A search for low mass resonances decaying into two jets and produced in association with a photon and development of pattern recognition algorithms for identifying track candidates in the ATLAS ITk detector.

This thesis presents a search for new subatomic particles by looking for localized excesses in the dijet mass distribution of low-mass dijet events produced in association with a high transverse momentum initial stated radiated photon. The search uses the full LHC Run 2 dataset (140 fb$^{-1}$) from Large Hadron Collider proton-proton collisions collected by the ATLAS experiment between 2015 and 2018 at a center-of-mass energy of 13 TeV. Two variants are presented: one which makes no jet flavor requirements and one which requires both jets originate from b-quarks. In the absence of a statistically significant excess in the dijet invariant mass spectrum in either channel, limits were set on the production cross-section for a benchmark Z' model and on generic beyond the Standard Model scenarios that produce Gaussian-shaped signals with a width of up to $15\%$ of the resonance mass in the dijet invariant mass spectrum. The analysis improves the sensitivity to the coupling of the Z' to quarks by up to $50\%$ compared to previously published results. The future HL-LHC is expected to deliver an integrated luminosity of $3000$ fb$^{-1}$ during its operation, increasing the sensitivity to new physics that could elucidate the interaction between dark matter and matter or explain the underlying mechanism of electroweak symmetry breaking. The high instantaneous luminosity creates multiple proton-proton interactions per proton beam crossing that pose significant challenges for object and event reconstruction algorithms, particularly for tracking algorithms used in trigger selections. On online system known as the Event Filter implements a track reconstruction chain and reduces the data rate from 40 MHz to 10kHz. In this thesis, I present a new method for data preparation as part of track reconstruction and a novel application of pattern recognition on FPGAs used in the online trigger to efficiently identify track candidates within the future ATLAS Inner Tracker. This algorithm is found to significantly reduce the number of fake candidates found, making the process of track fitting less computationally intensive.