Dynamical dark energy and the week that broke cosmology

August 20, 2026 The seminar announcement called it “the week that broke cosmology.” In March of 2025, a paper from the Dark Energy Spectroscopic Instrument (DESI) rocked the cosmology world by presenting tantalizing evidence for a deep, fundamental change to our understanding of the universe. Since the late 1990s, cosmologists have known that the universe is not only expanding, but speeding up. The discovery of expansion was first made way back in the 1920s, as astronomers saw that more distant galaxies seemed to be moving away from us more quickly – the exact behavior you would expect if the cosmos were getting bigger in a uniform way everywhere. The assumption at the time was that the big bang set off the expansion, and now gravity should be causing that expansion to slow down. When two groups of observers showed in 1998 that the expansion was, in fact, accelerating, it created something of a revolution in our understanding of the cosmos. Instead of a universe filled with just the usual matter and energy, both of which would slow expansion via gravity, there must be some extra component actively stretching the universe out. The name given to that new feature was dark energy, and despite it being completely unexpected, astronomers had a pretty good hypothesis for what it might be – one that could be traced all the way back to Albert Einstein. Before astronomers had good enough observations to know for sure that other galaxies existed, it appeared that the universe must be eternal and unchanging. But that came with a problem: why hadn’t gravity pulled everything together into one big clump? Einstein, working out his equations of general relativity that would describe gravity’s influence on matter and space (and vice versa), proposed an explanation for what could be out there holding the stars apart. He added a new term to balance out the inward pull of gravity that he called the cosmological constant. This term described an inherent property of spacetime that imbued it with just a bit of a tendency to stretch, so that the space in between clumps of matter could resist their gravitational attraction. When 1920s-era observers eventually discovered that the universe was actually expanding, rather than remaining static, Einstein realized the term was unnecessary. The cosmos wasn’t collapsing on itself because it was still in the process of expanding from the big bang. He threw the term out as a mistake. Flashing forward again to the discovery of acceleration in 1998, the cosmological constant was widely accepted as an obvious candidate for the mysterious dark energy. All the observations seemed to be consistent with the idea that the expansion started accelerating in the relatively recent cosmic past, because that’s when matter got diffuse enough that the cosmological constant started to win the gravitational tug-of-war. As the overall density of matter reduced with the increasing size of space, the cosmological constant continued to be a feature of all of space, so its density stayed the same. That meant it could keep stretching space while matter’s gravity became less and less important. Now, instead of just keeping galaxies from falling together, it was actively making the expansion speed up. For nearly three decades, cosmologists have been testing this hypothesis from every angle. If dark energy is a cosmological constant, it shouldn’t get more or less powerful over time – its influence on the evolution of the cosmos should be determined completely by the overall density of matter and energy as the cosmos expands, via the Friedmann equations. In the face of the data, the cosmological constant hypothesis worked so well that it became a cornerstone of the concordance model of cosmology, commonly referred to in the field as ΛCDM. Here, Λ is the Greek letter (Lambda) used to label the cosmological constant in Einstein’s equations, and CDM stands for cold dark matter – the invisible matter that makes up the scaffolding upon which galaxies and the large-scale structure of the universe appear to be built. In recent years, however, apparent disagreements between ΛCDM and new cosmological data have become impossible to ignore. First came the Hubble tension. Different measurements of the current expansion rate of the universe – a parameter we call H0 – have been giving different results, for reasons that no one has yet been able to satisfactorily explain. Using data from the cosmic microwave background (CMB), the background light coming to us from the final stages of the big bang, combined with our cosmological equations, we can infer a higher present-day expansion rate than what we get looking at data from surveys of supernovae in the nearby universe. The jury is still out on how to resolve this discrepancy, but if it’s not due to some kind of previously unknown bias in the data, it’s a major challenge to the robustness of ΛCDM itself. Then, there’s the neutrino mass problem. Ghostly neutrinos, which are tiny particles produced in stellar fusion and other high-energy astrophysical processes, contribute a minuscule amount to the total matter in the universe, but have measurable effects on the structure and distribution of