Determination of the Discovery Potential of Leptoquarks Decaying into a Tau-lepton and a Top-quark Using the ATLAS Detector at CERN

Leptoquarks are BSM particles that couple leptons and quarks at tree level. They naturally arise in GUT unification models comprising of higher symmetries like $SU(5)$ or $SO(10)$ that unify quarks and leptons. Recently, interest in leptoquarks has increased after the observations of B-physics anomalies. This thesis focuses on the development of an experimental pipeline with high sensitivity towards the novel asymmetric pair production of scalar leptoquarks in the $2\ell SS1\tau$ channel in high energy proton-proton collisions at $\sqrt{s}=13$ TeV. Run 2 data with a luminosity of $140\,fb^{-1}$ was used to evaluate data-MC agreement. Three different sets of signal MC samples were produced for modeling the signal, in total comprising of 28 mass points $-$ 11 mass points for same mass hypothesis and Yukawa coupling $y=0.5$, 11 mass points for same mass hypothesis and Yukawa coupling $y=1.0$ and 6 mass points for different mass hypothesis and Yukawa coupling $y=1.0$. The same mass samples had the leptoquark mass range from $1500-2500$ GeV in steps of $100$ GeV and the different mass samples had mass points of $1500\,\&\,2500$, $1700\,\&\,2300$, $1900\,\&\,2100$, $2100\,\&\,1900$, $2300\,\&\,1700$ and $2500\,\&\,1500$ GeV. Pre-Produced MC samples were used to model background. An extensive object selection and event selection was performed and the events were re-classified into prompt and fake leptons. Four orthogonal control regions were defined to model the agreement between background and real data and best fit values of norm factors for major fake lepton contributions were obtained using the template fit method. The obtained values were similar across all the signal models, ensuring stability of background modeling. A TabNet model was trained in the signal region to calculate signal probability with signal defined as all the signal smaples and background defined as all the SM processes. The classifier showed an impressive performance in separating background and signal with an $AUC=0.9995$. Expected 95% CL upper limits were obtained by performing a profile likelihood fit in the signal region on the effective mass and the signal probability. The results obtained using signal probability were observed to be slightly better for lower masses for the same mass hypothesis and better by a factor of about $0.6$ for different mass hypotheses. The analysis resulted in an expected exclusion of leptoquark masses less than about $2000$ GeV for $y=2.0$ and an expected exclusion of all the masses in the range $1500-2500$ GeV for different mass hypothesis and $y=2.0$. It was also established by the analysis that the kinematics of the signal samples is independent of the value of the Yukawa coupling.