---
title: Direct Collapse Black Holes
url: https://www.emergentmind.com/topics/direct-collapse-black-holes-dcbhs-de8a981d-4d18-404b-805f-5d75df1895f9
type: topic
---

# Direct Collapse Black Holes

to=arxiv_search.search  乐亚json
{"query":"Direct collapse black holes radio signals from early direct collapse black holes 2107.11307", "max_results": 5}
to=arxiv_search.search ＿天天json
{"query":"Role of magnetic fields in the formation of direct collapse black holes 2210.05611", "max_results": 5}
to=arxiv_search.search  手机天天彩票json
{"query":"3-cm Fine Structure Masers unique signature of supermassive black hole formation via direct collapse early universe 1601.04712", "max_results": 5}
to=arxiv_search.search  彩神争霸苹果json
{"query":"The Formation of Direct Collapse Black Holes at Cosmic Dawn and 21 cm Global Spectrum 2503.22130", "max_results": 5}
to=arxiv_search.search  天天中彩票是不是json
{"query":"Little Red Dots as Direct-collapse Black Hole Nurseries 2508.14897", "max_results": 5}
Direct-collapse black holes (DCBHs) are massive black-hole seeds, typically with initial masses of about $10^4$–$10^6\,M_\odot$, that are hypothesized to form at high redshift in atomic-cooling halos where $\mathrm{H}_2$ cooling is suppressed, the gas remains near $T \approx 8000$–$10^4\,\mathrm{K}$, and collapse proceeds through a supermassive-star or closely related intermediate stage rather than through ordinary star formation. In this framework, DCBHs are attractive progenitors of the first supermassive black holes because they start much heavier than stellar remnants, but their viability depends on a tightly coupled set of chemical, radiative, dynamical, and environmental conditions, and their post-birth growth is not guaranteed [2210.05611] [1205.6464] [2008.09120].

## 1. Formation pathway and defining conditions

In the direct-collapse pathway, the host is an atomic-cooling halo with virial temperature above the atomic threshold, conventionally $T_{\rm vir} \gtrsim 10^4\,\mathrm{K}$. A widely used scaling is
$$
T_{\rm vir} \approx 1.98\times10^4\,{\rm K}\,\Big(\frac{\mu}{0.6}\Big)\Big(\frac{M}{10^8\,h^{-1}\,M_\odot}\Big)^{2/3}\Big(\frac{\Omega_m \Delta_c}{18\pi^2}\Big)^{1/3}\Big(\frac{1+z}{10}\Big),
$$
which places the relevant halos at characteristic masses of $\gtrsim 10^7\,M_\odot$ at high redshift. The gas must remain pristine or nearly so, because metal-line cooling and dust-assisted fragmentation otherwise redirect collapse toward conventional star formation. In the idealized limit of suppressed molecular cooling, the collapse is nearly isothermal at $T \approx 8000\,\mathrm{K}$, and the inflow rate is set by the sound speed,
$$
\dot M \approx \frac{c_s^3}{G} \approx 0.1\,M_\odot\,{\rm yr}^{-1},
$$
which is sufficient to assemble a supermassive star of $M \approx 10^5\,M_\odot$ in $\lesssim 1$–$2\,\mathrm{Myr}$ before collapse to a massive seed black hole [2210.05611] [1205.6464].

The chemical bottleneck is the formation of $\mathrm{H}_2$, primarily through the $\mathrm{H}^-$ channel,
$$
\mathrm{H} + e^- \rightarrow \mathrm{H}^- + \gamma,\qquad
\mathrm{H}^- + \mathrm{H} \rightarrow \mathrm{H}_2 + e^-,
$$
and DCBH formation requires that this pathway be suppressed long enough for the halo to enter the atomic-cooling regime. This may be achieved by Lyman–Werner dissociation of $\mathrm{H}_2$, by photodetachment of $\mathrm{H}^-$, or by combinations of radiative and dynamical effects that keep the $\mathrm{H}_2$ cooling time above the relevant collapse timescale. A recent single-zone formulation makes this explicit by requiring the $\mathrm{H}_2$ cooling time to exceed the Hubble time until the gas reaches $T=10^4\,\mathrm{K}$, thereby defining an atomic-cooling-halo condition rather than an already fully monolithic collapse state [2509.25325].

