Freya Blekman

Physics

Currently being updated.

CMS

Top physics at Dzero

All-hadronic top physics

The top (t) quark is the heaviest known elementary particle, it is about twice as massive as gauge bosons and over an order of magnitude more heavy than the bottom (b) quark, which is the second most massive quark. Top quark decays are well understood in the Standard Model, over 99% of them create a b quark and a W boson when they decay. Because of this well-defined branching fraction, top quark measurements are typically classified by the considered decay modes of the W boson.

Multi-collision events

When using a dedicated trigger for selecting events with many jets, you are automatically biasing your sample to select events that are odd. One of the problems I had to solve in the all-hadronic top quark pair analysis had to do with exactly this sort of thing: My selection criteria were not only good at finding events with six jets, they were also very good at selecting events where two collisions took place, one creating two or three jets, and the second creating the remaining jets.

After cross checks with simulation, I rejected around 25 percent of all collision events we selected. This rejection was done based on the fact that the jets in the event were actually coming from different primary vertices (these represent the original collisions) and when the double-collision events were more than 5 centimeters apart I rejected them.

This is a problem that we will also have to deal with at the LHC, not every collision will consist of two hard proton collisions but some will! As we bias ourselves to keep events of this type we have to make sure we do not include these in our results.

Observation of gluon splitting in six-jet events

Energetic gluons are known to create quark pairs, also b quark pairs. This is one of the main backgrounds for many analyses that rely on the identification of b quarks as a way to reduce background. In hadronic top decays this is also the case, and unfoturnately the gluon splitting is enhanced in this type of topology due to the high energy scale of the events.

To make things even more difficult, gluon splitting is very difficult to model in Monte Carlo simulation. Show is one of the first distributions of the observed behaviour of gluon split b quark anti-quark pairs that was observed in Dzero Run 2. In the top to all jets measurements it is essential that this important background process is correctly modeled. SHown are distributions with and without gluon splitting included in the background prediction, and the result of two different methods to model this process.

Multivariate Methods

In the fully hadronic decay of top quark pairs, the separation of top signal from QCD multijet background is so difficult that, even after b-quark jet identification using secondary vertex tagging, it is necessary to use complex multivariate analysis techniques to isolate the signal. In this particular analysis I used a three-level artificial neural network to describe how 'top-like' observed events were.

Multivariate techniques and particularly neural networks are very powerful tools that can only be used when the input data is extremely well understood. This meant that the use of neural networks, particularly the systematical studies required to confirm the analysis is unbiased, can be very cumbersome. However, when used with caution and some healthy scepticism, the use of a neural net can actually yield considerable gains in purity or efficiency over cut-based analysis techniques.

Top cross section results

Shown: Summary of all Dzero top quark pair production cross section measurements (August 2007).