Summary of Run 9 W reco efficiency studies

To try to understand the data/theory discrepancy for the W x-sections in recently. This post focuses on corrections to the efficiency estimation, and there is a separate post for the luminosity corrections.  Below the corrections to the previous efficiency estimations are discussed.

1) ZDC rate dependence of W efficiency

In this study the embedding was divided into two samples based on the ZDC rate of the background ZB event in which the PYTHIA W event was embedded.  There is a significant dependence on the ZDC rate for the tracking efficiency likely due to pileup and distortion contributions which are larger at higher ZDC rate.  This displayed clearly the need to correctly choose the ZB events for the embedding to sample the ZDC rate in the embedding the same as the data.  The previous efficiency estimations undersampled some of the high ZDC rate bins.  After correcting this undersampling the efficiency was only very slightly reduced (~1%).

2) W Track QA cut efficiency vs Reconstructed Track PT

When comparing some distributions between embedding and data (for good W candidates) it was noticed that the data reconstructed track PT distribution was broader than the embedding.  The broader distribution in the data indicates that there are more events in the data with an incorrect reconstructed PT, and the embedding shows that tracks with an incorrectly reconstructed PT are less likely to satisfy our track QA cuts.  Thus the embedding needs to be reweighted to sample these tails of the PT (or in this case 1/PT) distribution.  The result is a lower average track reconstruction efficiency by ~4.5% and ~4.0% for W+ and W- respectively.

3) Removing ZB events from the abort gaps

For the previous efficiency calculations I was including embedding events from all bunch X-ings.  Since ZB events sample time randomly they sample bunch X-ings evenly (even in the abort gaps).  Because pile-up effects may be different in the abort gaps and the W events in the data obviously don't come from there I removed the abort gaps from the embedding sample used to estimate the efficiencies.  The result of removing these ZB events is the total efficiency is lower by ~2% relative to the previous values.