SN H0pe Arrived Three Times, 116 Days Apart

Chapters

1. 0:00 Three arrivals
2. 0:36 The lens
3. 1:27 Refsdal, 1964
4. 2:02 The supernova named Refsdal
5. 3:13 What Webb returned
6. 4:06 The delays
7. 4:40 From a delay to a constant
8. 5:30 Seven blind models
9. 5:49 The live unblinding: 81.3
10. 6:59 The standard candle reweighs the models
11. 7:49 The mass-sheet degeneracy
12. 8:40 What it settles
13. 9:18 SN Encore, and three answers
14. 10:48 A fourth image with a date
15. 11:35 The number that was measured

Transcript

Three arrivals 0:00

SN H0pe arrived three times, a hundred and sixteen days apart. A cluster of galaxies can bend one distant source onto three paths, as Webb shows here with another galaxy. From the gaps between the arrivals came a Hubble constant: seventy five point seven.

We redid the division the paper describes, and with the mass model held fixed, a delay ten per cent longer turns seventy into sixty three point six. This episode asks which part of the answer was measured, and which part was modelled.

The lens 0:36

The supernova sits at a redshift of one point seven eight. In front of it, at a redshift of point three five, is a cluster called PLCK G one six five. Its mass bends the light, and each path has a different length.

Nobody was looking for it. In late March twenty twenty three, a Webb survey imaged the cluster for another purpose. Subtracting an older Hubble image from twenty sixteen left a new point of light, in all three images of one lensed galaxy.

The discovery team named it SN H0pe, with a zero: H nought is the usual symbol for the Hubble constant. They wrote that the name was for its potential to measure the time delays between the images, and from those, the constant.

Refsdal, 1964 1:27

The idea is sixty years old. In nineteen sixty four, a Norwegian astronomer, Sjur Refsdal, described a supernova far behind a galaxy, whose light could follow two paths. The difference in travel time, he wrote, could be a couple of months or more.

He listed what that difference would give: the Hubble constant and the mass of the lensing galaxy, from the delay, the two redshifts, the brightness of each image, and the angle between them. Every term on that list returns in this episode.

The supernova named Refsdal 2:02

It took fifty years. In November twenty fourteen, Hubble found four images of one supernova in a cross around a galaxy of the cluster MACS J eleven forty nine. The team named it Refsdal. He had died in two thousand nine.

That cluster made more images of the host galaxy, and the models said the supernova would appear again, elsewhere in the field. On the eleventh of December twenty fifteen, it did, at the predicted position. A supernova, forecast before it was seen.

The gap between the cross and the late image was measured at three hundred and seventy six days, to one and a half per cent. With the two lens models that best matched the observations, the Hubble constant came out at sixty six point six.

And the team wrote down the catch. A delay that precise would give an equally precise Hubble constant, if the cluster model were perfect. With all eight models, the answer was sixty four point eight. The delay did not change. The models did.

What Webb returned 3:13

SN H0pe was the second chance. Webb's director gave time at once: two more visits with the near infrared camera, on the twenty second of April and the ninth of May, twenty twenty three, in six filters, plus spectra of all three images.

What the camera returns is a count of light, per filter, per image, per visit. Three visits make a short light curve for each image. A type one A supernova brightens and fades in a known shape, so each image can be placed on that shape.

Placed on the shape, the three images disagree on the date. Image b is the youngest in its explosion, near its peak. Image a is far past it. The light along path a arrived first, then c, and b came last.

The delays 4:06

The measured gaps: image a arrived about a hundred and sixteen days before image b, and image c about forty eight days before it. Each delay carries an uncertainty of eight or nine per cent, in days, for an observer on Earth.

The spectra gave a second, separate reading. The age of the explosion in each spectrum put image a about a hundred and twenty two days ahead. It agrees, with an uncertainty about four times larger. The photometry carries the weight.

From a delay to a constant 4:40

Now the conversion. The delay between two images is a time delay distance divided by the speed of light, times a term set by the shape of the mass. The paper says it plainly: that distance is inversely proportional to the Hubble constant.

So, with a mass model fixed, double the delay and you halve the constant. Our arithmetic: a delay ten per cent longer than the model predicts gives sixty three point six. Ten per cent shorter gives seventy seven point eight.

Which puts everything on the mass model. Most of a cluster's mass is dark matter, seen only through what it bends. A model is built from the positions of every lensed image in the field, and predicts the delay for any Hubble constant.

Seven blind models 5:30

For SN H0pe, seven teams built seven models, independently. None of them knew the measured delays. The people measuring the delays did not know the models. The spectroscopy team and the photometry team did not see each other's numbers.

