CBM Performance for Λ0 Hyperon Yield Measurements Using Machine Learning Techniques

Abstract

The phase diagram of the QCD matter in the baryon chemical potential region 500 MeV ≤ μB ≤ 800 MeV will be studied by the future Compressed Baryonic Matter (CBM) experiment in the beam energy range corresponding to √sNN = 2.9 - 4.9 GeV. The experiment will be carried out at the Facility for Anti-Proton and Ion Research. A prerequisite for determining the properties of dense baryonic matter is the multi-differential measurement of the yield of (multi-) strange hadrons.

This work evaluates the performance of CBM in measuring the multi-differential (charged tracks multiplicity, transverse momentum, and rapidity) yield of the most abundantly produced Λ0 baryon for Au-Au collisions at a beam momentum of 12 AGeV/c. The Kalman Filter algorithm is employed to reconstruct the Λ0 baryon through its weak decay to a proton and π- topology, which is selected non-linearly using the machine learning algorithm eXtreme Gradient Boosting (XGBoost). The selection is performed multi-differentially in transverse momentum and rapidity for the multiplicity interval [200,400] of charged tracks to achieve a high signal-to-background ratio.

After the selection, raw-yield extraction is performed multi-differentially through a multi-step fitting routine. The extracted raw-yield is corrected for the efficiency of the reconstruction and selection procedure and the geometrical acceptance of the experiment. The corrected yield is compared to the true yield to validate the reconstruction, selection, yield extraction, and correction procedure. The systematic uncertainties are evaluated by varying the selection parameters and they are typically below 3% but can go up to 6% for high transverse momentum intervals.