ONE SUPERNOVA, FIVE IMAGES, THE UNIVERSE'S EXPANSION RATE
How fast is the Universe expanding today? The number that answers it is the Hubble constant, H0. Two of the main ways of measuring it — a “ladder” of distances to nearby stars and galaxies, and the faint afterglow of the young Universe, the cosmic microwave background — currently give significantly different values. This gap, the Hubble tension, is one of the major open questions in cosmology.
A third route exists, and it does not depend on either.
A cosmic lens with a built-in clock
When a very massive object — a galaxy or a whole cluster of galaxies — sits in front of a distant source, its gravity bends the light and can split the source into several images. This is strong gravitational lensing. The light rays follow paths of different lengths. If the source changes in brightness, the images change one after the other.
Those time delays depend on the mass of the lens and on cosmic distances — and therefore on H0. The idea was proposed by Refsdal in 1964, with supernovae seen in several images.
The first such supernova, now known as SN Refsdal, was discovered in 2014 behind the galaxy cluster MACS J1149.5+2223. It first appeared as four images around one galaxy of the cluster; a fifth image, predicted by models, was expected about a year later. The measured delays include one of 376 days between two images.
77,000 pixels instead of a few dots
The hard part is knowing exactly how the cluster’s mass is distributed. Until now, models used only the positions of the multiple images of background galaxies. Stefan Schuldt, of the University of Turku and the University of Milan, and colleagues in Italy, Germany, Spain and Finland, went further.
Their model already included 308 galaxies of the cluster, dark-matter halos, and more than 100 multiple images of 17 background sources. On top of that, they added the full shape and brightness of the galaxy that hosted the supernova: 77,000 pixels from the Hubble Space Telescope. It is the first measurement of cosmological parameters with this kind of model for a galaxy cluster.
To be safe, they tested four ways of weighting the data and counted the differences between them as an extra uncertainty.
A 3% measurement
- H0 = 65.0 (+2.0 / −1.8) kilometres per second per megaparsec, with the matter content of the Universe, Ωm = 0.30, measured at the same time.
- The uncertainty is 3.0%: the most precise value ever obtained from a single strong lens, ahead of a 3.9% result from a lensed quasar.
- Adding the pixels cut the uncertainty by a factor of 2.0 compared with the same model using image positions alone.
- All versions of the model agree with each other and with earlier studies of SN Refsdal.
The value agrees with the one from the cosmic microwave background measured by the Planck mission, and with another lensed supernova, SN Encore. It sits further from the distance-ladder value of the SH0ES team.
The uncertainty from the mass model is now comparable to the uncertainty on the longest delay. In other words, the clock, not the map, has become the limiting factor.
What could still shift the number
The authors list what their 3% does not include. They kept the cluster’s mass recipe fixed so the comparison with older models would be fair, so details of that recipe are not counted as uncertainties. A known blind spot of lensing, the “mass-sheet degeneracy”, is only partly tamed. And if the lines of sight through galaxy clusters were emptier than average, cluster lenses would underestimate H0.
The next test is close: another lensed supernova, SN Requiem, should reappear in 2026 or 2027, about ten years after it was first seen. Its delay should be measured very precisely, and with a model like this one could give H0 to within a few percent. The Rubin Observatory is expected to find many more such supernovae.