How fast is the universe expanding? How much does matter clump up in our cosmic neighborhood? Scientists use two methods to answer these questions. One involves observing the early cosmos and extrapolating to present times, and the other makes direct observations of the nearby universe. But there is a problem. The two methods consistently yield different answers. The simplest explanation for these discrepancies is merely that our measurements are somehow erroneous, but researchers are increasingly entertaining another, more breathtaking possibility: These twin tensions—between expectation and observation, between the early and late universe—may reflect some deep flaw in the standard model of cosmology, which encapsulates our knowledge and assumptions about the universe. Finding and fixing that flaw could transform our understanding of the cosmos.
One way or another, an answer seems certain to emerge over the coming decade, as new space and terrestrial telescopes give astronomers clearer cosmic views. “Pursuing these tensions is a great way to learn about the universe,” says astrophysicist and Nobel laureate Adam Riess of Johns Hopkins University. “They give us the ability to focus our experiments on very specific tests, rather than just making it a general fishing expedition.”
These new telescopes, Riess anticipates, are about to usher in the third generation of precision cosmology. The first generation came in the 1990s and early 2000s with the Hubble Space Telescope and with NASA's WMAP satellite, which sharpened our measurements of the universe's oldest light, the cosmic microwave background (CMB). This first generation was also shaped by eight-meter-class telescopes in Chile and the twin 10-meter Keck behemoths in Hawaii.
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
Collectively, these observatories helped scientists formulate the standard model of cosmology, which holds that the universe is a cocktail of 5 percent ordinary matter, 27 percent dark matter and 68 percent dark energy. This model can account with uncanny accuracy for most of what we observe about galaxies, galaxy clusters, and other large-scale structures and their evolution over cosmic time. Ironically, by its very success, the model highlights what we do not know: the exact nature of 95 percent of the universe.
Driven by even more precise measurements of the CMB from the European Space Agency's Planck satellite and various ground-based telescopes, the second generation of precision cosmology supported the standard model but also brought to light the tensions. The focus shifted to reducing so-called systematics: repeatable errors that creep in because of faults in the design of experiments or equipment.
The third generation is only now starting to take the stage with the successful launch and deep-space deployment of Hubble's successor, the James Webb Space Telescope (JWST). On Earth, radio telescope arrays such as the Simons Observatory in the Atacama Desert in Chile and the CMB-S4, a future assemblage of 21 dishes and half a million cryogenically cooled detectors that will be divided between sites in the Atacama and at the South Pole, should take CMB measurements with Planck-surpassing levels of precision.
The centerpieces of the third generation will be telescopes that stare at wide swaths of the sky. The first of these is the ESA's 1.2-meter Euclid space telescope, which launched in July 2023. Euclid will study the shapes and distributions of billions of galaxies with a gaze that spans about a third of the sky. Its observations will dovetail with those of NASA's Nancy Grace Roman Space Telescope, a 2.4-meter telescope with a field of view about 100 times bigger than Hubble's, which is slated for launch in 2026 or 2027. Finally, when it begins operations in the mid-2020s, the ground-based Vera C. Rubin Observatory in Chile will map the entire overhead sky every few nights with its 8.4-meter mirror and a three-billion-pixel camera, the largest ever built for astronomy. “We're not going to be limited by noise and by systematics, because these are independent observatories,” says astrophysicist Priyamvada Natarajan of Yale University. “Even if we have a systematic in our framework, we should [be able to] figure it out.”
Scaling the Distance Ladder
Riess would like to see a resolution of the Hubble tension, which arises from differing estimates of the value of the Hubble constant, H0—the rate at which the universe is expanding. Riess leads a project called Supernovae, H0, for the Equation of State of Dark Energy (SH0ES). The goal is to measure H0, starting with the first rung of the so-called cosmic distance ladder, a hierarchy of methods to gauge ever greater celestial expanses.
The first rung—the one concerning the nearest cosmic objects—relies on determining the distance to special stars called Cepheid variables, which pulsate in proportion to their intrinsic luminosity. The longer the pulsation, the brighter the Cepheid. This relation between variability and luminosity makes Cepheids benchmark “standard candles” for determining distances around the Milky Way and nearby galaxies. They also form the basis of the cosmic distance ladder's second rung, in which astronomers gauge distances to more remote galaxies by comparing Cepheid-derived estimates with those from another, more powerful set of standard candles called type Ia (pronounced “one A”) supernovae, or SNe Ia.
A nocturnal view of the South Pole Telescope, one of several radio observatories mapping patterns in the cosmic microwave background. Credit: Danvis Collection/Alamy Stock Photo
Ascending farther, astronomers locate SNe Ia in even more far-flung galaxies, using them to establish a relation between distance and a galaxy's redshift, a measure of how fast it is moving away from us. The result is an estimate of H0.
