Unveiling the First Black Holes With JWST:Multi-wavelength Spectral Predictions

The discovery of a population of bright quasars at z ≳ 6, powered by accretion onto ${10}^{8-10}\,{M}_{\odot }$ black holes (BHs, e.g., Fan et al. 2001; Mortlock et al. 2011; Wu et al. 2015), has presented a challenge to our current understanding of supermassive black hole (SMBH) formation and growth: in particular, how these SMBHs could have assembled such masses so rapidly, within the first billion years after the Big Bang. The initial seed black holes from which these high-z quasars are predicted to form include several distinct channels (for reviews, see Volonteri 2012; Natarajan 2014; Latif & Ferrara 2016) that principally involve the production of either light or massive seeds.

In one scenario, the initial black hole seeds are essentially stellar remnants of the first stars that formed and evolved in the gas-rich early universe. The first stars in the universe, referred to as Population III (Pop III hereafter), formed out of H₂-cooled, metal-free gas when the universe was about 200 Myr old (z ∼ 25) onward. Pop III stars may leave behind black holes with masses ${M}_{\bullet ,{III}}\sim 100\,{M}_{\odot }$, which might grow, via Eddington accretion, to a mass of $\sim {10}^{9}\,{M}_{\odot }$ by z ∼ 6 powering the bright quasars that we now detect. However, this growth history requires optimistic post-seed formation conditions, such as steady accretion powered by uninterrupted gas supply, and inefficient radiation feedback from the growing seed on the ambient gas (see e.g., Tanaka et al. 2012; Pacucci et al. 2015b; Park et al. 2016). There is also considerable uncertainty in the initial masses of the Pop III stars: if the initial mass function of Pop III is skewed low, a dense star cluster is expected to form, in which the mass growth of a single remnant could potentially be boosted via super-Eddington accretion as it random walks in the cluster. After that, it could coalesce rapidly in the environment with other individual low-mass black holes, to yield a seed of $\sim {10}^{3}\mbox{--}{10}^{4}\,{M}_{\odot }$ (Devecchi & Volonteri 2009; Alexander & Natarajan 2014).

Several alternate scenarios envisage the production of massive black hole seeds with initial masses that lie between ${10}^{3-5}\,{M}_{\odot }$ (Eisenstein & Loeb 1995; Omukai 2001; Oh & Haiman 2002; Bromm & Loeb 2003; Koushiappas et al. 2004; Lodato & Natarajan 2006). The formation of direct collapse black holes (DCBHs) is one channel that is of great current interest. This is due to the fact that the physical conditions to activate this channel are naturally available in the early universe in halos where fragmentation and star formation can be curtailed. These sites where DCBH collapse can proceed are satellite halos of copiously star-forming (SF) galaxies. Massive initial star formation leads to the production of Lyman–Werner radiation (LW) radiation with energy $11.2\leqslant h\nu \leqslant 13.6\ \mathrm{eV}$. Photons in this band can photo–dissociate H₂ via ${{\rm{H}}}_{2}+{\gamma }_{\mathrm{LW}}\to {\rm{H}}+{\rm{H}}$ (see e.g., Omukai 2001). Atomic H is an inefficient coolant that can only cool gas down to 8000 K; H₂, on the other hand, can cool the gas down to ∼200 K. Therefore, the presence of an external critical LW radiation field from a neighboring early SF galaxy can prevent a pristine halo from forming Pop III stars. This delay, in turn, can instead lead to the isothermal collapse of the gas at 8000 K to extremely high densities, causing a runaway process that leads to the formation of a DCBH with ${M}_{\bullet ,\mathrm{DC}}\sim {10}^{4-5}\,{M}_{\odot }$ (Bromm & Loeb 2003; Begelman et al. 2006; Lodato & Natarajan 2006; Volonteri et al. 2008; Shang et al. 2010; Johnson et al. 2012; Ferrara et al. 2014). These more massive initial seeds can rapidly grow to SMBHs, even via sub-Eddington accretion, thus circumventing the timing challenges faced by the Pop III seed scenario.

