How fast is the universe expanding? Cosmic ‘illusions’ may hold answer

Sherry Suyu can’t wait for the Sun to get out of the way. Sometime after June, when Earth’s turn around the Sun brings a distant galaxy into view, she expects to see a giant star explode at the end of its life. Supernovae usually don’t send out invitations in advance. But this one, called SN Requiem, has already gone off three times before. Suyu, a cosmologist at the Max Planck Institute for Astro-physics, predicts it will detonate yet again—as if the star is refusing to stay dead.

It’s an illusion. The star exploded just once, but its death is replayed through the magic of gravitational lensing. A huge cluster of galaxies sits between the shattered star and Earth, and its potent gravity is bending the supernova’s light, forcing it to follow different paths with different travel times. As a result, the flash of the explosion appears repeatedly, at different locations and times. The lens produced three bursts of light in 2016. After precisely mapping the mass in the galaxy cluster, Suyu and her colleagues calculated it would also steer light along a longer path—with a fourth reappearance coming sometime in the next year or two. “We made our predictions, we published them,” Suyu says. Now, “We just wait.”

Starting in June, they will check on SN Requiem’s galaxy once a month with the Hubble Space Telescope. If it spots anything, they will enlist the sharper vision of JWST, NASA’s infrared observatory. They hope to watch the light show unfold in real time, and not only to confirm their prediction. By clocking the time delay, they can accurately measure the galaxy’s distance. And by combining its distance with how quickly it is receding from Earth, indicated by the “redshift” of its light, they will get a reading of the Hubble constant: the expansion rate of the universe.

The Hubble constant is a key measure of a universe in which everything is flying apart, floating on a sea of expanding space. It is the number that implies the universe was born in a Big Bang, and its reciprocal gives the age of the universe, about 13.8 billion years.

It is also the subject of a fierce dispute. The two main approaches to measuring the Hubble constant deliver different answers, and researchers have been locked at an impasse for more than a decade. “We really need something more, like a third way to measure the Hubble constant,” says cosmologist Eleonora Di Valentino of the University of Sheffield. Suyu’s lensed supernova, and the dozens more that powerful new survey telescopes are expected to find, could provide that third way and deliver a decisive verdict. “We’re at a really interesting sort of inflection point,” says Saul Perlmutter, an astrophysicist at the University of California (UC), Berkeley. “We have all the pieces in place.”

For nearly a century—ever since Edwin Hubble in 1929 offered his first assessment—astronomers have calculated the Hubble constant by working outward from the Milky Way. The easy part is measuring speed, by gauging the shift of an object’s light toward the red end of the spectrum. Far more difficult are distances, which require “standard candles”: objects in distant galaxies that pulse or flash in a way that reliably encodes how far away they are. Astronomers have found enough of these beacons to measure the Hubble constant out to about 1 billion light-years, settling on a value of about 73 kilometers per second per megaparsec. (One parsec is 3.26 light-years.)

More recently, cosmologists have gone after the Hubble constant in the opposite direction, from the edge of the observable universe inward. They start with the snapshot of the infant universe preserved in the cosmic microwave background (CMB)—the afterglow of the Big Bang. Ripples in the CMB, created by sound waves coursing through the newborn universe, offer a “standard yardstick” with which to calculate the Hubble constant at that primordial moment.

To adjust that value to the present day, however, two other factors must be accounted for. One is dark matter, the as-yet-unidentified stuff whose gravity slows down the outward spread of the galaxies. The other is dark energy, a repulsive force that is accelerating the expansion of space itself. Baking in those assumptions, the cosmologists say the Hubble constant should now be about 67.

Both results have been calculated to a precision of about 1%, leaving a yawning gap known as the Hubble tension. If the cosmologists are right, some systematic bias—an unknown unknown—must be skewing astronomers’ traditional distance measurements. But if the astronomers are right, the problem could lie deeper, in cosmologists’ standard picture of the universe. Resolving the Hubble tension in favor of the astronomers would suggest that cosmologists’ understanding of dark matter and dark energy is incomplete, Di Valentino says. “We are clinging to this model … we don’t want to let it go.”

Suyu and others note that their gravitational lensing method, also known as time-delay cosmography, is immune to the many finicky issues that surround standard candles. For example, stars called Cepheid variables pulse at a rate that signals their intrinsic brightness, making it possible to infer their distance from their brightness in the sky. But the chemical makeup of each star and the amount of dust surrounding it can skew the pulse-brightness relationship. Another standard candle, so-called type Ia supernovae, can be seen at much greater distances. They are thought to occur when new fuel added to a burnt-out star reaches a critical threshold, triggering an explosion with a predictable peak brightness. But astronomers keep finding other triggers for the explosions, casting doubt on how uniform they really are.

