Home / Learn / Science / Cosmology A to Z / Lesson 9.1
All lessons

Module 9: Dark energy · Lesson 9.1 · about 16 min

1998: the expansion is speeding up

In 1998 two rival teams found distant exploding stars fainter than expected: the expansion is speeding up. How they caught them and ruled out the doubts.

Key points, meanings and formulas on one page: choose "Save as PDF" in the print box.

Module 9 · about 16 minutes. Read first: 2.4 Type Ia supernovae and 5.6 The acceleration equation. Lesson 5.6 gave the headline: the expansion is speeding up. This lesson opens up the experiment itself: how you catch an exploding star billions of light-years away on a schedule, what “fainter” means in numbers, and how the teams ruled out the boring explanations. By the end you will be able to read the famous supernova plot and say why it convinced almost everyone within a year.

If the expansion had been slowing, distant supernovae would look a little brighter than in a steady universe. They looked dimmer. Their light had travelled farther than any slowing universe allows, so the expansion must have sped up while the light was on its way.

1. A search for two numbers

In 1970 the astronomer Allan Sandage described cosmology as “a search for two numbers”: the expansion rate today, H0, and how fast the expansion is slowing down, the deceleration parameter q0 (Lesson 5.6). Everybody expected q0 to be positive, since gravity can only pull. The only question was how big it was, and so whether the universe would expand for ever or fall back.

To measure slowing, you compare the expansion rate long ago with the rate today. That means a Hubble diagram reaching billions of light-years: a distance and a redshift for objects so far away that their light left when the universe was noticeably smaller. Cepheids give out long before that. Type Ia supernovae, which briefly outshine a whole galaxy and can be corrected to a standard brightness (Lesson 2.4), can reach it.

2. Catching supernovae on demand

The problem: a galaxy hosts a Type Ia supernova only about once a century, and each lasts a few weeks. You cannot book telescope time for an explosion that has not happened yet. In the early 1990s Saul Perlmutter’s team worked out the trick, now called a batch search.

Just after new moon, when the sky is darkest, photograph dozens of patches of sky, each holding hundreds of faint, distant galaxies. Thousands of galaxies in all.
About three weeks later, just before the next new moon, photograph the same patches again. A handful of galaxies now have a new point of light: a supernova still on its way up to its peak.
Subtract the first picture from the second. Everything that did not change cancels out and the new supernovae stand alone. Because there are thousands of galaxies, you are almost certain to find several.
Big telescopes, booked weeks earlier, take a spectrum to prove it is a Type Ia and measure the redshift; the Hubble Space Telescope and others follow its light curve. The peak, corrected by how fast it fades, gives the distance.

Both teams used this method: the Supernova Cosmology Project (led by Perlmutter) and the High-z Supernova Search Team (led by Brian Schmidt, with Adam Riess as lead author of its 1998 paper). Riess’s paper used 16 distant supernovae and Perlmutter’s 42, at redshifts up to about 0.8 to 1, each compared with dozens of nearby ones.

3. The plot that changed cosmology

Astronomers measure brightness in magnitudes: bigger numbers are fainter, and every 5 magnitudes is a factor of 100 in brightness (the Astronomy course explains). The distance of a standard candle is coded in its distance modulus:

μ=m−M=5log10⁡(dL10 pc)
m
apparent magnitude: how bright the supernova looks to us
M
absolute magnitude: how bright it really is (about −19.3 at peak, after correction)
dL
the luminosity distance (Lesson 5.8), which depends on the whole expansion history between the supernova and us
Physical meaning
A difference of Δm in magnitude is a brightness ratio of 100.4Δm and a distance ratio of 100.2Δm. So 0.25 magnitudes fainter means about 21% less light, and about 12% farther away.

The trick for seeing small differences is to plot each supernova’s distance modulus minus what an empty, coasting universe would give. Then a steady expansion is a flat line at zero, a slowing universe dips below it (brighter supernovae: a slowing universe was expanding faster in the past, so it reached the same stretch, the same redshift, in less time, and the light had a shorter trip), and a speeding-up universe rises above it.