There has been much recent progress in exploring the formation of DCBH seeds, their environment, and their early growth history (Volonteri et al. 2008; Agarwal et al. 2012, 2016a; Natarajan 2014; Pacucci & Ferrara 2015; Pacucci et al. 2015b, 2017; Pallottini et al. 2015). Work thus far has suggested that light and massive seeding models can, in principle, be distinguished with multi-wavelength data (Treister et al. 2013; Weigel et al. 2015). Attempts to do so have revealed that the most massive black holes at every epoch (that is, the massive end of the black hole mass function, were likely seeded by DCBHs (Volonteri et al. 2008; Natarajan & Volonteri 2012).

At high redshifts, it has been shown that there are distinct phases in the assembly history of light versus massive seeds (Volonteri et al. 2008; Agarwal et al. 2013). Due to the initial high masses of DCBH seeds, these black holes would have masses comparable to the stellar population in their host dark matter halos during their early mergers and evolution⁶ , in sharp contrast to hosts that harbor a light seed wherein the stellar mass would exceed the central BH mass at early times. This implies a distinctly different relationship between properties of the stellar population and the central black hole for DCBH seeds during early epochs, compared to what is seen locally (Ferrarese & Merritt 2000; Tremaine et al. 2002). Therefore, for DCBH seeds, we infer the occurrence of an early transient stage, the obese black hole galaxy stage (OBG stage; Agarwal et al. 2013), during which accretion onto the DCBH is expected to outshine the stellar population, resulting in distinct, detectable spectral signatures in the infrared and X-ray bands. Direct observational detection of rapidly growing black holes at early epochs, we argue here, is critical to probe the nature of the seeds, and direct evidence for DCBHs with a sub-dominant stellar component would be particularly telling. Here, we present the first multi-wavelength calculations of the detailed spectra for this OBG stage of growing DCBH and Pop III seeds, with particular attention focused on features that will be detectable by instruments aboard the upcoming James Webb Space Telescope (JWST) mission. The Near-Infrared Camera (NIRCam) is sensitive in the 1–5 μm range, whereas the Mid-Infrared Instrument (MIRI) is sensitive to detecting wavelengths from 5 to 28 μm. We note that, at the same epochs, these two seed populations have significantly different accretion histories revealed via their distinct signatures in the X-ray and IR wavelengths.

This paper is organized as follows. In Section 2, we present a brief overview of the massive seed formation channel and the unique properties during the OBG stage. In Section 3, we present the methodology adopted to compute the multi-wavelength spectra for growing light and massive initial seeds. Our results are presented in Section 4. We conclude with a discussion of the implications and future prospects for detection of growing BH seeds in the high-redshift universe.

One motivation to explore the DCBH seed formation channel is that viable physical conditions for their formation appear to be amply available in the early universe, as shown by Agarwal et al. (2012). The dark matter halo hosts of DCBHs, as per standard structure formation in cosmology, initially start off as mini-halos with a virial temperature of $2000\lt T\lt {10}^{4}$ K. The first episode of star formation in these halos could be delayed by a modest level of external LW radiation (Machacek et al. 2001; O'Shea & Norman 2008), keeping them pristine until they grow to the atomic cooling limit $T\sim {10}^{4}$ K. At this point, if the external LW radiation field prohibits molecular hydrogen formation by dissociation of any H₂ that may be present into H, DCBH formation would ensue due to the suppression of H₂ cooling and the inhibition of fragmentation into Pop III stars (Agarwal et al. 2012, 2014). Such conducive sites were found to be satellite halos in the vicinity of the first generation of SF galaxies in the early universe (Agarwal et al. 2013, 2014). At these satellite sites, gas-rich proto-galactic disks could form that do not fragment or cool but rather go dynamically unstable (Q-unstable, see Toomre 1964), leading to rapid runaway accretion of gas to the center and the formation of a massive DCBH seed. The physics and evolution of this process have been calculated analytically (Lodato & Natarajan 2006, 2007) and conducive conditions (assembly of massive proto-galactic disks) are increasingly seen to occur in state-of-the-art high-resolution cosmological simulations of early structure formation (Regan & Haehnelt 2009; Choi et al. 2013; Latif et al. 2013). It is estimated that the typical masses of these DCBH seeds would lie in the range ${10}^{4-5}\,{M}_{\odot }$, and the process is limited entirely by the gas reservoir (e.g., Johnson et al. 2012; Park et al. 2016; for a theoretical derivation, see Ferrara et al. 2014). Furthermore, radiative feedback from these assembling DCBH seeds could halt SF in these halos (Aykutalp et al. 2014). Pacucci et al. (2015a) have shown that $\gt 90$% of the gas in these proto-galactic core regions is actually accreted, and the rest blown away in ${T}_{\mathrm{vir}}\sim {10}^{4}$ K DCBH host halos, under radiatively inefficient conditions. Therefore, once assembled, the radiative feedback from these growing DCBHs expels gas from the host halo, preventing star formation in the host halo.

