Back to Research Theme 04 · ELT Simulations

Computational Astrophysics, Simulations, and Modelling

The Extremely Large Telescope (ELT) will be the largest optical telescope ever built. Before it starts observing, I create realistic virtual observations to find out which black holes it will be able to weigh, from small black holes in nearby dwarf galaxies to giant ones in the distant Universe.

Why Simulate a Telescope That Is Not Finished Yet?

The ELT is being built by the European Southern Observatory in the Atacama Desert in Chile. Its main mirror is 39 metres across, and its adaptive optics will correct for the blurring of Earth's atmosphere. Together these will give images far sharper than any telescope today. Two of its first instruments are at the heart of my work:

HARMONI
A spectrum for every pixel

An integral-field spectrograph. It records a full spectrum at every point of the image, so we can map how fast the stars are moving across the centre of a galaxy.

MICADO
Razor-sharp images

A high-resolution camera. Its images show precisely how the stars are distributed, which tells us how much of the mass at the centre is in stars.

Time on the world's largest telescopes is extremely competitive, and every hour counts. Simulations let us answer the key questions in advance: which galaxies are worth observing, which instrument settings to use, how long to observe, and, most importantly, whether the black hole mass we measure can be trusted.

How It Works

The idea is simple: we put a black hole of a known mass into a model galaxy, “observe” it with a virtual ELT, and check whether we can get the same mass back out.

  1. Start from a real galaxy
    Hubble Space Telescope images tell us how the stars are spread out in the galaxy.
    HST imaging · MGE
  2. Add a black hole
    We place a black hole of chosen mass at the centre and compute how the stars should move.
    JAM dynamical models
  3. Observe with a virtual ELT
    Simulators add realistic blurring, atmosphere, noise and exposure time.
    HSIM · SimCADO
  4. Measure and compare
    We analyse the fake data exactly like real data and check whether the mass we recover matches the one we put in.
    pPXF · JAM + MCMC

What I Have Found

Hidden black holes in dwarf galaxies

Many small galaxies have a dense, bright ball of millions of stars at their very centre, called a nuclear star cluster (Fig. 1). These clusters are among the best places to look for intermediate-mass black holes, the “missing link” between the black holes left behind by single stars and the supermassive giants in large galaxies. They are hard to find because their gravity dominates only a tiny region, too small for today's telescopes to resolve.

Hubble image of the dwarf galaxy FCC 119, with a zoom into its bright nuclear star cluster
Fig. 1 — A dwarf galaxy and its nuclear star cluster. Hubble image of FCC 119, a faint dwarf galaxy in the Fornax Cluster about 20 Mpc (65 million light-years) away. The zoom on the right shows its compact nuclear star cluster (NSC), where an intermediate-mass black hole could hide.From Ngo, Nguyen, Le et al. 2025, Universe, 11, 360.

In our virtual HARMONI observations, a black hole reveals itself as a “hot spot” at the centre of the map: close to the black hole, stars move faster (Fig. 2). The heavier the black hole, the stronger the signal.

Simulated HARMONI maps of stellar motions at the centre of NGC 300 for three black hole masses; a central hot spot appears as the black hole mass increases
Fig. 2 — What HARMONI would see. Simulated maps of stellar motions in the centre of the nearby galaxy NGC 300, for no black hole (top), and a black hole weighing 0.5% (middle) or 1% (bottom) of the star cluster's mass. From left to right: rotation, random motions ($\sigma$), and their combination ($V_{\rm rms}$). The black hole shows up as the growing red spot at the centre.From Nguyen, Cappellari, Ngo et al. 2025, AJ.

Giant black holes in the distant Universe

Today, this kind of direct black hole weighing only works for galaxies relatively close to us, within about 100 Mpc. We asked whether the ELT can push it much further, to galaxies at redshift 1–2, whose light left them roughly 8 to 10 billion years ago, when galaxies and their black holes were growing fastest.

A distant galaxy at redshift 0.82 seen by Hubble and in a simulated MICADO image that zooms ten times closer into its centre
Fig. 3 — Hubble today vs. the ELT tomorrow. The massive galaxy LEGAC-86906, seen when the Universe was about half its present age. Left: the real Hubble image (scale bar: 1 kpc). Right: a simulated MICADO image of just the central region (scale bar: 100 pc), with pixels about ten times finer than Hubble's.From Nguyen, Cappellari, Le, Ngo et al. 2026, MNRAS, 546, 4.

Does it work?

Yes. In both cases, the black hole masses we recover from the virtual observations match the masses we put in: the points in Fig. 4 lie along the dashed one-to-one line. When there is no black hole, the analysis correctly returns only an upper limit.

Input versus recovered black hole masses: small black holes in two dwarf galaxies (left) and giant black holes in five distant galaxies (right) lie on the one-to-one line
Fig. 4 — Putting in a mass, getting it back. Black hole mass put into the simulation (vertical axis) versus the mass recovered from the virtual observations (horizontal axis). Left: intermediate-mass black holes in the dwarf galaxies NGC 300 and NGC 3115 dw01. Right: billion-solar-mass black holes in five galaxies at redshift 1–2.Left: Nguyen, Cappellari, Ngo et al. 2025, AJ. Right: Nguyen, Cappellari, Le, Ngo et al. 2026, MNRAS, 546, 4.

Tools I Use

I write and run these pipelines mainly in Python, combining community tools for instrument simulation, dynamical modelling and Bayesian fitting.

PythonNumPyAstropyMatplotlibHSIM (HARMONI)SimCADO (MICADO)MGEJAMpPXFMCMC (AdaMet)

Papers in This Theme

Black Hole and Galaxy Coevolution Back to Research