Abstract

We predict the existence and observational signatures of a new class of objects that assembled early, during the first billion years of cosmic time: obese black hole galaxies (OBGs). OBGs are objects in which the mass of the central black hole (BH) initially exceeds that of the stellar component of the host galaxy, and the luminosity from BH accretion dominates the starlight. Conventional wisdom dictates that the first galaxies light up with the formation of the first stars; we show here that, in fact, there could exist a population of astrophysical objects in which this is not the case. From a cosmological simulation, we demonstrate that there are sites where star formation is initially inhibited and direct-collapse black holes (DCBHs) form due to the photodissociating effect of Lyman–Werner radiation on molecular hydrogen. We show that the formation of OBGs is inevitable, because the probability of finding the required extragalactic environment and the right physical conditions in a halo conducive to DCBH formation is quite high in the early Universe. We estimate an OBG number density of 0.009 Mpc−3 at z ∼ 8 and 0.03 Mpc−3 at z ∼ 6. Extrapolating from our simulation volume, we infer that the most luminous quasars detected at z ≥ 6 likely transited through an earlier OBG phase. Following the growth history of DCBHs and their host galaxies in an evolving dark matter halo shows that these primordial galaxies start off with an overmassive BH and acquire their stellar component from subsequent merging as well as in situ star formation. In doing so, they inevitably go through an OBG phase dominated by the accretion luminosity at the Eddington rate or below, released from the growing BH. The OBG phase is characterized by an ultraviolet (UV) spectrum fλ ∝ λβ with a slope of β ∼ −2.3 and the absence of a Balmer break. OBGs should also be spatially unresolved, and are expected to be brighter than the majority of known high-redshift galaxies. They could also display broad high-excitation emission lines, as expected from type I active galactic nuclei, although the strength of lines such as N v and C iv will obviously depend on the chemical enrichment of the host galaxy. OBGs could potentially be revealed via Hubble Space Telescope follow-up imaging of samples of brighter Lyman-break galaxies provided by wide-area ground-based surveys such as UltraVISTA, and should be easily uncovered and studied with instruments aboard the James Webb Space Telescope. The discovery and characterization of OBGs would provide important insights into the formation of the first BH, and their influence on early galaxy formation.

INTRODUCTION

It is now well established that most present-day galaxies harbour a quiescent supermassive black hole (SMBH), with a mass approximately one thousandth of the mass of stars in the bulge (Ferrarese & Merritt 2000; Häring & Rix 2004). Such a correlation is strongly suggestive of coupled growth of the SMBH and the stellar component, likely via regulation of the gas supply in galactic nuclei from the earliest times (e.g. Silk & Rees 1998; Haehnelt & Kauffmann 2000; Fabian, Wilman, & Crawford 2002; King 2003; Thompson, Quataert, & Murray 2005; Robertson et al. 2006; Hopkins, Murray, & Thompson 2009; Natarajan & Treister 2009; Treister et al. 2011, see however Hirschmann et al. 2010).

Since the same gas reservoir fuels star formation and feeds the black hole (BH), a connection between these two astrophysical processes regulated by the evolving gravitational potential of the dark matter (DM) halo is arguably expected. However, understanding when and how this interplay commences has been both a theoretical and observational challenge for current theories of structure formation.

In this paper, we explore the formation and evolution of the first massive BH seeds and the first stars during the earliest epochs in order to explore the onset of coupling between the BH and the stellar component. Our calculation incorporates two new physical processes that have only been recently recognized as critical to understanding the fate of collapsing gas in the early universe. The first is the computation of the Lyman–Werner (LW) radiation (11.2–13.6 eV) that impacts gas collapse in the first DM haloes [as it is able to efficiently dissociate the H2 molecules, thereby preventing cooling via molecular hydrogen; e.g. Haiman, Abel, & Rees (2000)]. The second is the implementation of our growing understanding of the role of the angular momentum of the baryonic gas in the collapse process (Lodato & Natarajan 2006; Davis & Natarajan 2010). Including these two processes within the context of the standard paradigm of structure formation predicts the possible existence of a new class of object in the high-redshift Universe in which BH growth commences before and continues to lead the build-up of the galaxy stellar population for a significant period of time. We define an Obese Black hole Galaxy (OBG) as a phase in a galaxy's evolution where post-direct-collapse black hole (DCBH) formation, the BH at least initially dominates over the stellar mass and is accreting at a rate sufficiently high enough to outshine the stellar component.

OBGs may provide a natural stage of early BH/galaxy evolution en route to the most luminous quasars already observed to be in place at z ≃ 6 with estimated BH masses MBH ≃ 109 M. Observationally, such objects should appear similar to moderate-luminosity AGN, but with very low luminosity host galaxies and low metallicities.

METHODOLOGY

Our model is a modified version of Agarwal et al. (2012, hereafter A12), where they identify the sites of DCBH formation by calculating the total amount (spatial and global) of LW radiation seen by any given halo within a cosmological N-body DM only simulation using a semi-analytic model for the star formation in these haloes. The key features of A12 are summarized below.