Tracking merging histories of dark matter halos, we find that these DCBHs' host satellite halos rapidly merge (within ${10}^{6}\,$ year) with the parent SF halo in their vicinity. This leads to the swift acquisition of a stellar component by the DCBH halo. This stage corresponds to the formation of a special transitory class of objects that we have referred to as OBGs (Agarwal et al. 2013). The OBG stage is the result of the DCBH halo merging with the LW source galaxy and acquiring its principal stellar component. This is under the assumption that star formation is halted or highly suppressed in the DCBH halo itself, due to radiative feedback from the assembling seed. Meanwhile, stars and the DCBH continue to grow post-merger, self-consistently in the sense that the same gas reservoir likely feeds the BH and forms stars in the merged remnant during this transient stage. These transitory OBGs at high redshift will lie far above the locally (z = 0) measured correlation between the mass of the central black hole and stars in the bulge, where this ratio is of the order of 0.1% (e.g., Häring & Rix 2004). OBGs are expected to start off well above the local ${M}_{\mathrm{BH}}\mbox{--}\sigma$ relation, and eventually proceed onto it via mergers and accretion (Agarwal et al. 2013). This is once again distinct compared to halos that harbor lighter seeds, i.e., systems hosting Pop III black holes, where the BH seed must grow substantially to start with, and then perhaps grow in tandem with the stellar component, to end up on the $M\mbox{--}\sigma$ relation by z = 0. The life-cycle of Pop III remnant seeds follows a different growth pathway as the black hole accretes mass, and a stellar component forms and grows in tandem, all likely modulated by feedback. In this paper, we follow the growth of an OBG and a growing Pop III seed, both at high redshift (z 9). As outlined above, the DCBH seeds form at the center of pristine, as-yet unenriched, satellite halos that are illuminated by Lyman–Werner radiation from a nearby SF galaxy (see Figure 1). Star formation is quenched in these DCBH host halos, and a stellar component is accrued promptly via merging with nearby SF galaxies. In the models studied here, we examine the subsequent growth following the merger of the DCBH (initial seed mass of ${10}^{5}\,{M}_{\odot }$) with a neighboring high- and low-metallicity stellar population. The gas in the post-merger halo is assumed to have the same metallicity as the stellar population. The same scenario is explored for a growing Pop III remnant seed, which starts out with an initial mass of 100 ${M}_{\odot }$. For both cases, we also explore growth via accretion from standard thin-disk accretion, as well as via a radiatively inefficient slim disk. In this treatment, we neglect the formation, merger. and growth of any BHs of lower stellar mass that might form in these Pop III halos.

Zoom In Zoom Out Reset image size

During the OBG phase for galaxies that host a rapidly growing DCBH seed, the energy output from BH accretion vastly exceeds the emission from the stellar component. In the next section, we outline our computation of the multi-wavelength SED of an OBG under these specific circumstances. We compare and contrast the growth histories and energy output during the same epochs derived for growing Pop III seeds, and home in on the features in the multi-wavelength spectra that permit discrimination between these two seeding models.

