A violation of this premise would require rethinking the foundations of cosmology
For nearly a century, the universe has been understood as a place of deep symmetry — the same in every direction, uniform at the grandest scales. Now, a team of physicists led by Asta Heinesen of the Niels Bohr Institute has found subtle but statistically meaningful cracks in that assumption, using machine learning to let observational data from supernovae and dark energy surveys speak without theoretical constraint. The findings, still preliminary, do not yet meet the threshold of confirmed discovery — but they whisper a possibility that has long unsettled cosmology: that the universe may not be as evenly woven as we have believed.
- A foundational pillar of modern cosmology — the assumption that the universe is uniform at its largest scales — is showing signs of strain under new observational scrutiny.
- Deviations of 2 to 4 sigma from expected cosmic behavior were detected across multiple datasets, a signal too persistent to dismiss but not yet strong enough to confirm.
- The research team deployed machine learning to strip away theoretical assumptions and let the raw data reconstruct the universe's expansion history on its own terms.
- Two competing physical effects — the Dyer-Roeder illusion of emptier space and the gravitational backreaction of massive cosmic structures — are the leading candidates to explain what is being seen.
- If the signal survives future surveys, entire frameworks built atop the FLRW model — dark energy theories, modified gravity, exotic matter proposals — may need to be rebuilt from the ground up.
For nearly a century, cosmology has rested on a single sweeping assumption: zoom out far enough, and the universe looks the same in every direction. This principle of homogeneity and isotropy is so foundational that it underpins the Friedmann-Lemaître-Robertson-Walker model — the mathematical skeleton of everything we believe about cosmic expansion, dark energy, and the universe's fate. New research now suggests that skeleton may have a fracture.
Asta Heinesen, a physicist at the Niels Bohr Institute and Queen Mary University, led a team that tested whether the universe actually behaves as FLRW predicts. Drawing on supernova catalogs, Dark Energy Spectroscopic Instrument measurements, and baryon acoustic oscillation surveys, they reconstructed the history of cosmic expansion without first imposing any theoretical model — then compared the result to what FLRW would expect. The deviations they found were small but statistically meaningful, ranging from 2 to 4 sigma depending on the dataset.
Physics sets its discovery threshold at 5 sigma, meaning these findings remain preliminary. But Heinesen's team considers them significant enough to warrant serious attention, noting a "surprising violation of a FLRW curvature consistency test" that could point toward new physics. Two effects are under consideration: the Dyer-Roeder phenomenon, in which light traveling through voids makes matter appear sparser than it is, and cosmological backreaction, in which the growth of galaxy clusters and cosmic voids may itself alter the universe's expansion rate. To isolate these signals, the team used symbolic regression — a form of machine learning that allows the data to find its own patterns rather than conforming to a predetermined shape.
The stakes of confirmation would be enormous. Most leading proposals for resolving current cosmological tensions — evolving dark energy, modified gravity, exotic matter — are built on the assumption that FLRW is correct. A fundamental violation of that premise would send cosmologists back to first principles. The datasets available today are still sparse, but the team notes their methods can be applied to existing observations immediately, and more precise surveys in the coming years should bring a clearer answer. Whether the universe is truly uniform at its largest scales remains an open question — one whose resolution may reshape our understanding of the cosmos itself.
For nearly a century, cosmology has rested on a single foundational assumption: that when you zoom out far enough, the universe looks the same everywhere. Matter is distributed evenly. No direction is special. This principle of homogeneity and isotropy is so central to modern physics that it's baked into the Friedmann-Lemaître-Robertson-Walker model, the mathematical framework underlying everything we think we know about cosmic expansion, dark energy, and the universe's past and future. But new research suggests this bedrock assumption may be cracked.
