High-Redshift Evidence for Early Black Holes

A high-redshift black-hole candidate is not validated by redness, one broad line, or an impressive inferred mass. It must survive a chain from detector calibration and line identification through lens reconstruction, spectral decomposition, mass modeling, and selection correction. Some links in that chain can already be unusually strong: a spectroscopic redshift, repeated variability, or a spatially resolved rotation field. The seed channel remains a population-level inference because accretion history can map very different initial masses to the same observed object. Undated claims that one seed channel is established are outside scope.

Helpful background. Black-Hole Geometry and Mechanics supplies the black-hole mass concept; Relativistic Matter and Compact Objects supplies stellar-remnant formation; Black-Hole Astrophysics supplies accretion, hosts, and selection; and Relativistic Cosmology supplies redshift, lensing, and cosmic-time inference.

Build the candidate from calibrated observables

Begin with the exact JWST program, detector configuration, pipeline and calibration reference files, extraction aperture, and public data-product version. A line at observed wavelength $\lambda_{\rm obs}$ supports an identification at

$$z=\frac{\lambda_{\rm obs}}{\lambda_0}-1$$

only if a common $z$ aligns several expected features and plausible low-redshift alternatives fail. The model must forward-convolve continua and lines with the wavelength-dependent line-spread function (LSF); for a compact source, the effective NIRSpec resolution can differ from the nominal uniformly illuminated-slit value. Images likewise require point-spread-function (PSF) matching before a source is declared unresolved or assigned a host size.

For a lensed source, report the image-plane position, counterimages, lens-model version, magnification $\mu$, shear, and source-plane covariance. Fluxes scale as $\mu^{-1}$, while sizes and velocity-map geometry depend on the local magnification tensor rather than on one scalar. A cluster member or foreground galaxy close to the trace can contaminate both photometry and spectroscopy. Quoted wavelengths and velocities below are in the source rest frame; luminosities and physical sizes are lens-corrected under each paper’s stated cosmology unless noted otherwise.

This separates durable measurements from mutable interpretations. The spectroscopic redshift $z_{\rm spec}=8.6319\pm0.0005$ of CANUCS-LRD-z8.6, its broad Hβ detection, and its C IV and N IV] features are close to the data layer. Its $1.0^{+0.6}_{-0.4}\times10^8M_\odot$ single-epoch mass uses a local virial calibration and a decomposed line width; the conclusion that it favors a particular seed is farther downstream (Tripodi et al. 2025, §“The target,” Fig. 1, and §“Evidence of AGN activity”).

Treat every mass as a model comparison

A single-epoch broad-line estimate has the structure

$$M_{\rm vir}=f_{\rm vir}\frac{R_{\rm BLR}(L)\,\Delta v_{\rm Dop}^{\,2}}{G}.$$

Here $R_{\rm BLR}$ is the radius assigned to the broad-line region and $\Delta v_{\rm Dop}$ is the width attributed to orbital Doppler motion. The luminosity-to-radius calibration, geometry factor $f_{\rm vir}$, and Doppler width are assumptions, not detector outputs. Electron scattering can create broad, nearly exponential wings even when the orbitally broadened core is narrow. Rusakov and collaborators fitted high-signal-to-noise NIRSpec spectra for 12 objects at $3.4\lt z\lt6.7$ and found intrinsic core widths near $300\ {\rm km\,s^{-1}}$ in nine; replacing full widths by cores roughly ten times narrower lowers a virial mass by roughly $10^2$ (Rusakov et al. 2026, pp. 575–578, Figs. 2–4, and Extended Data Table 1). This is a credible alternative to interpreting every Balmer wing as orbital motion, but it too depends on radiative-transfer geometry, electron temperature and column, extinction, and the premise that the residual core is virial.

Resolved dynamics use different assumptions. For a thin axisymmetric rotating tracer, let $\phi$ be azimuth in the disk measured from the projected major-axis direction. After subtracting the systemic velocity, the intrinsic local second moment is

$$v_{\rm rms,los}^2(r,\phi) =v_{\rm rot}^2(r)\sin^2 i\cos^2\phi+\sigma_{\rm los}^2(r,\phi).$$

This is before PSF and line-spread convolution; unresolved velocity gradients add apparent line width and cannot simply be labeled random motion. Mean rotation is also distinct from the circular speed $v_c$. In a stationary gas disk with isotropic pressure $P$ and negligible radial acceleration, the radial force equation gives