We compute the multi-wavelength SED for a growing DCBH seed in the OBG stage, as well as for a growing Pop III seed, using a synthetic model to calculate the combined contribution of fluxes from the stellar and the accreting black hole components. We assemble these two components for both the DCBH and the Pop III remnant seeding models in the context of the merging hierarchy of LCDM halos, using the self-consistent prescription developed in Agarwal et al. (2012). For instance, once a DCBH forms, its subsequent growth and energy output from accretion are simulated using a state-of-the-art 1D radiation-hydrodynamic code GEMS. The code simulates the spherical accretion onto a high-redshift black hole seed, computing the emitted luminosity self-consistently from the accretion rate. The full spectral analysis is then performed in a post-processing step with the code Cloudy (Ferland et al. 2013). A full description of GEMS is available in Pacucci & Ferrara (2015), Pacucci et al. (2015a, 2015b). In this work, we neglect the formation, merger, and growth of any stellar mass BHs that might form in addition to the primary seed in a Pop III halo.

The growth and resulting energy output for the DCBH seed and lighter Pop III remnant seed are modeled in two limiting cases. In the first, which we refer to as the standard case, the growth involves accretion via a classic α-disk (Shakura & Sunyaev 1976), which is geometrically thin and optically thick. This accretion disk is radiatively efficient and accretion occurs, on average, at the Eddington rate. As explained in Pacucci et al. (2015b), this regime is entirely feedback-limited and growth is stunted once the feedback from the accretion process heats the gas, thus disrupting the inflow. In addition to this BH growth mode, we also model the case of radiatively inefficient accretion via a slim disk (Paczynski & Abramowicz 1982; Abramowicz et al. 1988; Sadowski 2009; McKinney et al. 2014), in which radiation pressure is less efficient in quenching the gas inflow, due to radiation trapping. In this case, highly super-Eddington accretion rates may be reached for both massive and light seeds, and the mass growth is now entirely gas-supply limited. The slim disk solutions that we use capture the essential features of these complex super-Eddington flows, but not the details that require full GR-MHD treatment on the smallest scales.

The properties of the SF halo and the DCBH seed host halo that merge together are each computed using theoretical models that are built upon a suite of cosmological simulations (N-body: Agarwal et al. 2012, and hydrodynamical: Agarwal et al. 2014) that include physically well-motivated and self-consistent prescriptions for star formation and metal pollution from supernovae, enabling the construction of detailed star-formation histories (further details of the recipes used here can be found in Agarwal et al. 2014). These prescriptions are used to obtain an estimate of the global and local LW radiation field needed for DCBH seeding. Here, we study the accretion of metal-enriched gas onto the DCBH during the OBG stage, where the metallicity is indexed to that of the merged stellar population. At this juncture, the DCBH restarts its growth, driven by the merger. In addition to the two modes of accretion outlined above, we also explore two distinct metallicities for the OBG stage: (i) $Z=5\times {10}^{-2}\,{Z}_{\odot }$ and (ii) $Z\,=5\times {10}^{-4}\,{Z}_{\odot }$. The stellar populations that merge to form the OBG are also aged accordingly, with a lower metallicity of $5\times {10}^{-4}\,{Z}_{\odot }$ and an age of 350 Myr attributed to a younger Pop II, and a higher metallicity of $5\times {10}^{-2}\,{Z}_{\odot }$ and age of 800 Myr to an older Pop II stellar population. To model the stellar SED component of the OBG stage, we used Yggdrasil, a stellar population synthesis code that employs Raiter et al. (2010) for the lower metallicity, and Starburst99 (Leitherer et al. 1999) with Geneva high mass-loss tracks for the higher metallicity case. The nebular emission component was computed using Cloudy, adopting the parameters described in Zackrisson et al. (2011). The SEDs in all cases have been rescaled to the Kroupa (2001) stellar initial mass function.

