Below is a group of plots of the r-phi distortion corrections in a typical production file for hits from the inner sector of sectors 3 and 9, as a function of z. The top plot shows that distortion corrections may be notably different on the two halves of the TPC. For example, the Twist distortion has a sharp split at the central membrane (CM). This plot is completely reasonable for triggered events, but is inappropriate for pile-up hits. This is demonstrated by the bottom pair of plots which show the hits from sectors 21 and 3 separately:

Why is this a bad thing? We don't care if pile-up tracks are improperly corrected, right? If we want to identify tracks as pile-up by tagging the ones which have post membrane (PMB) hits, then we do care. These distortion corrections can cause pile-up tracks to split at reconstructed z=0, leaving a portion of the track which may be reconstructed without any PMB hits, so it cannot be tagged.

StMagUtilities was applying distortion corrections based solely on reconstructed hit positions. So there was no special consideration for the PMB hits. But the code already took special care of hits which were reconstructed within 0.2 cm of the CM, applying the distortion corrections from z = +0.2 cm to all hits reconstructed with 0 < z < 0.2 cm, and the corrections from -0.2 cm to those in -0.2 cm < z < 0.

To prevent track splitting, it was decided to take into account knowledge of the sector on which a hit was reconstructed (when available), and use these distortion corrections from z = ± 0.2 cm for all hits to the central membrane and beyond. The same three plots from above look now like this:

The bottom two plots show this extension of one correction to all PMB hits in each sector. The upper plot no longer looks as clean as it did before because PMB hits are not being treated as hits found on the correct sector for their reconstructed z.

These changes involved the following codes: StMagUtilities.h, StMagUtilities.cxx, StTpcHitMoverMaker.cxx

It is not obvious whether this is the best thing that can be done for Pile-up in z-dependent TPC distortions, thought it is almost certainly an improvement.

-Gene