When the universe was barely one billion years old, galaxies were already stringing themselves along vast filaments of matter, separated by enormous empty voids. Now, for the first time, astronomers can see that architecture in detail. A team led by researchers at the University of California, Riverside has used the James Webb Space Telescope to build what the university describes as the most detailed three-dimensional map of the early cosmic web ever assembled, tracing 164,000 galaxies back to an era when the first stars were still burning away the primordial fog of hydrogen that filled intergalactic space.
“This is the first time we’ve been able to map the cosmic web in such detail at these early epochs,” said Santosh Harish, a postdoctoral researcher at UC Riverside and lead author of the study, in a university press release. The results, published in June 2026 in The Astrophysical Journal, draw on the COSMOS-Web survey, the largest contiguous imaging program ever carried out with JWST. The map stretches from the relatively nearby universe all the way out to a redshift of roughly 7, offering a direct look at the scaffolding that would eventually support the galaxy clusters and superclusters visible today.
Inside the largest JWST imaging survey
COSMOS-Web consumed 255 hours of precious telescope time and covered 0.6 square degrees of sky with JWST’s NIRCam instrument, an area equivalent to about three full moons. An additional 0.2 square degrees were observed with the MIRI mid-infrared camera. That wide, unbroken field of view is what makes the cosmic-web reconstruction possible: mapping large-scale structure requires observing enough volume to capture filaments that can stretch tens of millions of light-years.
The broader galaxy catalog underpinning the work contains more than 700,000 galaxies with measured photometry, morphology, photometric redshifts, and derived physical properties such as stellar mass and star-formation rate, as described in a companion catalog paper. From that pool, the team selected the 164,000 galaxies whose redshift estimates met strict quality thresholds, ensuring the map reflects genuine clustering rather than measurement noise.
To convert those galaxy positions into a three-dimensional picture, the researchers divided the sample into a sequence of narrow redshift bins. Within each slice, they smoothed the galaxy distribution using a technique called weighted kernel density estimation, highlighting where galaxies cluster into filaments and nodes and where they thin out into voids. Stacking the slices creates something like a time-lapse of cosmic construction: the highest-redshift layers capture the earliest, faintest strands of the web, while lower-redshift layers show it growing denser and more complex as the universe ages.
Because the survey area is contiguous, the reconstruction can follow individual filaments across large stretches of the field rather than seeing them only in fragmented pieces. That continuity is essential for tracking how proto-clusters sit at the intersections of filaments and for measuring the sizes of the voids carved out between them. The team has also released a video visualization that sweeps through redshift, giving an intuitive sense of how the web evolves, along with the full pipeline, catalog, and density products, according to the UC Riverside press release.
Why the reionization era matters
The map’s deepest layers peer into the reionization era, the period roughly 500 million to one billion years after the Big Bang when ultraviolet radiation from the first stars and galaxies was ionizing the neutral hydrogen that had pervaded the universe since shortly after the Big Bang. Understanding how galaxies were distributed during reionization is critical because their arrangement likely determined how quickly and unevenly that process unfolded. Dense filaments packed with young galaxies would have carved ionized bubbles faster than isolated regions, creating a patchwork pattern that left imprints still detectable in the cosmic microwave background.
Earlier attempts to map the cosmic web at these distances relied heavily on gravitational lensing to trace dark matter or on spectroscopic surveys that could only reach moderate redshifts. Hubble-era reconstructions were limited by that telescope’s infrared sensitivity and narrower field of view. JWST’s NIRCam detects galaxies at higher redshifts and fainter magnitudes, which is why the new map reaches epochs that were previously inaccessible to direct galaxy-based reconstruction. While previous surveys could trace large-scale structure with high fidelity out to redshifts of roughly 1 to 2, COSMOS-Web pushes that frontier to z~7 using a single, uniform dataset, reducing the systematic differences that creep in when stitching together multiple surveys and instruments.
What the map does not yet tell us
For all its depth, the map traces where galaxies sit, not where most of the mass resides. Galaxies are visible tracers of a web dominated by dark matter, and distinguishing the distribution of ordinary matter from the underlying dark-matter skeleton will require additional data, particularly weak-lensing measurements and spectroscopic confirmation of redshifts.
The current reconstruction relies on photometric redshifts, which are estimated from a galaxy’s brightness across multiple filters rather than from precise spectral lines. These carry larger uncertainties than spectroscopic measurements. While the catalog paper provides quantitative performance metrics for the photo-z estimates, the degree to which residual redshift errors blur the density field at the highest redshift slices has not been fully characterized in published documentation.
It is also not yet clear whether the observed filament densities at z~7 match, exceed, or fall below predictions from standard cosmological simulations. Denser-than-expected filaments could suggest that gravitational collapse proceeded faster than cold-dark-matter models predict. More diffuse filaments might point to feedback processes, such as supernovae or early black-hole activity, dispersing gas and slowing the buildup of massive systems. Answering that question will require running simulations against the observed density field and, ideally, incorporating spectroscopic follow-up to sharpen the redshift measurements.
Cosmic variance adds another layer of uncertainty. The COSMOS-Web field, while large by JWST standards, still samples only a single patch of sky. If that region happens to be slightly over-dense or under-dense compared with the cosmic average, measurements of filament thickness, void sizes, and proto-cluster abundance could be skewed. Expanding similar mapping techniques to additional JWST fields or to upcoming wide-area infrared missions, such as NASA’s Nancy Grace Roman Space Telescope and ESA’s Euclid survey, will be necessary to determine how representative this slice of the early web truly is.
A public laboratory for early-universe science
By releasing their full pipeline, density products, and visualization, the COSMOS-Web team has effectively turned the map into a shared laboratory. Other research groups can now apply independent analysis methods, cross-match the structures with spectroscopic surveys, and overlay weak-lensing data where available. Over time, that layering of evidence should sharpen the picture considerably, transforming what is now an impressive visualization into a quantitative testing ground for theories of how the universe built its first large-scale structures.
For now, the map stands as the clearest view yet of cosmic architecture at the edge of observational reach. The filaments threading through those 164,000 galaxies are not just a snapshot of the distant past. They are the blueprint for everything that followed: the clusters, the voids, and the galaxy-studded web that the universe still inhabits today.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