Once the satellite DCBH host halo merges with the larger host halo of the SF galaxy, the black hole thereafter is assumed to accrete from the same gas reservoir that fuels star formation. This ensures that we do not exceed the total baryon budget at any point in the growth history of the OBG. The nebular emission is also computed assuming the same ambient gas metallicity as that of the stars. We model the post-merger star formation in the OBG host halo with a constant star-formation rate, leading to a final stellar mass of ${M}_{* }={10}^{7}\,{M}_{\odot }$ at the end of the OBG phase. The total duration of the OBG phase varies from ∼10 Myr in the slim disk cases to ∼100 Myr in standard accretion cases. Motivated by our prior theoretical work on the cosmological environments of OBGs (Agarwal et al. 2013), we then combine the stellar and the accreting DCBH components to obtain the self-consistent multi-wavelength template spectra for the various cases presented here, in Figures 2–5.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We also compute the multi-wavelength spectra, as well as the template spectra, in JWST MIRI and NIRCam bands for the lighter Pop III seed black holes under similar conditions; namely, growth via both the standard and the slim-disk mode, and in both high- and low-metallicity environments. Similar to the DCBH case, for the Pop III remnant starting with an initial seed mass of 100 ${M}_{\odot }$, we compute BH growth and star formation self-consistently until the stellar population grows to ${10}^{7}\,{M}_{\odot }$ as well. We follow the growth in tandem for the BH and the stellar component, using the GEMS code, and compute the output spectra as outlined above for the high- and low-metallicity cases, as well as the standard and slim-disk accretion modes. Once again, we model the stellar SED component of these growing Pop III seeds, using Yggdrasil for the lower-metallicity case, and Starburst99 with Geneva high mass-loss tracks for the higher-metallicity case. The nebular emission component was also computed once again for this scenario, using Cloudy and adopting the parameters described in Zackrisson et al. (2011). The SEDs were rescaled to the Kroupa (2001) stellar initial mass function. Here as well, once the Pop III remnant seed host halo merges with a larger host halo of a nearby SF galaxy, the black hole thereafter is assumed to accrete from the same gas reservoir that fuels star formation. This again ensures that we do not exceed the total baryon budget. The nebular emission is computed assuming the same ambient gas metallicity as that of the stars for this case as well. We model the post-merger star formation with a constant star-formation rate, yielding a final stellar mass of ${M}_{* }={10}^{7}\,{M}_{\odot }$ and comparable final BH mass. We do this to effectively compare the two seeding-model cases. The results of the computed multi-wavelength spectra for these Pop III remnants seeds for the various parameter choices are presented in Figures 4 and 5.

As shown in Figure 6, JWST has the power to discriminate between these two seeding models. In our analysis, we generate JWST photometry by integrating under each of the filters' sensitivity curves. These are available online for NIRCam and MIRI.⁷ We say that a source is detectable if it would have been detected at a signal-to-noise ratio greater than 10, after 10⁴ seconds of observation. Exposure time and sensitivity calculators used are also available online.⁸

Zoom In Zoom Out Reset image size

4.1. Observations of Early Black Holes

4.2. Searching for JWST OBG Candidates with Current Chandra, Spitzer, and Hubble Data

We explore data from current multi-wavelength surveys to derive selection criteria that we can use to data-mine for DCBH candidates in the OBG stage, as well as for potential contaminants. These were developed for sifting viable candidates from a landscape of template spectra in the viable OBG parameter space. In the previous section, we saw that DCBHs growing via slim-disk accretion, due to their suppressed X-ray fluxes (as can be seen in Figure 3) and Pop III sources, no matter how they accrete, are simply not luminous enough, and therefore cannot be found in existing X-ray data, even in the deepest exposures currently on-hand. Once again, looking at X-ray wavelengths in Figure 2, we note that DCBHs growing via standard disks in low-metallicity halos in the OBG stage are the most feasibly detected candidates and could be lurking in available deep Chandra X-ray data. We reiterate that these sources are characterized by high bolometric luminosities of the order of $\sim {10}^{44}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$, and by mild absorption in the IR. We therefore devise three-tiered criteria involving the demarcation of a color–color cut that takes into account the Lyman break (for redshift selection) and the shape of the SED in the IR, as well as the ratio of X-ray to rest-frame optical flux (which corresponds to observed wavelength of 2–4.5 μm), to isolate these candidates in JWST observations. Selecting a ratio of x-ray to rest-frame optical flux much greater than one picks out the presence of an actively accreting source, and as we show below, also helps the removal of confounding SF galaxies with no central AGN. As shown earlier, the color–color cut also ensures the removal of contaminating passive, dusty, SF galaxies at low redshift. Of course, we pre-select all candidates with no detection in the blue-bands (U and B-bands).

