How fast is the universe expanding? It depends on who you ask. Cast your gaze to the relatively nearby stars and galaxies that surround us in space, and you’ll arrive at a certain number for this value, known as the Hubble constant. But look into the far more distant universe, and you’ll get a slightly different number. This discrepancy, known as the Hubble tension, is small but carries weighty ramifications. The tension might simply be caused by flaws in our measurements—or it could be pointing to fundamental gaps in our understanding of cosmic structure. Admittedly, even without any tension whatsoever, there are deep mysteries tied up in the universe’s rate of expansion—namely, the fact that it is being accelerated by dark energy, an as-yet-unexplained force about which we know almost nothing. Now a new measurement of the Hubble constant, made by observing a mirror image of a distant exploding star, or supernova, is complicating matters further.

In research published today in the journal Science, Patrick Kelly of the University of Minnesota and his colleagues used the time delay from a distant supernova known as Refsdal to measure the Hubble constant. They arrived at an expansion rate of 66.6 kilometers per second per million parsecs (km/s/Mpc), or 66.6 km per second per 3.26 million light-years, with an uncertainty of 7 percent. (A previous study of the supernova, from 2017, reached a similar result but with significantly greater statistical uncertainty.)

Keeping in mind the associated uncertainties, this number—66.6 km/s/Mpc—may be in disagreement with other supernova-based measurements in the so-called local universe. These tend to yield a higher value for the Hubble constant: around 73 km/s/Mpc. Yet 66.6 km/s/Mpc is strikingly similar to Hubble constant measurements from far more distant sources in the “early” universe, which deliver values of around 67 km/s/Mpc. “Our measurement is in better agreement with that from the cosmic microwave background [CMB],” essentially the big bang’s remnant heat from when the universe was scarcely more than a 400,000-year-old fireball, Kelly says. “Although, given the uncertainties, it does not rule out the measurement from the local distance ladder.”

The Hubble constant can be measured in a number of ways. For the local universe, most rely on various standard candles—certain types of supernovae and other astrophysical objects that possess a known, scarcely varying intrinsic brightness, allowing their distances and motions with respect to us to be more easily ascertained. Measurements from multiple sorts of standard candles can be strung together to allow astronomers to gauge the Hubble constant out to ever greater distances, with each standard candle being one “rung” on what is known as the “cosmic distance ladder.” But the cosmic distance ladder begins to teeter and tumble over across truly vast distances. To measure the Hubble constant that prevailed in the early universe, researchers mostly use the CMB. Sound waves rippling through the fiery plasma that filled the early universe imprinted telltale patterns on the CMB that astronomers can use as standard rulers for charting the universe’s subsequent expansion.

In 1964 the Norwegian astrophysicist Sjur Refsdal first suggested another way that supernovae could be used to measure the Hubble constant. If, on its way to Earth, a distant supernova’s light happened to pass around the gravitational grip of a massive object—such as a galaxy cluster—the light could be “gravitationally lensed,” or warped and bent to follow multiple divergent paths to Earth, some longer and some shorter. The end result would be a single supernova appearing multiple times in slightly offset positions in the sky, with the delay between each apparition corresponding to the total distance its light had traveled. Combining such delays with the knowledge of how fast the supernova was moving away from us—obtained by measuring a property called redshift—and the mass of the lensing cluster would provide a value of the Hubble constant.

In November 2014 Kelly, then at the University of California, Berkeley, and his colleagues discovered the first known example of such an event—the supernova Refsdal, which took place some 14 billion light-years from Earth. They correctly predicted the arrival of a lensed image from the supernova, which reached our planet some 360 days later, at the end of 2015. Now the team has at last managed to use Refsdal to measure the expansion rate of the universe. “This is different from anything that’s been done before,” Kelly says. To arrive at a value, the team worked in groups that independently assessed blinded data, ultimately settling on its unexpected result of roughly 66.6 km/s/Mpc.

The result is “a great addition” to our knowledge of the Hubble constant, says Wendy Freedman, a University of Chicago astronomer, who specializes in studies of the universe’s expansion rate and was not involved in the new paper. “It’s completely independent of any other kind of method.”

Astronomers have used lensing to measure the expansion of the universe before but with quasars—the extremely bright cores of certain galaxies—rather than supernovae. In 2017 a team called H0LiCOW used this method to arrive at a value of around 72 km/s/Mpc. Lensed quasars are “more abundant” in the sky, giving this method some advantages, says H0LiCOW lead Sherry Suyu of the Max Planck Institute for Astrophysics in Garching, Germany. But supernovae display more obvious changes in brightness, meaning the exact time delay in images can be more precisely measured, perhaps giving a higher level of accuracy. “You really see this drastic variation,” Suyu says.

But while quasars may shine for millions of years—essentially forever for us—supernovae are short-lived, shining brightly for just weeks or months. “You have to be able to find them early on,” Suyu says. “If you miss it, they’re gone.” To date, only a handful of time-delayed supernovae are known. The most recent one, named H0pe, was found by the James Webb Space Telescope (JWST) earlier this year. Thus, while Refsdal is the first such event to be used to measure the expansion of the universe, it’s undoubtedly not the last.

If Kelly and his team’s value stands up, this would suggest that perhaps we need to tweak our best guesses about the nature of dark matter—the enigmatic, invisible stuff that seems to give galaxies and galaxy clusters most of their mass and thus modulates gravitational lensing. If true, Kelly says, their result “implies there must be a flaw in our models of the dark matter in galaxy clusters.” Updating those models could cascade in turn to demand changes to the so-called standard model of cosmology, which presumes that a certain, rather inert “cold” form of dark matter and a specific type of dark energy act together to guide the growth and evolution of galaxies and clusters across cosmic time.

“We don’t yet understand what dark matter and dark energy is,” Freedman says. “Measuring the Hubble constant locally is a way to directly test that model. If this shows there’s some fundamental piece of physics that’s missing from the standard model, that’s going to be very exciting.”

Not everyone is convinced just yet, however, that such cosmological sea changes are in store. Daniel Scolnic of Duke University says the result’s 7 percent uncertainty is still large enough at the margins to place it within the bounds of other local results. “If they got much smaller uncertainties, then everyone should be looking at themselves hard in the mirror right now,” says Scolnic, who was not involved in the study. “This would be really confusing because all of the local measurements seem to agree on higher values.”

To find out for certain, more time-delayed supernovae will need to be studied, and their values of the Hubble constant will need to be ascertained. Such results could arise sooner rather than later: a measurement of H0pe is expected from JWST in the coming months, and the upcoming Vera Rubin Observatory in Chile, set to switch on next year, should greatly increase the population of known time-delayed supernovae. “We’ll find many more of these,” Kelly says. “If they all favor a lower value of the Hubble constant, that would strengthen the disagreement. Hopefully we can figure out where the issue is.”

Editor’s Note (5/11/23): This article was edited after posting to better clarify Patrick Kelly’s comment on his team's measurement being in better agreement with that from the cosmic microwave background. The text had previously been amended on May 11 to correct the uncertainty of the result of 66.6 kilometers per second per million parsecs and to better clarify the potential conflict with previous measurements.