Compatibility of CDF single top results February 19, 2007Posted by dorigo in news, physics, science.
The CDF experiment has been looking for single top production for quite a while now, and the latest results, based on larger statistics and more refined analysis methods than ever before, have been puzzling.
Single top production occurs at the Tevatron proton-antiproton collisions mainly through two electroweak processes, which have a combined cross section of about 2.9 picobarns – that is to say, they happen once every 20 billion collisions, or twice per hour at the highest Tevatron luminosity. The process is quite hard to distinguish from background processes, and sophisticated analysis techniques have to be devised to see a signal.
A priori calculations show that the data collected in Run II is by now enough to show single top production, and indeed our cousins at the competitor D0 experiment claimed to see a nice signal of single top production a couple of months ago, but CDF struggled, hard-pressed to explain why its three analyses (now four) show conflicting results: while one analysis sees a signal compatible with expectations, the others see nothing at all, at the point of being close to ruling out the existence of the process!
The mesmerizing paradox spurred a deep investigation of the correlation of the four analyses, performed with pseudo-experiments: you simulate all known processes contributing to your data samples in large amounts with Monte Carlo generators, and then fish randomly from the simulated samples respecting the proportions of the various processes, creating a large number of sets of events (“pseudo-data”) each of which represent the possible outcome of your data selection process.
Once you have pseudo-data sets, you can forget about their origin, and concentrate on the search for single top with your four methods in each of the sets, coming up with four independent measurements of single top production cross section for each pseudo-experiment. The sets of results allows you to see how likely it was to see such a poor compatibility in the four cross section measurements as the one you saw in the real data.
CDF finds that they have been unlucky, although not overly so: the probability of the puzzling result in the data is 0.65%. That is to say, you should expect to get such poor agreement between the four analyses only twice in three hundred trials.
I am not sure what is the plan now. I suppose we will collect more data, planning on a measurement with 2 inverse femtobarns, which should allow to come up with a measurement after all.