A line from a fresh space digest arrived looking like an anniversary triviality: on September 23, NASA announced that Hubble completed its 200,000th orbit around Earth on September 19. However, the observation made on the anniversary day turned out to be not about the record at all—that day the telescope was looking at the galaxy cluster MACS J0417 and waiting for the supernova Athena, discovered by James Webb in 2025.
I latched onto not the round number, but the strangeness of the promise: this star has already exploded, and yet astronomers are naming the timeframe when they'll see it again with accuracy down to months. In previous issues, gravitational lensing flickered in the background—as a way to make out distant galaxies through clusters. But the story of "one and the same explosion, sliced by gravity into several copies with different delays" hasn't been unpacked in Curiosity yet. A repeat visit of an event that already happened—a rare case when the cosmos allows you to look at one catastrophe twice.
In 36 years in orbit, Hubble has traveled more than 5 billion miles and performed over 1.7 million observations, while astronomers annually request about seven times more time than is available. The shot on the day of the two-hundred-thousandth orbit was made precisely for the sake of waiting: the telescope returns to the same fields again and again, because in this story periodicity matters, not a single frame.
Athena was discovered by the Webb telescope in 2025, and the find was announced in January 2026—as part of the VENUS program, a large survey that images 60 rich galaxy clusters and uses them as optics. According to NASA's calculations, the supernova must appear again between September 2026 and early March 2027. The cluster MACS J0417 itself works as a magnifying glass: it bends and amplifies the light of objects behind it.
When a massive cluster ends up between us and a distant star, spacetime near it is curved so much that light from one and the same explosion reaches us along several routes. Photons travel different arcs, accumulating different path lengths and different gravitational delays—and several imprints of one event appear in the sky, separated in time by months and years, not a reflection like in a mirror.
The closer the route passes to the center of the lens, the longer the delay: gravity slows more strongly. In the example of Athena's neighbor, the supernova Ares, this is visible in pure form—three images are expected there, of which only one has arrived, and the remaining two should appear in about sixty years. Ares exploded when the Universe was about four billion years old, and it wasn't alone on the waiting list: Athena was found in the same program. One extinguished star sent us several copies of its finale, spread across the calendar—the second did the same.
The first such calendar was SN Refsdal. On November 11, 2014, Hubble saw four images of a supernova in the cluster MACS J1149.5+2223, arranged in a cross—that very configuration called an Einstein Cross. The host galaxy shone at a distance of about 14.4 billion light-years, and its light traveled to us for almost ten billion years; the flash happened long before the Solar System assembled from the protoplanetary disk.
The cluster, however, showed not only the cross: the same host galaxy is duplicated in it by at least three images, and each has its own path length. Seven independent teams built lens models, and all seven converged on approximately the same window for when a new image should appear. From late October 2015, Hubble kept looking at this field again and again, and on December 11 the supernova finally appeared exactly where and when the models promised.
This is the first case in history when the moment and place of a star's appearance were calculated in advance.
Another detail turned into disappointment: calculations said that in the third image of the galaxy the same supernova flared up in 1998, and no one saw it then, and possibly one image ran past forty-fifty years ago. Even more interesting is that the new image by brightness satisfied only some of the models—the blind test showed the scatter they organized it for. The Norwegian Sjur Refsdal was honored, who proposed exactly this idea back in 1964: to measure the expansion of the Universe by the delay of lensed images. His surname is carried by the supernova as well.
With quasars everything started earlier and more contentiously. In May 1979, Walsh, Carswell, and Weymann published in Nature an image of the "twin" Q0957+561—two identical quasars, later explained by a lens galaxy at redshift 0.39. Two images have one source, so brightness should repeat with a delay, and here real astronomical squabbling began: by the mid-1990s two candidates were in circulation, 417 and 424 days, and a paper by a group where Refsdal himself is among the co-authors was titled directly—"The time delay controversy not yet decided." Only analysis of photometry from fifteen years of observations from 1984 to 1999 allowed choosing between the numbers.
Another packaging of the same physics is the Einstein Cross (Q2237+0305), discovered by Huchra and colleagues in 1985: a quasar at redshift 1.695 stands almost perfectly behind the center of a spiral galaxy at redshift 0.039, and four images surround its nucleus. Because of the nearly perfect alignment, delays there are measured in hours, not years—and a separate game is already being played by individual stars of the lens, winking at each image in turn. One principle, laid out across timescales from hours to decades.
All this machinery stands on one rift—the dispute about the rate of expansion of the Universe. Cosmic microwave background radiation gives about 67 km/s per megaparsec, the nearby distance ladder with Type Ia supernovae—about 73, and the difference hasn't converged for decades. The time interval between copies of one source depends on the expansion history directly, so lensing works as a third arbiter.
In 2025, the TDCOSMO collaboration, relying on eight lensed quasars and stellar kinematics from Webb, Keck, and VLT, obtained H₀ = 71.6 with accuracy of about four-five percent, neatly between the two camps. And in January 2025, the first estimate of H₀ from a multiply lensed Type Ia supernova came out—and the second from lensed supernovae in general: SN H0pe, discovered by Webb in cluster PLCK G165.7+67.0, gave 71.8 by some lens models and 75.7—after the standard candle was applied to image amplification, meaning the calibration was baked right into the model.
And here's the catch: Refsdal itself in the 2023 recalculation gave H₀ = 64.8, meaning it shifted toward the cosmic microwave background value, while H0pe went toward the ladder. Two supernovae, one methodology—and they diverged in the middle of that very gap they were supposed to patch. The main uncertainty sits in mass distribution models, not in the supernovae themselves: that very mass sheet degeneracy from which the TDCOSMO collaboration protects itself with conservative errors and blind protocol. The method works; it just measures ambiguity more honestly than the constant itself for now.
Athena will return in the coming months—though sources diverge here: the January VENUS press release promised a repeat within one-two years, survey materials from the same January spoke of two-three, and NASA in September narrowed the window to early March 2027. The spread is small, but it's a reminder that "predicted timeframe" in astronomy is a confidence interval, not a bus schedule.
Ares will test the calendar not for us: for its two images it set a date about sixty years out. As Conor Larkin from STScI fairly noted, we'll most likely no longer be formulating the key questions about dark energy at that point, but the very fact of the repeat arrival will remain a measure of the Universe's expansion for those who will wait.
Athena's March window will close before we grow new questions. If the supernova doesn't arrive on time—the calculation will get another reason to recalculate, and that will be a result, not a failure.