 Norman, M. L. 2004, talk presented at Aspen Summer Workshop on the
De Villiers, J. P., & Hawley, J. F. 2003, ApJ, 589, 458 Formation of Supermassive Black Holes, June 2004, Aspen Center for
De Villiers, J. P., Hawley, J. F., & Krolick, J. H. 2003, ApJ, 599, 1238 Physics
De Villiers, J. P., Hawley, J. F., Krolick, J. H., & Hirose, S. 2004, ApJ, sub- Novikov, I. D., & Thorne, K. S. 1973, in Black Holes, Les Houches 1972, ed.
mitted (astro-ph /0407092) C. DeWitt & B. DeWitt (New York: Gordon & Breach), 343
Elvis, M., Risaliti, G., & Zamorani, G. 2002, ApJ, 565, L75 Omukai, K., & Palla, F. 2003, ApJ, 589, 677
Fan, X., et al. 2003, AJ, 125, 1649 Ostriker, J. P. 2000, Phys. Rev. Lett., 84, 5258
Favata, M., Hughes, S. A., & Holtz, D. E. 2004, ApJ, 607, L5 Pringle, J. E., & Rees, M. J. 1973, A&A, 29, 179
Ferrarese, L., & Merritt, D. 2000, ApJ, 539, L9 Rees, M. J. 1978, Nature, 275, 516
Figer, D. F., Najarro, F., Morris, M., McLean, I. S., Geballe, T. R., Ghez. A. M., ———. 1984, ARA&A, 22, 471
& Langer, N. 1998, ApJ, 506, 384 ———. 1998, in Black Holes and Relativistic Stars, ed. R. M. Wald (Chicago:
Fishbone, L. G., & Moncrief, V. 1976, ApJ, 207, 962 Univ. Chicago Press), 79
Fowler, W. A. 1964, Rev. Mod. Phys., 36, 1104 ———. 2001, in Black Holes in Binaries and Galactic Nuclei, ed. L. Kaper,
Gammie, C. F., Shapiro, S. L., & McKinney, J. C. 2004, ApJ, 602, 312 E. P. J. van den Heurel, & P. A. Woudt ( New York: Springer), 351
Gebhardt, K., et al. 2000, ApJ, 539, L13 Richards, G. T., et al. 2004, AJ, 127, 1305
Genzel, R., Eckart, A., Ott, T., & Eisenhauer, F. 1997, MNRAS, 291, 219 Richstone, D., et al. 1998, Nature, 395, 14
Ghez, A. M., Becklin, E., Duchene, G., Hornstein, S., Morris, M., Salim, S., & Ruszkowski, M., & Begelman, M. C. 2003, ApJ, 586, 384
Tanner, A. 2003, Astron. Nachr. Supp., 324, 3 Schödel, R., et al. 2002, Nature, 419, 694
Ghez, A. M., Morris, M., Becklin, E. E., Tanner, A., & Kremenek, T. 2000, Shapiro, S. L. 2004a, in Carnegie Observatories Astrophysics Series, Vol. 1:
Nature, 407, 349 Coevolution of Black Holes and Galaxies, ed. L. C. Ho (Cambridge:
Gnedin, O. Y. 2001, Classical Quantum Gravity, 18, 3983 Cambridge Univ. Press), 103
Haehnelt, M. G. 2004, in Carnegie Observatories Astrophysics Series, Vol. 1: ———. 2004b, ApJ, 610, 913
Coevolution of Black Holes and Galaxies, ed. L. C. Ho (Cambridge: Shapiro, S. L., & Shibata, M. 2002, ApJ, 577, 904
Cambridge Univ. Press), 406 Shapiro, S. L., & Teukolsky, S. A. 1979, ApJ, 234, L177
Haehnelt, M. G., & Kauffmann, G. 2000, MNRAS, 318, L35 ———. 1983, Black Holes, White Dwarfs, and Neutron Stars: The Physics of
Haiman, Z. 2004, ApJ, 613, 36 Compact Objects (New York: Wiley)
Haiman, Z., & Cen, R. 2002, ApJ, 578, 702 Shibata, M., & Shapiro, S. L. 2002, ApJ, 572, L39
Haiman, Z., & Loeb, A. 2001, ApJ, 552, 459 Soltan, A. 1982, MNRAS, 200, 115
Haiman, Z., & Quataert, E. 2005, in Supermassive Black Holes in the Distant Spergel, D. N., et al. 2003, ApJS, 148, 175
Universe, ed. A. J. Barger (Dordrecht: Kluwer), in press (astro-ph /0403225) Thorne, K. S. 1974, ApJ, 191, 507
68 SHAPIRO
Tremaine, S., et al. 2002, ApJ, 574, 740 Yoo, J., & Miralda-Escudé, J. 2004, ApJ, 614, L25
Volonteri, M., Haardt, F., & Madau, P. 2003, ApJ, 582, 559 Yoshida, N., Abel, T., Hernquist, L., & Sugiyama, N. 2003, ApJ, 592, 645
Volonteri, M., Madau, P., Quataert, E., & Rees, M. J. 2005, ApJ, 620, 69 Yu, Q., & Tremaine, S. 2002, MNRAS, 335, 965
Wagoner, R. V. 1969, ARA&A, 7, 553
Willott, C. J., McLure, R. J., & Jarvis, M. J. 2003, ApJ, 587, L15