The live unblinding: 81.3 5:49

Then the result was unblinded, live. The first number, from photometry alone, was eighty one point three, plus fifteen point nine, minus eight point five. That is the number the protocol produced on the day.

Errors were found after that, and the paper lists them. Magnification uncertainties had been too narrow. A bias correction had the wrong sign. The photometry-only value moved to seventy nine point six, to seventy six point nine, to seventy five point nine.

One lens model had its own error. Once corrected, its weight rose from zero to sixteen per cent, and the value from the delays alone moved from seventy three point two to seventy one point eight. The rules allowed those changes, and printed each one.

So, from the two delays and the seven models: seventy one point eight, plus nine point two, minus eight point one. Five of the models share most of the weight. That is the answer the delays give.

The standard candle reweighs the models 6:59

Here the type one A matters a second time. Its true brightness is known, so the measured brightness of each image says how much the lens magnified it. Refsdal had that term on his list in nineteen sixty four.

Each model also predicts those magnifications. Most get the ratios between the images right and the absolute values wrong. Scored on the absolute values, one model takes sixty five per cent of the weight, and the answer becomes seventy five point seven.

Seventy one point eight to seventy five point seven is five per cent. No delay was remeasured. No model was rebuilt. The same seven predictions were given new weights, by a second measurement of the same supernova.

The mass-sheet degeneracy 7:49

There is a reason the ratios are not enough. Given one mass model that reproduces the images, a simple transformation produces a family of others that reproduce them equally well, each with a different Hubble constant. It is called the mass sheet degeneracy.

The transformation stretches every delay by the same factor, so ratios of delays cannot see it. It also rescales every magnification. A standard candle, whose absolute magnification can be measured, is one of the few things that can break it.

The delays alone contribute about five and a half per cent to the uncertainty on the Hubble constant. The final margin is plus eleven, minus seven per cent. The rest belongs to the models of the cluster, not to the clock.

What it settles 8:40

Planck predicts sixty seven point four. The Hubble ladder measures seventy three. SN H0pe sits about one and a half sigma above Planck, and within one sigma of the ladder. One supernova leans. It does not decide.

Refsdal shows the same thing from the other side. Its three hundred and seventy six days have been fed to other models since. One reanalysis gives seventy point zero. Another gives sixty five point one. Same delay, same cluster.

SN Encore, and three answers 9:18

A third supernova joined in twenty twenty five. SN Encore, at a redshift of one point nine five, behind the cluster MACS J zero one three eight. Its delay, about forty days, gave sixty six point nine, plus eleven point two, minus eight point one.

Encore's margin is wider than H0pe's for a reason the team names. To measure a faded image, you subtract an image of the same field without the supernova, taken later. Encore does not have one yet, and the paper says that alone could improve it significantly.

Three lensed supernovae, three published values: sixty six point six, seventy five point seven, sixty six point nine. Each margin is wide enough to hold the others. Every delay is precise. Every model is where they part.

Lensed quasars do the same sum with flickering light. A collaboration called TDCOSMO combined eight of them in twenty twenty five, and got seventy one point six, plus three point nine, minus three point three.

They also measured the cost of the mass sheet. On an earlier set of seven lenses, a mass profile free enough to allow the full degeneracy took the precision from two per cent to nine. Same lenses, same delays.

A fourth image with a date 10:48

The next test has a date. SN Requiem, in the same cluster as Encore, has a fourth image still on its way. If the constant is seventy three, it should appear between June and November twenty twenty six.

Requiem was found in archive images, years after its three visible images had faded, too late to time them. Its fourth image comes about a decade behind the others. It would be the longest lensing delay ever measured.

If the constant is sixty seven point four, between April and September twenty twenty seven. The model turns the arrival date into a Hubble constant. As of this episode, no fourth image has been reported.

The number that was measured 11:35

Which leaves the number this episode ends on. Not seventy five point seven: that was converted, and it moved by five per cent when the weights changed. What Webb returned was light, in three places, on three dates.

The delay between the first image and the last: one hundred and sixteen days, plus ten point eight, minus nine point three. Measured by the clock of one exploding star.

Next time, the same kind of lens, pointed at one faint dot. Earendel was reported in twenty twenty two as the most distant single star, magnified thousands of times. A lens model from twenty twenty five says forty three to sixty seven.

