iTPC Software Integration
into STAR Software Library

Dmitri Smirnov

Updated

Intro

• In 2018 the number of tracking layers in TPC is tripled in a single inner sector 20
• In 2019 all inner sectors of TPC are expected to be upgraded
• TPC group provided the code written by Yuri F., Irakli C., ...
• Irakli passed the changes to the S&C tracking experts for review and integration in the STAR library

Goals

• Focus on real data reconstruction first and simulation later
• New code should produce the same results when run over old data and geometry models
• Make sure effects on track reconstruction is minimized when iTPC data not used in 2018 samples
• Quantify the effect (if any) from the new geometry

Summary

• In 2018 tracking layers in sectors 20 (East) and 4 (West) are at different radii and do not span the full length of TPC
  • STAR tracking was not optimized for such configuration so, some degradation in efficiency is expected near the membrane ($z = 0$)
• The new additions are fully backward compatible. Identical results observed when running new code on old data
• iTPC hits can be used in track reconstruction
  • By default iTPC hits are not used in tracking
    Use iTpcIT option to enable iTPC tracking
    This is similar to PxlIT, IstIT, and SstIT options for tracking with HFT detectors
• Full iTPC support is introduced in release SL18e
  • The tracking code is able to accomodate any configuration of installed iTPC sectors
  • The following results are based on SL18e release

Expected Inefficiency in Run 18 Data

• We studied the effect of half-length layers in TPC by reconstructing tracks with split Sti layers in non-iTPC data
• The effect shown below is exaggerated due to all layers being split whereas in Run 18 only layers in sectors 4 and 20 will span half length of TPC
• Red: Standard reconstruction (StiCA) with full-length TPC layers
  Blue and Green: TPC layers split at $z = 0$ with StiCA and Sti respectively

2017 Sti TPC geometry

• In 2017 (and earlier) none of the Sti layers is split
• Full-length Sti layers shown with grey color
• Half-length layers with an edge at $z = 0$ shown with magenta and purple colors

2018 Sti TPC geometry

• In 2018 only inner layers in sectors 4 and 20 are split
• Full-length Sti layers shown with grey color
• Half-length layers with an edge at $z = 0$ shown with magenta and purple colors

2019 Sti TPC geometry

• In 2019 all inner sectors instrumented with iTPC, none of the Sti layers is split
• Full-length Sti layers shown with grey color
• Half-length layers with an edge at $z = 0$ shown with magenta and purple colors

Further Notes

• TPC tracking layers extend from $-400$ to $+400$ cm in $z$ to accomodate prompt hits. We plan to reduce the length without loosing the prompt hits in tracking
• Simulation with iTPC is work in progress and not yet available in the STAR library
• The event display pictures were produced with the star-travex package available at
  https://github.com/star-bnl/star-travex
  • It can accept events in StEvent format from STAR event.root files

Number of Hits on Tracks

• iTPC hits can be used in tracking when iTpcIT option is enabled
• Some fraction of tracks in 2018 data has upto 72 assigned hits as expected

Event Display

• TPC geometry in 2018

Event Display

• TPC geometry in 2018

Event Display

• TPC geometry in 2018 with real data hits

Event Display

• TPC geometry in 2018 with several reconstructed tracks and their associated hits

Event Display

• TPC geometry in 2018

Event Display

• TPC geometry in 2018 with real data hits

Event Display

• TPC geometry in 2018 with several reconstructed tracks and their associated hits

Possibility to Split Arbitrary Sti TPC Layer

• We added a new feature allowing to split arbitrary TPC layer in the Sti geometry
• It is intended mainly for debugging and evaluating the effect of split layers on tracking with Sti/StiCA
• The feature is activated by placing a text file in certain format in the current directory
• The next slides illustrate split layers in three different geometries of 2017, 2018, and 2019. For example, we split the following:
  • Layers 9 and 12 through 42 in sectors 3 and 21
  • Layers 6 through 10 in sectors 12 and 24