Asta Heinesen, a physicist at the Niels Bohr Institute and Queen Mary University, led a team that developed novel methods to test whether the universe actually behaves as the FLRW model predicts. Using real observational data—supernova catalogs, measurements from the Dark Energy Spectroscopic Instrument, and baryon acoustic oscillation surveys—they reconstructed the history of cosmic expansion without imposing any theoretical model first. Then they compared what the universe actually did to what FLRW said it should do. The results, still in preprint on arXiv, revealed small but statistically meaningful deviations from the expected pattern.
The significance of these deviations ranged from 2 to 4 sigma depending on the dataset and method used. In physics, the gold standard for declaring a discovery is 5 sigma—a threshold so strict it means you'd see a false result like this only once in 3.5 million trials. These findings fall short of that bar, which is why Heinesen and her colleagues are careful to call them preliminary. But they are intriguing enough to demand serious attention. "We observed a surprising violation of a FLRW curvature consistency test, suggesting new physics beyond the standard model," Heinesen told Live Science. "This could be due to several effects, but more research is needed to determine the exact cause."
Two phenomena could explain the deviations. The first is the Dyer-Roeder effect: light from distant objects travels mostly through nearly empty regions of space, making the actual density of matter appear lower than it really is. The second is cosmological backreaction—the idea that the growth of massive structures like galaxy clusters and cosmic voids might actually alter the universe's expansion rate itself. To isolate these effects, the team used machine learning, specifically symbolic regression, to let the data speak for itself rather than forcing it into a predetermined shape.
If these deviations hold up under scrutiny, the implications would be seismic. Many of the leading proposals for solving current cosmological puzzles—evolving dark energy, modified gravity theories, exotic forms of matter—all assume the FLRW framework is correct. A fundamental violation of that framework would require cosmologists to go back to first principles and rebuild. "If the deviations are confirmed with future data, the consequences would be profound," the researchers note. "Many proposals for resolving current cosmological tensions are based on the FLRW framework. A violation of this premise would require rethinking the foundations of cosmology."
The team is careful to acknowledge the limitations of current data. The datasets available today are sparse, especially when it comes to measuring expansion rates across different cosmic epochs. But they also note that their method can be applied immediately to existing observations, and within a few years, more precise surveys should provide a definitive answer. The question of whether the universe is truly uniform at the largest scales remains open. The answer, when it comes, may force a fundamental reckoning with how we understand the cosmos itself.
Citas Notables
We observed a surprising violation of a FLRW curvature consistency test, suggesting new physics beyond the standard model. This could be due to several effects, but more research is needed to determine the exact cause.— Asta Heinesen, physicist at Niels Bohr Institute and Queen Mary University
La Conversación del Hearth Otra perspectiva de la historia
So this research is saying the universe isn't actually uniform? That seems like it would be huge news.
It is potentially huge—but the key word is "potentially." The deviations they found are real and statistically meaningful, but they're not yet at the threshold physicists require to call something a confirmed discovery. Think of it as a strong signal that something might be wrong, not proof that it is.
What would it mean if they're right? What breaks?
The entire mathematical framework we use to understand cosmic expansion, dark energy, the age of the universe—it all assumes the universe is fundamentally uniform at large scales. If that's not true, then all the solutions we've proposed for the mysteries we can't solve right now might be solving the wrong problem.
Like what mysteries?
Why is the universe accelerating? What is dark energy? Why do different measurements of the expansion rate give slightly different answers? We've built elaborate theories to explain these puzzles, but they all assume uniformity. If uniformity is the problem, not the solution, we'd have to start over.
But they said the evidence is preliminary. How preliminary are we talking?
They're at 2 to 4 sigma. The standard is 5 sigma. So imagine you're looking for a pattern in noise, and you think you see it—but you need to look at the data five times more carefully before you can be sure it's real and not just a trick of the light.
So what happens next?
Better telescopes, bigger surveys, more precise measurements. In a few years, we'll have much more data. Either the deviations will disappear and we'll know the universe is uniform after all, or they'll get stronger and we'll have to completely reimagine cosmology.