  • The DM only N-body simulation is run from z = 30 to z = 6 with a box size of ∼3.4 Mpc h−1 and DM particle mass of 6500 Mh−1. This was chosen so that we can resolve a minimum halo mass ∼105 M with 20 particles, similar to the minimum halo mass that can host a Population III (Pop III) star at z = 30 (Tegmark et al. 1997).

  • Both Pop III and Pop II star formation are allowed, and halo histories are tracked in order to determine if a halo is metal free.

  • Many realizations of the model are run to study the effect of different LW escape fractions, Pop II star formation efficiencies, gas outflow rates due to supernova (SN) feedback, number of Pop III stars forming per halo and reionization feedback. The results presented here are based on the run with an LW escape fraction of 1.0 and a Pop II star formation efficiency (SFE) of 0.005 with a burst mode of star formation. We create Pop II stars in a single burst that is placed randomly between the two time-steps for which we use the burst mode template (fig. 7e) from starburst99 (Leitherer et al. 1999). Using the burst mode leads to a peak in LW emission at 10−42 erg s−1 for a 1 Myr old, 106 M Pop II star cluster which drops to 10−38 erg s−1 at 700 Myr.

  • The LW specific intensity in units of 10−21 erg s−1 cm−2 Hz−1 sr−1, JLW is computed self-consistently depending on the type, mass and age of the stellar population, and with two components: a global and a local contribution. A pristine halo is considered for Pop III star formation or treated as a DCBH candidate depending on halo's virial temperature and JLW that it is exposed to.

  • The number density of DCBH sites can be up to 0.1 Mpc− 3 at z = 6, much higher than previously anticipated (Dijkstra et al. 2008).

In the present study, we refine the model of DCBH formation and also follow the subsequent growth of these seeds as identified in A12. We discuss new additions to the A12 model in the subsections below.

DCBH forming haloes

A DCBH forms in our model if a pristine massive halo is exposed to JJcrit1 (Wolcott-Green et al. 2011) and satisfies both the spin and size criterion required for the disc to withstand fragmentation (Lodato & Natarajan 2006, 2007; LN06 and LN07 hereafter). Note that Jcrit is always produced by stellar sources external to the pristine DCBH candidate halo as internal sources would pollute the gas inside the halo. We assume that the collapsing gas in pristine haloes will settle into a disc whose stability against fragmentation determines whether it will be able to collapse to a BH or will fragment into star-forming clumps (LN06; LN07). Assuming that initially the baryons have the same specific angular momentum as the halo, the halo must have a spin, λ, lower than a characteristic value, λmax, for a given Toomre stability parameter Qc, for which the pristine gaseous disc is exactly marginally stable and above which no accretion can take place on to the central region (LN06). The critical spin is given as  

(1)
\begin{equation} \lambda _{\rm max}= \frac{m_{\rm d}^2 Q_{\rm c}}{8j_{\rm d}} \sqrt{\frac{T_{\rm gas}}{T_{\rm vir}}} \ , \end{equation}
where md is the disc mass expressed as a fixed fraction (0.05) of the total baryonic mass in the halo (Mo, Mao, & White 1998), jd is the specific angular momentum of the disc that is also a fixed fraction (0.05) of halo's overall angular momentum (LN06), Tvir is the virial temperature of the halo and Tgas is the temperature of the gas in the disc which depends on whether atomic or molecular hydrogen is the dominant cooling species. In our case, since the halo is exposed to JJcrit, the dominant coolant is atomic hydrogen and Tgas is set to 8000 K. The second condition comes from the limit that the disc must be cooler than a characteristic temperature above which the gravitational torques required to redistribute the angular momentum become too large and can disrupt the disc. Tmax is used as a proxy for size of the disc and is defined as  
(2)
\begin{equation} T_{\rm max}=T_{\rm gas}\left(\frac{4\alpha _{\rm c}}{m_{\rm d}} \frac{1}{1 + M_{\rm BH}/m_{\rm d}M}\right)^{2/3}\ , \end{equation}
where αc is a dimensionless parameter (0.06) relating the critical viscosity to the gravitational torques in a halo with DM mass, M. This provides a mass estimate for the assembling DCBH (LN07):  
(3)
\begin{equation} M_{\rm BH} = m_{\rm d}M \left( 1 - \sqrt{ \frac{8 \lambda j_{\rm d}}{m_{\rm d}^2 Q_{\rm c} } \left(\frac{T_{\rm gas}}{T_{\rm vir}}\right)^{1/2}}\right)\ , \end{equation}
for λ < λmax and Tvir < Tmax.

We find that the inclusion of these two criteria for efficient angular momentum transport and accretion within the disc, in addition to our existing framework for the treatment of LW radiation feedback, fundamentally alters the progression of structure formation in these haloes and impacts the observable characteristics of stars and the BHs within them.