Description and sources

SN H0pe is a type Ia supernova at redshift 1.78, found by chance in Webb images of the galaxy cluster PLCK G165.7+67.0 in late March 2023. The cluster's mass bent its light onto three paths of different length: image a arrived about 116 days before image b, and image c about 48 days before it, each delay good to eight or nine per cent. A delay is not a Hubble constant. It becomes one through a model of the cluster's mass, mostly dark matter, and with the model fixed the constant scales as one over the delay. Seven teams built seven models blind to the delays. Unblinded live, photometry alone gave 81.3; errors found afterwards, and listed in the paper, moved it down. From the delays and the seven models: 71.8 +9.2/-8.1. Weighting the models by how well they predict the supernova's absolute magnification, which only a standard candle allows, gave 75.7 +8.1/-5.5. The same idea was proposed by Sjur Refsdal in 1964; the supernova named after him gave 66.6 from a 376-day delay, and SN Encore gave 66.9. Every delay is precise. The models are where the answers part.

PRINT-READY, FROM THIS CHANNEL

The Conversion Chain, Vol. 01 — the method, and four numbers taken apart with it

https://therepository.gumroad.com/l/the-conversion-chain

PRIMARY SOURCES

• Pascale, M. et al. (2025). SN H0pe: The First Measurement of H0 from a Multiply Imaged Type Ia Supernova, Discovered by JWST. ApJ 979, 13; arXiv:2403.18902 -- H0 = 71.8 +9.2/-8.1 (time delays, seven blind lens models); 75.7 +8.1/-5.5 (with absolute magnifications); live unblinding 81.3 +15.9/-8.5 and the post-unblinding corrections (section 8.2).
• Pierel, J. D. R. et al. (2024). JWST Photometric Time-Delay and Magnification Measurements for the Triply Imaged Type Ia 'Supernova H0pe'. ApJ 967, 50; arXiv:2403.18954 -- delays -116.6 +10.8/-9.3 and -48.6 +3.6/-4.0 days; about 5.6% contribution to H0.
• Chen, W. et al. (2024). ApJ 970, 102; arXiv:2403.19029 -- spectroscopic delays, -122.3 +43.7/-43.8 days.
• Frye, B. L. et al. (2024). The JWST Discovery of the Triply-imaged Type Ia 'Supernova H0pe'. ApJ 961, 171; arXiv:2309.07326 -- serendipitous discovery; DDT program 4446.
• Refsdal, S. (1964). On the Possibility of Determining Hubble's Parameter and the Masses of Galaxies from the Gravitational Lens Effect. MNRAS 128, 307.
• Kelly, P. L. et al. (2015). Multiple images of a highly magnified supernova formed by an early-type cluster galaxy lens. Science 347, 1123; arXiv:1411.6009. Kelly, P. L. et al. (2016). arXiv:1512.04654 -- reappearance on 11 December 2015.
• Kelly, P. L. et al. (2023). Constraints on the Hubble constant from supernova Refsdal's reappearance. Science 380, eabh1322; arXiv:2305.06367 -- 376.0 +5.6/-5.5 days; 66.6 +4.1/-3.3 (two models); 64.8 +4.4/-4.3 (eight).
• Pierel, J. D. R. et al. (2026). SN Encore. ApJ 998, 219; arXiv:2509.12301 -- delay -39.8 +3.9/-3.3 days; H0 = 66.9 +11.2/-8.1.
• TDCOSMO Collaboration (2025). A&A 704, A63; arXiv:2506.03023 -- 71.6 +3.9/-3.3 from eight lensed quasars; the mass-sheet degeneracy.
• HST program 18069 (PI Pierel), Requiem's Return -- fourth image expected June-November 2026 if H0 = 73, April-September 2027 if 67.4.
• Planck Collaboration (2020). A&A 641, A6 -- 67.4 +/- 0.5. Riess, A. G. et al. (2022). arXiv:2112.04510 -- 73.04 +/- 1.04.
• Welch, B. et al. (2022). Nature 603, 815 -- Earendel. Scofield et al. (2025). ApJ 993, 226 -- magnification 43-67 (next episode).

ARCHIVE PLATES — public-domain documents, with author, date, archive and licence.

• Pan: Seeing Triple | ESA/Webb | cc-by-4.0

... and 17 more plates, each credited on screen in the video and listed in this project's manifest.

Absolute Magnitude takes one number the internet repeats and asks what an instrument actually returned. Every episode ends on a measured figure with its source on screen, and says plainly which part of it was converted afterwards.

No AI-generated imagery is used on this channel. Every frame of archive footage is real, publicly released material, credited on screen with its licence; every diagram is drawn in code from the numbers in the cited papers.

There is no released video of SN H0pe itself. The cluster footage in this episode shows other gravitational-lens clusters imaged by Webb and Hubble, named in each on-screen credit.

#JWST #HubbleTension #GravitationalLensing #Cosmology #AbsoluteMagnitude