galaxies across the cosmos. Yet as soon as we developed galaxy surveys sensitive enough to measure these effects – which should have been able to tell us something about neutrino masses – we started getting results that appeared completely nonsensical. Including neutrinos in the data analysis seemed to have the opposite effect that it should, causing cosmologists to refer to the result as “negative neutrino masses.” To be clear, as far as I’m aware, no one actually thinks the mass of a neutrino can be negative. Instead, the shorthand demonstrates that the analysis isn’t working, for one reason or another. Maybe there’s some new physics hiding in the data that mimics neutrinos misbehaving. One culprit could be a change in dark energy. This all leads us back to that explosive DESI result in 2025. In their second data release (DR2) results, the DESI team showed that their analysis strongly preferred a version of dark energy inconsistent with a cosmological constant. Instead, the best fit to the data was a kind of dynamical dark energy – something that can be the cause of cosmic acceleration, but appears to be getting less powerful over time. While the result didn’t quite meet the gold standard benchmark for a discovery (a threshold of evidence we call 5 sigma), it was enough of a disagreement with ΛCDM to send shockwaves through the global cosmology community. DESI is an international project run by astronomers from around the world and funded primarily by the US Department of Energy’s Office of Science. While the telescope itself is located in the US, Canadian involvement is strong. Perimeter Institute Associate Faculty Will Percival, DESI’s co-spokesperson, leads a research group at the University of Waterloo’s Waterloo Centre for Astrophysics carrying out scientific analysis on DESI results. Also at Perimeter, Computational Scientist Dustin Lang works with the DESI team on survey operations and data analysis. As for how the instrument works, its primary job is to create a three-dimensional map of the galaxies in our (fairly large) corner of the universe. It does this by taking spectra of tens of millions of galaxies and using the clues embedded in that spectral light to determine exactly where the galaxies are in space and how they’re moving relative to us. This last part is important: as the universe expands, the light from distant objects is stretched out in a process called cosmological redshifting. (For optical light, such as DESI sees, the whole spectrum is shifted toward the red part of the electromagnetic spectrum.) The more distant the object in space, the more the expansion of space is stretching its light, and therefore the higher its measured redshift. Nearby things have a redshift of 0, and the most distant galaxies ever observed – some of the earliest galaxies to form in the universe – have redshifts around 14. DESI can observe galaxies with redshifts up to about 1.6 and quasars (the luminous cores of active galaxies) to redshifts of 3.5, meaning that their light is stretched all the way through the visible spectrum and into the infrared. These objects are so distant that their light has been travelling for around 12 billion years, meaning they’re showing us a time when the universe was only a fraction of its present age. While the galaxies and quasars themselves are an incredible data set, they’re not the main attraction of the DESI survey. DESI’s most exciting results come from its measurement of baryon acoustic oscillations. In the early universe, deep inside the radiation era, the hot plasma filling the cosmos was so dense it was humming. Some parts of the cosmos contained gravitational wells with an abundance of dark matter, while others were more diffuse. The dark matter in the denser regions tried to hold the plasma in, but the pressure of the fast-moving particles resisted the collapse. The result was an oscillation, where the hot “baryonic” ordinary matter would be pulled in but rebound away from the dark matter clumps due to the plasma’s pressure. When the universe expanded enough to let the plasma cool and the ordinary matter particles could slow down, the pressure wave of rebounding plasma left an imprint on the distribution of matter throughout the cosmos. Most matter can still be found clumped close to where the dark matter density peaked initially, but the pressure wave of the plasma created a residual shell of matter at a set distance from each peak, frozen in place when the universe transitioned from being radiation-dominated to a calmer, cooler universe of mostly matter. The physical size of that shell is set by the physics of gas and gravity. Viewed by us today, it gives us a kind of cosmic metre-stick embedded right in the data that we can use to create incredibly precise maps of the large-scale distribution of matter in the universe. By measuring the positions of galaxies at different distances (and therefore different times), astronomers can use this baryon acoustic oscillation (BAO) signature to measure how the cosmos was expanding in the past, out to billions of years ago. It’s important to keep in mind that there’s a lot of analysis involved in converting DESI’s measurements of galaxy spectra to inferences about the expansion history of