Split Arbitrary Sti TPC Layers: 2017

Split Arbitrary Sti TPC Layers: 2018

Split Arbitrary Sti TPC Layers: 2019

Historic Updates and Status Reports

Update February 22, 2018

Testing New Code: pre-2018 Case

• New code is expected to be backward compatible with pre-2018 data
• We test it by running the code over old data
  • Use data from Runs 15 (pp 100 GeV), 16 (AuAu 100 Gev), and
    17 (pp 250 GeV and AuAu 27 GeV), standard BFC options, 100 events
  • Observed differences in number of tracks when comparing iTPC_IRAK_20180216b vs SL18b
  • Small difference in the number of tracks (~10 tracks per event (1000s of tracks))
  • Small difference in the vertex position (~100s microns) and number of vertices
• Investigated the problem and identified the change in StiTpc/StiTpcDetectorBuilder.cxx causing the difference:
  • The length of TPC layers in Sti geometry was changed

TPC Geometry in SL18b and earlier

TPC Geometry for pre-2018 data on iTPC branch

TPC Geometry for pre-2018 data on iTPC branch

Update February 29, 2018

Testing and Solving Remaining Issues

• Pre-2018 geometry
  • The difference in track counts disappears when we keep the double length for TPC layers
  • Gene identified the hits on tracks causing change in track counts as prompt hits
  • It is suggested to keep the length of the TPC layers unchanged
• 2018 geometry
  • We test the case when all TPC layers are split in half in the middle to accomodate the increased number of layers in inner iTPC sectors 4 and 20

TPC Geometry 2018

TPC Geometry 2018

Revert Change on iTPC_IRAK_20180216b

--- a/StRoot/StiTpc/StiTpcDetectorBuilder.cxx
+++ b/StRoot/StiTpc/StiTpcDetectorBuilder.cxx
@@ -115,23 +115,19 @@ void StiTpcDetectorBuilder::useVMCGeometry() {
       }
       else {
        pShape->setThickness(St_tpcPadConfigC::instance()->outerSectorPadLength(sector));
        dZ = St_tpcPadConfigC::instance()->outerSectorPadPlaneZ(sector);
       }
-      if (NoStiSectors == 24) {
-       pShape->setHalfDepth(dZ/2);
-      } else {
-       pShape->setHalfDepth(dZ);
-      }
+      pShape->setHalfDepth(dZ*24/NoStiSectors);
       pShape->setHalfWidth(St_tpcPadConfigC::instance()->PadPitchAtRow(sector,row) * St_tpcPadConfigC::instance()->numberOfPadsAtRow(sector,row) / 2.);
       pShape->setName(name.Data()); if (debug>1) cout << *pShape << endl;
       //Retrieve position and orientation of the TPC pad rows from the database.
       StTpcLocalSectorDirection  dirLS[3];

Update March 14, 2018

Testing New Code: 2018 Case

• To estimate the effect of TPC layer splitting we implement the same in SL18b
• Test using old TPC data
  • Use data from Runs 15 (pp 100 GeV), 16 (AuAu 100 Gev), and
    17 (pp 250 GeV and AuAu 27 GeV), standard BFC options, 100 events
• We observe track inefficiency in all data sets (see next slides)

TPC Geometry pre-2018

TPC Geometry pre-2018

Inefficiency in Run 18168046

• Red: Non-split TPC layers, Blue: TPC layers split at z=0, Green: Sti only

Inefficiency in Run 18060107

• Red: Non-split TPC layers, Blue: TPC layers split at z=0, Green: Sti only

Inefficiency in Run 17072001

• Red: Non-split TPC layers, Blue: TPC layers split at z=0, Green: Sti only

Inefficiency in Run 16067017

• Red: Non-split TPC layers, Blue: TPC layers split at z=0, Green: Sti only

Proposals for Solving Issue with Tracking in 2018

• The tracks crossing TPC layers near the boundary will never see hits in the other half at the same radial layer
• Two solutions proposed:
• Solution 1: Re-assign hits close to volume boundary to the neighboring volume
  • This is a good direct way of addressing the issue
  • Victor is working on this solution
  • Dmitri will test when available
• Solution 2: Don't split TPC layers except inner sectors 4 and 20
  • Can be quickly implemented
  • Will not solve the problem in the iTPC sector
  • May be considered as a back-up plan to Solution 1