Seed masses in this literature span roughly $10^4$–$10^6\,M_\odot$, with $10^5\,M_\odot$ often serving as the fiducial direct-collapse product. That range recurs across radiative-transfer, MHD, near-infrared, radio, and gravitational-wave studies, and is central to why DCBHs are considered heavy seeds rather than merely efficient stellar remnants [2107.11307] [1601.04712] [1502.04125].

## 2. Radiative regulation and the non-universality of the critical field

A persistent theme in DCBH theory is that there is no single universal critical radiation threshold. Earlier work often parameterized the condition for direct collapse by a scalar $J_{\rm crit}$ in units of $J_{21}$, but detailed SED-dependent calculations show that the relevant control parameters are the $\mathrm{H}_2$ photodissociation rate and the $\mathrm{H}^-$ photodetachment rate, and that the same $J_{21}$ can correspond to very different chemistry depending on the source spectrum. For realistic stellar populations, the allowed threshold spans
$$
J_{\rm crit} \sim 0.5\text{--}10^3,
$$
rather than a fixed number, and the direct-collapse boundary is better represented as a critical curve in the $(k_{\rm H_2},k_{\rm H^-})$ plane than as a universal flux level [1504.04042] [2503.22130].

This SED sensitivity is reinforced by environmental models. In cosmological clustering calculations, the local LW field from nearby Pop II sources can exceed the mean background by up to $\sim 10^6$, and the onset of Pop II star formation at $z\sim 16$ coincides with the onset of DCBH formation because Pop II spectra are more efficient than Pop III spectra at pushing pristine atomic-cooling halos above the relevant threshold. In that framework, Pop III sources alone do not reach the required conditions, whereas local Pop II irradiation can do so, especially in clustered environments where source–target separations are only a few kpc [1205.6464].

A more restrictive but physically explicit version is the synchronized-halo scenario. High-resolution radiation-hydrodynamics simulations show that a low-level background field of
$$
J_{\rm BG} \gtrsim 100\,J_{21}
$$
can delay prior star formation, while a nearby star-bursting primary halo at separation $R_{\rm sep}\lesssim 300\,\mathrm{pc}$, turning on within $T_{\rm sync}\lesssim 4\,\mathrm{Myr}$ of secondary collapse, can drive the secondary onto the nearly isothermal atomic-cooling track without photoevaporating it or polluting it with metals, provided the primary remains luminous for $\lesssim 10\,\mathrm{Myr}$ [1703.03805].

Obscured DCBHs themselves can also act as radiative triggers for further DCBH formation. One-zone calculations using Compton-thick DCBH SEDs filtered through columns of $N_{\rm H}\sim 5.0\times10^{24}$–$1.3\times10^{25}\,\mathrm{cm^{-2}}$ find remarkably low thresholds,
$$
J_{\rm LW}^{\rm crit}\approx 22,\ 35,\ 54,
$$
which rise only to
$$
\approx 80,\ 170,\ 390
$$
even when the DCBH-sourced X-ray background is pushed to the maximum allowed by the present-day unresolved X-ray background. The reason is spectral: obscuration converts much of the ionizing output into sub-$13.6\,\mathrm{eV}$ continuum that efficiently photodetaches $\mathrm{H}^-$ without generating as much X-ray catalysis of $\mathrm{H}_2$ formation as normal galaxies do [1612.07885].

Internal radiation generated by the collapsing cloud further complicates the threshold picture. Trapped Ly$\alpha$ cooling photons can photodetach $\mathrm{H}^-$ and lower the external field required for direct collapse by up to a factor of a few, with the largest effect for hard background spectra characteristic of hot young stellar populations. In one-zone models this shifts $J_{21,\rm crit}$ from $\simeq 1.1\times10^3$ to $\simeq 900$ or even $\simeq 200$ for $T_{\rm rad}=10^5\,\mathrm{K}$, depending on the assumed central concentration of the Ly$\alpha$ source [1611.00780].

