The Cosmic Web: How Galaxies Hang Together

For most of the twentieth century, the working assumption about the universe at the largest scales was that galaxies were spread roughly uniformly through space. Slight clumping was expected from gravity, but the overall picture was something like a homogeneous fluid of galaxies.

The first hint that this was wrong came from optical surveys in the 1970s. The picture that emerged through the 1980s and was confirmed dramatically by the Sloan Digital Sky Survey in the 2000s is much stranger and more beautiful: galaxies are arranged along enormous filaments and walls, separated by near-empty voids, with massive clusters at the intersections. The whole thing is called the cosmic web.

This note summarises what the surveys have shown and how confidently we can call it a "network."

The first map: Geller and Huchra, 1986

The Center for Astrophysics Redshift Survey, published by Margaret Geller and John Huchra in 1986, was the first survey large enough to reveal the structure clearly. They measured redshifts (and therefore approximate distances) for several thousand galaxies in a thin strip of sky. When plotted in a wedge diagram (right ascension on one axis, distance on the other), the galaxies did not fill the wedge uniformly. They formed long curving features. The most striking was a roughly 500-million-light-year-long sheet of galaxies that became known as the Great Wall.

It was the first time anyone had seen large-scale cosmic structure as a picture. The CfA redshift survey only covered a small slice of sky, but the implication was clear: do this for the whole sky and the picture would not become uniform. It would become a network.

The current picture: SDSS and the era of millions of galaxies

The Sloan Digital Sky Survey (SDSS), operating since 2000, has now measured redshifts for over five million galaxies and quasars. The maps it produces show the cosmic web at high resolution out to redshifts of roughly z = 0.7 (about 7 billion light-years of look-back time). The structures the early surveys hinted at are now characterised quantitatively:

• Filaments: thread-like structures, typically 10 to 100 megaparsecs long, with diameters of a few megaparsecs. Galaxies are densely strung along them. Most galaxies in the universe are inside filaments.
• Walls (sheets): two-dimensional structures where multiple filaments meet. The Great Wall is one; the Sloan Great Wall (discovered 2003, about 1.4 billion light-years long) is among the largest known.
• Voids: regions almost devoid of galaxies, typically 30 to 100 megaparsecs across, occupying most of the universe's volume. The largest known void (the Boötes Void, discovered 1981) is about 330 million light-years across.
• Clusters and superclusters: dense knots at the intersections of filaments. The largest are several megaparsecs across and contain thousands of galaxies. Examples include the Virgo, Coma, and Shapley clusters.

The web is not random in any naive sense. It is a deterministic consequence of small density fluctuations in the very early universe, amplified by gravitational instability over 13.8 billion years.

The simulation match

One of the strongest pieces of evidence that we understand the cosmic web is that N-body simulations (Millennium, IllustrisTNG, FLAMINGO, and others) produce structures that match observations to a remarkable degree. Run the standard ΛCDM cosmological model forward from primordial density fluctuations imprinted on the cosmic microwave background, let gravity work for 13.8 billion years on a grid of particles, and the resulting filament-and-void pattern matches the surveys quantitatively in:

• Filament length distribution
• Void size distribution
• Two-point galaxy correlation function
• Cluster mass function
• The galaxy bias parameter

This is one of the cleanest confirmations of ΛCDM at large scales. Where ΛCDM is contested (the Hubble tension at low redshift, possible tensions in the matter-clustering parameter σ₈) the cosmic web's structural agreement is not in dispute. The web's existence and its gross statistics are settled.

What's still being argued

Several things about the cosmic web remain open:

1. The "missing baryons" problem. Inventories of visible matter at low redshift come up short compared to predictions from the CMB. The leading hypothesis is that a substantial fraction of the missing baryons is in the warm-hot intergalactic medium (WHIM) along filaments, at temperatures of 10⁵ to 10⁷ K, too hot to glow visibly but too cool to show up in standard X-ray surveys. Recent measurements (eROSITA, IllustrisTNG comparisons, and stacked-filament analyses) have detected the WHIM in absorption and stacking; whether the full missing-baryon budget is accounted for is still being worked out.

2. The largest scales. Are structures like the Hercules-Corona Borealis Great Wall (claimed in 2013, ~10 billion light-years in extent) real gravitationally-coherent structures, or are they statistical fluctuations in survey selection? At these scales, ΛCDM predicts the universe should look homogeneous; structures the size of GRB-defined walls would be in tension with that. The community is still arguing about the statistical significance.

3. Filament substructure and galaxy formation. How a galaxy's properties (mass, type, gas content) depend on its position within the cosmic web (inside a void, on a filament, at a cluster outskirt) is an active area. Galaxies near filaments tend to be aligned with them; this is a real effect but the mechanism is debated.

4. Web evolution. The cosmic web at z = 2 looks different from the web at z = 0: filaments are denser, voids are smaller. Quantifying this evolution is a major science goal for upcoming surveys (DESI, Euclid, Roman Space Telescope, LSST).

Why this matters for the lab

LokLab's pipeline operates almost entirely at solar-system scales, so the cosmic web is not a calibration target for us. But two things are worth keeping in mind:

1. The galactic-rotation correction we apply when reducing high-precision astrometric work depends on the local kinematic flow, which is itself a small-scale projection of the supercluster-scale flow toward the Great Attractor. We use the standard local-group rest-frame correction; it is good to roughly 0.5%.
2. When we publish anything that mentions "redshift" or "Hubble's law", we are implicitly using a cosmological model that is calibrated to the cosmic web's statistics. The Hubble tension is one of the active threads in this calibration.

We will return to the cosmic web with a more detailed note on the WHIM detection methods sometime in the next year. The methodology is genuinely interesting and worth a full treatment.

References

Primary sources for the figures and claims above:

• Geller & Huchra, Science, 1989, "Mapping the Universe" (the Great Wall paper).
• York et al., AJ, 2000 (SDSS overview and ongoing data releases).
• Gott et al., ApJ, 2005 (Sloan Great Wall discovery).
• Kirshner, Oemler, Schechter, Shectman, ApJ, 1981 (the Boötes Void).
• Tegmark et al., ApJ, 2004 (SDSS three-dimensional power spectrum).
• Springel et al., Nature, 2005 (Millennium Simulation).
• Nicastro et al., Nature, 2018 (WHIM filament detection).
• Zhang et al., 2024 (eROSITA + stacking WHIM constraints).

The full bibliography lives at /bibliography.

---

All numerical figures are from peer-reviewed publications referenced above. No original observations are claimed in this note; it is a synthesis with sourced references.