In December 2021, Riess says, “after a couple of years of taking a deep dive on the subject,” the SH0ES team and the Pantheon+ team, which has compiled a large data set of type Ia supernovae, announced the results of nearly 70 different analyses of their combined data. The data included observations of Cepheid variables in 37 host galaxies that contained 42 type Ia supernovae, more than double the number of supernovae studied by SH0ES in 2016. Riess and his co-authors suspect this study represents the Hubble's last stand, the outer limits of that hallowed telescope's ability to help them climb higher up the cosmic scale. The set of supernovae now includes “all suitable SNe Ia—of which we are aware—observed between 1980 and 2021” in the nearby universe. In their analysis, H0 comes out to be 73.04 ± 1.04 kilometers per second per megaparsec.
That number is way off the value obtained by an entirely different method that looks at the other end of cosmic history—the so-called epoch of recombination, when the universe became transparent to light, about 380,000 years after the big bang. The light from this epoch, now stretched to microwave wavelengths because of the universe's subsequent expansion, is detectable as the all-pervading cosmic microwave background. Tiny fluctuations in temperature and polarization of the CMB capture an important signal: the distance a sound wave travels from almost the beginning of the universe to the epoch of recombination.
This length is a useful metric for precision cosmology and can be used to estimate the value of H0 by extrapolating to the present-day universe using the standard LCDM model. (L stands for lambda or dark energy, and CDM for cold dark matter; “cold” refers to the assumption that dark matter particles are relatively slow-moving.) An analysis published in 2021 combined data from the Planck satellite and two ground-based instruments, the Atacama Cosmology Telescope and the South Pole Telescope, to arrive at an H0 of 67.49 ± 0.53.
The discrepancy between the two estimates has a statistical significance of five sigma, meaning there is only about a one-in-a-million chance of its being a statistical fluke. “It's certainly at the level that people should take seriously,” Riess says. “And they have.”
An artist's conception of the James Webb Space Telescope, which has just begun to perform breakthrough studies of both the early and current universe. Credit: NASA/GSFC/CIL/Adriana Manrique Gutierrez
HOW CLUMPY IS THE COSMOS?
The other tension that researchers are starting to take seriously concerns a cosmic parameter called S8, which depends on the density of matter in the universe and the extent to which it is clumped up rather than evenly distributed. Estimates of S8 also involve, on one end, measurements of the CMB and, on the other, measurements of the local universe. The CMB-derived value of S8 in the early universe, extrapolated using LCDM, generates a present-day value of about 0.834.
The local universe measurements of S8 involve a host of different methods. Among the most stringent are so-called weak gravitational lensing observations, which measure how the average shape of millions of galaxies across large patches of the sky is distorted by the gravitational influence of intervening concentrations of dark and normal matter. Astronomers used data from the Kilo-Degree Survey, which more than doubled its sky coverage from 350 to 777 square degrees of the sky (the full moon, by comparison, spans a mere half a degree) and estimated S8 to be about 0.759. The tension between the early- and late-universe estimates of S8 grew from 2.5 sigma in 2019 to three sigma (or a one-in-740 chance of being a fluke). “This tension isn't going away,” says astronomer Hendrik Hildebrandt of Ruhr University Bochum in Germany. “It has hardened.”
There is yet another way to arrive at the value of S8: by counting the number of the most massive galaxy clusters in some volume of space. Astronomers can do that directly—for example, by using gravitational lensing. They can also count clusters by studying their imprint on the cosmic microwave background, thanks to something called the Sunyaev-Zeldovich effect. (This effect causes CMB photons to scatter off the hot electrons in clusters of galaxies, creating shadows in the CMB that are proportional to the mass of the cluster.)
A detailed 2019 study that used data from the South Pole Telescope estimated S8 to be 0.749—again, way off from the CMB+LCDM-based estimates. These numbers could be reconciled if the estimates of the masses of these clusters were wrong by about 40 to 50 percent, Natarajan says, although she thinks such substantial revisions are unlikely. “We are not that badly off in the measurement game,” she says. “So that's another kind of internal inconsistency, another anomaly pointing to something else.”
BREAKING THE TENSIONS
Given these tensions, it is no surprise that cosmologists are anxiously awaiting fresh data from the new generation of observatories. For instance, David Spergel of Princeton University is eager for astronomers to use JWST to study the brightest of the so-called red-giant-branch stars. These stars have a well-known luminosity and can be used as standard candles to measure galactic distances—an independent rung on the cosmic ladder, if you will. In 2019 Wendy Freedman of the University of Chicago and her colleagues used this technique to estimate H0, finding that their value sat smack in the middle of the early- and late-universe estimates. “The error bars on the current tip of the red-giant-branch data are such that they're consistent with both possibilities,” Spergel says. Astronomers are also planning to use JWST to recalibrate the Cepheids surveyed by Hubble, and separately the telescope will help create another new rung for the distance ladder by targeting Mira stars (which, like Cepheids, have a luminosity-periodicity relation useful for cosmic cartography).