the cosmos. Some of that analysis involves understanding the galaxies themselves and correcting for various potential systematic errors in the data. But other datasets get brought in as well, to help calibrate the measurements and fill in some of the blanks about physical or astrophysical assumptions. In their DR2 results, the DESI team included data from CMB probes (to better understand the early time behaviour of the universe and get additional information about BAOs), and supernova surveys (to fill in the expansion history at low redshifts where BAOs are hard to detect). And while it’s true that with the right modelling, the expansion rate can be inferred from the BAO measurements, the connection to dark energy specifically takes a couple of additional steps. These caveats are important to keep in mind because as cosmologists scrutinize and re-evaluate DESI’s bombshell results, these are the places we’ll be looking to determine whether the conclusions are truly robust. There’s also some subtlety to the framing of the scientific question itself. The main thing DESI set out to answer was whether dark energy is really a cosmological constant, as opposed to something that can vary over time. That’s a bit tricky, though, because while we know exactly what a cosmological constant should do to the expansion rate (assuming we understand how the matter behaves; more on that later), something that is not a cosmological constant might do anything. If you want actual quantitative results, therefore, you have to parameterize the behavior of dark energy somehow and then measure the deviation in some direction from a constant. The typical characterization we use for dark energy is based on the equation of state parameter. This number, labeled w, is the ratio of a substance’s pressure to its energy density. For dark matter, which is most of the matter in the universe, w=0, because it has no pressure at all. For a cosmological constant, which has negative pressure, in the relevant units, w=-1. By making some standard assumptions about the equation of state of the matter and radiation and how it all fits into a cosmological model, it’s possible to infer the equation of state for dark energy, specifically. Before DESI, this w has tended to be broadly consistent with -1, with some variations among different observations. And taken alone, DESI’s results are also consistent with w=-1 for dark energy. If dark energy varies with time, though, it’s conceivable that the equation of state parameter could be w=-1 today, but different in the past. To account for that possibility, cosmologists consider a kind of dark energy characterized by both the value of w today, which we call w0, and a number that tells us how it varies with time, called wa. If wa=0, w is constant at whatever w0 value we measure today. If wa is positive, that means dark energy is getting more powerful over time; if it’s negative, it’s diminishing (also called “thawing”). This is called the CPL parameterization, and it’s usually written as w(a) = w0 + wa(1-a), where a=1/(1+z) is the scale factor of the universe, related to its size over time, and z is the redshift. What DESI’s results showed was that when variation of w was considered, the data preferred a value of w0 broadly consistent with -1 but wa well below zero, suggesting a version of dark energy that looks a lot like a cosmological constant now but is in fact getting less powerful as time goes on. The CPL parameterization is a fairly simple model that we use mainly to have something to test a cosmological constant against – it’s not meant to be describing the detailed behavior of dark energy over time or explaining why it’s evolving. But if we take the broad brushstrokes evolution it suggests when applied to the DESI data seriously, then we end up painted into a very weird corner. It appears to be pointing to an equation of state parameter of dark energy that was less than -1 sometime in the cosmic past. And this is a problem because while a cosmological constant is already weird enough, having a component of the universe with w<-1 would break some seemingly foundational rules of physics. York University Faculty and Perimeter Associate Matthew Johnson, who studies cosmology, early universe theory, and dark energy, explains that this w<-1 version of dark energy – known as phantom dark energy – would be a real problem. “These DESI results point to the possibility that the amount or density of dark energy actually grows as the universe expands,” he says. “It's as though the amount of stuff in this growing box is growing faster than the box is growing, which is a really strange property that we don't really have any examples of in the natural world. And it kind of flies in the face of our understanding of what is allowed in the theory of general relativity.” Whether dark energy could have gone through a phantom phase in the past has been a hot topic in the cosmology literature since the DESI result came out. With such a confusing and potentially revolutionary result, there’s a huge incentive among researchers to explain it, either by finding previously unseen issues in the data or analysis, or by finding some new physics model that could make it all make sense. “There is no