Optimization of top-tagging algorithms at the LHC

Author: 
Arne Reimers
Date: 
Oct 2013

Thesis Type:

Since many extensions of the Standard Model of particle physics predict heavy, so far undiscovered particles, which preferentially decay into top quarks, the top quark plays an important role in detecting the production of such new particles. The top quarks originating from these decays have high transverse momentum and hence their decay products are highly collimated. Therefore jet substructure information is used to distinguish between hadronically decaying top quarks and background processes which produce jets abundantly.
In this thesis the CMS-Top-Tagger is improved by using multivariate analysis methods and implementing new variables that can be used for the separation between top quark decays and background processes mentioned above. As a reference a multivariate algorithm is used, deve- loped in former studies.
Choosing the same signal efficiency of approx. 30% for the CMS-Top-Tagger and the multivariate algorithms, the method developed in this thesis reduces the background efficiency to 1,04% ± 0,01% compared to 1,90% ± 0,01%. The already existing algorithm operates at 1,27% ± 0,01%. Furthermore the definition of signal events is improved. It is found that the usage of multivariate methods can reduce the background efficiency by more than 50% with respect to the CMS-Top-Tagger while working at similar signal efficiencies of approx. 80%.