## 3. Collapse dynamics, magnetic support, and fragmentation

The adjective “direct” does not imply a featureless or strictly fragmentation-free collapse. The most detailed cosmological MHD calculations evolved for a full $\approx 1.6\,\mathrm{Myr}$ show that magnetic fields are rapidly amplified by strong accretion shocks at disk edges and by turbulence, far beyond simple flux freezing. In these runs the magnetic field grows by $\approx 14$ orders of magnitude for a fiducial seed field and by up to $\approx 20$ orders for a weaker seed, with a growth time of about $10^3\,\mathrm{yr}$, and saturates near equipartition with turbulent energy over the inner tens of parsecs. By the end of the simulations, the field reaches $\approx 1\,\mathrm{G}$ within the central $\lesssim 2000\,\mathrm{AU}$ and $\approx 10^{-5}\,\mathrm{G}$ at $r\approx 20$–$40\,\mathrm{pc}$ [2210.05611].

In that regime, magnetic support is dynamically important. Using
$$
v_A = \frac{B}{\sqrt{4\pi\rho}},\qquad
\beta = \frac{8\pi P_{\rm gas}}{B^2},
$$
the inner disk satisfies $v_A \gtrsim c_s$ and $\beta \approx 0.3$–$1$, so magnetic pressure is comparable to thermal pressure. A convenient description is to define an effective sound speed
$$
c_{\rm eff}^2 \approx c_s^2 + v_A^2,
$$
which raises both the effective Toomre stability parameter and the magnetically supported Jeans mass,
$$
Q_{\rm eff} = \frac{c_{\rm eff}\kappa}{\pi G\Sigma},\qquad
M_{J,B} \approx \frac{\pi^{5/2}}{6}\frac{c_{\rm eff}^3}{G^{3/2}\rho^{1/2}}.
$$
The numerical consequence is not the elimination of substructure but a systematic reduction in multiplicity: MHD disks are larger and smoother, fragmentation is significantly reduced, and the surviving central clump typically still reaches $\sim 10^5\,M_\odot$ by $1.6\,\mathrm{Myr}$ with mean inflow rates of $\sim 0.1\,M_\odot\,\mathrm{yr^{-1}}$, broadly similar to non-MHD runs because rapid coalescence in purely hydrodynamic disks compensates for their larger number of fragments [2210.05611].

This matters because it corrects a common oversimplification. DCBH formation is not equivalent to the absolute absence of fragmentation. Rather, the pathway tolerates fragmentation so long as the global thermal state, inflow rate, and merger/coalescence history preserve the build-up of a dominant $\sim 10^5\,M_\odot$ clump or supermassive star. Magnetic fields therefore appear to stabilize the pathway mainly by reducing multiplicity and supporting coherent inflow, not by shutting off all small-scale structure [2210.05611].

## 4. Seed birth, early growth, and post-formation dynamics

Direct collapse solves the initial-mass problem, but not automatically the growth problem. In the most optimistic environments, a newly born $\sim 10^5\,M_\odot$ seed can sit at the nexus of cold accretion flows and remain near the Eddington limit for long enough to reach quasar scale. A radiation-hydrodynamics calculation that “switches on” a $10^5\,M_\odot$ DCBH at $z=19.2$ inside a $3\times10^8\,M_\odot$ atomically cooling halo finds a birth luminosity of
$$
L_{\rm bol}=2.42\times10^{44}\ {\rm erg\ s^{-1}},
$$
corresponding to $\approx 0.85\,L_{\rm Edd}$, and follows the host’s growth to $1.2\times10^{12}\,M_\odot$ by $z=7.1$ under sustained cold-flow feeding [2005.03018].

Other cosmological simulations reach the opposite conclusion. When the Bondi scale is resolved during long-term post-formation evolution, accretion can remain far below Eddington because the seed forms in a metal-free environment typically $\sim 1\,\mathrm{kpc}$ from the first galaxy, falls into the potential well, and acquires a relative velocity of order
$$
\sim 100\ {\rm km\ s^{-1}}
$$
with respect to the gas. In that regime the effective Bondi rate
$$
\dot M_{\rm Bondi} = 4\pi \lambda \frac{G^2 M_{\rm BH}^2 \rho}{(c_s^2+v_{\rm rel}^2)^{3/2}}
$$
is strongly suppressed. An analytic estimate then implies that DCBH formation must occur within
$$
\sim 100\ {\rm pc}
$$
of the galactic center for dynamical friction to decelerate the black hole before $z=7$, but such locations are expected to have metallicities of
$$
Z \sim 10^{-5}\text{--}10^{-3}\,Z_\odot,
$$
which is in tension with the pristine requirement of classical direct collapse [2008.09120].

The earliest post-birth phase may itself be violent. In one model, the newborn DCBH is surrounded by a self-gravitating nuclear disk that fragments at
$$
r_f \simeq 3.3\times10^{-2}\ {\rm pc}
$$
into clumps of initial mass
$$
M_{c,0}\sim 28\,M_\odot,
$$
which evolve into massive Pop III stars and assemble a compact nuclear star cluster. Stellar relaxation can scatter some of these stars into the black hole’s loss cone on timescales $\lesssim 10^6\,\mathrm{yr}$, producing tidal disruption events with jet luminosities
$$
L_{\rm j} \gtrsim 10^{50}\ {\rm erg\ s^{-1}}
$$
and observed prompt X-ray durations
$$
\delta t_{\rm obs}\sim 10^{5-6}(1+z)\ {\rm s},
$$
followed by radio afterglows potentially detectable even from $z\sim 20$ [1602.04293].

Taken together, these results imply that DCBHs are heavy seeds, not guaranteed quasars. Their later fate depends on where they form, how rapidly they couple to the densest gas, and whether cold inflows, mergers, or central metal-enriched pathways can keep them fueled [2005.03018] [2008.09120].

## 5. Observational signatures from Ly$\alpha$ to radio continua

The simplest intrinsic line signature is Ly$\alpha$. During collapse powered only by gravitational heating, the Ly$\alpha$ cooling luminosity is limited to
$$
L_{\alpha}^{g\text{--}cool}\lesssim 10^{38}\ {\rm erg\ s^{-1}}
\left(\frac{M_{\rm gas}}{10^6\,M_\odot}\right)^2,
$$
because collisional de-excitation eventually suppresses escape. Photoionization by a central accreting source is far more luminous and can produce
$$
L_{\alpha}^{\rm rec}\sim 10^{43}\ {\rm erg\ s^{-1}}
\left(\frac{M_{\rm BH}}{10^6\,M_\odot}\right)
$$
during favorable evolutionary stages. The emergent Ly$\alpha$ profile is highly sensitive to $\mathrm{H\,I}$ column density, geometry, and velocity field: predicted widths and offsets range from a few tens to a few thousands of $\mathrm{km\,s^{-1}}$. Applied to CR7, these models imply that if the source is black-hole powered, the Ly$\alpha$ luminosity alone requires
$$
M_{\rm BH}>10^7\,M_\odot,
$$
while the observed line width favors
$$
\log [N_{\rm HI}/{\rm cm^{-2}}]\sim 19\text{--}20
$$
and an outflowing medium, indicating that the original high-column formation conditions have already been erased [1602.07695].

A more specialized but exceptionally distinctive signature is the fine-structure maser at rest wavelength $\lambda_0 \approx 3.04\,\mathrm{cm}$, arising when trapped Ly$\alpha$ photons overpopulate the $2p$ level of atomic hydrogen. The inversion condition is
$$
n_{2p}>3n_{2s},
$$
which produces negative opacity in the $2s_{1/2}$–$2p_{3/2}$ transition. In simplified DCBH collapse models, this can amplify the CMB by up to $\approx 10^5$ before saturation, generating a broad asymmetric line with hyperfine structure, a flux of about $0.3$–$3\,\mu\mathrm{Jy}$, an angular scale of $\sim 1$–$10\,\mathrm{mas}$, and an observed frequency in the range $\sim 0.6$–$1.4\,\mathrm{GHz}$ for $z\sim 6$–$15$. The required combination of dust-free gas, very high $N_{\rm HI}$, and extremely low $\mathrm{H}_2$ makes this a particularly specific DCBH marker [1601.04712].

Near-infrared searches target the birth phase and the first tens of megayears thereafter. Radiation-hydrodynamics plus Cloudy post-processing predicts that newborn DCBHs in cold accretion flows are brightest in the reddest JWST/NIRCam bands, with apparent magnitudes
$$
m_{\rm AB}\approx 27.5\text{--}30.1
$$
across $z=8$–$20$ in the $4.44$ and $4.60\,\mu{\rm m}$ filters. In $100\,\mathrm{hr}$ exposures they are detectable in all four long-wavelength NIRCam filters out to $z\approx 19$, and redward of $3.56\,\mu{\rm m}$ detectability extends to $z\approx 25$. Because Euclid and Roman are shallower but wider, strong lensing becomes decisive there: for reasonable host-halo abundances, Roman, Euclid, and JWST could potentially find hundreds of strongly lensed DCBHs at $z=7$–$20$, with Roman performing best at $z\lesssim 10$ and JWST at higher redshift [2005.03018] [2205.14163].

