For decades, astronomers have known that galaxies are not scattered randomly through space. They cling to an enormous, mostly invisible network of dark-matter filaments, a structure so vast it dwarfs anything visible to the naked eye. Scientists call it the cosmic web. Now, for the first time, a team using the James Webb Space Telescope has mapped that hidden architecture in unprecedented detail, tracing its threads back to when the universe was barely one billion years old.
The results, published in May 2026 across two peer-reviewed studies, draw on roughly 164,000 galaxies cataloged through COSMOS-Web, the largest survey ever conducted with JWST. The dark-matter map they produced is twice as sharp as anything the Hubble Space Telescope achieved in the same patch of sky, according to NASA. It is the clearest picture yet of how unseen matter guided the earliest stages of galaxy formation.
A web you cannot see but everything depends on
Dark matter makes up roughly 27 percent of the universe’s total energy content, yet it emits no light and interacts with ordinary matter only through gravity. Its presence is inferred from the way galaxies rotate, the way light bends around massive clusters, and the patterns imprinted on the cosmic microwave background. The cosmic web is its largest expression: a sprawling lattice of filaments, dense nodes, and near-empty voids that stretches across billions of light-years and acts as the scaffolding along which galaxies form, grow, and migrate.
Mapping that scaffolding directly has been one of observational cosmology’s most persistent challenges. Previous efforts using the Hubble Space Telescope produced the first lensing-based dark-matter maps of the COSMOS field, but Hubble’s optical sensors could not detect enough faint, distant galaxies to resolve fine filament structure at high redshift. JWST’s infrared instruments changed the equation by revealing galaxies that Hubble simply could not see.
How the map was built
The two new studies attacked the problem from complementary directions. The first, published in The Astrophysical Journal, used a technique called weighted kernel density estimation to trace how the 164,000 cataloged galaxies cluster along dark-matter filaments. By sorting galaxies into slices of photometric redshift, the team reconstructed the large-scale density field out to a redshift of roughly 7, corresponding to structures that existed about one billion years after the Big Bang. The reconstruction spans roughly six billion years of cosmic history, showing how filaments thickened and voids deepened as the universe aged.
The companion paper, published in Nature Astronomy, took a different approach: weak gravitational lensing. Every massive structure between us and a distant galaxy bends that galaxy’s light by a tiny amount, distorting its apparent shape. By measuring those distortions across the survey field, the team built a mass map that does not depend on where galaxies happen to sit but instead reveals where all matter, visible and dark, actually concentrates. That paper is available through the journal at doi:10.1038/s41550-025-02507-x.
That lensing map covers a footprint of 0.77 by 0.70 degrees with an angular resolution of about one arcminute and a source galaxy density of roughly 129 galaxies per square arcminute. Both the footprint size and the source density represent clear advances over earlier Hubble-era lensing maps of the same field. The improvement allowed the team to recover not just the dense ridges of filaments but also the underdense voids between them.
NASA published a side-by-side comparison of the Webb and Hubble maps of the same sky region, illustrating how the higher source density from Webb’s infrared sensors pulls finer filament geometry into focus.
Both teams drew on a shared galaxy catalog whose selection methods, photometric redshift calibrations, and systematic controls are detailed in an open-access survey preprint. That catalog acts as the backbone for the density-field reconstruction and a key input for the lensing shape measurements.
The people behind the map and why it matters to them
“We are essentially seeing the skeleton of the universe for the first time at this level of detail,” said Caitlin Casey, a University of Texas at Austin astronomer who serves as co-principal investigator of COSMOS-Web, in the NASA announcement. The survey mobilized more than 100 researchers across dozens of institutions, and Casey described the collaboration as years in the making, beginning well before JWST’s first science observations.
Kartik Sheth, a JWST program scientist at NASA headquarters, framed the result in broader terms. “These maps show us how dark matter has been pulling the strings of galaxy formation since the universe was very young,” he said in the same release. For the researchers involved, the payoff is not just a sharper image but a tool that can be cross-matched against simulations, X-ray observations, and spectroscopic surveys to test fundamental predictions about how structure grows.
Why it matters beyond a pretty picture
A sharper map of the cosmic web is not just an aesthetic upgrade. The distribution of dark matter at different epochs is one of the most direct tests of the standard cosmological model, known as Lambda-CDM. If filaments at high redshift turn out to be thicker, thinner, or more connected than simulations predict, it could signal that something about our understanding of dark matter or dark energy needs revision.
The new maps also open a window into galaxy evolution. Galaxies living at the intersections of filaments, the densest nodes of the web, experience different gravitational environments than those drifting through voids. Tracking how galaxy properties such as star-formation rate and morphology vary with position in the web could reveal whether environment or internal processes play the larger role in shutting down star formation, a question that has divided the field for years.
COSMOS-Web’s original design document listed the study of rare quiescent galaxies at high redshift among its core science goals. The new maps lay the groundwork for that analysis, though neither paper has yet published a direct statistical comparison of quiescent galaxy fractions at filament nodes against predictions from hydrodynamical simulations.
What the maps cannot yet tell us
Several important caveats temper the findings. The galaxy-density reconstruction relies on photometric redshifts, which estimate distance from broadband colors rather than precise spectral lines. That introduces scatter along the line of sight, meaning some filaments could be smeared or blended with neighboring structures. No independent spectroscopic follow-up of the full 164,000-galaxy sample has been published, so the exact degree of smearing remains model-dependent.
The weak-lensing map, while sharper than Hubble’s, still covers less than one square degree of sky. Cosmologists studying the cosmic web on the largest scales typically need survey areas hundreds of times larger to draw robust statistical conclusions about how representative any single field is. The COSMOS field is one of the most cross-checked patches of sky in astronomy, which helps, but the results may not generalize to regions with very different large-scale density.
There is also the question of galaxy bias: on large scales, galaxies are expected to trace the underlying dark matter, but the precise relationship between the two varies with galaxy type, mass, and redshift. The current analyses show qualitative agreement between galaxy overdensities and lensing peaks, but a fully quantitative, scale-dependent bias measurement across the full redshift range has not been finalized in the published work.
Even with JWST’s sensitivity, the galaxy catalog becomes incomplete at the faintest, lowest-mass end, particularly at high redshift. Narrow filaments traced mainly by small galaxies could be underrepresented, subtly biasing measurements of filament thickness, connectivity, or the contrast between filaments and voids.
What spectroscopic follow-up and wider surveys must still resolve
The COSMOS-Web team’s next steps will likely include spectroscopic follow-up to sharpen redshift estimates, cross-matching with X-ray and radio surveys to trace hot gas along filaments, and direct comparisons with large-scale cosmological simulations. Other JWST programs and upcoming ground-based surveys, including the Vera C. Rubin Observatory’s Legacy Survey of Space and Time, are expected to extend cosmic-web mapping across far wider areas of sky, providing the statistical power that a single deep field cannot.
For now, the strongest verified numbers from these studies are clear: 164,000 galaxies, 129 source galaxies per square arcminute, a 0.77-by-0.70-degree footprint, and a redshift reach of roughly 7. Together, they form the most detailed observational portrait of the universe’s hidden framework, a structure that, while invisible, shapes every galaxy we can see.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
