Interview with Geoff Marcy
November 8, 2005
CI: People who look at the history of planet hunting lock in on 1995 as an important year. As a pioneer in the field, you have an insider’s view on how arduous the whole process of discovering extrasolar planets was. How long did it take to be in the position to make the first discovery? When did it really start for you?
GM: That question has two or three layers to it. The superficial layer is that I was a double major in physics and astronomy at UCLA in the early seventies. I loved physics and astronomy, and I remember as an undergraduate at UCLA thinking that the grand picture of the universe put humans in the proper perspective as small cogs on a great wheel. I remember feeling that there was some kind of importance associated with knowing our place—knowing how we came to be here on Earth. It was kind of starry-eyed, but I still think there’s a grain of that in me, a thread that’s excited about the notion of the vast space and time of the universe.
I knew as an undergraduate that I couldn’t make a living
as an astronomer. I only knew a few professors at UCLA, and I certainly wasn’t
as smart as they were. I thought that I
would be a physicist and enjoy astronomy as an amateur endeavor. When I
graduated from UCLA, I applied to a few graduate schools on a whim, and surprisingly
got into many of them. I was accepted at CalTech and
CI: There’s no shame in that.
GM: It’s funny to go through a good program like UCLA’s physics program and realize that you’re in the top ten, but not the top five. I went to graduate school at UC Santa Cruz.
CI: It sounds like you were already inclined to look at bigger questions instead of focusing on a narrow research topic with the aim of becoming a world expert.
GM: No, I honestly wouldn’t say that. I know this is might be hard to believe, but I felt like I was struggling most of the time. I didn’t feel comfortable enough to think that I could actually participate in answering big questions. I was feeling lucky just to be a part of the astronomy world as a graduate student. I was delighted and excited to be a part of astronomy research, but I didn’t have grand notions about actually making a contribution. For most of my career my mantra was, “Gee, if I can just carry out a career and make some tiny contribution, some little increment to the knowledge of humanity about the universe, I will be satisfied.” My perspective was feeling just barely lucky enough to make a little difference.
CI: That’s a reasonable perspective for anyone who’s a research scientist. We’re lucky enough just to be able to earn a living. To have a grandiose and ambitious goal is icing on top, and possibly even unrealistic.
GM: Yeah, that was how I felt about it. It was just icing to be able to participate, and if I could actually make a little difference, that would be the best.
CI: Just to rewind a little to UCLA—George Abell was there at that point?
GM: George Abell played a very critical role in my development. First of all, he was my official advisor, so I went to him for academic advice. But he also taught the most inspiring class that I took when I was at UCLA. It was a class on celestial mechanics. We learned about coordinate systems and how to transform from one spherical system to another. We learned how Gauss developed a technique to determine the entire set of orbital parameters for an asteroid from just three observations. Abell made us actually work the problems out. He gave us just three observations for an asteroid, and we had to do the calculations by hand. It’s a little hard to imagine; it makes me feel old that we didn’t have calculators. It was 1973, and we had to work the problems with books that had sines and cosines to seven or eight significant digits. It sounds like I walked in the snow, ten miles to school, uphill both ways.
But Abell taught two remarkable things. One was an incredible love of science. What it meant to be inquisitive, to have the ability to determine the orbit of an asteroid from three observations. I can almost see him smiling and standing on his toes, as he described how glorious it was that even in the eighteenth and nineteenth centuries, people could work out orbits and so could we.
The other thing he taught was care. You had to be extraordinarily careful. Using the Gauss technique to get the orbit required a hundred different arithmetic operations. You were taking differences of differences to get second derivatives, and if you made even the slightest error in eighth digit in one of your hundred steps, the whole thing was wrong, and you would have no idea where the error was. He emphasized that attention to detail, and I imagine that the precision that he promoted in that exercise is the same kind of precision that we need now to measure Doppler shifts to eight or nine significant digits. It’s a labor of picayunish attention that does pay off in the end
CI: You’re describing an apprenticeship—learning by direct experience and example.
