How fast is the universe expanding?
One might assume scientists long ago settled this basic question, first explored nearly a century ago by Edwin Hubble. But right now the answer depends on who you ask. Cosmologists using the Planck satellite to study the cosmic microwave background—light from the “early” universe, only about 380,000 years after the big bang—have arrived at a high-precision value of the expansion rate, known as the Hubble constant (H0). Astronomers observing stars and galaxies closer to home—in the “late” universe—have also measured H0 with extreme precision. The two numbers, however, disagree. According to Planck, H0 should be about 67—shorthand for the universe expanding some 67 kilometers per second faster every 3.26 million light-years. The most influential measurements of the late universe, coming from a project called Supernova H0 for the Equation of State (SH0ES), peg the Hubble constant at about 74.
This discrepancy—the so-called Hubble tension—has been growing for years, increasing as study after study of both the early and late universe yield ever more precise results and leave scientists on both sides anxious and bewildered. After all, it could be that either faction is somehow just mismeasuring the universe. But the tension may be a true reflection of reality, requiring exotic new physics and a dramatic revision to our understanding of cosmic evolution.
On July 4 fresh results from the late universe were released that reinforced the SH0ES figure, pushing the tension past a threshold of statistical significance that physicists use as a benchmark for genuine discoveries. For a moment, the prospect of new physics loomed larger than ever before. Yet days later, another independent batch of late-universe measurements muddled the debate, delivering an H0 value of 69.8, midway between the canonical values from Planck and SH0ES. Much of the drama unfolded in real time at the Tensions Between the Early and the Late Universe conference, held from July 15 to 17 at the Kavli Institute for Theoretical Physics in Santa Barbara, Calif.
“This week is too much. Go home H0, you’re drunk,” tweeted Dan Scolnic, a SH0ES member at Duke University, after yet another befuddling new result for H0 was revealed at the conference.
Initiated more than a decade ago by Adam Riess, an astrophysicist at Johns Hopkins University, SH0ES estimates the universe’s expansion rate by measuring distances to other galaxies. Such measurements are often made for our sun’s planets and for nearby stars by simply charting their motion through the sky over time. But the remote galaxies SH0ES studies are so far away that their position hardly changes— they appear “fixed” on the sky. So Riess and his colleagues instead use the apparent brightness of supernovae produced by exploding white dwarf stars, which they compare to the supernovae’s intrinsic luminosity to find distance. SH0ES then obtains H0 from the supernova’s distance and redshift—the stretching of its emitted light by the universe’s expansion. To relate brightness to luminosity, the data must be calibrated against brightness measurements from standard candles, astrophysical objects with known luminosities. As standard candles, SH0ES uses Cepheid variables, stars that pulsate in a manner that reveals their true luminosity. The SH0ES team finds Cepheids that exist close enough to Earth to measure their distance directly to obtain luminosity. In this way, they gauge the distance to faraway galaxies harboring Cepheids and white-dwarf supernovae—and with it, the value of H0.
(In contrast, Planck’s estimate of H0 relies on measuring features in the cosmic microwave background and predicting their evolution using the standard model of cosmology, called Lambda-CDM. Lambda-CDM uses estimates about the composition of the early universe to provide a high-precision description of its expansion over time.)
The July 4 results that so amplified the tension came from another project called H0 Lenses in COSMOGRAIL’s Wellspring (H0LiCOW), which sought to achieve a late-universe result, independent of SH0ES, using a technique called gravitational lensing. Gravitational lensing is a phenomenon predicted by Einstein’s general theory of relativity, in which a massive foreground object can act as a lens, warping the light from a background object much farther away. The H0LiCOW collaboration focused on quasars—extremely bright cores of very distant galaxies harboring active supermassive black holes—gravitationally lensed by intervening galaxies, as seen from Earth. The lensed light from a quasar can follow several paths to Earth—some shorter, some longer—resulting in multiple time-delayed images of that quasar appearing at the edges of the foreground galaxy.
“The different paths that the light takes have different lengths, so there’s a time delay due to that. And they pass through different parts of the [lensing] galaxy, so they experience a different amount of gravity,” says H0LiCOW member and University of California, Davis, astrophysicist Christopher Fassnacht. Using such time delays—as well as other properties, such as a quasar’s redshift and the estimated distribution of mass along a lensed image’s light path—the H0LiCOW team computed an H0 value of 76.8. Combined with the latest SH0ES results, this figure raised the tension with Planck’s early-universe H0 measurements to five times the uncertainty, or five sigma—the gold standard for discovery in physics, equivalent to a result having a one-in-3.5-million chance of being a statistical fluke. “For many people, that’s like a magic number,” Riess says. “You know, I was already convinced.” The five-sigma standard, he notes, was developed by particle physicists, whose myriad experiments routinely create petabytes of data. Because cosmological observations usually generate much less data, Riess says, the chances of a statistical fluke are lower.
Giant Steps, New Standards
The H0LiCOW result left researchers on the late-universe side of the tension feeling vindicated—but not for long. On July 16, in the middle of the Kavli conference in Santa Barbara, researchers at the Carnegie Institution for Science and the University of Chicago dropped a bombshell: another measurement of H0, derived from a newly developed standard candle distinct from SH0ES’s Cepheids. Clocking in at 69.8, this independent reckoning of H0 brought the tension’s statistical significance back down to slightly below the exalted five sigma.
Chiefly developed by University of Chicago astrophysicist Wendy Freedman, a pioneer in Hubble constant studies, this new standard candle uses red giant stars—aging stars that balloon in size and brightness over millions of years as they transition to burning helium at their cores after exhausting most of their less energetic hydrogen fuel. This process culminates in a “helium flash,” a rapid brightening due to a surge in helium burning, after which a star dramatically dims. Astrophysical models can accurately predict the highest temperatures and pressures at which a helium flash takes place, allowing researchers to discern the peak brightness a red giant can reach. Red giants shining at that theoretical limit can thus become standard candles. Freedman and her colleagues studied such stars in the sparse stellar halos surrounding 18 large galaxies, using them to calibrate supernovae, which led to the middle-of-the-road H0 estimate.
In addition to providing an independent check, Freedman’s method may also sidestep some vexing problems associated with other common standard candles. Most notably, thanks to their location in a galaxy’s uncrowded outskirts, red giants are less prone to the contaminating effects of interstellar dust and other nearby stars than Cepheids. “We really struggle with [Cepheids] as we go out in distance, because they’re young stars. So we can only find them in the disks of spiral galaxies where star formation is ongoing, and so we have to find them against the background of the entire galaxy. The red giant stars are in the halo, isolated, and it’s much easier to measure accurate luminosities for those stars,” Freedman says.
Then again, according to Riess, the new approach’s intermediate H0 value may itself be a product of systematic errors, arising not from high-fidelity measurements but rather from Freedman and her colleagues’ guess at how much dust is in the Large Magellanic Cloud, a dwarf galaxy near the Milky Way that they use for calibrating red-giant observations. “Their new method is giving much more dust,” Riess says—a situation that could cause Freedman’s team to underestimate the distances to galaxies and thus the actual value of H0. Additionally, three other H0 results unveiled at the Kavli conference—based on studies of new varieties of variable stars, of microwave emissions from galactic centers and of core-to-cusp brightness measurements of galaxies—all hew closer to the SH0ES value. The tension, it seems, is not going away anytime soon.
If the tension is real, then the model physicists use to describe the early universe is incomplete. But Lloyd Knox, an early-universe cosmologist at University of California, Davis, says the model has already predicted so much data correctly that many researchers are reluctant to give it up. “The Lambda-CDM model has been amazingly successful,” he says. “ If there’s a major overhaul of the model, it’s hard to see how it wouldn’t look like a conspiracy. Somehow this ‘wrong’ model got it all right.”
So far, all the proposed modifications of Lambda-CDM have drawbacks. Some shrink the tension but worsen the model’s prediction of other data. Solutions that maintain its successes exist, but the physics community is loath to accept anything that does not feel intuitive. “There’s not that kind of magic bullet that does it,” Scolnic says. That fact, of course, does not mean no answer exists.
There is one thing everyone agrees on—more data are necessary. Two independent methods of measuring H0 using data from the early universe are in close agreement, and the opposing late-universe literature is growing. “I think the possibility of new physics remains real. It’s an exciting question,” Freedman says. But, she adds, her results suggest that astronomers’ most reliable standard candles may be subject to systematic errors that have gone unnoticed until now.
“We’re right at this five-sigma threshold, and that’s what makes this moment so special,” Scolnic says. “We have this puzzle, and we’re just missing a few pieces.”