JWST might resolve or strengthen the H0 tension, and the wide-field survey data from the Euclid, Roman and Rubin observatories could do the same for the S8 tension by studying the clustering and clumping of matter. The sheer amount of data expected from this trio of telescopes will reduce S8 error bars enormously. “The statistics are going to go through the roof,” Natarajan says.
Meanwhile theoreticians are already having a field day with the twin tensions. “This is a playground for theorists,” Riess says. “You throw in some actual observed tensions, and they are having more fun than we are.”
One of the most recent theoretical ideas to receive a great deal of interest is something called early dark energy (EDE). In the canonical LCDM model, dark energy started dominating the universe relatively late in cosmic history, only about five billion years ago. But, Spergel says, “we don't know why dark energy is the dominant component of the universe today. Because we don't know why it's important today, it could have also been important early on.” That is partly the rationale for invoking dark energy's effects much earlier, before the epoch of recombination. Even if dark energy was just 10 percent of the universe's energy budget during those times, that would be enough to accelerate the early phases of cosmic expansion, causing recombination to occur sooner and shrinking the distance traversed by primordial sound waves. The net effect would be to ease the H0 tension.
“What I find most interesting about these models is that they can be wrong,” Spergel says. Cosmologists' EDE models make predictions about the resulting EDE-modulated patterns in the photons of the CMB. In February 2022 Silvia Galli, a member of the Planck collaboration at the Sorbonne University in Paris, and her colleagues published an analysis of observations from Planck and ground-based CMB telescopes, suggesting that they collectively favor EDE over LCDM by a statistical smidgen. Confirming or refuting this tentative result will require more and better data—which could come from observations by the same ground-based CMB telescopes. But even if EDE models prove to be better fits and fix the H0 tension, they do little to alleviate the tension from S8.
Potential fixes for S8 exhibit a similarly vexing lack of overlap with H0. In March 2022 Guillermo Franco Abellán of the University of Amsterdam and his colleagues published a study in Physical Review D showing that the S8 tension could be eased by the hypothetical decay of cold dark matter particles into one massive particle and one “warm” massless particle. This mechanism would lower the value of S8 arising from CMB-based extrapolations, bringing it more in line with the late-universe measurements. Unfortunately, it doesn't solve the H0 tension. “It seems like a robust pattern: whatever model you come up with that solves the H0 tension makes the S8 tension worse, and the other way around,” Hildebrandt says. “There are a few models that at least don't make the other tension worse, but [they] also don't improve it a lot.”
“WE ARE MISSING SOMETHING”
Once fresh data arrive, Spergel foresees a few possible scenarios. First, the new CMB data could turn out to be consistent with early dark energy, resolving the H0 tension, and the upcoming survey telescope observations could separately ease the S8 tension. That would be a win for early dark energy models—and would constitute a major shift in our understanding of the opening chapters of cosmic history. It's also possible that both H0 and S8 tensions will resolve in favor of LCDM—a win for the cosmological standard model and a possibly bittersweet victory for cosmologists hoping for paradigm-shifting breakthroughs. Of course, it might turn out that neither tension is resolved. “Outcome three would be both tensions become increasingly significant as the data improve—and early dark energy isn't the answer,” Spergel says. Then, LCDM would presumably have to be reworked differently, although how is unclear.
Natarajan thinks that the tensions and discrepancies are probably telling us that LCDM is merely an “effective theory,” a technical term meaning that it accurately explains a certain subset of the current compendium of cosmic observations. “Perhaps what's really happening is that there is an underlying, more complex theory,” she says. “And that LCDM is this [effective] theory, which seems to have most of the key ingredients. For the level of observational probes we had previously, that effective theory was sufficient.” But times change, and the data deluge from precision cosmology's third generation of powerful observatories may demand more creative and elaborate theories.
Theorists, of course, are happy to oblige. For instance, Spergel speculates that if early dark energy could interact with dark matter (in LCDM, dark energy and dark matter do not interact), this arrangement could suppress the fluctuations of matter in the early universe in ways that would resolve the S8 tension while simultaneously taking care of the H0 tension. “It makes the models more baroque,” Spergel says, “but maybe that's what nature will demand.”
As an observational astronomer, Hildebrandt is circumspect. “If there was a convincing model that beautifully solves these two tensions, we'd already have the next standard model,” he says. “That we're instead still talking about these tensions and scratching our heads is just reflecting the fact that we don't have such a model yet.” Riess agrees. “After all, this is a problem of using a model based on an understanding of physics and the universe that is about 95 percent incomplete, in terms of the nature of dark matter and dark energy,” he says. “It wouldn't be crazy to think that we are missing something.”