We now demonstrate the robustness of our OBG selection criteria, using existing deep Hubble, Spitzer, and Chandra data. Potential OBG candidates could be lurking in the low-luminosity AGN in the CANDELS data set, and contaminating SF galaxies could also be present in the same data. We use the CANDELS-GOODS-S photometric data, release version 1.1 (Guo et al. 2013) for the first color–color cut. For each object in the CANDELS catalog, we interpolate the IR photometry into three JWST filters: F_090W, F_200W, and F_444W. We pick the F_090W photometric band, as it allows a redshift selection for all sources at z > 6. The two additional bands, F_200W and F_444W, are chosen to probe the SED in the NIR and MIR regions, because these sources are expected to have much redder colors than other high-redshift contaminating sources. Given the high expected redshifts for these sources and IGM absorption, these filters are explicitly chosen to detect the Lyman break via the drop-out technique. We then determine the X-ray fluxes for all CANDELS sources, including both detected and marginally detected faint sources. The marginally detected faint sources are assigned an X-ray flux that is derived using the prescription developed in Cappelluti et al. (2016) and Basu-Zych et al. (2013). As per this prescription, CANDELS sources that do not have X-ray counter-parts are assigned X-ray fluxes via the relation between SFR, redshift, and X-ray luminosity. We then apply our multiple selection criteria to this combined catalog; the color–color cut, followed by the high ${F}_{X}/{F}_{444W}$ cut and over-plot on these diagnostics, plots our OBG model template spectra in the same filters as shown in Figure 7. For the purpose of this exercise, we produce only high-redshift template sources ($z\in \{9,11,13,15\}$), due to the requirement of low metallicity and appropriate Lyman–Werner background, for a range of DCBH seed masses (${M}_{\bullet }/({10}^{5}\,{M}_{\odot })\in \{0.2,0.4,0.6,0.8,1.0\}$).

Zoom In Zoom Out Reset image size

We find that the OBG stage can be isolated with the following criteria, which extend those reported in Pacucci et al. (2016):

• 1.  
  Pre-select sources with bolometric luminosity in excess of ${10}^{44}\,\mathrm{erg}\,{{\rm{s}}}^{-1}.$
• 2.  
  After pre-selection and eliminating blue sources, find dropouts in F_090W that are detected in both F_200W and F_444W with $({F}_{090W}-{F}_{220W})\gt 0$, and $-0.3\lt ({F}_{200W}\,-{F}_{444W})\lt 0.3.$
• 3.  
  Find an X-ray counterpart and its associated X-ray flux.
• 4.  
  Isolate sources with a high ratio of X-ray to rest-frame optical flux (${F}_{X}/{F}_{444W}\gg 1$).

In Figure 7, we apply the selection criteria to our template sources, as well as CANDELS sources with real (measured) and estimated (derived using correlations) X-ray fluxes. In the left panel, we display a color–color plot of our models, along with each of the CANDELS sources that would satisfy the first criterion. Depending on the metallicity of the gas, and the time at which the spectrum is computed, the IR spectral slope may vary slightly. In this work, we have selected the time slice that produces the highest X-ray flux. In the right panel, we display all sources that would have been detected in CDF-S (${F}_{x}\gt 6\times {10}^{-18}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}\,{{\rm{s}}}^{-1}$), but do not necessarily satisfy our first criterion. Using the ratio of X-ray flux to rest-frame optical flux, our OBG sources easily stand out among SF galaxies and AGN. There are no CANDELS sources in the GOODS-S field that would satisfy all our criteria. We note that the Pacucci et al. (2016) DCBH candidates are not in our CANDELS catalog, as we select sources in the OBG stage during which the X-ray flux is the highest. Our selection of the OBG stage is consistent with the merger-driven scenario that produces a bright quasar phase. We also explore the dependence of our selection criteria on initial DCBH seed mass and redshift, and report that decreasing the DCBH seed mass simply causes the overall flux to decrease, whereas the X-ray to rest-frame optical flux ratio is independent of seed mass. There is very little redshift-dependence on the measured colors, due to the fact that these spectra are flat in the IR, until about z = 15, at which point the Lyman break falls within F_200W. Sources that satisfy our selection criteria are optimal candidates for spectral follow-up with NIRSPEC; from our modeling work, we predict up to 10 sources within z = 6–10 in the CANDELS field. For instance, to detect the fiducial OBG continuum at ${M}_{\mathrm{AB}}\sim 26$ (as shown in the top panel of Figure 6), with NIRSPEC at 3 μm with a S/N ∼ 10, would take about 3–4 hr, and the detection of lines would take ∼1 hr of integration, as per the available exposure time calculators.