We plot the Tvir-λ distribution of pristine atomic cooling haloes that are exposed to JLWJcrit, for different values of Qc in Fig. 1. The haloes with spin in the range λ < λmax are marked in red, with the almost-vertical solid curve representing λmax, whereas the ones with λ > λmax are plotted in black. The size constraint in order for the disc to withstand fragmentation is denoted by the dashed curve. These limits together constrain DCBH formation to a small allowed domain in the Tvir-λ plane marked by the yellow region.

Figure 1.

The temperature/spin distribution of pristine massive DM haloes that are exposed to JLWJcrit. The virial temperature, Tvir, is plotted against the halo spin parameter, λ, for all DCBH candidates with Toomre stability parameter Qc = 1.5, 2, 3. The nearly vertical solid curve represents the value of λmax and all haloes with spin less than this critical value lie to the left of this line (red points), whereas those with larger spin lie to the right (black points). The upper limit on the scalelength of the hosted disc given the allowed Tvirmax combination is marked as the dashed line. Any halo in the yellow region, i.e. below the dashed line and to the left of the solid vertical curve will host a DCBH (blue points) in our model. Note that the yellow region shrinks as Qc decreases, thereby reducing the probability of finding a halo that can form a DCBH. Inset: a zoom-in on the Tvir-λ distribution of the four DCBH candidates in our fiducial case with Qc = 3.

Figure 1.

The temperature/spin distribution of pristine massive DM haloes that are exposed to JLWJcrit. The virial temperature, Tvir, is plotted against the halo spin parameter, λ, for all DCBH candidates with Toomre stability parameter Qc = 1.5, 2, 3. The nearly vertical solid curve represents the value of λmax and all haloes with spin less than this critical value lie to the left of this line (red points), whereas those with larger spin lie to the right (black points). The upper limit on the scalelength of the hosted disc given the allowed Tvirmax combination is marked as the dashed line. Any halo in the yellow region, i.e. below the dashed line and to the left of the solid vertical curve will host a DCBH (blue points) in our model. Note that the yellow region shrinks as Qc decreases, thereby reducing the probability of finding a halo that can form a DCBH. Inset: a zoom-in on the Tvir-λ distribution of the four DCBH candidates in our fiducial case with Qc = 3.

Note that LN06 and LN07 require the gas disc to be marginally stable, i.e. Qc ∼ O(1). Given that the actual high-redshift disc parameters are uncertain, we choose values of Qc close to unity and use Qc = 3 in our fiducial model, which sets an upper limit on the number of DCBHs with reasonable disc parameters and for which the disc sizes are not too large. This yields DCBHs (blue points) with a comoving number density of 0.03 Mpc− 3 in our fiducial case with Qc = 3 and flim = 0.75 (see the following section for flim and Table 1 for a summary of parameters used for the various runs).

Star formation

In our model, Pop III stars form in pristine haloes subject to the following physical prescriptions/effects, discussed in A12 in more detail.

  • Pop III star formation is prohibited due to LW feedback in pristine haloes with $$2000 \le \rm T_{\rm vir}< 10^4\,\rm K$$ even when JLW < Jcrit (Machacek, Bryan, & Abel 2001; O’Shea & Norman 2008). A pristine minihalo that is subject to even a small value of LW radiation needs to be above a characteristic mass to host a Pop III star due to the partial dissociation (and hence inefficient cooling) of H2 molecules.

  • Pop III stars form following a top-heavy Salpeter IMF with mass limits dependent on halo's virial temperature, i.e. a single star with mass cut-offs at [100, 500] M in haloes with $$2000 \le \rm T_{\rm vir}< 10^4\,\rm K$$ and 10 stars with mass cut-offs at [10, 100] M in haloes with Tvir ≥ 104 K.

We consider a halo polluted if it has hosted a star or merged with a halo hosting a star. We set a mass threshold of M > 108 M for polluted haloes to form Pop II stars (Kitayama et al. 2004; Whalen et al. 2008; Muratov et al. 2012), following the reasoning that a halo needs to be massive enough to allow for the fall back or the retention of metals ejected from a previous Pop III star formation episode. In these polluted haloes, baryons are allowed to coexist in the form of hot non-star-forming gas, cold star-forming gas, stars, or those locked into a DCBH that might have formed in or ended up in the halo through a merger.

We assume in our model that a DM halo is initially comprised of hot gas, Mhot = fbMDM, where fb is the universal baryon fraction and MDM is the halo's current DM mass2. We add non-star-forming gas to the halo by calculating the accretion rate, $$\dot{M}_{\rm acc}$$, defined as  

(4)
\begin{equation} \dot{M}_{\rm acc} \equiv \frac{f_{\rm b} \Delta M_{\rm DM} - M_{\rm *,p} - M_{\rm out, p} - M_{{\rm BH}}}{\Delta t} \ . \end{equation}
In this model ΔMDM is the amount by which the DM halo grows between two snapshots separated by Δt years. M*, p and Mout, p represent the total stellar mass and net mass lost (from both cold and hot gas reservoir) in previous SN outflows at the beginning of the time-step, respectively. MBH is the total mass of the DCBH in the halo.

