This thesis presents a search for evidence of W±W∓γ production from p−p scattering with s = 13 TeV at the Large Hadron Collider using 140 fb−1 of integrated luminosity. The case where the W bosons decay to opposite-flavor light leptons is considered, as other decay channels are dominated by backgrounds. Monte Carlo simulations are used to estimate the contributions of the W±W∓γ process as well as various background processes to the e±μ∓γ channel. The contribution to the e±μ∓γ channel of processes with a misidentified or non-prompt photon in the final state are estimated with data-driven methods. A machine learning algorithm trained on Monte Carlo simulations is used to further increase the purity of \wwy events in the dataset. A maximum likelihood fit to the binned distribution of the machine learning discriminant is performed to determine the best-fit value of the \wwy contribution to the \emuy channel. From this, the expected \wwy production cross section in a fiducial region is determined to be 10.5−11%+17% (theory)±15% (experiment) fb. The observed cross section in the fiducial region will be measured after the \emuy dataset is unblinded.

Additionally, to study potential deviations from the Standard Model prediction in the \emuy channel, the ATLAS run-2 dataset is used to set upper and lower limits at the 95% confidence level on 13 Wilson coefficients of an effective field theory extending the Standard Model with dimension-8 operators. Two methods to restore unitarity to the effective field theory are investigated: the clipping method as well as a dipole form factor model. The dependence of the Wilson coefficients' expected upper and lower limits on these two methods' parameters are presented.