Back to Research
LHCbFlavour PhysicsMachine Learning

Charmed Λ Baryon Decays (LHCb)

Boosted-decision-tree signal/background separation for charmed lambda baryon decay channels.

Institute of Nuclear Physics, Polish Academy of Sciences (IFJ PAN), Kraków
Summer 2026PPSS summer internship
ROOT Python XGBoost

The PPSS (Programme for Polish Summer Studentships) placed this project inside the LHCb collaboration's charm-physics effort, under Prof. Dr hab. Mariusz Witek at IFJ PAN in Kraków — one of LHCb's home institutes and a major contributor to the experiment's tracking and physics-analysis software.

Charmed Λ baryon decays are a useful probe of both the strong-interaction dynamics that bind quarks into baryons and, in some channels, of CP violation in the charm sector — but the genuine signal sits under a large combinatorial background from unrelated tracks that happen to reconstruct into a similar invariant mass. Separating the two by hand-tuned cuts alone leaves signal on the table; a multivariate classifier does better.

The core of the project was building and validating a Boosted Decision Tree (via XGBoost) trained on kinematic and vertex-quality variables from ROOT ntuples, to assign each candidate decay a signal-likeness score rather than a hard pass/fail cut — the standard approach for squeezing more usable signal out of LHCb data without inflating the background.