The hot gas, Mhot, converts into cold gas, Mcold, by collapsing over the dynamical time3 of the halo, tdyn. Pop II star formation can then occur via a Kennicutt-type relation (Kennicutt 1998)  

(5)
\begin{equation} \dot{M}_{*,\rm II}=\frac{\alpha }{0.1 t_{\rm dyn}}M_{\rm cold} \ , \end{equation}
where α is the SFE set to 0.005 and the factor 0.1tdyn is motivated by the angular momentum conservation condition for the central galaxy in a DM halo (Mo et al. 1998; Kauffmann et al. 1999, and see A12 for a descriptions of the parameters used).

In Pop II star-forming haloes, the outflow rate due to SN feedback is computed via the relation$$\dot{M}_{\rm out} =\gamma \ \dot{M}_{*,\rm II}$$,   where $$\gamma = (\frac{V_{\rm c}}{V_{\rm out}})^{-\beta }$$ (Cole et al. 2000).

We set Vout = 110 km s− 1 and β = −1.74 resulting in typical values of γ ≈ 20 following the results of the high-resolution hydrodynamical simulations of the high-redshift Universe (Dalla Vecchia and Khochfar, in preparation).

We track the evolution of baryons with the following set of coupled differential equations for the individual baryonic components:  

(6)
\begin{equation} \dot{M}_{\rm cold} = \frac{M_{\rm hot}}{t_{\rm dyn}} - \dot{M}_{*,\rm II} - \dot{M}_{\rm out} - \dot{M}_{\rm BH, cold}\ , \end{equation}
 
(7)
\begin{equation} \dot{M}_{\rm hot} = \dot{M}_{\rm acc} -\frac{M_{\rm hot}}{t_{\rm dyn}} - \dot{M}_{\rm BH, hot} \ . \end{equation}
Equations (4)–(7) are solved numerically over 100 smaller time-steps between two snapshots.

Growth of a DCBH

Haloes hosting DCBHs are initially not massive enough and are not polluted enough to lead to Pop II star formation (Schneider et al. 2002). It is reasonable to assume that prior to the introduction of a stellar component, the gas reservoir is still massive enough to feed the central BH. At this stage, the accreting DCBH might appear as a miniquasar, but of essentially zero metallicity. Following this epoch, however, the BH grows by accreting gas available in the halo, unchallenged by any further star formation until Pop II stars start forming in the halo, or until the halo merges with another halo hosting stars. The stellar component and the BH from this point on begin to grow in tandem, marking the onset of the OBG phase. How the two components evolve in detail is sensitive to the accretion rate and subsequent merging history – a parameter space that we have explored extensively.

Once a DCBH forms in a halo, it is allowed to grow at a fixed fraction flim of the Eddington accretion rate, assuming that both the cold gas and hot gas can be accreted by the BH. The upper limit of the accretion rate is set by the parameter flim that we vary between individual runs. If the total gas available during our integration time-steps for accretion is less than this fraction, the total mass available sets the accretion rate. We run our model for flim = [1.5, 1.0, 0.75, 0.1] to explore the parameter space. From the model, at any given time-step, Δt (in Myr), the accretion efficiency computed from the gas reservoir is  

(8)
\begin{equation} f_{\rm model} = \ln \left( 1 + M_{\rm g}/M_{\rm BH}\right) \times \frac{ \epsilon }{(1-\epsilon )} \frac{\ 450\,{\rm Myr}}{\Delta t} \ , \end{equation}
where Mg is the total gas available in the halo at time-step Δt and MBH is the DCBH mass with the radiative efficiency, ϵ, set to 10 per cent. The accretion efficiency then used for the actual computation of the increase in the DCBH mass is  
(9)
\begin{equation} f_{\rm acc} = \min [f_{\rm model}, f_{\rm lim}]. \end{equation}

Finally, we write  

(10)
\begin{equation} M_{\rm BH, final}=M_{\rm BH, ini} \exp \left(f_{\rm acc} \frac{1-\epsilon }{\epsilon }\frac{\Delta t}{450\,{\rm Myr}}\right). \end{equation}

Our fiducial case corresponds to an LW escape fraction of 1.0, Pop II SFE of 0.005, Qc = 3 and flim = 0.75. Note that the number of direct collapse (DC) sites are directly dependant on fesc and α, where increasing the values of those parameters leads to a higher number of DC sites (A12). The number of DCBHs that form from those sites directly depends on Qc, where a higher value of Qc leads to a higher number of DCBHs. The BH accretion parameters only affect the mass accreted by the DCBH as seen in Fig. 2 (see Section 3). In haloes which host a DCBH but are not massive enough to form Pop II stars, the gas is assumed to be hot and diffuse (i.e. has not condensed over the dynamical time of the halo). In haloes which host a DCBH and a Pop II stellar component, both the hot and cold phases of gas are assumed to contribute to the accretion process. The total mass accreted by the DCBH is split into hot and cold components depending on the ratio of the hot and the cold gas reservoirs. We do not assume any feedback from the accreting DCBH affecting star formation in the galaxy. We do this to avoid inserting a correlation between the BH and stars by assuming such a feedback loop since the precise nature of accretion and feedback in galactic nuclei is largely unknown at z > 6.

