Various Approaches to Detect New Physics Signals at Colliders and Cosmological Observations

In this thesis, we describe usages of effective field theory (EFT) and machine learning approaches in both particle collider and cosmological observation. EFT framework consists of the Standard Model (SM) supplemented by a series of higher dimensional operators comprises of the SM fields and their covariant derivatives. This can navigate new physics phenomenon shifting the SM particle final states: dilepton and diboson. In addition, new approaches such as deep neural network are powerful tool without introducing invasive cuts on physical observables. We employ neural network approaches to seek for new physics signals. This thesis covers three major topics.

In the first parts chapter 2 and 3, we implement Standard Model EFT calculations on Drell-Yan process and Wγ final state. We show specific calculations with full SMEFT extension up to O(1/Λ⁴). In Drell-Yan, we calculate the four-fermion interaction contributions from dimension six and eight operators, which dominates at large center of mass energy. Further generalization of such interactions allows us to estimate the effects from even higher dimensional terms (>8). As we shift to the bosonic final states, Wγ, we analyze polarization contents of the final states. We show different SMEFT operators affecting specific polarizations.

In chapter 4, we take a step further from the SMEFT based calculation and focus on building pipeline using machine learning in order to tag polarization of massive vector boson. Longitudinally and transversely polarized massive bosons yields different angular distributions yet the signature in hadronically decaying W's gets smeared away. Analyzing polarization of massive vector boson, we show deep neural network approach to measure the polarization fraction from hadronically decaying W's which can assist unveiling new physics hidden in higher mass than the collider's reach.

Finally, in chapter 5, we extend the usage of neural network approach to find localized cosmological defects. We implement inflationary production of particle with mass much larger than the inflationary Hubble scale that can be pair-produced non-adiabatically during inflation. Due to their large mass, the produced particles modify the curvature perturbation around their locations. This perturbation becomes a localized spot on cosmic microwave background. In this work, we show that Convolutional Neural Networks (CNN) provide a powerful tool for identifying PHS on the CMB