A bit more than a year ago the SETI@home team was granted 24 hours at the Arecibo observatory to reobserve the best candidate locations detected by volunteers running the SETI@home screen saver. (For details, check out the Planetary Society's reports on the reobservations here.)

At Arecibo, we were able to observe 226 points on the sky, containing many of the best SETI@home candidates, several candidates found by the SERENDIP project, and some interesting astronomical objects, including a few known planetary systems and external galaxies.

We sent out the data resulting from these observations to our SETI@home volunteers (that means you). To people who got this data, there was no visible difference between it, and ordinary SETI@home data. For each of these work-units, we scanned the verified results for signals similar to the candidates.

One of the more difficult aspects of the analysis was determining whether a match to the candidate was good enough to be considered significant. We expected about a 10% chance that due to random noise a candidate would get a better score than it had originally.

In the process we found some bugs in our scoring algorithms. Now that we've sorted those out we can talk about our results. As we expected, for most candidates, no matching signals were found and therefore, their scores got worse. The exception was Gaussian candidates with a wide frequency window (non-barycentric Gaussians). Most of these got better scores. It turns out that because of the way SETI@home detects Gaussians, their candidate score is missing a mathematical term that all the other signal types have. We're looking for ways to derive the missing term so we can fix the problems with scores of these Gaussians.

For all of the other signal types, only one candidate was found whose score improved. (Remember, we expected about a 10% chance that we'd find one of these.) It is a Gaussian-type signal with a narrow frequency window. Normally, that would get us excited, but unfortunately, the properties of the signal don't seem consistent with a real signal. In a narrow frequency window, we would expect to find Gaussians with low Doppler drift rates (ones whose frequency is not changing rapidly with time). Unfortunately this Gaussian candidate consists of signals whose Doppler Drift rates are between 10 and 50 Hz per second. These would drift out of our 125 Hz matching window in a few seconds, so if we had looked at that part of the sky even a few seconds later (in any of our observations of this part of the sky), we wouldn't have found a match. Even so, we'll keep an eye on this spot on the sky.

The candidates we reobserved came from the first 2 years of SETI@home observations, so we've still got some work to do. We're working to identify a new set of candidates, and we will ask to be allowed to look at them with the Arecibo telescope. SETI@home will continue with a new software architecture (BOINC) and new hardware at Arecibo (ALFA, SERENDIP V) that will perform a comprehensive survey of the sky.

Return to Science Newsletter main menu