Interactions at hadron colliders, such as the LHC, produce an immense number of heavy particles. In turn, these particles often decay into much lighter quarks and gluons. Because of the nature of the strong force, these quarks and gluons fragment into showers of particles which are collectively grouped as “jets.” When the initial heavy particle has very high energy, relativistic effects can cause the jets associated with its decay products to become distorted and overlap. Furthermore, identification of these boosted jets is affected by the presence of pileup and the QCD background. Pileup refers to the external radiation resulting from colliding protons in bunches and detecting particles from multiple collisions at once. The QCD background refers to the jets produced from qq/qg/gg scattering processes which are difficult to distinguish from electroweak boson di-jet decays. Many algorithms have been proposed to identify and analyze hadronic jets; however, their effectiveness and methodology for these boosted particles is questionable. By exploiting properties unique to a jet in its center of mass (C.O.M.) frame, this project focused on developing and testing a new algorithm for studying boosted intermediate boson decays.
The algorithm proposed in this research study was developed as a tool for jet clustering, grooming, and background rejection. Clustering was accomplished using the anti-kt algorithm, a standard jet clustering algorithm, in the detector frame, boosting the jet’s constituents into their C.O.M. rest frame, and re-clustering into sub-jets using an anti-kt variant to account for the coordinate transformation. Treatment of external radiation in the form of pileup was handled by analyzing various jet selection methods in the C.O.M. frame. Methods for background rejection were developed by determining substructure variables with high rejection power. Our current results indicate that C.O.M. clustering has significant potential as a tool for pileup suppression and background rejection.
C.O.M. clustering was tested by reconstructing electroweak bosons in thousands of Monte Carlo truth level simulated collisions and analyzing invariant mass resolution along with other shower shape variables. C.O.M. clustering more accurately differentiates between true jet constituents and pileup particles than the extensively used jet grooming algorithm, trimming. Furthermore, several variables unique to the C.O.M. frame have been identified which provide substantial background rejection with high signal acceptance. Although we are still trying to fully understand and optimally implement this algorithm, these preliminary results are very promising.
Throughout this research experience I was able to develop my computer programming and data analytics skills by writing algorithms in C++ using CERN’s data analysis framework, ROOT. Furthermore, I obtained a better understanding of the Standard Model and the inherent geometry and symmetry underlying the Lorentz group. Most importantly, I was given the opportunity to work with experts in the field of experimental high-energy physics, who were able to provide mentorship throughout the project. Working with Prof. Elliot Lipeles, Theodor Christian Herwig, and the rest of the Penn ATLAS group proved to be an incredible opportunity. Ultimately, I am very thankful to the PURM program for providing me with this research experience.