Supernova brightness compared with an empty, coasting universe. Up means fainter. The dots are made-up supernovae scattered like a real sample around today's best model. Slide matter and dark energy to fit them: only models with dark energy pass through the distant points.

The distant supernovae sat above the line: about 0.25 magnitudes fainter than a slowing, low-density universe with no dark energy (Ωm ≈ 0.3), and fainter still than the matter-only Einstein-de Sitter universe. Riess’s team found that a positive cosmological constant was needed at more than 99% confidence. For a flat universe, Perlmutter’s team found Ωm ≈ 0.28, leaving about 72% for dark energy.

4. Ruling out the boring answers

A result this strange needed every simple explanation checked:

  • Dust. Dust in the host galaxy dims light, but it also reddens it, since it scatters blue more than red. The teams measured the supernovae’s colours and found no extra reddening. “Grey” dust that dims all colours equally was proposed, but it would make supernovae keep getting fainter the farther you look.
  • Evolution. Maybe supernovae long ago were simply dimmer. But their spectra and light curves at high redshift matched nearby ones.
  • The decisive test. In a universe that slowed down first and only later sped up, the most distant supernovae, from the slowing era, should turn back down towards “brighter”. In 2004 Riess’s team used the Hubble Space Telescope to find supernovae beyond redshift 1, and they were brighter than dust or evolution would allow: the turnover was there.

The switch from slowing to speeding up is set by the acceleration equation: matter’s pull is ρm/2 and a cosmological constant’s push is ρΛ, so the switch comes when matter has thinned to twice the density of dark energy.

For our universe that is redshift about 0.65, around 6 billion years ago. Since then, independent evidence has piled up: the cosmic microwave background (Lesson 7.8), the baryon acoustic “ruler” in galaxy maps (Lesson 10.7) and the growth of galaxy clusters all need dark energy, even with no supernovae at all.

From “two numbers” to a Nobel Prize.

  • 1970 Allan Sandage calls cosmology a search for H0 and q0.
  • 1988 Saul Perlmutter's Supernova Cosmology Project starts at Lawrence Berkeley National Laboratory.
  • 1994 Brian Schmidt and Nicholas Suntzeff form the rival High-z Supernova Search Team.
  • January 1998 Both teams show early results at the American Astronomical Society meeting; Riess et al. is submitted in March. Science magazine names it the Breakthrough of the Year.
  • 2004 Supernovae beyond z = 1 show the earlier slowing era.
  • 2011 Perlmutter, Schmidt and Riess share the Nobel Prize in Physics.

“The supernovae were fainter, so maybe they are just a bit dimmer than we thought.” That was the first worry, and the teams tested it hard. The key is the shape of the curve, not a single offset: a change in the supernovae themselves, or in dust, would not make the most distant ones come back up again. Today the same conclusion comes from methods that use no supernovae at all.

  • Two teams in 1998 measured distances to dozens of distant Type Ia supernovae, found with a “batch search”: picture thousands of galaxies twice, three weeks apart, and subtract.
  • Distant supernovae were about 0.25 magnitudes (about 21% in light, 12% in distance) fainter than a slowing, low-density universe allows.
  • A Hubble diagram relative to an empty universe shows it plainly: the data rise above zero, as only a universe with dark energy does.
  • Dust and supernova evolution were ruled out; supernovae beyond z = 1 show the earlier slowing era.
  • Speeding up began when matter had thinned to twice dark energy’s density, at z ≈ 0.65, about 6 billion years ago. Nobel Prize 2011.

Check yourself

1What did the 1998 teams actually measure that pointed to acceleration?
2How did the teams find supernovae in distant galaxies on a schedule?
3Why could dust not explain the faint supernovae?

Sources

Your ticks stay in this browser. Save them as a file on the course home.

Updated 2026-11-22. Numbers are rounded for teaching; measured values come from the sources listed in each lesson. Simulations run in your browser; nothing you type is sent anywhere.