single explanation for what dark energy could be,” Will Percival tells me. “It's open season now for people to test their own models and to see what works best and what fits.” The literature is now bursting with proposed explanations for the DESI results. Alongside exotic physics models that allow dark energy to genuinely cross the “phantom divide,” researchers have found numerous ways that the data we have could be fooling us. Some have suggested that systematic effects in the data, or artifacts of the parameterization chosen, could have tipped the scales in favor of dynamical dark energy, making dark energy appear to be changing when it really isn’t. Others propose that the phantom behavior of dark energy is just an illusion, due to some unexpected property of dark matter that changes the relationship between the global cosmic w and dark energy’s specific w. There have also been proposals blaming a possible unaccounted-for spatial curvature, or an unexpectedly high value of the electron optical depth, a parameter of the cosmological model that is hard to pin down with other observations. There is also a possibility that the DESI dark energy anomaly is connected to either the Hubble tension or the neutrino mass problem. In both cases, cosmologists have explored whether allowing for something unexpected with dark energy could alleviate those other tensions, which might help us to find a consistent model of cosmic evolution. Meanwhile, researchers are continuing to gather more data and look for consistency (or not) with the DESI results. A recent analysis by the Dark Energy Survey Supernova Program also finds a preference for evolving dark energy, in the same direction as DESI DR2, but with a slightly lower statistical significance (bringing the combined significance down from 4.2 sigma to only 3.2). At the same time, the DESI team are continuing their ongoing analysis of galaxy clustering in their dataset, which should come out in the next few months, along with new BAO measurements in 2027. Also in 2027, we can look forward to the start of a supernova survey from the new Vera Rubin Observatory, which has been estimated to be capable of reaching the 5 sigma level of significance for dynamical dark energy in just one year of observing, if dark energy really does evolve. As of today, it’s a matter of ongoing debate whether or not cosmology is actually “broken.” We have an intriguing, potentially game-changing result, supported by multiple observations, but just complicated and preliminary enough for us to find explanations that can be accommodated within the tried-and-tested concordance model. While astronomers continue to poke at every available dataset and each step in the analysis to try to make the answer more definitive, it’s likely that no solid conclusions will be drawn until we have more data to work with from a wider range of observational tools and techniques. In the meantime, those of us who are more theory-oriented will continue to propose new models in the hope that maybe, this time, we’ll conclusively defeat the concordance model of cosmology. Because as nice as it is to have a model that seems to work reasonably well with the data we have, it’s always better to have an opportunity to push the boundaries of our understanding, and to find the clue that tells us what comes next. Credit for "the week that broke cosmology" goes to my colleague Niayesh Afshordi. Further exploration DESI DR2 results. II. Measurements of baryon acoustic oscillations and cosmological constraints Abdul Karim, M. et al. 2025, Phys. Rev. D 112, 083515 Extended Dark Energy analysis using DESI DR2 BAO measurements Lodha, K. et al. 2025, arXiv:2503.14743 Cosmological neutrino mass: a frequentist overview in light of DESI Chebat, D. et al. 2026, JCAP01(2026)041 Constraints on Neutrino Physics from DESI DR2 BAO and DR1 Full Shape Elbers, W. et al., 2025, Phys. Rev. D 112, 083513 The Dark Energy Survey Supernova Program: A Reanalysis Of Cosmology Results And Evidence For Evolving Dark Energy With An Updated Type Ia Supernova Calibration Popovic, B. et al., 2026, MNRAS 548, stag632 A measurement of H0 from DESI DR1 using energy densities Krowleski, A. et al., 2026, arxiv:2511.23432 High-redshift physics from the acoustic scale Weiner, Z. 2026, arXiv:2603.18131 Testing ΛCDM versus dynamical dark energy in one year: A DESI spectroscopic follow-up program for Rubin supernovae Truong, J. et al., 2026, arxiv:2604.18859 About PI Perimeter Institute is the world’s largest research hub devoted to theoretical physics. The independent Institute was founded in 1999 to foster breakthroughs in the fundamental understanding of our universe, from the smallest particles to the entire cosmos. Research at Perimeter is motivated by the understanding that fundamental science advances human knowledge and catalyzes innovation, and that today’s theoretical physics is tomorrow’s technology. Located in the Region of Waterloo, the not-for-profit Institute is a unique public-private endeavour, including the Governments of Ontario and Canada, that enables cutting-edge research, trains the next generation of scientific pioneers, and shares the power of physics through award-winning educational outreach and public engagement.