Figure 2.

Predicted redshift evolution of MBH and M* for the four OBGs in our simulation down to z = 6. We show the tracks for different fractions of the Eddington accretion rate flim = 1.5, 0.75 (fiducial), 0.1. The parameter zg denotes the redshift when the galaxy becomes an OBG and first appears on this plot. The shaded portion represents the region where the BH's accretion rate would have to be larger than the Eddington limit for the galaxy to qualify as an OBG. Using solid black lines, we show the local MBH–Mbulge relation and the 1σ error in the fit (Häring & Rix 2004).

Figure 2.

Predicted redshift evolution of MBH and M* for the four OBGs in our simulation down to z = 6. We show the tracks for different fractions of the Eddington accretion rate flim = 1.5, 0.75 (fiducial), 0.1. The parameter zg denotes the redshift when the galaxy becomes an OBG and first appears on this plot. The shaded portion represents the region where the BH's accretion rate would have to be larger than the Eddington limit for the galaxy to qualify as an OBG. Using solid black lines, we show the local MBH–Mbulge relation and the 1σ error in the fit (Häring & Rix 2004).

To summarize, the total gas mass available for accretion, and hence the total mass accreted by the DCBH, depends on whether Pop II stars are forming in the halo or not. If there is no assembling Pop II stellar component, the DCBH is assumed to accrete from the hot gas reservoir, i.e. the limiting accretion efficiency, $$f_{\rm acc}^{\rm hot}$$, is determined via equations (8) and (9) using the hot gas (Mhot in equation 8) in the halo  

(11)
\begin{equation} M_{\rm BH, final}=M_{\rm BH, ini} \exp \left(f_{\rm acc}^{\rm hot} \frac{1-\epsilon }{\epsilon }\frac{\Delta t}{450\,{\rm Myr}}\right). \end{equation}
If the halo hosts a Pop II stellar component, the limiting accretion efficiency, $$f_{\rm acc}^{\rm hot + cold}$$, is determined via equations (8) and (9) using the hot and cold gas (Mhot + Mcold in equation 8) in the halo  
(12)
\begin{equation} M_{\rm BH, final}=M_{\rm BH, ini} \exp \left(f_{\rm acc}^{\rm hot+cold} \frac{1-\epsilon }{\epsilon }\frac{\Delta t}{450\,{\rm Myr}}\right). \end{equation}
The net BH mass accreted in a time-step, MBH, acc = MBH, finalMBH, ini, can then be written as a sum of the hot ($$M_{\rm BH, acc}^{\rm hot}$$) and cold ($$M_{\rm BH, acc}^{\rm cold}$$) components  
(13)
\begin{equation} M_{\rm BH, acc} = M_{\rm BH, acc}^{\rm hot} + M_{\rm BH, acc}^{\rm cold}. \end{equation}
The individual masses are computed as  
(14)
\begin{equation} M_{\rm BH, acc}^{\rm cold}=RM_{\rm cold}, \end{equation}
 
(15)
\begin{equation} M_{\rm BH, acc}^{\rm hot}=(1-R)M_{\rm hot}, \end{equation}
where $$R = \frac{M_{\rm cold}}{M_{\rm hot}}$$, if Mcold < Mhot, else $$R = \frac{M_{\rm hot}}{M_{\rm cold}}$$. $$M_{\rm BH, acc}^{\rm cold}$$ and $$M_{\rm BH, acc}^{\rm hot}$$ are then used in equation (6) and (7) to compute the updated hot and cold gas fractions.

Table 1.

Summary of cases considered in our work.

Name Symbol Value 
Pop II SFE α 0.005 
LW escape fraction fesc 1.0 
Radiative efficiency ϵ 0.1 
Limiting Eddington accretion fraction flim 0.1–1.5 
Toomre parameter Qc 1.5–3 
Name Symbol Value 
Pop II SFE α 0.005 
LW escape fraction fesc 1.0 
Radiative efficiency ϵ 0.1 
Limiting Eddington accretion fraction flim 0.1–1.5 
Toomre parameter Qc 1.5–3 

RESULTS

For Qc = 3, we find four OBGs, named O1–O4, in our simulation box. The stellar and BH growth tracks of these OBGs are shown in Fig. 2, colour coded as O1 – blue, O2 – red, O3 – green and O4 – purple, respectively. The fiducial case (flim = 0.75) is marked by the solid curves and the open triangles and circles denote the time-steps at flim = 0.1, 1.5. The grey shaded region is where the BH would need to accrete at super-Eddington rates for the galaxy to appear as an OBG. Since these objects have MBH > M*, the nuclear emission can dominate the starlight even when accreting at significantly sub-Eddington rates in the non-shaded region.