GM: Absolutely, and you can see that I haven’t forgotten it. It’s thirty-two years later, and I can still remember sitting in the astronomy library, poring over those interpolations and trigonometric tables, and thinking, “My God, what if I make a mistake in the eighth digit? I’m doomed!” It’s still that way with planet hunting.
CI: What was the topic of your Ph.D. at
GM: I did it on the Zeeman effect in Sun-like stars. The idea was to take spectra at very high resolution—even to this day it would be considered high resolution—and look for the broadening of spectral lines that were sensitive to the Zeeman effect. It was the first attempt to survey the magnetic field on other Sun-like F, G, and K stars.
In 1982, I became a Carnegie fellow, which was my first
and only post-doc. I could tell that the Zeeman work wasn’t going to go very
far, because it was too difficult. I was feeling personally down about my
future as a scientist, because I could see that my one area of expertise was not
going to be very fruitful. I thought that I had already done what I could do,
and there were many uncertainties. I remember lingering in the shower one
morning in early 1983, with this Carnegie fellowship, thinking, “What in the
world am I going to do for the rest of my career? Am I going to make it as an
astronomer?” I didn’t think I that I was,
but I thought that if I wasn’t going to be very successful, the best thing I could
do was go for broke. I would try to answer a question that was meaningful to me
on a human, personal level, never mind what conventional science thought was a
proper question for a stellar spectroscopist. I remember thinking that with the
high-resolution spectroscopy experience I had in my pocket from the Zeeman work,
I might be able to measure Doppler shifts very precisely. I remember walking
out of the shower thinking, “That’s what I’m going to do. It may be my last
gasp, but I’m going to try to measure Doppler shifts very precisely.” I did
that using the
I got a lot of time, because no one else wanted the
100-inch except for Allan Sandage, who was doing metallicity work on halo
stars. I got ten nights a month on the
For the first time it made sense for me as a young person
to ask, “What is limiting the Doppler shift? Why can’t you measure Doppler
shifts to arbitrary precision, and therefore detect Jupiters?” People were
embarrassed for me when I started looking for low-mass planets there at
CI: How much of that was due to the checkered history of finding brown dwarves in the preceding decade?
GM: I think it was woven together. For one thing, people basically knew that you couldn’t find planets. They were down by a factor of ten to the nine in brightness, which meant you just couldn’t see them. The other aspect that you point out was this rather stuffed graveyard of false claims. By the mid-seventies, it was becoming more known that the Barnard star claim which had been prominently displayed in textbooks of the late fifties, sixties, and early seventies, was probably wrong, although it’s not clear yet that it is wrong. Even to this day, the rate of false claims hasn’t really declined. The rate has actually picked up, and there are still false claims out there. That added to the morose climate of planet hunting.
CI: It sounds like you were exhibiting a wonderful mixture of bravery and foolishness in pursuing this at that time.
GM: Again, I had nothing to lose. It maybe sounds odd to say, but I
didn’t think I was cut out to be a very good astronomer. As proof of how profoundly I felt this way, I
went into teaching in 1984 when my two years of the nominal Carnegie fellow
were up. Instead of taking a job at some other institution like the space
telescope, I decided to go to
CI: You were teaching ten times as much as a CalTech professor!
GM: Exactly, and I still love
CI: I’m sure the process of pushing the limits of what you could do with radial velocities was a fairly long tunnel of technical work. When did you begin to feel and think that the hunt was really on—that you were within spitting distance of the precision and errors that would get you what you wanted?
GM: What happened there is important for understanding how you do a project like this. In 1986 or 1987, Paul Butler and I started trying various things out. We realized that we needed some kind of wavelength calibration device, so we borrowed from the Canadian team that had used hydrogen fluoride. Then we decided to find our own molecule that would impose a wavelength standard right on the spectrum. To make a long story short, we found iodine. It was around the late eighties, using the iodine technique that we conjured up, that we started to get pretty good results. We would take ten measurements of the same star, over and over again over the course of a few months, and the rms was near twenty meters per second.
