Forward Tracking in a Linear Collider Detector Robin
Forward Tracking in a Linear Collider Detector Robin Glattauer Rudolf Frühwirth Winfried A. Mitaroff Annual Meeting of ÖPG-FAKT Univ. Graz, 18– 21 Sept. 2012
• Physics motivation • Experimental environment: – International Linear Collider (ILC) – International Large Detector (ILD) • Track reconstruction: – Strategy – Forward tracking – Performance • Summary and outlook 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 2
Collisions at the Te. V scale 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 3
Cross sections at the Te. V scale pp 21 Sept. 2012 e– e+ Winni Mitaroff: ÖPG-FAKT 4
Example: simulated Higgs event LHC 21 Sept. 2012 e– e+ Z H – Z e– e+ , H b b … Winni Mitaroff: ÖPG-FAKT ILC 5
The International Linear Collider (ILC) The ILC basic design is a worldwide consent since autumn 2004; Technology is based on superconducting RF cavities at 1. 3 GHz, average field gradient is 31. 5 MV/m in the first stages ≤ 500 Ge. V; The project is pursued by the “Global Design Effort” since 2005. Beam crossing angle 14 mrad; Only 1 experimental zone with 2 detectors operated in “push/pull”. 21 Sept. 2012 – 3 stages: collision energies 250 Ge. V (“Higgs Factory”), 500 Ge. V, eventually 1 Te. V (adjustable for scans in range 200 – 500 Ge. V); – Stability and precision of the beam energies to be below 0. 1 %; – Peak luminosity of ≈ 2× 1034 cm-2 s-1, with an integrated luminosity of 500 fb-1 to be achieved within the first 4 years of operation; – Electron polarization at least 80%, positron polarization an option; – Options: Z 0 factory (“Giga. Z”), e–e–, e–γ, γγ (“photon collider”). Winni Mitaroff: ÖPG-FAKT 6
The International Large Detector (ILD) Basic design parameters (ILD_00): • • • Central tracking detector: large TPC – excellent pattern recognition in a dense track environment, – proven technology; Silicon tracker: pixels and ss/ds strips – extended tracking coverage, – improved track momentum resolution; High-precision Si vertex detector – close (16 mm) to the beam interaction point, – best possible heavy flavour tagging; Fine-granularity calorimeters – particle flow (PFA) calorimetry is an asset, – provides necessary jet energy resolution; Solenoid magnetic field of 3. 5 T – upgradable to 4 T (for the ILC 1 Te. V stage); Almost 4π geometric acceptance – to the benefit of tracking & calorimetry. HEPHY Vienna is founding member of the ILD proto-collaboration. ILD is one of two ILC detector concepts “validated” by IDAG in April 2009. 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 7
The ILD silicon trackers Domain of HEPHY Vienna’s hardware contributions ! 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 8
ILD Forward Tracking Detector (FTD) 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 9
Forward track reconstruction in ILD New stand-alone software package Forward. Tracking: Stage 1: Cellular Automaton (CA), Stage 2: Kalman Filter (KF), Stage 3: Hopfield Neural Network (HNN). Embedded in ILD’s software framework Marlin. 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 10
Stage 1: the Cellular Automaton (CA) Fast semi-global track finding method: 21 Sept. 2012 Takes all hits into account simultaneously, but situation evolves based on local rules; Track segments interact with connected ones and are tested for compatibility. Winni Mitaroff: ÖPG-FAKT 11
Stage 2: the Kalman Filter (KF) Two main goals: Track parameter determination, Chi-squared probability gives feedback about the track quality; Chi-squared probability cut value = 0. 005; Algorithms called: Kal. Test + Kal. Det + Marlin. Trk. Note: an ultimate track fit by a KF + smoother will also be performed, after track search, on the final sample ! 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 12
Stage 3: Hopfield Neural Network (HNN) Ambiguity resolving: this is the last stage in forward track search; Tracks sharing hits are incompatible: overlap comes from combinatorics in reconstruction ⇒ ghosts and clones: 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 13
How does the HNN work ? Tracks are assigned a quality and an activation state, and they do dynamically interact; Compatible tracks amplify each other, whereas incompatible ones weaken each other; In order to prevent oscillation between states, updating is done asynchronously; A global extremum is searched for – in order to avoid falling into a local one, an annealing scheme is used (by assigning the system a “temperature” being cooled down). 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 14
Performance: efficiency Forward. Tracking: new forward tracking package, Silicon. Tracking: old package (still used in barrel), Track. Subset. Processor: combines results of both. 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 15
Performance: ghost rate Forward. Tracking: new forward tracking package, Silicon. Tracking: old package (still used in barrel), Track. Subset. Processor: combines results of both. 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 16
Performance: processing time Forward. Tracking: new forward tracking package, Silicon. Tracking: old package (still used in barrel). Background scaled conforming to the Lo. I with 500 Ge. V ! 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 17
Conclusions and Outlook A new software package for stand-alone track reconstruction in the forward region of ILD has been successfully developed and implemented; It shows superior performance w. r. t. the old ILD software (originally developed for the barrel); Our Forward. Tracking package is on board for benchmark processing for ILD’s “Detailed Baseline Design” (DBD) report, due by Dec. 2012; Our package will also be used for a modified ILD detector at the “Compact Linear Collider” (CLIC). 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 18
Backup Slides 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 19
Toy detector: true tracks 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 20
True hits (green) + background hits (red) 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 21
CA: building segments (“cells”) 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 22
CA: iteration #1 (red states) 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 23
CA: iteration #2 (orange states) 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 24
CA: iteration #3 (green states) 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 25
CA: iteration #4 (blue states) 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 26
CA: after clean-up of bad states 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 27
KF + HNN: final tracks found 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 28
Two machine studies: ILC and CLIC The CERN Linear Collider Detector Project: adapting the ILC detector concepts for the higher CLIC energies (CLIC_ILD, CLIC_Si. D), and using the software developed by the ILC collaborations for simulation and optimization studies. The decision ILC vs. CLIC will be based on “new physics” results from LHC. If it will be in favour of CLIC, the ILC detector collaborations will move. 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 29
Detector performance requirements of ILC / CLIC vs. those of LHC ○ Inner vertex layer ~ 3 - 6 times closer to IP ○ Vertex pixel size ~ 30 times smaller ○ Vertex detector layer ~ 30 times thinner Impact param resolution: Δd ≤ 5 μm + 10 μm / [ (p/Ge. V) x sin 3/2 θ ] ○ Material in the tracker ~ 30 times less ○ Track momentum resolution ~ 10 times better Momentum resolution: Δp / p 2 ≤ 5 x 10 -5 / Ge. V “barrel region”, Δp / p 2 ≤ 3 x 10 -5 / Ge. V “forward region” ○ Granularity of EM calorimeter ~ 200 times better Jet energy resolution: ΔE / E ≤ 0. 3 /√E o Forward hermeticity down to θ ≥ 5 - 10 mrad 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 30
Forward region of ILD_00 layout Very forward region: – 5. 00 < ϑ < 11. 50: only FTD measuremts. contributing, 900 ends. – Range of FTD 1 (2) starts where that FTDϑ 6= (7) Intermediate region: 36. 70 – 11. 50 < ϑ < 25. 50: complex mix of VTX + FTD + TPC, – FTD: only FTD 1 … 3, plus FTD 4 until ϑ < 16. 5 0, – 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) – 25. 50 9. 80 < ϑ < 36. 90: PR link to fwd. ECAL, useful in PFA. 16. 50 10 padrows → 11. 50 80 50 FTD 1 21 Sept. 2012 2 3 Pixel disks Winni Mitaroff: ÖPG-FAKT 4 5 6 7 Double-sided (stereo angle) strip disks 31
Spurrekonstruktion für FTD Jeder Detektor ist anders Hintergrund Paarbildung (Photonen) Pixels: aufintegrierte Events Strips: Ghost Hits Geschwindigkeit Efficiency und Ghost Rate Wart- und Lesbarkeit 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT FTD TPC 32
Processors in ILD’s framework “Marlin” 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 33
For compatibility cut off criteria are needed Cut offs rely on analysis of true tracks Efficiency vs. ghost rate & computing time 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 34
Conways Spiel des Lebens 21 Sept. 2012 Winni Mitaroff: ÖPG-FAKT 35
- Slides: 35