$$v_c^2\equiv r\,\partial_r\Phi =v_{\rm rot}^2-\frac{r}{\rho}\frac{dP}{dr}.$$

An outward-decreasing pressure makes the correction positive. Only for a spherical or point-mass-dominated gravitational potential is $v_c^2=GM_{\rm enc}(\lt r)/r$; a flattened massive disk needs its actual potential. Stellar asymmetric drift has a related but different Jeans-equation correction involving anisotropy. Thus inclination $i$, pressure support, PSF smearing, lens shear, and the division among a point mass, stars, gas, and dark matter must be fitted. Abell 2744-QSO1 at $z=7.04$ provides a demanding example. Its narrow-Hα velocity field extends to about $200$ pc, its spectroastrometric red and blue centroids are separated by $24.9\pm9.4$ pc, and a full PSF-aware three-dimensional model gives $\log_{10}(M/M_\odot)=7.7\pm0.3$ for inclination $52\pm2^\circ$. Extended nuclear-cluster models leave larger residuals or collapse to an unresolved point mass (Juodžbalis et al. 2026, pp. 1018–1020, Figs. 1–2, and Supplementary Information §§1.1–1.8). This independently supports a compact gravitating mass, but calling it a seed caught at birth additionally assumes an age, environment, and growth history.

Compare objects without erasing their differences

GN-z11 has NIRSpec detections of [Ne IV] and C II*, high-density semi-forbidden emission, and an $800$–$1{,}000\ {\rm km\,s^{-1}}$ C IV absorption outflow. These mutually reinforcing diagnostics favor an accreting source. The often-quoted $1.6\times10^6M_\odot$ mass and accretion rate near five times Eddington assume local virial and bolometric calibrations; the authors explicitly show that light, intermediate, and heavy seeds can reach the measured state under different growth histories (Maiolino et al. 2024, pp. 59–62, Figs. 1 and 3, and Methods).

UHZ1 illustrates the value and risk of combining instruments. NIRSpec prism spectroscopy fixes $z=10.073\pm0.002$ from the Lyman break and emission features, while the proposed accreting source rests on a $4.2\sigma$ Chandra signal associated with the lensed JWST galaxy. The spectrum itself is dominated by star-formation features and does not reveal the X-ray source. Consequently the redshift is durable, whereas the X-ray association, obscuring column, bolometric correction, $10^7$–$10^8M_\odot$ Eddington-based mass, and heavy-seed preference remain linked inferences (Goulding et al. 2023, §§3.1–3.3 and Figs. 2–3; Bogdán et al. 2024, pp. 126–130 and Figs. 1–3). A robust reanalysis therefore propagates Chandra astrometry, cluster-lens uncertainty, background trials, and counterpart probability together rather than treating each as independent confirmation.

Correct the selection before interpreting a population

Photometric color cuts select only a projection of the underlying population and admit dusty starbursts, extreme nebular emitters, cool stars, supernovae, and reduction artifacts. Spectroscopic follow-up must publish failures as well as successes. To see how selection enters, let $p(d\mid x)$ be the unconditioned measurement likelihood and let $\mathcal A$ be the accepted region of data space. The detection probability for latent source parameters $x$ is $S(x)=\int_{\mathcal A}p(d\mid x)\,dd$. For $N$ accepted objects and a normalized source population $p(x\mid\theta)$, the likelihood conditional on $N$ is

$$\begin{aligned} \alpha(\theta)&=\int S(x)p(x\mid\theta)\,dx,\\ p(\mathcal D\mid\theta,N,S) &\propto\frac{1}{\alpha(\theta)^N} \prod_{i=1}^N\int p(d_i\mid x)p(x\mid\theta)\,dx, \qquad d_i\in\mathcal A. \end{aligned}$$

The normalization $\alpha$ depends on the population model and cannot be dropped. There is no extra $S(x)$ in each numerator here: acceptance has already been imposed on the observed data. If one instead uses a measurement likelihood conditioned on acceptance, the equivalent numerator is $\int p(d_i\mid x,S)S(x)p(x\mid\theta)\,dx$. Mixing these conventions counts selection twice. The derivation and the extension to event counts are given by Mandel, Farr & Gair 2019, eqs. (6)–(14). Staged targeting and follow-up require the corresponding joint measurement and selection model.

Number-density inference must also retain the count likelihood, including its expected number of detections, rather than condition on $N$. It additionally requires effective survey area, completeness injections, magnification maps, cosmic variance, and duplicated-image removal. X-ray and radio nondetections are model tests only after coverage, depth, absorption, and expected spectral energy distributions are specified. Variability, multiple ionization diagnostics, independent line-profile families, and resolved kinematics are valuable precisely because their failure modes are not identical.

Results and limitations through 2026-08-18

The evidence cutoff is 2026-08-18. Results first released after that date are outside this assessment. The matrix keeps measurements separate from mass and seed interpretations.

