The Cosmic Web

The shape of the universe

Galaxies bind into groups, groups into clusters, clusters into filaments — and the filaments wrap around the empty majority.

Matter in the universe is organised on a ladder: galaxies bind into groups, groups into clusters, clusters string along filaments, and the filaments wrap around voids that fill most of space. Every rung of that ladder is visible in this catalog, and this page walks up it using nothing but 3,957,865 DESI galaxies.

Density map of galaxies in a wedge, showing filaments and voids
A thin slab of sky, 198,302 galaxies, opened out from us at the apex to 641 Mpc. Nothing has been smoothed or idealised — the filaments, the walls and the dark voids between them are simply where the galaxies are. Excludes galaxies outside this declination slab, and z > 0.15.

Galaxies

The bottom rung. A galaxy is a gravitationally bound collection of stars, gas, dust and dark matter, and for the purposes of a map like this one it is a single point: a position on the sky and a redshift. This catalog holds 3,957,865 of them, out to 6.3 billion light-years.

They are not anonymous. Cross-matching against HyperLEDA gives 486,546 of them a catalogue identity, 134,843 of those a morphological type — spiral, elliptical, irregular — and 2,225 a designation you can type into the map's search box and fly to.

Groups and clusters

Galaxies are social. Left alone under gravity for thirteen billion years, they fall towards one another and end up in bound systems: a pair, a handful, a few dozen, occasionally a thousand. Finding those systems in a catalog is a data-mining problem, and the classical solution is friends-of-friends: link every pair of galaxies closer than some distance, then call each connected clump a group.

Run on this catalog between z = 0.01 and z = 0.15, it finds 39,328 systems containing 529,866 galaxies — 44% of every galaxy in that range. The median system has 7 members. The largest has 2,007. 1,278 have fifty or more, which is where astronomers stop saying "group" and start saying "cluster".

Stephan's Quintet
Stephan's Quintet — the classic compact group, and a good lesson in reading one. Four of these galaxies are genuinely bound and tearing at each other; the large bright spiral at lower left is a foreground object about five times closer, which happens to lie in the same direction. A group finder working from positions on the sky alone would put all five together; redshifts separate them. Stephan's Quintet — J. Rachlin, 14-inch PlaneWave, Lowell Observatory (© John Rachlin).

Groups are not always obvious. Some are compact knots like the Quintet; others announce themselves only by the damage — a bridge of stars pulled between two discs, or a tail flung out by a pass that happened hundreds of millions of years ago and is still unwinding.

UGC 11797 / 11798 / 11801, Cygnus
A less famous group in Cygnus: UGC 11797, 11798 and 11801, roughly 200 million light-years away. The long straight filament running between the galaxies is a tidal tail — stars pulled out of a disc by a close pass, still tracing the path they were thrown along. UGC 11797 / 11798 / 11801, Cygnus — J. Rachlin, 14-inch PlaneWave, Lowell Observatory (© John Rachlin).
Number of groups against number of member galaxies, log-log
The hierarchy, counted. Small systems outnumber large ones by orders of magnitude: 39,328 groups, of which only 1,278 reach fifty members. This steeply falling shape is what a universe built by gravitational collapse looks like. Excludes nothing — every group in the catalog is counted.

Weighing a cluster you cannot touch

Here is where a catalog stops being a list and starts being physics. The galaxies in a bound system orbit their common centre of mass, so their redshifts scatter around the system's average. That scatter — the velocity dispersion — is set by how much mass is pulling on them. A heavier system makes its members move faster.

So the prediction is simple: richer systems should show larger dispersions. The catalog agrees. Across 39,328 groups the median dispersion climbs from about 147 km/s for the smallest systems to 602 km/s for the richest — a relation nobody put into the data, recovered from redshifts alone.

Velocity dispersion against group richness, log-log, with median trend
Each grey dot is one group; the cyan line is the running median. The upward trend is the virial relation — more mass, faster orbits — and it is the basis of essentially every measurement of cluster mass ever made. Excludes bins holding fewer than 8 groups, and single-member systems.

Filaments, walls and voids

Above clusters the structure stops being bound. Filaments and walls are not objects orbiting a centre; they are the pattern left behind as matter drains out of the emptying regions and onto the dense ones. Look again at the slice at the top of this page: the bright ridges are filaments seen in projection, the blank regions are voids, and the voids are the majority of the volume.

This is the cosmic web, and its shape is a fossil. The pattern grew from density ripples already imprinted when the universe was a few hundred thousand years old — the same ripples visible in the cosmic microwave background. Gravity has been amplifying them ever since, draining the underdense regions and piling matter onto the overdense ones.

Two things the picture gets wrong

Both are worth knowing before reading structure off any redshift map, including this one.

Fingers of God. A galaxy's own motion adds to its redshift, so in a cluster — where motions are fastest — members are smeared along the line of sight into a spike pointing at the observer. Those radial streaks in the slice are not filaments. They are the same velocity dispersion that let us weigh the cluster, now corrupting its shape.

The survey thins with distance. A brightness-limited survey sees only the intrinsically luminous galaxies far away, so the apparent emptying of the map with distance is mostly the telescope running out of reach.

Galaxy number density against redshift, log scale
Galaxies per cubic megaparsec against redshift. The decline is the survey's selection function, not the universe: the distant universe is just as full as the nearby one. The map's density field corrects for this; the raw point cloud does not. Excludes bins holding 50 galaxies or fewer.
The group catalog on this page is produced by a straightforward friends-of-friends run and is meant for teaching, not for research. It over-merges systems joined by a single bridging galaxy, its membership depends on linking lengths that are a choice rather than a measurement, and it has not been validated against a published group catalog.

The map is the place to go next: the same 39,328 groups are drawn over the point cloud as an optional layer, and the richest are labelled.