How Do We Know How Bright a Supernova Was?

In 2025 the case for evolving dark energy weakened by a full standard deviation. The culprits: a dust formula left approximate since around 2013, and a file in which nine calibration numbers had been rounded from 0.33 to 0.3. No new telescope, no new supernovae, just a better accounting of how bright the old ones really were. That is worth sitting with. Type Ia supernovae are cosmology’s standard candles, and the entire supernova route to measuring the universe rests on one deceptively hard question. When a star explodes halfway across the universe, how do you know how much light it truly gave off?

Why an exploding star can be a ruler

The supernovae used for cosmology are a specific kind, Type Ia. Each begins as a white dwarf, the dense cinder left when a star like the Sun dies, made mostly of carbon and oxygen. Alone it would simply cool forever. In a binary system it does not: it gains matter until a thermonuclear runaway tears it apart in seconds.

The flash you see afterward is not the explosion itself. The blast forges about half a Sun’s worth of radioactive nickel-56. The light comes from that nickel decaying over the following weeks, first to cobalt, then to iron. Nuclear energy unbinds the star; radioactivity makes it shine. The amount of nickel forged sets the peak brightness, which is the single fact that makes these explosions useful.

The old story for their uniformity was tidy. A white dwarf cannot exceed about 1.4 times the Sun’s mass, the Chandrasekhar limit, so they were thought to detonate at the same mass and release the same energy. Same bomb, same yield, same brightness.

Not standard, but standardisable

Here is the catch the phrase “standard candle” hides. Type Ia supernovae are not identical. Their raw peak brightnesses scatter by enough that, used as they come, they would be useless for the precision cosmology needs. The trick is not that they are standard. It is that they can be made standard, by measured corrections.

The first correction is the one Mark Phillips found in 1993, and it is almost too convenient. Brighter Type Ia supernovae fade more slowly. Measure how fast the light drops in the fifteen days after peak, and you know how bright the peak truly was. The physics behind it is clean: more nickel means both more light and more material for that light to diffuse through, so a brighter explosion is also a slower one.

The second correction is colour. Redder supernovae are fainter. Dust in the host galaxy reddens and dims them, and part of the effect is intrinsic to the explosion, which is harder to pin down. Fold the decline rate and the colour into a standard formula and the scatter in brightness collapses from unusable to around 0.12 to 0.14 magnitudes, a few percent in distance. That collapse is what turned exploding stars into a cosmological tool.

The correction to the correction

Even standardised, the candles are not quite honest, and the residual bias is the sort of thing that keeps cosmologists awake.

After all the corrections, a Type Ia supernova in a large, old galaxy still comes out slightly brighter than one in a small galaxy, by around a tenth of a magnitude. This is the mass step, and it is real at more than three standard deviations. Nobody has fully explained it. It might trace the age or the chemistry of the stars that produced the explosion.

It matters because the universe changes with distance. Faraway supernovae, seen as they were billions of years ago, sit in galaxies that were younger and smaller on average than nearby ones. Model the mass step wrong and you bias the comparison between near and far. That comparison is precisely where dark energy hides. A correction you do not understand, applied to the exact quantity your conclusion depends on, is a dangerous thing to get slightly wrong.

Calibrating against the quiet cousins

All of these corrections are relative. They make supernovae agree with each other. Turning that agreement into real brightness, and real brightness into distance, needs an anchor to true physical flux, and the anchor is another kind of white dwarf entirely.

Just three hot, quiet white dwarfs define the entire brightness scale of the Hubble Space Telescope: GD 71, GD 153, and G191-B2B. These are not exploding. They sit there, and their light follows from first principles. A hydrogen-atmosphere white dwarf has a simple spectrum, its shape fixed by temperature and gravity. A physical model then says how much light it emits at each wavelength, to better than one percent. A newer network of thirty-two fainter white dwarfs extends the same scale across the whole sky at sub-percent precision. These are dim enough not to blind modern survey cameras.

When a rounding error moved the cosmos

Now the 2025 result the piece opened on. The Dark Energy Survey rebuilt its supernova calibration from the ground up, in an analysis called DES-Dovekie, and it is worth seeing what “recalibration” actually meant.

Four things changed. They cross-calibrated the surveys using those predictable white dwarfs. They retrained the light-curve model on the new scale. They replaced an approximation to a standard dust law, which had sat in the analysis software since around 2013, with the real thing. And they corrected an input file: nine calibration weights had been rounded to 0.3 instead of 0.33, quietly understating one systematic uncertainty by about twenty percent.

The effect on the cosmology was not subtle. The preference for dark energy that changes over time, which had stood at 4.2 standard deviations, fell to 3.2. A dust formula and a rounding error were worth a full sigma of what had looked like a discovery in the making.

One number can look like two discoveries

The Dovekie episode is a clean example of a general truth: a significance is a statement about the gap between two numbers, and both numbers carry assumptions. Here the supernovae barely moved. What moved was the calibration underneath them, and the headline shifted with it.

It pairs with the way the same result reads under two statistical frameworks. A 3.2-sigma preference and roughly five-to-one odds describe identical data, yet sound like opposite verdicts. If that split is unfamiliar, we take it apart in our explainer on what 5 sigma actually means. Calibration and statistics are two different places the same soft assumptions hide.

From relative brightness to the distance ladder

Standardised, flux-calibrated supernovae give a gorgeous relative map. This one is four times fainter than that one, so twice as far. Pinning the map to absolute distances, and to a value for the expansion rate, takes one more rung.

That rung is the Cepheid variable, a pulsing star whose period reveals its true brightness. Astronomers measure Cepheids in nearby galaxies that also hosted a Type Ia supernova, and anchor those to direct geometric distances. Supernovae extend that ladder into the deep universe. The ladder is also where the Hubble tension lives. That is the stubborn disagreement between this local distance scale and the value read off the early universe. Every one of these rungs inherits the calibration below it, which is why a rounded weights file three rungs down is not a footnote.

Established, contested, unproven

Established. Radioactive nickel-56 powers the light curve. The brighter-slower relation and the colour relation are real and shrink the scatter to around 0.12 to 0.14 magnitudes. The host mass step is real at more than three sigma. The white-dwarf flux scale is good to about one percent, and to sub-percent across the faint network.

Contested. What actually triggers a Type Ia, and whether these stars always explode near the Chandrasekhar mass. How much of the colour correction is dust and how much is intrinsic. The physical cause of the mass step. And how much of the evolving-dark-energy preference is leftover calibration systematics, though the recalibration that raised the question concluded that most of it is not.

Unproven, and the deepest worry. Whether Type Ia supernovae subtly change with cosmic time. The mix of progenitors and their chemistry differed in the young universe. If that shifts the standardised brightness even slightly, it mimics the signal cosmologists hunt for. Nobody can close this worry until the trigger is understood. An unsolved question in stellar physics sits underneath a headline about the fate of the universe.