The origin of initial black hole seeds that eventually grow to produce the SMBHs observed in the high-redshift universe is currently debated. Traditionally, it was believed that the end states of the first (believed to be massive) stars to form produced remnants that yielded black hole seeds with masses between ∼10 and 100 ${M}_{\odot }$. The discovery of several luminous high-redshift quasars at z > 7 powered by ${10}^{9}\,{M}_{\odot }$ black holes when the universe was less than 7% of its present age causes a timing crunch. In order to alleviate this, it was proposed that initial black hole seeds that might form from the direct collapse of un-enriched pre-galactic gas disks may yield seeds with initial masses of ${10}^{4-5}\,{M}_{\odot }$. These massive DCBH seeds could then grow rapidly and evolve into the bright quasars that are seen. Because black holes erase the initial conditions of their mass assembly, we look to the properties of the host galaxy and environment to test and discriminate between these two models for light and massive seed production in the early universe. In particular, we exploit the unique relationship between black hole mass and the properties of the host galaxy—a key feature of the DCBH formation process—to discriminate seed formation physics. In terms of the physics, what really distinguishes light BH seeds from massive BH seeds is the epoch at which they have super-boosted growth, if we are to explain the origin of the SMBHs that are powering the highest redshift quasars. In order for light seed models to work, brief periods of super-Eddington accretion are required after seed formation—whereas for DCBHs, the process of formation itself likely involves super-Eddington accretion. Interestingly, as we show here for DCBHs, there is a transitory phase at high redshift, the OBG stage, when there is a unique signature vis-à-vis host galaxy properties that might provide a clue to the initial seed mass.

Contrary to the local ${M}_{\mathrm{bh}}\mbox{--}{M}_{\mathrm{bulge}}$ relation, wherein the central black hole mass is a small fraction, up to ∼0.5%, of the stellar mass of the bulge, at high redshifts we find that if SMBHs are seeded with massive DCBHs, then they transiently enter an OBG phase, characterized by ${M}_{\mathrm{BH}}/{M}_{* }\sim 1$ (Agarwal et al. 2013; Kormendy & Ho 2013). In this case, a unique set of detectable observational features are produced. Here, we present template multi-wavelength calculations of OBG spectra, and show that they are distinguishable from the lower-mass Pop III seeded systems, as well as other strong infrared-emitting sources like normal stellar populations at z ≳ 6, and dusty, old, and red galaxies at low redshifts. One noteworthy feature is that the accreting DCBH in an OBG completely outshines the stellar component at X-ray wavelengths >0.1 keV, as well as in the infrared. This allows us to pre-select viable OBG candidates, as a normal stellar population with a Pop III black hole seed by contrast would be undetected in the X-ray with current data, and will also remain undetectable by JWST due to the significantly lower reprocessed emission in the infrared. This allows us to immediately sift out infrared-bright OBG candidates for JWST observations. In the MIRI bands, we find that OBGs can be easily distinguished from contaminants, such as nearby quiescent red galaxies and high-redshift SF galaxies without a DCBH nucleus. To summarize, we derive three criteria in color–color space $({F}_{090W}-{F}_{200W})\gt 0$ and $-0.3\,\lt ({F}_{200W}-{F}_{444W})\lt 0.3$, and high ratio of X-ray flux to rest-frame optical flux (${F}_{X}/{F}_{444W}\gg 1$), to select for OBG candidates.

