More on same-sign lepton pairs by CDF March 17, 2008Posted by dorigo in news, physics, science.
In the last post I promised a review of the brand-new result of an automated search for new physics in proton-antiproton collisions collected by the CDF experiment in Run II. I will make an attempt at keeping my word here, despite a full agenda for today…
VISTA@CDF is a program developed by George Choudalakis, Si Xie, Conor Henderson, Bruce Knuteson (all from MIT) and Ray Culbertson (FNAL), five colleagues interested in new physics searches in CDF. The program, rather than focusing on any particular new physics scenario, searches data for discrepancies with the Standard Model prediction. A refining of the former, SLEUTH, focuses more on high-Pt tails of final state objects, in the assumption that there is where a signal of new physics is most likely to show up; according to the authors, SLEUTH is “particularly sensitive to new electroweak-scale physics”.
VISTA concentrates on four different features of the data: numerical excesses of events in one of many exclusive final states, shape differences in kinematical distributions, numerical excesses in the high-Pt tail of particular final states, or localized excesses in a variety of invariant mass distributions (what is usually dubbed “mass bumps” by insiders).
In order to define exclusive final states, standardized kinematical distributions, and multi-body invariant masses, each final state observable object is well-defined according to some common-sense selection cuts. So, for instance, an electron must satisfy regular identification cuts and have Pt>25 GeV in the central detector, or Pt>40 GeV if it has high rapidity (where backgrounds are usually larger); photons, muons, b-tagged jets, and tau candidates are also considered above pre-defined thresholds. These vary depending on the classification of the event as a single-object, two-object, or three-object container.
A very detailed and complicated tuning is performed in order to create a Monte Carlo simulation that describes all Standard Model processes capable of producing any of the considered final states. Of course, this is the tricky part: k-factors have to be applied, fake rates need to be tuned, detection and trigger efficiencies beg to be modeled. There are a total of 45 parameters which are constrained to reproduce observed features of the data.
Once parameters are defined and fixed, the CDF “search region” data is partitioned in a multitude of exclusive final states, which are then ordered in decreasing order of numerical discrepancy with predictions. The most interesting of these are the numerical excesses found by SLEUTH in some high-Pt tails. Here are the most discrepant ones:
The probability figures quoted are of course computed by factoring in the so-called “trial factor”, which accounts for the fact that if you look for a numerical discrepancy in several places you are more likely to find one than if you look in just one box. The top one is the one originating from the plot I posted yesterday. I attach another plot from that same final state below, but let me first make a point here.
The ones listed above are exclusive final states, and all these discrepancies arise in some of the most sensitive signatures of new physics an experimental particle physicist might conceive: high-momentum leptons are, in fact, very clean probes of electroweak physics, being solely the result of electroweak interactions or decays at a hadron collider. But all the more striking is that same-sign lepton pairs are excessively rare, since proton-antiproton collisions have the nasty tendency to conserve electric charge, an additive quantum number.
Are we on the doorsteps of discovering new physics, at last ? I think the possibility is there, although still very improbable. Take the most significant (the smallest) of those probabilities: it is still about a thousand times more probable than a “less-than-one-in-a-million” effect. And I did see (and publish) such an effect in Run I data, in another, very controversial, non-a-priori search… (see the full story in this unabridged multi-part post).
And then, there are issues with the credibility of those figures – for regardless how accurate and precise the work of the MIT+FNAL scientists has been, one cannot dismiss the chance that some k-factor was not as well fine-tuned as it should have. In other words, the weak point of this automated search is the fact that no matter how well the Monte Carlo is tuned to CDF data overall, it is hard to say whether a particular systematic error affecting mostly same-sign leptons has been constrained enough.
The authors appear to have reached the same conclusion, since in the public page of their analysis they state quite clearly that
“This combined global search for new physics in 2.0/fb of collisions at reveals no indication of physics beyond the Standard Model.”
Good to know! My 1000$ “insurance bet” against new physics is still safe…
Below I attach a one more plots relative to the numerical excesses observed in same-sign dileptonic final states.
The above is again the muon transverse momentum (similar to the plot I posted yesterday) for a very similar final state – one where together with a muon and an electron of the same charge there is a jet. Again, some excess is noticeable in the region of muon momenta around 40 GeV…
The two plots above (click to enlarge) are further documentation for two additional final states among the five listed above. In these plots the variable found most sensitive by SLEUTH is the sum of transverse momenta of all final state objects – respectively, the two leptons and the two leptons, missing Pt, and two jets. The insets in these plots show the region chosen by the automated procedure as the one providing the largest discrepancy with backgrounds.
As I look at all these plots and numbers, I have a sudden suspicion. The authors claim they accounted for the trial factor – the fact that they search for a discrepancy of either sign (an excess or a deficit) in a number of distributions. I wonder if the fact that all of the above discrepancies is an excess rather than a random collection of excesses and deficits counts as “further evidence”…
Hmm. I will investigate the matter.