Analysis overview

In the analysis, kaons, pions and protons are identified through their specific energy loss $dE/dx$ in the main TPC. The yields are extracted by fitting $dE/dx$ distributions of quality-selected tracks, with a multiple-gaussian fit. Because energy loss is a function of total momentum only, the choice was made to do the fits in bins of total momentum $p$ and transverse momentum $p_T$. The yields are then corrected for acceptance and efficiency (including kaon decay) and a coordinate transformation is made to a rectangular grid in rapidity $y$ and $p_T$. To obtain rapidity spectra, the measured $p_T$ spectra are integrated by summing the measured yields in the $p_T$ interval where a reliable measurement is available and extrapolating to high (and in some cases low) $p_T$.

Technically, the process is divided into the following steps:

• Fill $dE/dx$ histograms from ROOT micro-DST
• Fit $dE/dx$ histograms to extract raw yield
• Calculate correction for acceptance and decay
• Calculate reconstruction efficiency correction
• Apply corrections and do transformation
• Make final plots

The most intricate step is the actual fitting and I will spend some time explaining the considerations that went into this. In terms of processing power, probably each of the four first steps are equal. Note also that the correction for acceptance, decay and efficiency is split into two parts for technical reasons. The reconstruction efficiency is calculated from embedding, which is inefficient when efficiencies are low, so the acceptance and decay correction is calculated separately from a larger statistics sample of pure simulation.

Not mentioned explicitly is the calculation of systematic uncertainties. These are calculated by varying some parameters of the $dE/dx$ fit that are not fitted to the data in every bin separately (see Section 3.3.4). In practice, this means that the fitting step is carried out multiple times and that the final steps include systematic error propagation.