How we almost discovered the first Kuiper Belt Objects

Although most of my research is purely theoretical, at one point I came very close to making a major discovery, while collaborating with Tony Tyson, Raja Guhathakurta and Gary Bernstein. Here is a summary of our close miss, from the book Beyond Pluto, by John Davies [Cambridge University Press, 2001], pp. 60-62:

One other search project carried out around this time is worthy of mention, even though it was initially carried out using images taken for a quite different reason. In September 1990, Piet Hut of the Institute for Advanced Study at Princeton, New Jersey, was listening to a talk about a proposed telescope which would survey a large region of sky (this later became the Sloan Digital Sky Survey). The primary purpose of the survey was to make a huge census of galaxies, but Hut suggested that it might also turn up some rather closer things. After the talk, Hut was approached by Tony Tyson from the nearby Bell Laboratories, who said that it was not necessary to wait for the survey to be done. He and his colleagues Raja Guhathakurta and Gary Bernstein already had some deep CCD images that might be suitable for just such a search. Hut talked with Guhathakurta about ways in which their data could be used and they developed some software to add together individual images in moving reference frames. Adding images in this way smears out the fixed stars and galaxies, but adds together the signal from moving objects. Of course, this procedure only works if you know in advance how fast the target is moving. Luckily, since the motion of objects beyond Neptune is dominated by the reflex motion of the Earth overtaking them, a good guess at the correct rate of motion is possible even without knowing exactly how far away the targets are. The search algorithms were then tested on frames which had been modified to include a fake moving object. The software seemed to work well, so it was applied to some real images to see if anything could be found.

The method they settled upon was to use a number of their images taken with relatively short integration times and combine them to make a deep image in the reference frame of the fixed stars. Then they electronically subtracted these fixed objects, and removed any defects and cosmic ray hits from each of the original exposures. This left a series of "blank" residual images. These were then added together with successive offsets to allow for the motion of any hypothetical faint objects traveling across the image in a certain direction with a given speed. This rate of movement across the image is called the object's apparent motion vector. The shifting and adding was then repeated for a range of slightly different apparent motion vectors to maximize the chance of finding faint moving objects.

Although this method will work for any part of the sky, the best chance of success will be for regions close to the ecliptic, where the density of potential targets is the highest. Tyson, Hut and colleagues tried their method on some images of a field 3.5 degree from the ecliptic which they had taken in 1991. The field was observed nine times over a two-nights period and covered a total area of 40 arcminutes a side. They added the images to cover a range of possible motions from 1 to 4 arcsec per hour. Nothing was found, even though they would have expected to see objects as faint as 25th magnitude. They were unlucky, the method was a good one and the sensitivity they reached was high enough to have detected something if it had been there. It just so happened that the field they picked was empty on the nights they chose to look there. They got, in the words of Piet Hut, "Close, but no cigar".

The first Kuiper Belt object would be discovered soon afterward, in August 1992, by David Jewitt and Jane Luu.