Note that these OBGs preferably form in low-mass atomic cooling haloes as seen in Fig. 1. Almost all the DC candidate haloes meet the spin cut, but only the lower mass haloes, close to 107 M, meet the size cut to allow for the formation of the DCBH. We also report that DCBH host haloes are in fact satellites of larger haloes hosting Pop II stars, in which the DCBH haloes eventually end up. This is expected as the critical value of the LW radiation is generally produced by the larger star-forming haloes forming a close pair with the DC candidate halo. The abundances of these objects at z ∼ 6 is similar to the ones reported by Volonteri & Begelman (2010), for their ‘low-threshold’ case of DC seed formation.

Observational predictions

After identifying the sites of DC (A12), inclusion of physical processes that leads to their formation (LN06) enables the calculation of the observational signatures of OBGs. Haloes that harbour growing DCBH seeds with no (or little) associated stellar component merge into haloes that have formed the first and second generation of stars. We compute the observed spectral energy distribution (SED) of these copiously accreting DCBH seeds and the population of stars within the OBGs. To model the stellar component of the SED, we use the stellar masses and ages from our merger tree and derive the spectrum using starburst99 (Leitherer et al. 1999). The accretion disc spectrum is modelled as a radiating blackbody with a temperature profile of the disc given by the alpha-disc model (Pringle 1981). Since OBGs are expected to lie in haloes of low metallicity, we do not include any dust absorption in our models. Note that an OBG is characterized by possessing an actively accreting BH and an underlying stellar population that is Pop III or Pop II or both; however, an OBG might appear as what has often been referred to as miniquasars in the very early stages of its evolution when no stellar component is found in the DCBH host galaxy.

The ultraviolet (UV)–optical SED of an OBG is inevitably dominated by the accretion on to the central BH. The predicted SEDs of the OBGs over the wavelength range observable with the Near-Infrared Camera (NIRCAM) and Mid-Infrared Instrument (MIRI) instruments aboard NASA's proposed James Webb Space Telescope (JWST) are shown in Fig. 3. The stellar spectrum dominates the OBG spectrum only when the BH is limited to < 0.1 fedd at all times. However, such a low rate would be incompatible with the dramatic mass growth rates expected of the most massive, early BHs (e.g. Sijacki, Springel & Haehnelt 2009).

Figure 3.

OBG candidates and their observability with JWST at z ∼ 6. We plot the observed flux density in AB magnitudes for all the OBGs, while varying the maximal allowed accretion rate, flim. When the BH accretion dominates the spectrum there is virtually no Balmer break, and the UV slope is obviously fixed by the accreting BH. The black points denote the flux limits and bandpass widths of NIRCam (circles) and MIRI (squares) wide filters, assuming a 10 000 s exposure with JWST and a signal-to-noise ratio of 10. The shaded area marks the wavelength region shortward Lyα, where intergalactic neutral hydrogen is expected to completely absorb the OBG signal.

Figure 3.

OBG candidates and their observability with JWST at z ∼ 6. We plot the observed flux density in AB magnitudes for all the OBGs, while varying the maximal allowed accretion rate, flim. When the BH accretion dominates the spectrum there is virtually no Balmer break, and the UV slope is obviously fixed by the accreting BH. The black points denote the flux limits and bandpass widths of NIRCam (circles) and MIRI (squares) wide filters, assuming a 10 000 s exposure with JWST and a signal-to-noise ratio of 10. The shaded area marks the wavelength region shortward Lyα, where intergalactic neutral hydrogen is expected to completely absorb the OBG signal.

We note that the magnitude of our brightest OBGs could be mAB ≈ 25, comparable to the brightest putative Lyman-break galaxies uncovered at z ≃ 7 in ground-based surveys such as UltraVISTA (Bowler et al. 2012). However, it will be hard to distinguish OBGs from Lyman-break galaxies with ground-based imaging because, while the predicted UV continuum BH emission is expected to be relatively blue (with a UV slope β ∼ −2.3, where fλ ∝ λβ), it is not significantly bluer than that displayed by the general galaxy population at these early times (e.g. Dunlop et al. 2012). OBGs are also of course expected to display a negligible Balmer break. However, while it is interesting that the stack of the brightest Lyman-break galaxies in UltraVISTA shows at most a very weak Balmer break, it will still require extremely high-quality mid-infrared photometry to conclusively rule out the possibility that the UV–optical SED of a putative Lyman-break galaxy is incompatible with that produced by a very young stellar population.

Identification of OBGs amidst the observed high-redshift galaxy population will therefore require high-resolution imaging with Hubble Space Telescope (HST) and ultimately JWST. With MBH > M*, an OBG accreting at a reasonable fraction of the Eddington limit should certainly appear unresolved and point-like in high-resolution rest-frame UV (observed near-infrared) imaging. As illustrated in Fig. 4, it is already known that the vast majority of faint high-redshift galaxies uncovered via deep HST imaging are resolved (see Oesch et al. 2010; Ono et al. 2012), but this does not rule out the existence of an OBG population with a surface density <1 arcmin−2, and HST follow-up of the brighter and rarer high-redshift (z ≃ 7) objects uncovered by the near-infrared ground-based surveys is required to established whether or not they are dominated by central BH emission.

