Searching For Dark Photon Production in Simulated Proton-Proton Collisions

Dark matter has been a key area of research in physics for over ninety years, yet the scientific community has not reached a consensus on its nature. The dark photon is a proposed force mediator that extends the Standard Model, enabling interactions between ordinary matter and dark matter. This study investigates the use of modern computational techniques to simulate and detect dark photon production in proton-proton (p-p) collisions at $\sqrt{s}$ = 14 TeV. To simulate dark photon production in p-p collisions, the Pythia library was used to create a dataset of five hundred thousand points, out of which only twenty-five were associated with dark photon production. Due to the extreme imbalance in the dataset, SMOTE-ENN, an algorithm for handling data imbalance, was used to resample the simulated data. An XGBoost (eXtreme Gradient Boosting) model was trained on the resampled data and achieved a high ROC-AUC of 0.98, but exhibited challenges with false positives. These results provide a framework for integrating computational techniques into high energy experimental setups, potentially advancing our understanding of dark matter and guiding experimental validations of the dark photon model.