A day at work

The occasional reader of this blog could think that, apart from attending a meeting, my day has just gone into trying unsuccessfully to get a date with a former student, chatting with colleagues, and blogging out of my boredom. I will try to convince you, dear reader, that this is not the case: in fact, I did have a very productive day at work.

Apart from leading the jet energy and resolution working group, I usually split my time between four research projects. The first aims at extracting a Z->bb signal to calibrate the response of our detector to b-quark jets. The second is a search for top quark decay into tau leptons and jets, a channel which has not been exploited for top identification yet; I have a PhD student working at it and mostly sit back and watch the steady progress of that analysis with a fulfilled smile. The third is the extraction of the top pair production cross section measurement from the decay of top quark pairs to six hadronic jets: I was part of the group that did it with the CDF detector in Run I (1992-95), and we are perfecting the measurement with the larger dataset we are sitting on now. The fourth is a project for the CMS experiment at CERN; with another PhD student, we are trying to use the silicon pixel detector of CMS (a detector which can detect the passage of charged particles with a position precision of few microns) to design a fast trigger at Level 1.

Ok, so what did I do today ? I basically kept modifying and running a program that analyzes the data collected by the ZBB trigger, to perfect the extraction of the Z signal. We have collected more than 20 million events, among which we know there are about 10000 Z->bb decays. To see that signal, one needs to weed out the data, by applying cuts that retain the signal but discard the background.
Once that is done, one needs to study the dijet mass distribution of the remaining data (typically, 50000 events among which there are 2000 signal events left). The dijet mass is the reconstructed mass of the hypothesized object that produced the two observed jets in its disintegration. If the data selection has been smart enough, one hopes to see a bump at 91 GeV in a smooth distribution: the background events do not have a well defined mass, and their dijet mass follows an exponential distribution; but when the Z resonance is produced, the two jets from Z decay will always correspond to a total energy release of about 91 GeV (which is the Z mass value). Energy and mass indeed are the same thing.

The plot below shows the signal I extracted from Run I data in 1998, the main result of my PhD thesis. The red points represent the experimental data; they are fit to the sum of a background distribution and a gaussian-like resonance, peaking at the Z mass value as they should.

Today, I obtained good results by using a selection variable whose effectiveness I've known for a long time, but which for some reason had kept aside. This is the transverse energy of an additional jet in the event. When the Z boson is produced, it seldom produces more than two back-to-back jets; on the contrary, most of the backgrounds will often have a third jet. So the energy of this third jet becomes a good discriminant. My program takes about 3 hours to run on all the data, but I've made it smart enough that I do not have to wait so long to see its results: the program fills histograms which are dumped on a file every 100,000 processed events, so that I can look at them while they are being filled, and get an update every minute or so. I've left it running for the nth time as I went home this evening, and I know that tomorrow I will be looking at some fine results.