d0 and z0 parameters at the extrapolated track state at the ECAL surface

[dudarboh]

The d0 and z0 parameters of the track state at the ECAL should always be equal to 0 by definition.

What is the purpose of the if condition here?

MarlinTrk/src/MarlinTrkUtils.cc

Lines 852 to 859 in e636fdd

	//fg: for curling tracks the propagated track has the wrong z0 whereas it should be 0. really
	if( std::abs( trkStateCalo->getZ0() ) > std::abs( 2.*M_PI/trkStateCalo->getOmega() * trkStateCalo->getTanLambda() ) ){
	streamlog_out( DEBUG2 ) << " >>>>>>>>>>> createTrackStateAtCaloFace : setting z0 to 0. for track state at calorimeteter : "
	<< toString(trkStateCalo ) << std::endl ;
	trkStateCalo->setZ0( 0. ) ;
	}

Do we need it? Shouldn't all track states at calo have d0 and z0 set to 0?

There are curly tracks that consistently pass this if condition and result in the huge z0.
This later on affects track length calculation significantly.

Location                                                AtCalorimeter 
Id                                                            00265079
D0                                                       -3.410605e-13
Phi                                                      -2.908563e+00
Omega                                                    -1.132190e-03
Z0                                                        1.901083e+03
Tan Lambda                                               -3.425631e-01
ReferencePoint                              -1082.73, 756.357, -2411.8
Cov matrix:  +2.126779e+01
             -1.300012e-02, +9.355696e-06
             -1.971645e-06, +3.736557e-09, +4.949642e-12
             +9.402990e-01, -7.208810e-03, -1.218709e-05, +3.466207e+01
             -5.503403e-03, +1.006174e-05, +1.273179e-08, -3.512686e-02, +3.673301e-05

Location                                                AtCalorimeter 
Id                                                            00436880
D0                                                       -2.273737e-13
Phi                                                       3.066983e+00
Omega                                                     1.229360e-03
Z0                                                       -1.496372e+03
Tan Lambda                                                2.927782e-01
ReferencePoint                               411.078, -1472.59, 2411.8
Cov matrix:  +3.070539e+01
             +7.454400e-02, +1.824613e-04
             -2.149480e-06, -1.030419e-08, +2.739041e-11
             -1.121788e-01, -6.276377e-03, +3.091315e-05, +3.622616e+01
             -5.844261e-03, -2.985082e-05, +8.095816e-08, +9.340481e-02, +2.442500e-04
             

[gaede]

Good question. From my vague memory and looking at the code: there seems to be an ambiguity in the track state computation for curlers - likely those that are coming back in when hitting the endcap - where the z0 is off by one full turn, i.e. $2*\pi/\omega*tan(\lambda)$. Your example tracks show this feature. Why I put in a condition that the z0 should be larger than this I don't remember - most likely a bug. Can you try w/ removing the condition ?

[dudarboh]

In my analysis, during the track length calculation, I manually set d0/z0 of the trackStateAtCalo to 0.
Non-physical lines disappear, so results got better.

Left plot: d0/z0 from production (bugged)
Right plot: d0/z0 set to 0

P.S. To better see the improvement, open plots in new tabs and swap between them.