Using files available as of 2013-03-06 01:00 eastern U.S. time:

• Sti: /star/data99/reco/AuAu200/hijing_382/b0_20/minbias/y2010c/gheisha_on/eval_Sti
• Stv: /star/data99/reco/AuAu200/hijing_382/b0_20/minbias/y2010c/gheisha_on/eval13_Stv

I obtain the following results simply by opening a few MuDsts...

TChain MuDst("MuDst");
MuDst.Add("rcf11022_28*.MuDst*.root");

gStyle->SetPalette(1);gStyle->SetGridColor(kGray);

MuDst.Draw("GlobalTracks.mEta:GlobalTracks.mFirstPoint.mX3>>hh1(400,-200,200,250,-2.5,2.5)","","zcol");
gPad->SetLogz();

MuDst.Draw("GlobalTracks.mLastPoint.mX3:GlobalTracks.mFirstPoint.mX3>>hh2(400,-200,200,400,-200,200)","","zcol");

MuDst.Draw("atan2(GlobalTracks.mOuterHelix.mP.mX3,sqrt(sq(GlobalTracks.mOuterHelix.mP.mX2)+sq(GlobalTracks.mOuterHelix.mP.mX1))):GlobalTracks.mOuterHelix.mOrigin.mX3>>hh5(400,-200,200,250,-1.5,1.5)","","zcol");

MuDst.Draw("atan2(GlobalTracks.mHelix.mP.mX3,sqrt(sq(GlobalTracks.mHelix.mP.mX2)+sq(GlobalTracks.mHelix.mP.mX1)))-atan2(GlobalTracks.mOuterHelix.mOrigin.mX3,sqrt(sq(GlobalTracks.mOuterHelix.mOrigin.mX2)+sq(GlobalTracks.mOuterHelix.mOrigin.mX1))):atan2(GlobalTracks.mOuterHelix.mOrigin.mX3,sqrt(sq(GlobalTracks.mOuterHelix.mOrigin.mX2)+sq(GlobalTracks.mOuterHelix.mOrigin.mX1)))>>hh8(300,-1.5,1.5,200,-2,2)","","zcol");

Here are the results. Plots are Stv first, then Sti.


η vs z_first:

This indicates tracks are lost with have a window of positive η for negative z_first, and a window of negative η for positive z_first.


z_last vs. z_first:

This indicates that the lost tracks have same sign first and last z, so they are not central-membrane-crossing tracks;
they are tracks with z_last closer to the central membrane than z_first, but both z_last and z_first are on the same side of the TPC.


atan2(p_z,p_T) vs. z_last [i.e. direction of the track, using p_z,T at the last (outermost) hit]:

This indicates that at z=0 there is a discontinuity in the window of allowed momentum vectors (track directions).


[atan2(p_z,p_T) - atan2(z_last,r_last)] vs. atan2(z_last,r_last) [i.e. difference between the track direction and the direction towards (0,0,0)
at the last (outermost) hit; a schematic of this angular difference is included at the bottom of this post]:

One can clearly see the two kinds of tracks here: primaries lie near zero on the vertical axis with a spread defined
by the z_vertex distribution, while secondaries (and very low p_T primaries) form a diagonal band which is defined by
the acceptance of the TPC. My understanding is that [atan2(p_z,p_T) - atan2(z_last,r_last)] is supposed to be constrained
to ±π/2, but I have these observations of the constraint as it appears in these last two pairs of plots:

1. It is not constrained to ±π/2.
2. It is not flat in [atan2(p_z,p_T) - atan2(z_last,r_last)]...
3. ...nor is it flat in atan2(p_z,p_T), so it isn't cutting at some fixed η, e.g. it is not η<0 for z<0.
4. It is discontinuous across z=0.

I have made a sketch of the apparent constraint below (exaggerated to help see the variation). My guess is that this
constraint is calculated incorrectly.

-Gene