There is a bug in the latest release (tpcegeo3 and 3a) of the TPC geometry.  The volumes which comprise the support structure (pictured below) were flagged as MANY volumes.  These volumes overlap another MANY volume, TWAS -- the wheel and sector  volume which contains the FEE cards (not shown) -- which occurs at the same depth as the structure in the geometry tree.  The TWAS volume is an "air" volume, serving as a logical volume which contains the FEE cards.  The overlap of these volumes is problematic, because GEANT will choose which volume a track is in based on which volume appears first in the list of volumes it maintains in memory.  Any reording of this list would change the material distribution in the TPC wheel.  Such reordering could happen if changes were made to the TPC geometry module (i.e. change the order in which volumes are positioned), or if the geant propagators are changed (i.e. we use ROOT/TGeo propagators in VMC).  With the geometry in its current state, changes to the TPC geometry module may result in unanticipated behavior.  So it is desirable to fix this bug if it does not have an impact on existing geometry definitions.

Figure 1 -- Aluminium support structure for one sector of the TPC wheel.  The 9 "bricks" are volume TBRW, the crossmember is TWMR, and the spokes of the wheel are TWTR.  The TWAS volume "fills in" the lower and upper halves of the support structure, overlapping with (at least) TBRW.

I propose to fix the tpcegeo3 and 3a geometries.  Since tpcegeo3 is in production, we need to validate that the change does not have a material impact on the STAR geometry.  Below you fill find figures which show the difference in material budget between the AgSTAR geometry with and without the bug fix.  I consider the geometry to "pass" the differential test if the difference is less than 1%.  But here we see differences << 1% (note the large negative exponent on the Y axis).  Starting with "all material contained within the CAVE", i.e. the full STAR detector, we observe no differences beyond roundoff error.

Next examining the TPC we again see no material difference beyond roundoff error.  The maximum difference is order 10E-9 %.

I skip the calorimters (they show no differences) and examine the TPCE in more detail.  The issue is that the Aluminum structural elements in the TPC wheel should be marked as ONLY instead of MANY.  These structural elements are all contained within the TSAW volume.  So a differential plot between the buggy and fixed version of the TPC should result in the same material distribution.  As the plot shows below, this is the case.

However, we have changed the tracking behavior through two volumes.  These volumes are TBRW and TWMR.  These are both aluminium structural elements.  TBRW are the 9 "bricks" of aluminum in the structure figure above, and TWMR are the cross beams and spokes of the wheel.  Because these volumes overlap and they are both marked as ONLY in the new geometry, geant will resolve which volume the track is in by finding the first volume in the list.  So it is possible that the material distributions in these volumes will change.  But because these volumes have the same material and width (in Z), the integrated distribution of material (in their mother TSAW) remains the same.

Conclusion:  We can mark the structural daughters in TSAW as ONLY without changing existing definitions of the STAR geometry.