GM: I was working at Lick Observatory. I was actually the last person
I then started using the small 0.6-meter telescope at Lick
Observatory. It was called the CAT, and was a 24-inch telescope. That’s what we
used, because at
CI: It may be counterintuitive, if you don’t know a little bit about the Doppler technique, to find things as difficult and faint as planets with such a small telescope. Why didn’t you need to use the biggest telescope you could get your hands on?
GM: There are two sides to that. One is that there are luckily some Sun-like stars nearby. The very nearest of them—within light years—are bright enough that you just don’t need a big telescope. They’re naked eye stars that you can name. You don’t even need binoculars. It is a lucky thing for humanity that galaxies are structured such that we can actually see stars at night. If stars were separated by a factor of ten more than they are, we wouldn’t be able to see any of them. There are stars that are close enough, but only a handful.
We started with a set of 120 stars, and that was a
stretch, because we couldn’t use twenty or thirty of them. We were down to
about eighty-five stars that we could use this small telescope to observe. Occasionally
we were able to get one night a semester on the 3-meter telescope right at full
moon. No one else wanted full moon, because they were all studying quasars at
that point. They would give us absolute full moon, but we didn’t care, because the
stars were so bright that neither the moon nor the
CI: Did you begin to feel that it was just a matter of time and patience? There’s some instant gratification to a pretty picture of a nebula, especially when you’re gathering data points the hard way, knowing that it might be years before any signal will be anticipated. What sustained you through that phase?
GM: There’s a plus and a minus. The plus is that everybody knew Jupiters were going to take ten to twenty years to go around their host star, so we had time to kill, and it was totally appropriate for us to take data as we were doing. I say that tongue-in-cheek of course. Additionally, the work on the Doppler spectroscopic analysis, which was very challenging, was the heart of our effort. We wanted to take spectra with the iodine lines in there and somehow put together an analysis, but we had time to burn. Until you get ten years of data, you can’t expect to see a Jupiter anyway. Paul and I spent almost all our time developing the spectroscopic analysis. We knew that a patience level of ten years was going to be part of the ballgame. There was just no other way. If Jupiter takes twelve years to go around the sun, we were going to have to wait twelve years to see analogs of Jupiter go around their stars. It was a plus and a minus to know of these long orbital periods.
CI: As we know, you didn’t have to be quite as patient as you thought you might have to be. You were involved in a long and painstaking process, but that doesn’t mean that there weren’t any “a-ha” moments. Were there any sudden epiphanies? Times when signals emerged and the analysis was incontrovertible?
GM: Let me briefly give you three. One was technical, and as you
know, it’s the technical achievements that are really the “a-ha” part. Once you’ve
accomplished a technical goal, the science is going to happen. That occurred
around 1992, when the Canadian team finished their effort. They had only
studied twenty-one stars, but their precision was ten meters per second, and
they had found no planets at all. Paul and I were on a train together in the
The next “a-ha” moment was in regards to the Swiss team’s discovery
of 51 Pegasus. It was quite a unique coincidence that literally six days after
they made the announcement in Florence, we happened to have four straight
nights on the Lick 3-meter telescope—a real rarity for us. We took four
consecutive measurements of 51 Pegasus and drove off the mountain with the
sinusoidal curve, knowing that the Swiss were right. Immediately it became huge
news; a bombastic splash that Time
Magazine and others reported on. Nightline came to
CI: What an exciting time! Was there even a tinge of disappointment?
GM: No, no, absolutely not. A lot of people ask that, or say things like “Oh, you lost the race.” But what people forget is that there were ten teams out there hunting for planets, and most of them had dropped by the wayside. In some ways we’re one of the two winners. Because at that time—early October—we still thought that the chances of finding a planet in our lifetime were pretty remote. It’s hard to picture going through ten years and wondering if we would find a shred of evidence that there really were bonafide planets out there. To have the Swiss make the announcement was really inspiring and exciting to us. We didn’t feel like we had lost at all; we felt like the door had just swung open.
