Cosmological constraints from the Minkowski functionals of the BOSS CMASS galaxy sample

The Big Picture

Imagine trying to describe a sponge by measuring only the average size of its holes. You’d miss the most interesting parts: how its tunnels twist and connect, whether its voids are isolated bubbles or a labyrinthine network. Scale that sponge to the size of the observable universe, and you have the cosmic web, the vast filamentary structure of galaxies and dark matter that cosmologists are trying to decode.

For decades, the standard tool for mapping this structure has been the two-point correlation function (2PCF), a measure of how often galaxy pairs appear at different separations. It’s powerful, but incomplete. It captures regular clustering patterns while missing the complex, irregular shapes gravity has sculpted over billions of years. Those shapes encode secrets about dark matter, dark energy, and the primordial density fluctuations that seeded everything we see.

A new study from Wei Liu and collaborators cracks open that hidden information. They’ve built the first complete, simulation-based model for Minkowski functionals, mathematical tools from topology that measure the full shape and connectivity of cosmic structure, and applied it to real galaxy data from the BOSS survey. The result: constraints on key cosmological parameters up to twice as tight as the standard approach.

Key Insight: By measuring not just how galaxies cluster but how the cosmic web is shaped — its volume, surface area, curvature, and connectivity — Minkowski functionals extract non-Gaussian information that two-point statistics simply cannot reach.

How It Works

The four Minkowski functionals each measure a distinct geometric property of the galaxy density field, a 3D map of where matter concentrates and where it thins out. They capture: (1) the volume of overdense regions above a density threshold, (2) the surface area bounding those regions, (3) the integrated mean curvature of that surface, and (4) the Euler characteristic, a single number encoding overall connectivity that distinguishes isolated blobs from tunnels from enclosed voids. Sweep across density thresholds and you trace how the cosmic web’s topology shifts from dense filaments to empty voids.

Earlier MF analyses smoothed heavily, deliberately washing out irregular features to avoid modeling them. This team took a different route: full simulation-based inference, built in four steps.

1. Simulate: Thousands of mock galaxy catalogs from AbacusSummit N-body simulations, tracking billions of gravitating particles across cosmic time, spanning a grid of cosmological parameters.
2. Populate: The halo occupation distribution (HOD) framework places galaxies inside dark matter halos, with varying HOD parameters to account for uncertainty in galaxy formation.
3. Add realism: Full observational effects are injected. Redshift-space distortions, survey geometry and veto masks, angular and radial selection effects, and the Alcock-Paczynski distortion (a geometric smearing that arises when you assume the wrong cosmology to convert observed positions into a 3D map).
4. Emulate: A neural network trained on all these mocks predicts MFs (and separately, the 2PCF) for any combination of cosmological and HOD parameters, fast enough to run Markov chain Monte Carlo inference across millions of parameter combinations.

The emulator covers six cosmological parameters plus six HOD parameters simultaneously. Before touching real data, the team ran extensive validation: feed mock catalogs through the pipeline and check whether inferred parameters match the known truth. They also tested on mocks from a completely different galaxy formation model, the Uchuu simulation with an independent HOD prescription, to probe model dependence. The pipeline passed.

Why It Matters

Applied to the BOSS DR12 CMASS sample (roughly 600,000 galaxies at redshifts 0.45–0.58), the team measured MFs and 2PCF from the same data, fitting them separately and jointly. The MFs alone outperformed the 2PCF alone on every key parameter. Their joint analysis delivered:

• ω_cdm (cold dark matter density): 2.0× tighter than 2PCF alone
• σ_8 = 0.783 ± 0.026 (amplitude of matter fluctuations): 1.9× tighter
• n_s = 0.966 (spectral index of primordial fluctuations): 1.6× tighter
• fσ_8 = 0.453 ± 0.016 (growth rate of cosmic structure): 1.9× tighter

These results agree well with Planck 2018 predictions and with other analyses of the same dataset. That consistency matters as much as the precision. It means the non-Gaussian information is being extracted cleanly, not corrupted by modeling assumptions.

We are entering an era of galaxy surveys (DESI, Euclid, the Roman Space Telescope) so vast and precise that squeezing every bit of statistical power from the data becomes essential. The two-point function leaves a significant fraction of that power on the table. Minkowski functionals are one of the most promising ways to recover it.

The methodology here is as significant as the result itself. Previous MF analyses either operated where non-Gaussian effects were negligible or applied approximate analytical corrections. This team built a full forward model: no approximations, no linearization, everything from the simulation up. That pipeline transfers directly to current DESI data and future surveys, where galaxy samples will be orders of magnitude larger.

Open questions remain. The HOD model, while tested against alternative prescriptions, is still an idealized description of galaxy formation. How well MF constraints hold up against more exotic halo-galaxy connection models, including assembly bias (where galaxy properties depend on when and how their dark matter halo formed), velocity bias, and environmental effects, is the next frontier. Early results look encouraging, but stress-testing against hydrodynamical simulations will be essential as precision demands grow.

Bottom Line: By building the first simulation-based model capturing the full topology of the cosmic web, including its non-Gaussian structure, this team doubled the constraining power on key cosmological parameters. The approach is ready for next-generation surveys to extract far more from large-scale structure than two-point statistics ever could.