Evidence or object	Durable result by the cutoff	Scope, status, or limitation
GN-z11 NIRSpec spectroscopy	$z=10.6$ source with several AGN-favoring ionic, density, and outflow diagnostics	Virial mass, bolometric correction, super-Eddington ratio, and seed history remain model-dependent
UHZ1 NIRSpec plus Chandra	Spectroscopic $z=10.073\pm0.002$; a reported X-ray signal is positionally associated with the lensed galaxy	Astrometry, lens model, obscuration and background control enter before Eddington mass or heavy-seed preference
CANUCS-LRD-z8.6	$z_{\rm spec}=8.6319\pm0.0005$, broad Hβ and high-ionization lines in a compact source	LSF decomposition, virial calibration, SED partition, and host mass control the claimed $10^8M_\odot$ mass
LRD electron-scattering sample	Exponential-wing fits often favor narrow intrinsic Hα cores over a single broad Gaussian	Selected sample, cocoon transfer model, extinction and virial-core assumption limit population and mass conclusions
Abell 2744-QSO1	Lensed, resolved Hα kinematics favor a compact point mass over tested extended clusters	Inclination, lens shear, PSF, tracer dynamics and untested mass profiles remain; one object does not identify a channel

Credible competing positions include virial broad-line masses with ordinary AGN geometry, electron-scattering wings around lower-mass accretors, compact starburst or nebular contributions to little-red-dot spectra, accreting black holes with unusual obscuration, and lensing-assisted dynamical masses. Photometric contamination can also create candidates before spectroscopy. Current data establish several early accreting black holes and one strong high-redshift dynamical case; they do not establish a unique seed mass function.

The interpretation could change when a reprocessed JWST release materially changes an LSF, PSF, line flux, or redshift; when a revised cluster model changes magnification or source-plane kinematics; when resolved dynamics, variability, or independent X-ray or radio data confirm or reject a candidate; or when a prospectively defined sample with published nondetections revises completeness-corrected number densities. Any such change can alter mass and channel inference even while the original detector counts remain unchanged.

Formulation-choice diagnostic

A lensed candidate has observed Hα FWHM $1{,}800\ {\rm km\,s^{-1}}$, an approximately Gaussian instrumental width $300\ {\rm km\,s^{-1}}$, and scalar magnification $\mu=4$. A Gaussian-only discovery analysis uses the observed continuum without correcting for lensing and attributes the deconvolved full width to orbital motion. A scattering fit instead gives an intrinsic Doppler core of $500\ {\rm km\,s^{-1}}$. Assume uniform magnification across the emitting regions and $R_{\rm BLR}\propto L^{1/2}$. By what factor does the scattering-and-lens-corrected virial mass differ from the discovery estimate, and which formulation is scientifically appropriate?

Solution

Under the deliberately Gaussian approximation, the discovery width is

$$\Delta v_{\rm disc}=\sqrt{1800^2-300^2}=1{,}775\ {\rm km\,s^{-1}}.$$

Lensing reduces intrinsic luminosity by $\mu$, so it reduces the inferred radius by $\mu^{1/2}=2$. Because $M_{\rm vir}\propto R\Delta v^2$,

$$\frac{M_{\rm corrected}}{M_{\rm disc}} =\mu^{-1/2}\left(\frac{500}{1775}\right)^2 =0.0397.$$

The corrected estimate is therefore about 25 times smaller. The scattering formulation is preferred only if its forward-convolved profile fits the calibrated line and alternatives better; one must not subtract widths in quadrature for a non-Gaussian LSF or scattering kernel. The answer demonstrates why magnification and broadening choices belong in the mass likelihood, not in an after-the-fact error bar.

References

• Bogdán, Á., Goulding, A. D., Natarajan, P., et al. “Evidence for Heavy-Seed Origin of Early Supermassive Black Holes from a $z\approx10$ X-Ray Quasar.” Nature Astronomy 8, 126–133 (2024). DOI.
• Goulding, A. D., Greene, J. E., Setton, D. J., et al. “UNCOVER: The Growth of the First Massive Black Holes from JWST/NIRSpec—Spectroscopic Redshift Confirmation of an X-Ray Luminous AGN at $z=10.1$.” The Astrophysical Journal Letters 955, L24 (2023). DOI.
• Juodžbalis, I., Marconcini, C., D’Eugenio, F., et al. “A Direct Black-Hole Mass Measurement in a Little Red Dot at High Redshift.” Nature 653, 1017–1021 (2026). DOI.
• Maiolino, R., Scholtz, J., Witstok, J., et al. “A Small and Vigorous Black Hole in the Early Universe.” Nature 627, 59–63 (2024). DOI.
• Mandel, I., Farr, W. M., & Gair, J. R. “Extracting Distribution Parameters from Multiple Uncertain Observations with Selection Biases.” Monthly Notices of the Royal Astronomical Society 486, 1086–1093 (2019). DOI; Open PDF.
• Rusakov, V., Watson, D., Nikopoulos, G. P., et al. “Little Red Dots as Young Supermassive Black Holes in Dense Ionized Cocoons.” Nature 649, 574–579 (2026). DOI.
• Tripodi, R., Martis, N., Markov, V., et al. “Extreme Properties of a Compact and Massive Accreting Black Hole Host in the First 500 Myr.” Nature Communications 16, 9830 (2025). DOI.