CI: And then you also knew about your own data set which hadn’t yet been analyzed.
GM: Exactly. We knew that we had just perfected a technique at five meters per second and had one hundred and twenty stars sitting on a hard disc. We were only being held back by the slowness of the computers of that time. It still took us six hours of CPU time on a good Sun Microsystems computer to just get one Doppler shift from one spectrum. Of course that improved very quickly in the next few years, but at that time we had to literally borrow Sun computers from all over Berkley campus, and start running jobs here and there just so we could start to crunch through the eighty or one hundred stars that we had sitting on hard disk.
As you know, within two months, we had the third of these “a-ha” moments, which was the discovery of two more planets. One was the planet around 70 Virginus, which was really exciting, because as you look at the graph velocity versus time, you find it has a period of about 116 days. Paul and I stared at the computer screen for about an hour, absolutely speechless, when we saw the planet around 70 Virginus.
We had found the planet around 47 Ursa Majoris previously, but it needed a couple more data points for us to be really sure; we’re very conservative. So basically, two months after 51 Pegasus, there were two more planets. These two planets played a very significant role. For about three years, 51 Peg became embroiled in controversy. We figured out very quickly that it was indeed a planet, and vigorously defended 51 Pegasus on webpages and public talks.
CI: As you push a technique like this even further, do you start to run into other astrophysical effects which may be limiting for the process?
GM: Yes, the dominant issue is mainly astrophysical. It’s the turbulence on the surfaces of stars that, in the parlance of solar physics, we call granulation. The turbulent motion of the gas over the hemisphere of the star has a velocity of hundreds of meters per second. Because each cell has a wildly crazy velocity that changes on acoustical time scales, it’s just lucky that there are enough of these cells over the surface that they average out. But it still now constitutes the astrophysical floor of the Doppler technique. We know that it’s going to be very difficult to measure Doppler shifts more precisely than about plus or minus one meter per second. Which, by the way, is where we are right now; we’re at one meter per second.
CI: That’s per measurement?
GM: Yes, we open the shutter, and five minutes later we have a measurement that agrees with the other measurements within one meter per second. We’re now using the Keck telescope and have dozens of stars for which the rms is right around one meter per second.
CI: As you potentially move away from solar stars, how do you find your sample size expanding? What are you looking for now?
GM: There are two areas. One is looking for short period, very low
mass planets of order ten Earth masses. That’s the big push right now in planet
detection. Planets of ten Earth masses having orbital periods under a month can
be detected if your Doppler precision is one meter per second. We’re hoping to
detect what we expect to be rocky planets, or miniature
CI: And this is using the same targets that you’ve been using all along? You’re essentially just looking for harmonics?
GM: Ironically, we’re going back to the original Lick sample of the one hundred and twenty stars we chose as the brightest nearby stars. I’m of course giving you a very short version of what’s occupying a lot of our time. But we have set aside a sample of stars—almost two hundred of them—that we are observing with Keck. A good fraction of those two hundred stars are simply the same ones we did from Lick observatory, but now at one meter per second precision. The goal is to understand whether or not rocky planets are common, at least down to a few Earth-masses.
The other goal is the detection of Jupiter analogs. Though that was the original goal; we amazingly haven’t achieved it yet. We haven’t yet found a Jupiter-mass planet with an orbit at five AU. We’ve found one planet around 55 Cancree but it’s still about three Jupiter-masses. Since we would say it’s a decent sign-post of a planetary system that has the same architecture as our own, it’s clearly of some astrobiological interest, both for the protecting role of a Jupiter for smaller terrestrial planets, and for its circular orbit. For very anthropocentric reasons, it’s still compelling to find Jupiters in nearly circular orbits out at five AU.
