One of the biggest scientific findings in recent years is the discovery that the universe is not only expanding, but it is also accelerating in its expansion. Under the influence of a mysterious dark energy, the universe will eventually thin out to nothingness and die a cold death. For the Insights story, "Dark Forces at Work," appearing in the May 2008 Scientific American, David Appell talked with Saul Perlmutter of the University of California, Berkeley, and one of the leaders of the group that came to the astonishing conclusion. Here is an edited excerpt of that interview.

In finding that the universe is on a path to runaway expansion, you had to find type Ia supernovae, which can act as distance markers. How did you get involved with supernova searching?
I was at the University of California at Berkeley for graduate school. One of the heroes here at Berkeley is Luis Alvarez. The tradition that he started is looking for interesting science no matter where it is and then finding tools to do those things. For example, he invented one of the first steady cams.

One of his protégés was my professor, Richard Muller. There was a project to do a superautomated supernova search that Luis Alvarez had suggested to Rich. They had just done one of the first adaptive-optics experiments.

So, as a grad student, you began developing the techniques for a robotic supernova search. But you also found that they could be used for other kinds of targets.
I had gotten interested in a theory that had been put forward by Rich Muller and Marc Davis and Piet Hut—namely, that you could explain periodic mass extinctions on the earth by a possible companion star around the sun. They called this star Nemesis. We realized that with the most common kind of star in the sky, the red dwarfs, you wouldn't know if it were orbiting around our sun. You could assume it was a much farther away red giant, just because it would have very little motion with respect to us, because it's bound to us. We realized that we could look in the catalogue of red stars for one that had this motion that you expect and that we could use the same telescope we were using for the robotic supernova project. So I developed the astrometric techniques to make this precise measurement of the parallax motion.

Have you done any work on Nemesis since?
No, it was one of the things I was doing, but my main project was working on the robotic nearby-supernova search. I had developed some new software. We were running the search and finding  supernovae, cataloging and keeping track of everything so you could compare it to everything we found.

Are you more of a theory guy, or are you more of a gearhead?
I'm really an experimental type. I really enjoy trying to catch the universe in the act of doing something very surprising and perhaps figuring out what it's doing by making measurements.

You were the team leader of the Supernova Cosmology Project in the 1990s, which consisted of a dozen members or so. How did you delegate responsibilities to your team members?
You're trying to figure out what are the critical paths to get to the result you want. And sometimes you think it's faster doing it yourself, and other things come along and you'd like to do them, but I'm trying to get this one done and somebody else comes along in your group and they start doing it. So it's much more organic in the sense that you find that you can't do everything. The best world would be one in which everyone you find to work with is better than you, because then you're doing better work than you could ever do.

What were the downsides to working on the project?
The worst thing about my research life was that I was always worried about something that had to happen in the next 24 hours, or sometimes, the next two hours. It was a terrible way to lead an ordinary life.

When you started the project, you were expecting to find that the universe was decelerating in its expansion, but the supernova data did not hew to that hypothesis. Was there ever a eureka moment where everyone said, "Maybe we have to consider acceleration"?
There were a number of things that primed that for us. Supernova 1992bi had error bars that were down into the range of acceleration. I asked, what do we do with that part of it? That would be a universe with negative mass. And someone said, there is that old cosmological constant we could throw in. So we put it in there formally as a way to express the answer. We didn't hate the cosmological constant, but everyone accepted that it would be set to zero. With any reasonably large deviation from zero, the universe would look nothing like it looks today.

Indeed, Albert Einstein called the cosmological constant his greatest mistake. It was a theoretical patch job to keep the universe stationary instead of collapsing on itself because of gravity. But the redshift measurements you took of supernovae—which is related to an assumption about the cosmological constant and the critical mass of the universe—led you to draw conclusions about the fate of the universe.
The idea that you could ask something so deep just by making this measurement—I was shocked that people weren't tripping over themselves to do it. You could find out if the universe was going to last forever, and you could find out if the universe was infinite or curved in on itself. What would be more enticing as an experiment to do?

How long did the data analysis take?
Analyzing all the stuff takes forever with a huge amounts of cross-checks. The last thing you do is plot it on the Hubble diagram [a plot that connects an object's distance and redshift], but even then we know that we still have six things left to check.