Radio predictions bifurcate according to the emission model. If accreting DCBHs launch blazar-like jets with power
$$
P_{\rm jet}\gtrsim 10^{42\text{--}43}\ {\rm erg\ s^{-1}},
$$
then at $z\approx 10$ SKA-mid and ngVLA can detect sources with
$$
M_{\rm BH}\gtrsim 10^5\,M_\odot,
$$
and an optimistic normalization of
$$
n_{\rm DC}\approx 2.5\times10^{-3}\ {\rm Mpc^{-3}}
$$
yields about
$$
100\ {\rm deg^{-2}\,z^{-1}}
$$
above the SKA1-mid threshold in $100\,\mathrm{hr}$, provided all DCBHs are active and jetted. The low-frequency spectrum is often strongly suppressed by synchrotron self-absorption and, if the emitting region remains inside the dense envelope, by free–free absorption, so joint SKA-low and SKA-mid observations become a discriminant against star-forming galaxies [2107.11307]. A more conservative approach, using several fundamental planes of black-hole accretion and assuming no jets, still finds that SKA-FIN could detect a $10^6\,M_\odot$ DCBH at $500\,\mathrm{MHz}$ out to $z\approx 11$, while SKA and ngVLA could probe $10^6$–$10^7\,M_\odot$ black holes to $z\sim 20$; by contrast, $10^5\,M_\odot$ DCBHs at $z\geq 8$ remain beyond current limits in that framework [2110.00012].

## 6. Population histories, cosmological probes, and related high-redshift populations

Semi-analytic and simulation-based population models span a wide dynamic range. In one cosmological treatment that explicitly follows merger trees, stellar populations, and local LW fields, the direct-collapse formation rate rises from
$$
\sim 10^{-3}\ {\rm Mpc^{-3}\,z^{-1}}
$$
at $z=12$ to
$$
\sim 10^{-2}\ {\rm Mpc^{-3}\,z^{-1}}
$$
at $z=6$, with the first DCBHs appearing by $z\approx 12$ and with local LW intensities that can exceed the spatially averaged background by up to $\sim 10^6$. In that picture, DCBH hosts are more clustered than similar-mass non-DCBH halos, especially at $z>10$ [1205.6464].

A different but influential formulation is the brief “DCBH era.” In that model, once a few Compton-thick DCBHs form, their reprocessed LW/NIR emission triggers a runaway rise in further DCBH formation. The universe enters a DCBH-dominated phase at
$$
z\approx 20,
$$
the comoving mass density rises from
$$
\sim 5\,M_\odot\,{\rm Mpc^{-3}}
$$
at $z\sim 30$ to
$$
\sim 5\times10^5\,M_\odot\,{\rm Mpc^{-3}}
$$
at $z\sim 14$ in the fiducial case, and new DCBH formation is then almost completely shut off by photoevaporation after
$$
z\sim 13.
$$
The entire era lasts only $\approx 150\,\mathrm{Myr}$ [1402.5675].

Several indirect cosmological probes have now been attached to DCBH phenomenology. One is the global 21-cm absorption trough. In a model that combines a JWST-consistent UV luminosity function with X-ray-dependent critical curves for direct collapse, the trough depth correlates with seed abundance: if
$$
\delta T_b^{\rm trough}\gtrsim -100\ {\rm mK},
$$
DCBHs are expected to be rare; if
$$
-150\ {\rm mK}\lesssim \delta T_b^{\rm trough}\lesssim -100\ {\rm mK},
$$
then
$$
n_{\rm DCBH}\sim \mathcal{O}(10^{-2}\text{--}10^{-3})\ {\rm cMpc^{-3}},
$$
while deeper troughs allow substantially larger abundances [2503.22130].

Another connection is to the JWST population of Little Red Dots. Preliminary MELIORA cosmological hydrodynamics runs that implement explicit DCBH criteria find that newly formed DCBHs are associated with a gas-compaction event, spend roughly $100$–$200\,\mathrm{Myr}$ in a bright near-Eddington phase, and inhabit dense compact reservoirs with
$$
R_{\rm half}^{\rm gas}\sim 100\text{--}200\ {\rm pc}.
$$
The abundance of these newborn seeds declines steeply at $z\lesssim 6$, paralleling the observed decline of LRDs, and the combination of a compact neutral reservoir, strong obscuration, and the onset of Pop III star formation has been proposed as a possible explanation for the weak X-ray and hot-dust emission of at least a fraction of LRDs [2508.14897].