Figure 4.

Size versus magnitude relation. The Hubble Ultra Deep Field (HUDF)4 galaxies, not corrected for point spread function (PSF), at z = 6, 7 and 8 are represented by upright triangles, downward triangles and squares, respectively. Observational limits and the PSF of the NIRCam are plotted as the straight lines. The OBGs O1–O4 in our sample, denoted by the arrows pointing left, would be unresolved objects that could be brighter than the galaxies. Note that we have excluded the flim = 0.1 case from this plot as the mAB for the OBGs is quite high.

Figure 4.

Size versus magnitude relation. The Hubble Ultra Deep Field (HUDF)4 galaxies, not corrected for point spread function (PSF), at z = 6, 7 and 8 are represented by upright triangles, downward triangles and squares, respectively. Observational limits and the PSF of the NIRCam are plotted as the straight lines. The OBGs O1–O4 in our sample, denoted by the arrows pointing left, would be unresolved objects that could be brighter than the galaxies. Note that we have excluded the flim = 0.1 case from this plot as the mAB for the OBGs is quite high.

DISCUSSION AND CONCLUSION

In this study, we report the possible existence of OBGs at z > 6 in which the DCBH precedes the epoch of stellar assembly and outshines the stellar component in for a considerable fraction of galaxy's lifetime. Our 3.41 Mpc h−1 box produces about four of these OBGs. Although this is not a cosmological average owing to the small box size, the main aim of this study is to discuss the physical conditions that could lead to the existence of OBGs, which are effects that operate on a scale of less than a few tens of physical kpc, mostly insensitive to our chosen box size.

Besides the observational features discussed, like all active galactic nuclei, OBGs are expected to display broad-line emission from highly excited species in the vicinity of the BH. However, as OBGs have very low metallicity, it is unclear whether lines such as N v and C iv are expected to be detectable even given high-quality near-infrared spectroscopy, and Lyman α is often severely quenched by neutral Hydrogen as we enter the epoch of reionization.

Thus, the best observational route to establishing whether OBGs exist, and if so constraining their number density (and ultimately their evolving luminosity function), appears to be via deep imaging of putative high-redshift Lyman-break galaxies, with sufficient angular resolution to prove they are unresolved, coupled with sufficiently accurate photometry to prove any point-like objects cannot be dwarf-star contaminants (see for example Dunlop 2013).

The discovery of an OBG could in principle settle the long standing debate on whether DCBHs can form and be the seeds of the first SMBHs (A12; Dijkstra et al. 2008; Volonteri, Lodato, & Natarajan 2008; Volonteri & Natarajan 2009; Bellovary et al. 2011; Johnson, Whalen, & Holz 2012). Uncovering this population holds great promise for understanding the onset of the BH and host galaxy growth.

The authors would like to thank Rychard Bouwens for providing the HUDF data used in Plot. 4. BA would like to thank Jarrett L. Johnson for his useful comments on the first draft. JSD acknowledges the support of the European Research Council via an Advanced Grant and the support of the Royal Society through a Wolfson Research Merit award.

1
Note that Jcrit is the critical level of LW radiation required by a pristine atomic cooling halo to undergo direct collapse. The critical level of extragalactic LW radiation required by a pristine atomic cooling halo from Pop III stars is ∼1000 and from Pop II stars ∼30–100 (Wolcott-Green, Haiman, & Bryan 2011).
2
Using a lower baryon fraction linearly affects the BH mass and evolution discussed in this work.
3
tdyn is defined as the ratio of the halo's virial radius to the circular velocity defined for the infall mass and infall redshift.
4
Private communication with Rychrad Bouwens.

REFERENCES

Agarwal
B.
Khochfar
S.
Johnson
J. L.
Neistein
E.
Vecchia
C. D.
Livio
M.
MNRAS
 , 
2012
, vol. 
425
 pg. 
2854
  
A12
Bellovary
J.
Volonteri
M.
Governato
F.
Shen
S.
Quinn
T.
Wadsley
J.
ApJ
 , 
2011
, vol. 
742
 pg. 
13
 
Bowler
R. A. A.
, et al.  . 
MNRAS
 , 
2012
, vol. 
426
 pg. 
2772
 
Cole
S.
Lacey
C. G.
Baugh
C. M.
Frenk
C. S.
MNRAS
 , 
2000
, vol. 
319
 pg. 
168
 
Davis
A. J.
Natarajan
P.
MNRAS
 , 
2010
, vol. 
407
 pg. 
691
 
Dijkstra
M.
Haiman
Z.
Mesinger
A.
Wyithe
J. S. B.
MNRAS
 , 
2008
, vol. 
391
 pg. 
1961
 