We do have candidates. Right now we’re surveying two thousand stars using telescopes all over the world. Among the two thousand stars, there are at least a dozen of them or so that show a clear Doppler signature—not yet confirmed—of a Jupiter-sized planet out at five AU. In the next two or three years I suspect we’ll have a handful of Jupiters. The real question becomes the distribution of orbital eccentricities among Jupiters and Saturns at five AU. In other words, what fraction of Jupiters have nearly circular orbits?
CI: At what point was the systematic discovery of closely orbiting, super-Jupiters starting to be strange and puzzling? How were theorists reacting to it?
GM: As soon as 51 Peg was discovered, three things happened within months. One was that Doug Lynn, Peter Bodenheimer, and Derrick Richardson wrote a paper suggesting that migration would bring Jupiters in, instead of dragging them outward as originally thought. Now, people more or less agree that migration is a fact of life.
The remaining theoretical question then was why they don’t migrate all the way in. Doug Lynn’s answer, and my answer, is that they do—planets migrate in and fall right into their star. Another round of planets form, migrate in, and at some point the musical chairs stop. When the proto-planetary disc goes away, the planets are frozen in the chairs they last found themselves. So that was point number one, that migration, planet dynamics, and the interaction of the planets is very important.
The other two things that
happened were more controversial. David Gray at
CI: That certainly doesn’t hold true once you have a few dozen cases.
GM: You know, that’s what you say, and that’s what I said at the time,
but not everybody could appreciate that. The reason that they couldn’t is that
most non-stellar scientists don’t realize that there’s only a reservoir of a
few hundred such stars to draw from, and we were observing all of them. We
weren’t leaving any out. They thought that somehow we were looking at face on
binary stars that just hadn’t been detected yet. So the
CI: It seems you’ve taken a step down the road of the Copernican revolution in showing that planets are a natural consequence of star formation. Maybe the flip side is you’ve found that even though other planets and other planetary systems exist, there is still something special about our own Solar System.
GM: I think that you’re right: there are two sides to this. On the one hand, we’re discovering that our Solar System is just one of many planetary systems, so we aren’t special. On the other hand, though we may just be one of many planetary systems, the architecture of our planetary system is special in that it’s a low entropy state. By that I mean a state where if you nudge one of the planets even the smallest amount—if you perturb Jupiter or Saturn or even Mars—the house of cards falls apart. I think that is something the astronomical community still doesn’t appreciate—that circular orbits designate a low-entropy, delicately balanced state; perturbations can vault it into a realm of no return. Which means that eccentric orbits and Darwinian planetary selection within a system is really the dominant activity. That planets, once perturbed, eject each other from the planetary system. At that point, you end up with eccentric orbits, maybe with only the most massive planets among them as a final product. We don’t know how often this happens. That’s why it’s so important for us to find Jupiters at five AU, and determine what fraction are in circular orbits. We just simply don’t know how often that happens.
CI: Does all this make you fairly susceptible to Rare Earth arguments?
GM: Of course I have friends at the
I think you can play it both ways, but the bottom line is that seven percent of the stars we’ve observed have Jupiters or Saturns. So, Jupiters and Saturns are fairly common. Most, but not all, are in eccentric orbits. These eccentric orbits wouldn’t be suitable conditions for creating stability for Earth-like planets. But even among the planets that we’ve found, some ten to twenty percent of them are circular enough that Earths would have no trouble surviving.
CI: This is still a field where patience is required, because as always in science, you detect the things that are easiest to detect.
GM: I think it’s neat. Obviously, what’s happening is that telescopes like the LBT, VLTI, and the Keck I instruments—with the help of adaptive optics—are going to begin nibbling away at Jupiters that are out around one hundred AU in the youngest systems. We’re going to begin to see Jupiters very far away from their host stars, but those that are at five or ten AU are simply going to require patience and the application of the Doppler technique. Honestly, there’s just no other way to do it.
