State of the Art Big Data Science for Unprecedented Searches for New Z’ Particles Motivated by the B Meson Anomaly and Dark Matter (DSI-SRP)
This DSI-SRP fellowship funded Elijah Sheridan to work in the laboratory of Professor Alfredo Gurrola in the Department of Physics & Astronomy during the summer of 2021. Elijah is a senior with majors in Physics and Mathematics and a minor in Computer Science.
The project funded by the fellowship sought to investigate the feasibility of experimentally detecting a new theorized heavy neutral particle. The Standard Model (SM) has served as humanity’s best fundamental theory of physics since the 1960s, and while the SM’s predictions have achieved unprecedented experimental verification, there are still glaring holes in the theory. To provide some specific examples, dark matter remains a mystery, and physicists do not understand why the different elementary particles have the masses they do, or why there is so much more matter than antimatter in the universe. Z′ (“Z prime”) models are popular theoretical extensions of the SM due to their potential to reconcile each of these specific enduring problems. In particular, Z′ models predict the existence of Z′ particles—named in reference to the Z boson, the SM particle it most closely resembles—which could be a mediator of dark matter interactions or give rise to a force which explain particle masses or matter-antimatter asymmetry. More recently, there have been observed disagreements between SM predictions and precision measurements of the decays of composite particles known as “b mesons.” These anomalies could be resolved by specific Z′ theories wherein the Z′ particle interacts most strongly with the heaviest quarks: top and bottom.
Elijah’s research this summer examines this particular Z′ theory and employs machine learning to investigate how this Z′ particle, if it exists, could be observed at the Large Hadron Collider in Switzerland: this process of experimentally investigating a theory is known as phenomenology. This task requires a method for differentiating collisions involving Z′s (signal) from closely-resemblant collisions not involving Z′s (background). Fortunately, this problem is well-suited for data scientific methodologies: in particular, binary classifiers are a class of machine learning models specifically designed to distinguish between two types of data. Elijah trained and optimized different binary classifier models—logistic regression, random forests, and gradient boosting—on computationally simulated particle collision data, allowing the models to learn how to discriminate between signal and background. This resulted in the discovery that a machine learning-driven approach would enable the discovery of a broader class of Z′ particles at the LHC than was previously thought possible. Moreover, given both the novelty and power of this machine learning approach, Elijah developed a publicly-available, fully documented Python library to facilitate the training and evaluation of binary classifiers in the context of particle phenomenology, providing a tool to be used by researchers for years to come. Elijah plans to continue this project into the 2021-22 school year, with the hopes of publishing his results in a peer-reviewed journal.
In addition to receiving support through a DSI-SRP fellowship, this project was supported and facilitated by the DSI Data Science Team through their regular summer workshops and demo sessions.