Every galaxy you see is resting in a well of something you can't.
A short, scroll-driven tour of the dark-matter scaffolding that underpins the visible universe — and how we measure it with deep extragalactic surveys.
First there were halos, and they were dark.
Gravity doesn't care whether matter shines. In the first hundred million years after the Big Bang, dark matter — five times more abundant than the ordinary kind — collapsed into a web of self-gravitating knots called halos.
Each halo is a roughly spherical overdensity of dark matter, with a mass that sets how deep its gravitational well runs. The bigger the halo, the deeper the well.
Then gas cooled, and stars lit up.
Ordinary matter — hydrogen, helium, a sprinkle of everything else — fell into those wells. When it cooled enough, it fragmented into stars. The stars formed galaxies, and the galaxies inherited the geography of their halos.
But the mapping is not one-to-one. A halo ten times more massive does not host a galaxy ten times more massive — feedback from supernovae and supermassive black holes makes the efficiency peak at a very particular halo mass.
The SHMR is the love letter between baryons and dark matter.
Plot stellar mass M* against halo mass Mh and you get a curve that bends at 1012 M☉. Below the peak, supernova feedback blows gas out. Above the peak, AGN feedback shuts cooling down.
In Shuntov et al. (2025) we measured how this curve evolves from z=0.2 to z=12 — the first coherent measurement spanning nearly the whole history of the Universe, made possible by COSMOS-Web's JWST imaging.
Turn the redshift dial. The universe re-paints itself.
Use the scrubber in the bottom right to change the redshift. As you move from z=0 backwards, halos shrink and merge back into smaller progenitors. Stellar masses fall, and the peak of the SHMR migrates.
The evolution of the peak efficiency is non-monotonic: it sits near ~20% today, dips to ~6% around z ~ 1.5, recovers gently through cosmic noon, and then rises again sharply beyond z ~ 3.5 — reaching values as high as 50–100% for ~1011.5 M☉ halos in the first few hundred Myr (Shuntov+ 2025). The early universe was a place where halos and galaxies grew almost in lockstep.
This is the kind of measurement that only becomes possible with surveys like COSMOS-Web (JWST, 0.54 deg²) and Euclid Q1, which reach the faint, high-redshift galaxies that populate small halos.
Galaxies don't live alone. They cluster.
Because massive halos are rarer, and form at the peaks of the dark-matter density field, galaxies inside them are more strongly clustered than the dark matter itself. The excess clustering — the "galaxy bias" — is a direct probe of which halos host which galaxies.
We measure this through the two-point correlation function: the excess probability, over random, of finding two galaxies separated by a given distance. Comparing the measurement to a Halo Occupation Distribution model pins down Mmin, M1 and the satellite slope α — and the SHMR drops out.
All of cosmic history, written in where galaxies sit and where they don't.
Thanks for reading.
← Back to site