CI: It’s true that bottom line, the Doppler technique still rules, right?
GM: Yes. I think the transit work is exciting, and we’re learning about the peculiarities of planets that happen to be close to their star. To learn about the chemical composition of planets that close in would be very interesting. But as you say, the Doppler technique is extremely valuable in that it will assess the Jupiters and the Saturns at solar system dimensions.
CI: What’s the body count so far?
GM: We’ve found 157, but the truth is we’ve got ten in the can that I haven’t written up yet.
CI: So you could be forgiven for just putting your feet up on the desk and smoking a big cigar! What’s the next step for you?
GM: There were two discoveries my team made this summer that I think are extraordinary. One is a planet of 7.5 Earth-masses that we found around Liza 876, which is far less than anything previously found. It clearly opens the door technically, as well as inspirationally, to finding planets of five Earth-masses or three Earth-masses. Finding rocky planets is tremendously exciting.
We also found a planet orbiting the star HG 149026. It may not make a big splash in The New York Times, but we’re very confident that it has a large, rocky core. It’s a Saturn-size planet orbiting close, but it transits. Debra Fischer was the lead on this. What Debra and her collaborators, including myself, found was that the planet is much too small to be pure hydrogen and helium. It has to have a significant core of rocky material. The reason that this is profound—at least from the astrophysical standpoint—is that this shows that planets form from the bottom up. Basically heavy atomic material coagulates into planetesimals with the gas then gravitationally accretes onto that core. It’s a very strong indication, though certainly not a proof, that the paradigm is right. The paradigm of how Saturn, Neptune, and Uranus formed is apparently operating for most of the giant planets that we’re finding. It’s things like that that still get me up in the morning; I’m not going to put my feet up on the table quite yet.
CI: So there’s the possibility that there’s a clone, proxy Earth that could have had a five or six billion year head start on us? That’s staggering to think about.
GM: I think the probability is very high. We know of many, many stars whose ages we can measure. Ages on the order of eight or nine million years. Some of those stars have metallicities near the sun. So, bottom line, there almost certainly are planets orbiting those stars. I can’t see any reason why there wouldn’t have been dust coagulation and formation of cores around those stars. Presumably they have the elements on the periodic table that would enable them to form complex amino acids. Maybe there are fewer such stars, but from the grand perspective, there are billions within our Milky Way galaxy. Billions of stars that are several billions of years older than our Sun with all the ingredients to make both planets and organic molecules.
CI: Do you then consider it just a matter of time before we find biomarkers around terrestrial planets elsewhere?
GM: You’re opening a huge can of worms with that question. We need more discussion, not just among scientists, but with the public, because it is going to take funding from Congress. There are some very serious political issues going on, as well as international issues. We should be collaborating with the Europeans, as many of your colleagues have articulated. And right now, we have a political climate that I don’t think is conducive to the kind of cooperation we need.
The quick version is that to hunt for biomarkers around rocky planets will require an imaging, space-borne telescope that can take a spectrum. Biomarkers such as methane, water vapor, or the concurrence of oxygen plus methane, or ozone plus methane—things that are chemically not stable together—are all very exciting. But the Terrestrial Planet Finder as an optical challenge has not been proven—either the interferometric version of it or the coronographic version of it. Those are the two architectures on the table. Right now, the coronographs are favored politically, but we are probably at least fifteen years away from any kind of Terrestrial Planet Finder launch. It’s still up in the air whether either one of them is feasible.
I think biomarkers are great science, great synergism, and strong because the study of them is so multidisciplinary. It’s even spiritually exciting, but again may be farther down the road. It’s analogous to the dream that we will someday put humans on Mars. It seems like everyone in the public knows that we want to put humans on Mars, but I don’t think they have any idea just how far away from that we are. The same is true of the Terrestrial Planet Finder. But I am excited; it’s a great goal. It might happen within our lifetimes, perhaps more likely our children’s lifetimes, but it’s going to happen.