Summary

• The new code appears to be ready for reconstruction of pre-2018 data without iTPC
• Tested 2018 case with split TPC layers and compared to default non-split case using old data
  • A significant loss in track efficiency is observed
  • Victor is working on the code to eliminate boundary effects in split Sti Layers
• Gene has working code for StEvent topology map extended for iTPC by Thomas and the code to fill it
• Also...
  • xml implementation of iTPC geometry (essential for iTPC simulation) is work in progress by Irakli and Jason

Update March 22, 2018

Run 18168046: With Victor's suggestion

• Red: Non-split TPC layers, Blue: TPC layers split at z=0, Green: Victor's fix

Run 18060107: With Victor's suggestion

• Red: Non-split TPC layers, Blue: TPC layers split at z=0, Green: Victor's fix

Run 16067017: With Victor's suggestion

• Red: Non-split TPC layers, Blue: TPC layers split at z=0, Green: Victor's fix

Summary

• Tested code to eliminate boundary effects provided by Victor
• The loss in efficiency is not recovered

Update April 12, 2018

Observations from Additional Tests

• Set Sti TPC padrow layers to 5% thickness
• Observe a non-critical difference in track numbers and parameters
• Seems 5%, 10%, and 50% thickness produce almost identical tracks (i.e. the difference is small)

Summary

• Option to exclude iTPC hits implemented: "NoiTpcIT"
• TPC hit double counting resolved in pre-2018 data
  • When different DAQ reader types are tried sequentially in StTpcHitMaker legacy TPC DAQ records could be found. This caused an extra call to the routine filling TPC hits.
  • May need further adjustments for data with iTPC records
• Yuri and Irakli provided an update for Sti TPC geometry with all Sti layers being long except for sectors 4 and 20 (see next pages)
  • There is still an issue with the 2018 geometry. iTPC layers in West half?
  • There is still an issue with total length in 2019 geometry causing loss of prompt hits
• Most of support files for iTPC are on the main branch, StTpcHitMaker, ...
  • All tests for pre-2018 data pass OK

Sti TPC Geometry 2018, East Half

Sti TPC Geometry 2018, West Half

TPC Geometry 2019

TPC Geometry 2019

Update June 14, 2018

[Solved] Issue with T0 for iTPC tracks

Update June 29, 2018

2017 Sti TPC geometry

• Compare two implementations: by TPC group (left) and by S&C group (right)
• Full-length Sti layers shown with grey color
• Half-length layers with an edge at $z = 0$ shown with magenta and purple colors
  
• None of the Sti layers should be split
• Problems with implementation by TPC group: Short layers exclude prompt hits; unexpectedly split layers in two sectors

2018 Sti TPC geometry

• Compare two implementations: by TPC group (left) and by S&C group (right)
• Full-length Sti layers shown with grey color
• Half-length layers with an edge at $z = 0$ shown with magenta and purple colors
  
• Only inner layers in sectors 4 and 20 should be split
• Problems with implementation by TPC group: Short layers exclude prompt hits; outside layers split in sectors 4 and 20 and split layers in sector 7

2019 Sti TPC geometry

• Compare two implementations: by TPC group (left) and by S&C group (right)
• Full-length Sti layers shown with grey color
• Half-length layers with an edge at $z = 0$ shown with magenta and purple colors
  
• None of the Sti layers should be split
• Problems with implementation by TPC group: Short layers exclude prompt hits; inner layers split in sectors 4 and 20; split layers in sector 7