More speculative variants extend the radiative trigger beyond galaxies and DCBHs themselves. One such model proposes that axion-like dark matter decays in the IGM inject $1$–$13.6\,\mathrm{eV}$ photons that populate finite slices of the LW band. In that single-zone analysis, atomic-cooling-halo formation at $z=10$ is achieved for
$$
m_a\simeq 24.5\text{--}26.5\ {\rm eV}
$$
and
$$
g_{a\gamma\gamma}\sim (3\text{--}4)\times10^{-12}\ {\rm GeV^{-1}},
$$
with the line-resolved treatment of the LW band identified as essential [2509.25325].

Gravitational waves provide yet another window. An early analytic estimate of the stochastic signal from DCBH formation bursts found a very low duty cycle,
$$
{\cal D}\sim 10^{-3},
$$
a peak background amplitude
$$
\Omega_{\rm gw}=1.1\times10^{-54}
$$
at
$$
\nu_{\rm max}=0.9\ {\rm mHz},
$$
and a peak signal-to-noise ratio of about $22$ at $\nu=20\,\mathrm{mHz}$ for Ultimate-DECIGO, albeit below the Galactic confusion foreground [1502.04125]. More recent numerical-relativity pipeline work aims to derive physically anchored collapse waveforms for LISA from cosmological DCBH initial conditions, but explicitly emphasizes that population properties such as masses, mass ratios, spins, and eccentricities remain poorly constrained [2512.09197].

## 7. Open issues, tensions, and common misconceptions

One common misconception is that DCBH formation is controlled by a single $J_{\rm crit}$. The accumulated evidence points the other way: the relevant threshold depends on the source SED, on whether one works with $J_{21}$ or with the two-rate plane $(k_{\rm H_2},k_{\rm H^-})$, on X-ray backgrounds, on trapped Ly$\alpha$ feedback, on obscuration, and on the geometry and timing of nearby sources [1504.04042] [2503.22130] [1611.00780].

A second misconception is that “direct collapse” means perfectly monolithic collapse with no fragmentation. Long-duration MHD simulations show that magnetic fields reduce multiplicity and stabilize disks, but do not erase all clump formation; similarly, the immediate post-birth disk around a DCBH may fragment into massive stars whose later dynamics feed back on the seed through tidal disruptions and photoevaporation. The operative criterion is therefore not zero fragmentation, but whether the thermal and inflow conditions still deliver a dominant central object of order $10^5\,M_\odot$ [2210.05611] [1602.04293].

A third misconception is that producing a heavy seed is enough to explain the first quasars. In fact, direct-collapse sites can be hostile to growth. Radiative feedback, large BH–gas relative velocities, and off-center birth locations can hold accretion far below Eddington for $\gtrsim 100\,\mathrm{Myr}$, whereas only special environments such as sustained cold accretion flows or perhaps central, mildly metal-enriched pathways appear able to maintain the required fueling [2008.09120] [2005.03018].

Finally, observational forecasts are model-contingent. Radio source counts depend on the unknown jetted fraction, jet duty cycle, spin, envelope absorption, and whether a fundamental-plane extrapolation is appropriate at all for newborn DCBHs; lensing yields depend on high-magnification tails and source sizes; 21-cm mappings retain at least dex-level uncertainty; and the abundance of viable pristine halos remains highly sensitive to clustering, metal transport, and radiative backgrounds [2107.11307] [2110.00012] [2503.22130] [1205.6464].

These tensions do not invalidate the DCBH hypothesis; they delimit it. A plausible implication is that “DCBH” is best treated not as a single sharply defined channel, but as a family of heavy-seed pathways centered on atomic-cooling collapse under suppressed molecular cooling, with outcomes controlled by the interplay of radiation fields, magnetic support, inflow history, obscuration, and subsequent environmental coupling.

Source: https://www.emergentmind.com/topics/direct-collapse-black-holes-dcbhs-de8a981d-4d18-404b-805f-5d75df1895f9