First we got seven supernovae on the Hubble diagram, but it was suggestive of an "omega mass = 1' universe.

In other words, the mass density of the universe was such that the universe is "flat": it would not expand forever, but it would not collapse back on itself, either.
Then we got data from the Hubble Space Telescope, and that made a huge difference, because it could peer much farther, to much higher redshift. That was the first time you started to see the data pull away from a flat universe, and it was suggestive.

The fact that we were not getting an omega mass of 1 started to imply to people that maybe there was an omega lambda [that is, a cosmological constant]. That was the first time we started to have that from our own data. This was the beginning of 1997.

Eventually we had 42 supernovae. The error bars were still big enough, and there's so may things you have to cross-check, you just don't take seriously something like this [acceleration result] until you check everything that could be a mistake. We knew a cosmological constant would be an amazing discovery, if true. But this is not one you want to get wrong.

So you reached the conclusion about a runaway universe gradually? There was no sudden realization?
If you want to say was there a eureka moment, it was the moment we finally decided this was true, and we started to go out and give talks about it. It was the most bizarre eureka moment you could imagine: E…U...R…E...K…A, spread out over nine months. In this case you've seen that data for almost a year, and it's not a surprise to you anymore, but you're a little bit surprised that you're starting to believe it.

For me, the closest thing to a eureka moment was when I gave a talk at Santa Cruz, and I showed the result and tried to make it clear. After the talk, the famous cosmologist Joel Primack got up and said, "Before anyone asks any questions, I want to make clear to the audience what the significance of the cosmological constant is. This is a shock, a huge deal." The cosmological constant just wasn't in people's brains at that point, and it was then that I realized it was a shock. That was late November to early December of 1997.

Then we decided to go really public and make a big splash and put the poster up at the January 9, 1998, American Astronomical Society meeting. We did a poster, a talk and a press conference. Then it became even more visible when the competing team came out with their results and used other terminology: "antigravity."

That other team was the High-Z Supernova Search team led by Brian Schmidt of the Australian National University. What was the sense of competition between your Supernova Cosmology Project and the High-Z team?
There were some advantages to both teams announcing so close to one another. I don't know of many times in science history where there was something that should have been a big surprise and that people should not have accepted. By the end of the year this was treated as a done deal in many circles. There are so many different things that happened that are relevant to this discovery. In the most parochial sense, members of my team feel we worked very hard to establish all the techniques that would make this thing work. Then the other team announced their results a month and a half later at a meeting at Marina del Rey where we also presented. It's a lot easier to present your results that are a real shock when you know another group has already got it. Our group would feel there's priority, that they established it. But it's also true that there's a final published paper that came out from the other group first.

Why didn't you publish in January of 1998, when you announced the finding?
We thought that establishing what the result was and getting that out to the community was the most important thing, and making sure that every last issue was checked and cross-checked. We didn't think there was an issue of priority.

I think most people feel that it plays as a simultaneous event. We could have been the winners if we had presented in a slightly different way. In the big picture, some days you get a little annoyed, but I can’t get too annoyed if everyone shares the credit. When you step back and ask, these kind of discoveries depend on so many things being put together. It took people spending years to understand different aspects. These results don't come about because one person did something, but because a community worked very hard to prove the result.

To what do you most attribute your scientific success?
I think the biggest thing is, first of all, being willing to learn things, being willing to pick up a new area, but also just being able to work with other people. Most of these jobs are too big for any one person. You end up trying to find a team of people who are as excited as you are and want to push the technique forward. I'm always struck by the fact that the image of the scientist is as a lone person wearing a lab jacket in the lab by themselves for hours, whereas my sense is that maybe the single most important thing for a scientist, aside from being able to think of good questions, is figuring out good people to work with and enjoying the process of inventing ideas together with other people.

Do you get excited in October, when the Nobel Prizes are announced?
These prizes are such a funny thing. The biggest concern is about taking credit. It's always nice to be someone that people thought did something nice. People can spend their whole lives waiting to see if they'll get the Nobel Prize, and it's not a very healthy psychological state.