Dunlop
J. S.
Astrophys. Space Sci. Libr.
 , 
2013
, vol. 
396
 pg. 
223
 
Dunlop
J. S.
, et al.  . 
2012
 
preprint (arXiv:1212.0860)
Fabian
A. C.
Wilman
R. J.
Crawford
C. S.
MNRAS
 , 
2002
, vol. 
329
 pg. 
L18
 
Ferrarese
L.
Merritt
D.
ApJ
 , 
2000
, vol. 
539
 pg. 
L9
 
Haehnelt
M. G.
Kauffmann
G.
MNRAS
 , 
2000
, vol. 
318
 pg. 
L35
 
Haiman
Z.
Abel
T.
Rees
M. J.
ApJ
 , 
2000
, vol. 
534
 pg. 
11
 
Häring
N.
Rix
H.-W.
ApJ
 , 
2004
, vol. 
604
 pg. 
L89
 
Hirschmann
M.
Khochfar
S.
Burkert
A.
Naab
T.
Genel
S.
Somerville
R. S.
MNRAS
 , 
2010
, vol. 
407
 pg. 
1016
 
Hopkins
P. F.
Murray
N.
Thompson
T. A.
MNRAS
 , 
2009
, vol. 
398
 pg. 
303
 
Johnson
J. L.
Whalen
D. J.
Holz
D. E.
2012
 
preprint (arXiv:1211.0548)
Kauffmann
G.
Colberg
J. M.
Diaferio
A.
White
S. D. M.
MNRAS
 , 
1999
, vol. 
303
 pg. 
188
 
Kennicutt
R. C.
ApJ
 , 
1998
, vol. 
498
 pg. 
541
 
King
A.
ApJ
 , 
2003
, vol. 
596
 pg. 
L27
 
Kitayama
T.
Yoshida
N.
Susa
H.
Umemura
M.
ApJ
 , 
2004
, vol. 
613
 pg. 
631
 
Leitherer
C.
, et al.  . 
ApJ
 , 
1999
, vol. 
123
 pg. 
3
 
Lodato
G.
Natarajan
P.
MNRAS
 , 
2006
, vol. 
371
 pg. 
1813
  
LN06
Lodato
G.
Natarajan
P.
MNRAS
 , 
2007
, vol. 
377
 pg. 
L64
  
LN07
Machacek
M. E.
Bryan
G. L.
Abel
T.
ApJ
 , 
2001
, vol. 
548
 pg. 
509
 
Mo
H. J.
Mao
S.
White
S. D. M.
MNRAS
 , 
1998
, vol. 
295
 pg. 
319
 
Muratov
A. L.
Gnedin
O. Y.
Gnedin
N. Y.
Zemp
M.
2012
 
preprint (arXiv:1212.0909)
Natarajan
P.
Treister
E.
MNRAS
 , 
2009
, vol. 
393
 pg. 
838
 
Oesch
P. A.
, et al.  . 
ApJ
 , 
2010
, vol. 
709
 pg. 
L21
 
Ono
Y.
, et al.  . 
2012
 
preprint (arXiv:1212.3869)
O’Shea
B. W.
Norman
M. L.
ApJ
 , 
2008
, vol. 
673
 pg. 
14
 
Pringle
J. E.
ARA&A
 , 
1981
, vol. 
19
 pg. 
137
 
Robertson
B.
Hernquist
L.
Cox
T. J.
Matteo
T. D.
Hopkins
P. F.
Martini
P.
Springel
V.
ApJ
 , 
2006
, vol. 
641
 pg. 
90
 
Schneider
R.
Ferrara
A.
Natarajan
P.
Omukai
K.
ApJ
 , 
2002
, vol. 
571
 pg. 
30
 
Sijacki
D.
Springel
V.
Haehnelt
M. G.
MNRAS
 , 
2009
, vol. 
400
 pg. 
100
 
Silk
J.
Rees
M. J.
A&A
 , 
1998
, vol. 
331
 pg. 
L1
 
Tegmark
M.
Silk
J.
Rees
M. J.
Blanchard
A.
Abel
T.
Palla
F.
ApJ
 , 
1997
, vol. 
474
 pg. 
1
 
Thompson
T. A.
Quataert
E.
Murray
N.
ApJ
 , 
2005
, vol. 
630
 pg. 
167
 
Treister
E.
Schawinski
K.
Volonteri
M.
Natarajan
P.
Gawiser
E.
Nat
 , 
2011
, vol. 
474
 pg. 
356
 
Volonteri
M.
Begelman
M. C.
MNRAS
 , 
2010
, vol. 
409
 pg. 
1022
 
Volonteri
M.
Lodato
G.
Natarajan
P.
MNRAS
 , 
2008
, vol. 
383
 pg. 
1079
 
Volonteri
M.
Natarajan
P.
MNRAS
 , 
2009
, vol. 
400
 pg. 
1911
 
Whalen
D.
O’Shea
B. W.
Smidt
J.
Norman
M. L.
ApJ
 , 
2008
, vol. 
679
 pg. 
925
 
Wolcott-Green
J.
Haiman
Z.
Bryan
G. L.
MNRAS
 , 
2011
, vol. 
1673