July 7, 2005
The Barnard’s Star Blunder
The existence of extrasolar planets has become common knowledge in the past decade. But in the mid-20th Century, the search for worlds orbiting other stars got off to a rocky start. Alan Boss, an astronomer with the Carnegie Institution of Washington, recounts the early days of extrasolar planet-hunting.
Astrobiology Magazine -- These days, it is generally accepted within the scientific community that there are planets orbiting nearby stars. In the past decade, more than 130 such extrasolar planets have been discovered. But the first such "discovery," of a planet allegedly orbiting Barnard's star, turned out to be a false alarm. In a talk at a recent symposium on extrasolar planets, astronomer Alan Boss, of the Carnegie Institution of Washington, told the tale of this scientific snafu.
The extrasolar planets field started in many ways with Peter Van de Kamp. Van de Kamp had been a professor at the University of Virginia for several years. In 1937 he went to Swarthmore College and became director of the Sproul Observatory there. The next year he began a long-term search for very low-mass companions to stars.
One of the first stars he put on the search program was a star called Barnard's star. Barnard's star is the second closest star system to our own. The only one closer to us is the Alpha Centauri triple system. Unfortunately, it's an M dwarf, so it can't be seen by the visible eye, but it can be easily seen with a small telescope.
Van de Kamp started taking data on Barnard's star in 1938, and continued taking data for roughly 25 years. In 1963, he finally felt confident enough to present his first results. These were pretty excruciatingly difficult measurements. He and his colleagues were looking for variations of plus or minus 1 micron in the position of the star on a photographic plate.
They were trying to measure the photo center of these little blurry dots on the photographic emulsions to 1 part in 100. They would have 10 people measure the same plates independently, and then try to average over whatever individual systematic errors they would introduce, to find the true photo center of the positions.
After looking at some 2400 plates, he found evidence that there was a bit of a wobble in Barnard's star, which fit with the curve that would result if Barnard's star was being orbited by a planet about 1.6 times the mass of Jupiter out at a distance of 4.4 AU.
The odd thing about it was, though, that it didn't fit with a nice sine curve, which would indicate a roughly circular orbit, like Jupiter's, but it had a little bit of a cusp to it. So it was a somewhat eccentric orbit, but people thought, Well, maybe that's not so bad. This basically became the textbook example of an extrasolar planet. People had believed for a long time that extrasolar planets should exist, and this one became literally the textbook example.
But then about 10 years later, in 1973, along came George Gatewood. He had been doing his Ph.D. on astrometry at the University of Pittsburgh. He didn't really want to study Barnard's star, but some of his professors told him, Yeah, sure you're going to study Barnard's star. And so he was dragged into it. He did his own measurements, using different telescopes, the Allegheny Observatory's Thaw Refractor, as well as some plates taken from the Van Vleck Observatory. He only had 240 plates, but they were taken with completely different telescopes. For his thesis project, he set about to reduce those plates.
Instead of having them reduced by individuals sitting at a plate-measuring machine, though, they were reduced by a newfangled plate-measuring machine that the U.S. Naval Observatory had come up with. So it was done automatically. Equally importantly, they reduced the data with a different technique than what had been used before. His thesis adviser, Heinrich Eichhorn, was one of the fathers of analytical astrometry, and so they used Eichhorn's technique for the data analysis.
In 1973 they were able to produce their results on Barnard's star. But they found that some of the points in which they had the most confidence did not fit the curve produced by Van de Kamp at all. So they very politely and gently said that they had found no evidence for the traditional planet that Peter van de Kamp thought he had evidence for.
Things got even worse for Barnard's star's planet that same year, because there was another paper published in Astronomical Journal by John Hershey, who was also working at the Swarthmore College Observatory. He had studied another star called Gliese 793, another low-mass M dwarf star, and he found that, if he plotted the astrometric wobbles of Barnard's star and Gliese 793 together, both of them took a jump in one direction in 1949, and in 1957 took another jump in the other direction, implying that both of them either had exactly the same planet going around them, or else there was something else going on, namely some systematic errors.
Of course, that was what turned out to have been the case. In 1949, there had been a major change in the telescope, they put in a new cast-iron cell to hold the Swarthmore College refracting lens. They also changed the photographic emulsions they were using, which makes a big difference when you're trying to measure things to 1/100th of the size of a blur. And in 1957, they made a lens adjustment. So van de Kamp had tried to correct for all these things, but clearly the corrections were not sufficient. And so at that point, van de Kamp threw away his old data and started taking new data, and still continued to believe that Barnard's star should have a planet. But most people didn't believe it. This was 1973, and the field fell into a deep sleep at that point for 20 years.
Since then, the HST (Hubble Space Telescope) fine-guidance sensor team, led by Fritz Benedict, of the University of Texas, has been following Barnard's star to try to find whatever planets it has, and to my knowledge they have not found any signal yet. Instead, Barnard's star is more useful for debugging mechanical problems on HST, because when Barnard's star seems to wobble, it usually means that something's happened with HST.
On the Net: