Forward Tracking Report of the Vienna iLCsoft Group Winfried A. Mitaroff “Discussion document” at the 5 th ILD Workshop and SiLC Meeting
Jan 11, 2016
Forward TrackingReport of the Vienna iLCsoft Group
Winfried A. Mitaroff“Discussion document” at the
5th ILD Workshop and SiLC Meeting
LAL Orsay / APC Paris VII, May 2011
Vienna’s commitment to ILD
• MarlinReco’s Track Search and Reconstruction based on wrapped legacy Fortran 77 code (from LEP era).
• Urgent need to re-write this from scratch in C++.• HEPHY agreed in fall 2009, as a “significant contribution
to ILD”, to provide the new processor for tracking in the forward region (stand-alone track search shared).
• Required for processing data for the “Detailed Baseline Design” (DBD), i.e. to be released by spring 2012 !
• Scheduled start in October 2010 delayed until May 2011, due to problems recruiting a diploma/PhD student.
• Started with a new student (Robin Glattauer) on 9 May.
2Mitaroff: Discussion document at ILD#5 and SiLC22 - 26 May 2011
Forward region (ILD_00 layout)
Mitaroff: Discussion document at ILD#5 and SiLC22 - 26 May 2011
ϑ = 900
36.70
25.50
16.50
11.50
80
50
10 padrows →
FTD 1 2 3 4 5 6 7
Very forward region:– 5.00 < ϑ < 11.50: only FTD measuremts. contributing,– Range of FTD 1 (2) starts where that FTD 6 (7) ends.Intermediate region:– 11.50 < ϑ < 25.50: complex mix of VTX + FTD + TPC,– FTD: only FTD 1 … 3, plus FTD 4 until ϑ < 16.50,– TPC: 10 pad-rows @ 11.50 … 100 pad-rows @ 25.50.Barrel + FTD 1 only:– 25.50 < ϑ < 36.70: VTX + FTD 1 + SIT + TPC.ETD: ignored by track fitting (no more precision)– 9.80 < ϑ < 36.90: PR link to fwd. ECAL, useful in PFA.
Pixel disks Double-sided (stereo angle) strip disks
3
Forward tracking sub-tasks
1. FTD geometry description (done), \2. FTD drivers in Mokka (done), } not by Vienna
3. FTD digitizations in Marlin (missing). /4. FTD stand-alone track search –– shared task
(very fwd. and intermediate regions, 50 < ϑ < 25.50);Valencia’s stand-alone silicon track search may be evaluated for FTD.
5. TPC-supported track search(optional in intermediate region, 11.50 < ϑ < 25.50).
6. DAF-based final hit associations.
7. Precision forward track fit.
8. Region 25.50 < ϑ < 36.70 is “mostly barrel”(VTX, SIT, TPC) with only one FTD 1 barrel or fwd. task ?➯
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 4
Fwd. track search strategies• Stand-alone in FTD:
– This is the only possible strategy in the very forward region,
– Various algorithms exist – which to chose needs careful study,
– For small ϑ, hits from beamstrahlung-induced background may cause further problems (we need a reliable estimate),
– Layout for optimized track resolution (e.g. strip orientation and stereo angle) not necessarily optimal for track search.
• Combined TPC–FTD:– This may be an optional strategy for the intermediate region:
– Inward extrapolation of tracks found by local PR in the TPC,FTD hits tested against and associated to them,
– Timing problems hopefully solved by “time stamps”.
• Soft hit association:– Hits may be shared among tracks, and the final association relegated
to track reconstruction based on the DAF.
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 5
Forward track reconstruction
• Algorithms used:– Based on the Kalman Filter, with robustification by the adaptive
Deterministic Annealing Filter (DAF):– (1) Testing and updating the track hypothesis (hit associations)
by identifying and removing “outliers”, and resolving ambiguous associations from the track search,
– (2) Performing a precision track fit.
• Special features:– Flexible track navigation in the complex intermediate region,
coping with inclined or staggered plaquettes of FTD.– Energy loss of electrons modeled by the Gaussian Sum Filter
(GSF) requires ➯ extension of the LCIO data model,– Magnetic field distortions by the “anti-DiD” taken into account
(small “Billoir corrections” on helices, or Runge-Kutta).
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 6
Design and implementation
• The user API should be common for barrel and forward tracks.• However, the implementation is suggested to be separate and
complimentary for the barrel and the forward regions:• Optimal track search algorithms will differ for barrel and fwd.,
• Internal track representations may differ (e.g. 1/pT vs. 1/P),
• Coordinated independence of the two programming teams.
• A small MarlinReco control processor for the required top-level steering “barrel vs. fwd. calls”, transparent to the user.
• Coordination is enhanced by a common pool of utility classes; however, the skeleton GenFit is preferred over KalTest.
• Both implementations will rely on common interfaces, e.g. for• Using available results from a previous track search in TPC,
• Persistency by the new LCIO data model, augmented for GSF,
• Interfacing the new GEAR geometry toolkit (written by CERN).
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 7
Sharing of responsibilities
Discussions with F. Gaede and A. Ruiz at ILD Meetings in July & Oct. 2010, consent about sharing of tasks for the new ILD tracking:
• DESY Hamburg:Overall coordination and all barrel tracking: work has started.
• Spain (Santander, Valencia) and HEPHY Vienna:All forward tracking, with the sub-tasks
• (1,2) FTD geometry description, drivers in Mokka: Spain,• (3) FTD digitizations: Spain with contributions by Prague,• (4) FTD stand-alone fwd. track search: Spain and Vienna,• (5) TPC-supported fwd. track search: Vienna,• (6,7) DAF-based fits, precision track fit: Vienna.
Meanwhile, Paris VII (Aurore et al.) had contributed sub-tasks 1 and 2,
and will take care of sub-tasks 3 (with Prague) and 4 (with Vienna).
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 8
iLCsoft manpower @ Vienna
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 9
• Winni Mitaroff (head, will retire Dec. 2011, unpaid consultant thereafter)
• New undergraduate student:
– Robin Glattauer: Project work started 9 May on Marlin / MarlinReco, will be continued as a diploma thesis at TU Vienna July ‘11 – June ’12 on the new Forward Tracking (supervision by Winni M., backed by Rudi F.)
• Part-time support expected:
– Rudi Frühwirth (senior staff, statistics guru: Kalman, DAF, GSF …)
– Manfred Valentan (PhD student, mastermind fastsim tool “LiC Toy”)
– Wolfgang Waltenberger (junior staff, mastermind vertex tool RAVE)
• Aim at a first version of the new Forward Tracking to be used for the DBD !
• Possibly follow-on study on backgrounds in the forward region ?
Vienna’s contract with AIDA“Advanced European Infrastructures for Detectors at Accelerators”
• Official start at Kickoff Meeting,16-18 February 2011 at CERN• HEPHY participates in 2 of 9 “Working Packages” (contact: M. Krammer)
• WP 2: “Common Software Tools” (coordinators: F. Gaede, P. Mato)
• Task 2.2: “Geometry Toolkit for HEP”• Task 2.3: “Reconstruction Toolkits for HEP” (coord.: F. Gaede, DESY)
– Sub-task 2.3.a: “Tracking Toolkit” (coordinator: St. Aplin, DESY) * Sub2-task “Barrel Tracking”: DESY, ... * Sub2-task “Forward Tracking”: Santander, Valencia, Vienna
– Sub-task 2.3.b: “Alignment”– Sub-task 2.3.c: “Pile-up Events”– Sub-task 2.3.d: “PFA Calorimetry”
• Diploma student’s supervisor eligible for 1/3 salary refunding by AIDA
• Deliverable: generic toolkit within 38 months (due by end April 2014)
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 10
ILD vs. AIDA implementation
• AIDA sub-task 2.3.a– Deliverable: generic toolkit for Track Reconstruction.
– Timescale: February 2011 – April 2014 (38 months).
• ILD track reconstruction– Processor in MarlinReco: wrapped legacy Fortran code,
used for the 2009 LoI. Urgent need of new C++ code !
– Timescale: now – early 2012, to be used for the DBD.
• Pragmatic solution– Coding for ILD may serve as a “pilot implementation”.
– Needs clear separation generic vs. detector-dependent.
– Experience gained will be used for the AIDA deliverable.
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 11
Backup Slides
12Mitaroff: Discussion document at ILD#5 and SiLC22 - 26 May 2011
Software expertise at Vienna
• “Vienna School” of detector-related data analysis– Founded in the late 1970s by Meinhard Regler at HEPHY Vienna.
• Contributions to experiments since ~ mid 1980s– LEP-DELPHI, framework DELANA (f77): Local PR of FCA. Forward
track fit incl. VFT (Kalman), later merged with barrel for reprocessing.
– LHC-CMS, framework CMSSW (C++): Track and vertex reconstruction(Kalman, pioneering DAF, GSF). Offspring: vertex reco toolkit RAVE.
– Detector-independent tool LDT (MatLab) for fast simulation and trackreconstruction (Kalman). Used for optimizing ILD and CLIC detectors.