Time-delay cosmography, by contrast, does not depend on the astrophysics of pulsing or exploding stars, nor does it make any assumptions about dark matter or dark energy. It relies only on a galaxy or galaxy cluster to act as a gravitational lens, and an object, such as a supernova, whose bright flash provides a time stamp. Astronomers only need to measure the delay between observed flashes and apply simple geometry: The delay gives the difference in path lengths, and the positions of the images on the sky allow astronomers to triangulate, giving an absolute distance to the supernova.

The idea goes back to the early 1960s, when gravitational lenses had only been theorized. But Sjur Refsdal, a graduate student at the University of Oslo, suggested a way to calculate how such a lens would bend light, using the same geometric tools used to model the paths of light through a glass lens. His thesis evaluators weren’t convinced—until he got the result published in the Monthly Notices of the Royal Astronomical Society in 1964. In the same issue, Refsdal also proposed a way to put his technique to work. He suggested time-delayed images of a lensed supernova could offer a handle on the Hubble constant.

No one had ever seen a gravitational lens, let alone multiple images of a supernova. But at the end of that second paper, Refsdal wondered whether quasars—strange bright objects that had recently been discovered—might also work. Quasars would soon be recognized as supermassive black holes, shining constantly from the hearts of distant galaxies as they gorge on matter. They sometimes burp on their meals and flare up, generating a spike in brightness that might provide the time stamp astronomers need to look for in lensed images.

Today, astronomers have found dozens of lensed quasars by spotting their twin images, close together in the sky. But using them to calculate the Hubble constant isn’t easy, says astrophysicist Tommaso Treu of UC Los Angeles. One problem is that quasar light is noisy, making it hard to recognize a specific flare-up when it reappears later in a different image. Waiting for those flickers also requires lots of telescope time, which is not easy to get from busy observatories. In 2002, Treu, then a postdoc at the California Institute of Technology, and a colleague published their first stab at calculating the Hubble constant, based on a single lensed quasar. “At that time, we were very happy to get [the precision] to about 20%,” he says.

A common challenge for both supernovae and quasars is working out how mass is distributed in the gravitational lens. Unlike a glass lens, a gravitational lens can be full of clumps and voids that produce images like a funhouse mirror, making it hard to predict how long light will take to wend its way through. And a further complication must be accounted for. According to Albert Einstein’s theory of gravity, light slows down as it passes through a strong gravitational field, which can significantly add to the delay.

Timing the cosmos

Observations of SN Requiem, a distant supernova, by the JWST space observatory could help resolve a dispute about the Hubble constant, the speed of the universe’s expansion. A massive galaxy cluster lies between the supernova and Earth, and its gravity bends light along multiple paths, causing the supernova to flash at different places and times. The time delays combined with the geometry of the light paths reveal the supernova’s distance—and a new measure of the Hubble constant.

As Treu and others searched for more lensed quasars, they largely focused on those with simpler gravitational lenses: individual galaxies rather than clusters. Astrophysicists have a good understanding of how the mass is distributed in individual galaxies, especially elliptical ones, which are relatively symmetric. By 2010, Treu and his teammates had narrowed the precision down below 7% for another single quasar. “Finally, we can go somewhere,” he says.

To go further, however, they had to find a better way to determine the mass distribution in the lens. The solution has come from instruments known as integral field spectrographs, attached to the world’s most powerful telescopes, such as the Keck telescopes in Hawaii, the Very Large Telescope in Chile, and JWST. The instruments divide an image of a galaxy into squares and glean a spectrum for each square, which reveals how fast the stars are moving within it. The greater a galaxy’s gravity, the faster stars swirl around it, so the observations improve models of the galaxy’s mass and how it’s distributed.

Various teams working on lensed quasars joined together in a collaboration called TDCOSMO. In December 2025, they published their latest result, based on eight lensed quasars, with most of the lenses scrutinized by an integral field spectrograph on JWST. The resulting Hubble constant was 71.6—a value between the astronomers’ and cosmologists’ camps. The precision improved to 4.6%—not yet good enough to decisively favor a high or low value. Treu says the collaboration’s next effort will include an analysis of up to 12 quasars, which the team calculates will take its precision to between 2% and 3%.

Meanwhile, astronomers hoping to find gravitationally lensed supernovae, the beacons Refsdal originally suggested, have not been idle.

In 2014, Patrick Kelly, then a postdoc at UC Berkeley, was peering at Hubble images of massive galaxy clusters, using them as cosmic magnifying glasses to study distant galaxies. He was looking at a fresh image, taken in November, of one of the clusters when he spotted four bright starbursts that hadn’t been there before. “It was pretty evident what it was right away,” says Kelly, now at the University of Minnesota Twin Cities.

