The connection
Galaxies don't form in isolation: each one sits inside a dark-matter halo whose mass, shape, and assembly history shape every observable property of the galaxy — its stellar mass, size, gas content, star-formation history, even its globular cluster population. The galaxy–halo connection is the mapping between the baryonic and dark sides of that pairing, and it is one of the tightest empirical constraints we have on galaxy formation physics.
Weak gravitational lensing is the cleanest direct probe of total halo mass — it cares only about gravity and so includes the dark component automatically. My group has used CFHTLenS and now UNIONS to chart how halo mass, shape, and substructure correlate with the stellar component, the globular cluster system, the central black hole, and the merger history.
Galaxy–halo connections through weak lensing
The stellar-to-halo mass ratio and its evolution
In Hudson, Gillis, Coupon et al. 2015 we used CFHTLenS galaxy–galaxy lensing to measure the stellar-to-halo mass ratio (SHMR) over 108.75–1011.3 M⊙ and 0.2 < z < 0.8. At z ~ 0.5 the SHMR peaks at 4.0 ± 0.2% at Mh ~ 1012.25 M⊙. For the first time from lensing alone, we detected evolution: the peak falls from 4.5 ± 0.3% at z ~ 0.7 to 3.4 ± 0.2% at z ~ 0.3, dominated by red galaxies — consistent with the stellar-mass scale at which quenching sets in downsizing with cosmic time. The blue (star-forming) SHMR is well fit by a non-evolving power law, implying blue galaxies grow stellar mass at a rate that tracks their dark-matter accretion.
How dark are the filaments?
The cosmic web's filaments connect halos; their mean mass-to-light ratio reveals how efficiently galaxies form along them. In Yang, Hudson & Afshordi 2020 we stacked weak-lensing mass maps and SDSS galaxy light around ~8 h−1 Mpc pairs of luminous red galaxies. The inferred filament mass-to-light ratio is M/L = 351 ± 137 (r-band), with a stellar mass fraction M☆/M = 0.0073 ± 0.0030 — consistent with the cosmic mean and its predicted redshift evolution. Filaments are not unusually dark.
Halo shapes around luminous red galaxies
Cold-dark-matter halos are predicted to be triaxial. In Robison, Hudson, Cuillandre et al. 2023 we used CFIS/UNIONS weak lensing of SDSS-DR7 and BOSS LRGs to measure the halo ellipticity, assuming alignment with the stellar major axis. For DR7 LRGs (M ~ 2.7 × 1013 M⊙ h−1) we find e = 0.46 ± 0.10, with the halo ~2.2 times more elliptical than the galaxy light. Combined with literature constraints, halo ellipticity increases by ~0.10 per decade in halo mass — matching hydrodynamical simulations.
Black hole–halo mass relation
In Li, Kilbinger, Luo et al. 2024 we presented the first direct weak-lensing constraint on the black-hole–halo mass relation. Cross-correlating ~36,000 SDSS Type I/II AGNs with UNIONS shear from 9.4 × 107 lensed background galaxies, we find more massive AGNs in more massive halos, well described by a power law of slope ~0.6 below MBH ≲ 108.5 M⊙, consistent with feedback-regulated models linking black-hole growth to baryon physics.
Globular clusters as halo tracers
A near-linear scaling between globular cluster mass and halo mass
In Hudson, Harris & Harris 2014 we combined a comprehensive globular-cluster (GC) population database with halo masses from weak lensing and found that η ≡ MGCS/Mh is essentially constant at ⟨η⟩ ~ 4 × 10−5 with intrinsic scatter ≤ 0.2 dex. Globular clusters are the only known stellar population that formed in essentially direct proportion to host halo mass. This implies most GCs formed very early, largely insensitive to the feedback that suppressed subsequent field-star formation; the mean η further implies that about a quarter of the initial protogalactic gas collected into clusters large enough to host GCs. As a by-product, applying the calibration gives halo masses for the Milky Way ((1.2 ± 0.5) × 1012 M⊙) and M31 ((3.9 ± 1.8) × 1012 M⊙).
Globular cluster system sizes track halo sizes — non-linearly
In Hudson & Robison 2018 we measured the radial density profiles of GC systems in nearby galaxy groups (well fit by a de Vaucouleurs profile) and combined them with literature data to relate the effective radius of the GC system to the virial radius of the halo. The relation is tight (~0.2 dex scatter) but steeply non-linear: Re,GCS ∝ R2002.5–3 ∝ M2000.8–1 for halos above ~1012 M⊙ — a constraint that encodes both GC assembly and halo merger history.
Mergers
Probing merger remnants with galaxy–galaxy lensing
In Cheng, Elvin-Poole, Hudson et al. 2025 we used UNIONS galaxy–galaxy lensing of 1,623 post-coalescence mergers and ~30,000 non-merging controls (matched in stellar mass, redshift, and environment) to test whether the merger process changes the host halo. The current data show no statistically significant difference in the excess surface density profile; both samples have halo masses Mhalo ~ 4 × 1012 M⊙. The non-detection lets us rule out, at 95% confidence, merger-induced starbursts in which more than 60% of the stellar mass is formed in the burst. Stage-IV surveys with ~10× larger samples should detect the lensing signature directly.
Open questions
- What sets the peak of the SHMR, and why does the peak shift with redshift? The downsizing in the quenching mass scale we detected with CFHTLenS is a strong constraint for feedback models; UNIONS depth will let us track it across the full mass range.
- What drives the steep Re,GCS–M200 scaling? Linear MGCS–Mh with steeply non-linear Re,GCS–R200 is an unusual combination — suggesting GC assembly is dominated by accretion of progenitor halos rather than in-situ formation.
- How does halo shape and orientation evolve with merger history? Our halo-ellipticity measurements show the misalignment between stellar and dark major axes decreasing with mass; tracking it across the merger sequence with UNIONS-deep and Euclid is the obvious next step.
- How long does a merger burst last, and where does the stellar mass go? The Reeves & Hudson burst fraction (~18%) is several times higher than hydrodynamical-simulation predictions; understanding whether this reflects burst duration, star-formation efficiency, or sample selection is an active question.
- Does the merger process leave a measurable imprint on the halo? Our current limit (no detection at ~1013 M⊙) is consistent with simulations; the Stage-IV (LSST, Euclid) era will be sensitive enough to test merger-driven halo concentration changes directly.
- The galaxy–halo connection at z > 1. Most of our constraints are at z ≲ 0.5. Pushing weak lensing and clustering to redshifts where Euclid will resolve individual sources lets us track halo masses and assembly histories into the regime where galaxies are actively forming.