Massive Black Holes Across the Mass Spectrum

Overview

Known massive black holes span roughly eight decades in mass, but the census is far from uniform. The interval between about 10² and 10⁵ M_☉ — the intermediate-mass black holes — remains almost empty, not necessarily because such objects are rare but because their gravitational sphere of influence subtends only a few tens of milliarcseconds even in the nearest dwarf galaxies, nuclear star clusters and globular clusters. At the opposite extreme, the most massive black holes in brightest cluster galaxies probe the limits of accretion and merger-driven growth. Because different mass ranges are traditionally studied with different techniques — stellar dynamics, cold-gas dynamics, reverberation mapping, single-epoch AGN scaling — systematic offsets between methods can easily masquerade as physical structure in the mass function. Applying one carefully controlled dynamical framework across the full spectrum is therefore essential.

Measuring Black Hole Masses Dynamically

A black hole can be weighed directly only where its gravity dominates the motions of the surrounding matter, inside the sphere of influence $r_{\rm SOI} = G M_{\rm BH}/\sigma^2$, where $\sigma$ is the stellar velocity dispersion of the host. The central observational challenge is angular resolution: the sphere of influence must be resolved, or nearly so, and it shrinks rapidly with decreasing black hole mass and increasing distance. Our programme uses two complementary tracers of the central potential, stars and cold molecular gas, and models each with a well-tested forward-modelling framework.

Stars and gas in a galaxy move in two ways: ordered rotation and random motion. Both leave an imprint on the spectrum through the Doppler effect. Rotation shifts spectral lines by an amount that changes with position across the galaxy, while random motions broaden the lines, and the line width measures the velocity dispersion $\sigma$ (Fig. 1). Close to a massive black hole, stars and gas must move faster to stay in orbit, so rotation and dispersion both rise sharply toward the centre (Fig. 2). Integral-field spectroscopy records a spectrum in every spatial pixel, turning this signature into two-dimensional maps that dynamical models can fit.

Method 01
Stellar Kinematics with Jeans Anisotropic Models

How it works

We start from high-resolution imaging, mainly from HST, and describe the galaxy's surface brightness with a Multi-Gaussian Expansion (MGE), which can be deprojected analytically into a three-dimensional stellar density. Jeans Anisotropic Models (JAM) solve the axisymmetric Jeans equations for this density plus a central point mass, and predict the second velocity moment $V_{\rm rms} = \sqrt{V^2 + \sigma^2}$ on the sky after convolution with the instrumental point-spread function. The model is then compared with the observed two-dimensional kinematics, and the free parameters (black hole mass $M_{\rm BH}$, stellar mass-to-light ratio $M/L$, orbital anisotropy and inclination) are constrained within a Bayesian framework using Markov chain Monte Carlo sampling.

For nearby nuclei we use the JWST/NIRSpec integral-field unit. In the near-infrared, the stellar CO band-head absorption features at ~2.3 µm, a “gold-standard” tracer, penetrate the obscuring nuclear dust, and the spectra let us separate the stellar light from the non-thermal AGN continuum before extracting the kinematics with pPXF. To treat systematic uncertainties honestly, we do not rely on the formal errors of a single preferred model. Instead we fit an ensemble of models that vary the point-spread function, the $M/L$ profile and the anisotropy assumptions, and quote the ensemble median and spread.

Results with JWST/NIRSpec

Looking ahead: the same method on the ELT

Because JAM only needs a mass model and resolved stellar kinematics, the same pipeline can be run end-to-end on simulated observations. We create mock ELT/HARMONI integral-field data cubes with HSIM and mock ELT/MICADO images with SimCADO, analyse them as if they were real data, and ask which black hole masses the ELT will be able to recover.

• Intermediate-mass black holes in nuclear star clusters. From a sample of 44 dwarf galaxies with nuclear star clusters within 10 Mpc, HARMONI can resolve the sphere of influence and recover black hole masses down to ≈0.5% of the cluster mass (AJ 2025).
• Out to 20 Mpc. For FCC 119, one of the faintest dwarfs in the Fornax Cluster, HARMONI can detect the kinematic signature of an IMBH and measure its mass for $M_{\rm BH} \gtrsim 10^5\,M_\odot$ (Universe 2025, lead author). For VCC 1861 in the Virgo Cluster, IMBHs of 5% of the nuclear-cluster mass are detectable (arXiv:2509.03364, lead author, submitted).
• The inner profile of the cluster. Combining MICADO-based mass models with HARMONI kinematics shows how the inner surface-brightness slope of the nuclear star cluster affects the recovered IMBH mass (Universe 2026).
• Beyond the local Universe. For massive quiescent galaxies at $1 \lesssim z \lesssim 2$, MICADO mass models plus HARMONI kinematics recover black hole masses to ~10%, with ~1 h of MICADO imaging and 5–7.5 h of HARMONI integration (MNRAS 2026).

Method 02
Molecular Gas Dynamics with ALMA CO(2–1)

How it works

Many early-type galaxies host a compact disk of cold molecular gas in their nucleus, often visible as a dust lane in HST images. Unlike stars, which move on a wide range of orbits, this gas settles onto nearly circular orbits, so its rotation traces the enclosed mass directly. We observe the $^{12}$CO(2–1) line with ALMA at high angular resolution, ideally with a synthesized beam comparable to or smaller than the black hole's sphere of influence.

Rather than fitting a one-dimensional rotation curve, we forward-model the full three-dimensional ALMA data cube with KinMS. The circular velocity comes from the stellar mass (an MGE model scaled by $M/L$), the molecular gas mass, and the central black hole. The model gas disk, with its own surface-brightness distribution, velocity dispersion, inclination and position angle, is then convolved with the synthesized beam and channelised like the real data. The black hole mass, $M/L$ and disk geometry are constrained with a Bayesian analysis, and we test different prescriptions for the gas distribution to quantify systematic uncertainties.

Results with ALMA

Further Methods

Summary of Findings

Taken together, these measurements show that a single, carefully controlled dynamical approach, with ensembles of models to capture systematic uncertainties, can deliver reliable masses from ~10⁷ to a few 10⁹ M_☉. Where an independent benchmark exists (the NGC 4258 maser), the stellar-dynamical mass agrees to within 5%. In several galaxies (NGC 4736, M81, NGC 4061), the new masses resolve long-standing discrepancies between earlier estimates. Our ELT simulations show that the same stellar-kinematic method can extend this census down into the intermediate-mass regime.

Literature Review

Progress hinges on how the masses are measured. Stellar-dynamical modelling (Schwarzschild orbit superposition and Jeans anisotropic models) and ionised-gas kinematics were the early workhorses, complemented by the exquisite geometry of nuclear water megamasers where they exist.

More recently, ALMA has opened cold molecular gas as a clean dynamical tracer, resolving regular circumnuclear disks whose rotation pins down the enclosed mass (e.g. Davis et al. 2013; Nguyen et al. 2021). JWST/NIRSpec now pushes spatially resolved stellar kinematics to fainter, more distant nuclei. My own work sits within this multi-tracer effort, cross-calibrating methods on the same targets to expose systematic differences.

Papers in This Theme