Discovered 50 years after Refsdal’s prediction, this first example of a lensed supernova was dubbed SN Refsdal. The supernova’s host galaxy was lensed into six images. Astronomers calculated that there had been an earlier supernova image, on top of the four 2014 discoveries, and that SN Refsdal would make a sixth appearance.

Observers went into overdrive, Kelly says, gathering data to help model the gravity of the lensing cluster. Some models predicted a reappearance within a year, but Hubble saw nothing in October or November 2015. Suyu, a postdoc at the time, took part in one of the modeling groups. She says she was “supernervous,” because their model predicted with 99.9% confidence that the supernova should reappear before the end of 2015. It turned up right on time, in December. “We got an early Christmas present that year,” Suyu says.

The delay between the first and last observed appearances yielded a Hubble constant value of 64.8, with 5.5% precision. Since then, astronomers have discovered a handful of other lensed supernovae. Some had very short time delays, which don’t produce as precise a Hubble constant value. Others had predicted reappearances that were too far in the future for impatient astronomers. That was initially the case for the fourth reappearance of SN Requiem which, based on early estimates, wasn’t due until 2037.

Then came a surprise: In 2023, observers using JWST found a second supernova, SN Encore, which exploded in the same background galaxy as SN Requiem. With two potential gauges of the Hubble constant in the same galaxy, Suyu and collaborators sought to build the best possible mass model of the cluster bending their light.

As lenses, clusters are far more powerful than individual galaxies, producing time delays of years or decades instead of days or weeks. But it is much more complicated to model their mass distributions: Each cluster includes tens or hundreds of member galaxies and each of those has its own halo of dark matter and hot gas that must be accounted for.

First, Suyu says, researchers had to figure out the cluster’s membership list. They used the Very Large Telescope to get redshifts for each galaxy in the image, identifying more than 80 that had similar redshifts and so probably belonged to the cluster. They estimated each galaxy’s mass, based on its brightness. Then they built a mass model and checked to see whether it would distort background objects other than the supernovae in a way that matched observations. “It’s an iterative process,” says Ana Acebron of the University of Cantabria. “We start simple and then add complexity. It’s kind of an art, actually.”

Suyu assembled seven teams, each working independently to construct its own cluster mass model using different methods. Finally, the models were tested to see how well they reproduced the supernova images. Suyu and her colleagues then created an overall model, with input from the seven models weighted according to their performance. The final prediction for SN Requiem’s fourth appearance, it turned out, was in 10 years’ time, not 20. In other words, sometime in the next 2 years. “I really think the supernova will reappear in these 2 years, it’s just when,” Suyu says.

The timing will be crucial: If SN Requiem reappears this year, it will lead to a higher value of the Hubble constant and validate the astronomers’ camp. If the supernova is not seen until next year, it will lead to a lower Hubble constant value that affirms the cosmologists. Suyu says she’s trying to remain neutral.

On its own, Suyu says, SN Requiem could offer a precision of 2% to 3%, but that won’t satisfy everyone. She eventually wants to get below 1% precision, the same as the two existing methods. Then, she says, “we put this to rest once and for all.”

A clear verdict will require many more lensed supernovae like SN Requiem. The second lensed supernova in SN Requiem’s galaxy isn’t expected to reappear until the early 2030s. But JWST is starting to turn up many others. Since its 2021 launch, it has spotted more than 10, eight of them through a program that is carrying out lengthy observations of 60 galaxy clusters for a variety of science goals. “These things that we would find maybe once every year or every couple of years, we have now found one every 2 weeks over the past 4 months of this program,” says Conor Larison of the Space Telescope Science Institute. “JWST’s resolution is just so much better for finding these things.”

A raft of new survey telescopes designed to peer deeply and widely across the heavens could add to the harvest. A European space telescope called Euclid, launched in 2023, will release its first tranche of data later this year. NASA may launch its Nancy Grace Roman Space Telescope before the end of the year. And the Vera C. Rubin Observatory in Chile will begin a 10-year cosmic survey in the coming weeks. In March, ahead of the start of its formal survey, Rubin in a single night gave astronomers a taste of what’s to come: 800,000 alerts to potential transient objects, any one of which could be a lensed supernova. “We’re still learning,” Suyu says. “This is really a new territory for us, but at the same time, I think it opens a lot of new opportunities.”

Perlmutter is eager to see whether time-delay cosmography can finally pin down the expanding universe’s key number. “Now we have a technique waiting in the wings … and we’re about to open the curtain.”

Correction, 21 April, 5:30 p.m.: A previous version of this story incorrectly attributed the discovery of SN Encore to Sherry Suyu and her colleagues.