With the selection criteria derived, the obvious next question to address is the expected abundance of OBGs. Estimates of the expected abundance of both Pop III seeds and DCBH seeds are currently highly uncertain, and this is a topic of active research that is yet to be fully settled among several groups working on the formation of early black holes. In a recent paper that utilizes simulations to assess and compute the abundance of the sites including the appropriate conditions—Lyman–Werner flux—required to form DCBH seeds, Habouzit et al. (2016) estimate that DCBH seeds could account for the origin of all the observed high-redshift quasars at z > 6. Meanwhile, Pezzulli et al. (2017) note that we might be lacking a proper census in the X-rays for early black holes, either due to the uncertainty in the occupation fraction or prevalence of obscured super-Eddington accretion. Therefore, it is patently clear that the IR wavelengths are the most promising, as we have shown, to obtain demographic properties of this population. The abundance of these sources will, of course, determine the precise observational strategy to be pursued—either by JWST, as we propose here, or with shallower, larger-area WFIRST surveys. Given the bright IR magnitudes that we estimate here for these template cases, it is clear that WFIRST surveys would be an efficient way to detect OBGs. In future work, we intend to explore these detection strategies in more detail.

A recent calculation by Ricotti (2016) reports an expected abundance for Pop III stars of approximately a few hundred per comoving Mpc³. These lighter Pop III black hole seeds are expected, however, to be significantly more abundant than DCBH seeds in the early universe. Using our state-of-the-art theoretical models that cover possible accretion modes and formation times, we predict an abundance with an upper limit range of $\approx 7.5\times {10}^{-4}\mbox{--}2.5\times {10}^{-5}$ for OBG systems per Mpc³. We estimate this abundance by extrapolating the fraction of pristine halos that are exposed to a critically high LW flux (Dijkstra et al. 2014). As a first-order estimate, using the peak of the critical flux distribution from Agarwal et al. (2016b) to estimate the number of viable hosts for DCBH formation and assembly, in the 800 arcmin² area of the CANDELS fields that corresponds to a $\sim 9\times {10}^{-3}$ Gpc³ volume, we expect about a hundred potential DCBH formation sites. However, not all of these sources will be DCBHs or in the OBG stage, and detectable by Chandra, because only a fraction of them are expected to be active. Guided by a recent theoretical estimate by Pacucci et al. (2015a) that takes into account the duty cycle of DCBHs, we predict up to 10 DCBH sources between z = 6–10 in the CANDELS fields. This is consistent with the recent study of Pacucci et al. (2016), in which the authors claimed the possibility of two DCBH candidates in CANDELS/GOODS-S. In future work, we plan to investigate the expected abundance using snapshots from cosmological simulations that include radiative transfer to track DCBH sites in the OBG stage.

As demonstrated in this paper, growing DCBH seeds in the OBG stage would be detectable primarily from their infrared excess in the JWST MIRI bands, compared to growing Pop III seeds and other contaminating sources. Detection and NIRSPEC follow-up, as well as ground-based spectroscopic follow-up, using upcoming 30 m class telescopes, of these sources during the OBG stage would confirm and improve our understanding of the direct collapse model as a viable channel for seed formation at high redshifts. We also anticipate that WFIRST surveys could feasibly help us detect this population of OBGs. Given the current uncertainties, the detection of an OBG population will suggest a higher formation efficiency of DCBH seeds and/or a more efficient accretion scenario. Our models suggest that sources comprising an accreting Pop III seed that has grown to a final black hole mass of ${10}^{7}\,{M}_{\odot }$ by z > 9 will be to too faint in the infrared wavelengths to be detectable by JWST. Therefore, we make the case that OBGs will be booming sources in the MIRI bands, and a sample of detected OBGs would resolve tension between the two SMBH seeding models by offering us a snapshot of quasars in the early stages of their lives.

P.N. acknowledges support from a Theoretical and Computational Astrophysics Network grant, award number 1332858, from the National Science Foundation. This work was completed at the Aspen Center for Physics, which is supported by National Science Foundation grant PHY-1066293. Useful conversations with Benny Trakhenbrot and Nicholas Stone at the Aspen Black Holes Workshop are gratefully acknowledged. B.A. and A.R. acknowledge support from the TCAN grant for a post-doctoral fellowship and a graduate fellowship, respectively. This research was supported in part (AF) by the National Science Foundation under Grant No. NSF PHY11-25915. E.Z. acknowledges funding from the Swedish Research Council (project 2011-5349). N.C. acknowledges the YCAA Prize Postdoctoral fellowship. N.C. and F.P. acknowledge the SAO Chandra grant AR6-17017B and NASA-ADAP grant MA160009.