– KEKB-Belle II, framework BASF 2 (C++): Track search in SVD. Trackfit in PXD-SVD-CDC (Kalman, DAF, GSF). Both are still in progress.
• Dept. for Algorithms and Software Development– Headed by Rudi Frühwirth. A recent publication of general interest:
A. Strandlie and R. Frühwirth, Rev.Mod.Phys. 82 (2010) 1419-1458.
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 13
Mitaroff: Discussion document at ILD#5 and SiLC22 - 26 May 2011
Contributions to ILD by Vienna• ILD detector optimization:
– Evaluation of various tracker layouts, with emphasis on tracking quality in the difficult forward and intermediate regions.
– Development of optimization algorithms, and of a detector-independent fast simulation tool (“LiC Detector Toy” = LDT).
– Major contributions to the ILD “Letter of Intent”, which was the basis for IDAG’s 2009 approval of this detector concept.
• iLCsoft (Software for ILC Detectors):– Development of a detector-independent vertex reconstruction toolkit
(RAVE), based on “robust” algorithms developed for CMS.– Development of a new Track Reconstruction in the forward region of
the ILD detector (a major contribution, work starting now).
• SiLC (Silicon for Linear Colliders):– Generic R&D programme to develop low material budget high-precision
tracker modules, based on silicon micro-strip technology.– Design and construction of a Si envelope around a large prototype TPC
inside a superconducting magnet at a test beam at DESY.
14
What is the “forward region” ?• Very forward region
– 50 < ϑ < 11.50: only FTD measurements contributing,
– Range of FTD 1 (2) starts where that FTD 6 (7) ends.
• Intermediate region– 11.50 < ϑ < 25.50: complex mix of VTX + FTD + TPC,
– FTD: only FTD 1 … 3, plus FTD 4 until ϑ < 16.50,
– TPC: 10 pad-rows @ 11.50 … 100 pad-rows @ 25.50.
• Barrel + FTD 1 only– 25.50 < ϑ < 36.70: VTX + FTD 1 + SIT + TPC.
• ETD: 9.80 < ϑ < 36.90
– Ignored by track fitting: cannot contribute to precision,
– Useful for PFA (pattern recognition link to fwd. ECAL).
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 15
Forward tracking summary
• Forward track search:– Stand-alone in the FTD: careful choice among various algorithms,
– Combined TPC – FTD in the intermediate region (“time stamps”),
– Soft hit associations resolved later by the DAF.➯
• Forward track reconstruction:– Algorithms based on the Kalman Filter, with robustification by the
adaptive Deterministic Annealing Filter (DAF).
– (1) Testing and updating the track hypothesis (hit associations) by identifying and removing “outliers”, and resolving ambiguous associations from the track search,
– (2) Performing a precision track fit.
– Energy loss of electrons modeled by a Gaussian Sum Filter (GSF),
– Magnetic field distortions by the “anti-DiD” to be taken into account.
– Usage of a skeleton toolkit (GenFit or KalTest) ?
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 16
LDT example: forward optimization
17Mitaroff: Discussion document at ILD#5 and SiLC22 - 26 May 2011
(M. Valentan: presented at the 8th SiLC Meeting, Santander, Dec. 2008)
Track model for the GSF
22 - 26 May 2011 Mitaroff: Discussion document at ILD#5 and SiLC 18
19Mitaroff: Discussion document at ILD#5 and SiLC22 - 26 May 2011
Off topic: RAVE vertexing• Detector-independent vertex recon-struction toolkit: RAVE
– Developed 2004–08 by W. Waltenberger (master mind), F. Moser, B. Pflugfelder and H.V. Riedel,
– Interfaced to topological vertex search ZvTop,– Standard vertexing for CMS, Belle II (and 4th),– Pilot implementations in ILD’s “Marlin” and SiD’s Java-based “org.lcsim”
frameworks exist, but …
• The “rival”: LCFIvertex (tools ZvTop, ZvRes, ZvKin, ZvMst)– Developed by D.J. Jackson, S. Hillert, B. Jeffery et al., mostly at Oxford and RAL
until 2009; then support taken by Japan (Tohoku, KEK, Tokyo),– De-facto standard vertexing for ILD and SiD.
• New effort to reconcile RAVE with LCFIvertex:– Comparison study with ZvTop/ZvRes: recent meeting in Bonn 12 May 2011
(W. Waltenberger, S. Hillert et al.).
• W. Waltenberger: RAVE – a detector-independent toolkit to reconstruct vertices, IEEE Transactions on Nuclear Science, 2011, DOI 10.1109/TNS.2011.2119492 (08.03.2011).