Many thanks to Martin Heck Belle & Neural Networks To Physics From Tracks Florian Bernlochner florian.bernlochner@cern.ch
Setting the Scene
The Precision Frontier of Particle Physics ` γ New Physics γ New Physics p . . . p New Physics New Physics Energy Frontier Ansatz Intensity Frontier Ansatz e.g. LHC, Tevatron e.g. BaBar, CLEO, Belle Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 3
The Precision Frontier of Particle Physics ` γ New Physics γ New Physics p . . . p New Physics New Physics Energy Frontier Ansatz Intensity Frontier Ansatz e.g. LHC, Tevatron e.g. BaBar, CLEO, Belle sures their decays into light flavours Large degree of complementarity Talks of Uli, Plot: Andreas Crivellin Direct searches Teppei and Wouter Belle At B-Factories with B-Mesons Flavour observables Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 4
The Experiment ‘ → el le ← ’ ‘B’ breaks the symmetry In elle, hence Belle :-) Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 5
The Experiment: Collision energy √ s = 10 . 58 GeV e + e − Υ (4 S ) Υ (1 S ) = h b ¯ b i h b ¯ b i Υ (4 S ) = h b ¯ b i ¯ q q q q quark-antiquark-pair h ¯ h b ¯ q i bq i produced from vacuum Fragmentation into two bound states: B-Mesons Plot: CLEO B ! � threshold Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 6
The Experiment: B-Meson Decays … B-Meson r h ¯ bq i m light q τ B ≈ 1 . 5 × 10 − 12 s Quark 7 inclusive charged particle b multiplicity: ~5.4 per B-Meson heavy or ~ 11 per B-Meson pair anti-b-Quark Was einfach zu messen What you can measure ist, kann nicht Murphy’s Law of without a problem, you ausgerechnet werden, cannot calculate. Flavour Physics was einfach zu rechnen What you can calculate ist, ist schwierig zu easily, you cannot measure messen. Stolen from Martin Heck Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 7 4
The Experiment: Asymmetric Beam Energies Asymmetric Beam energies: allow to directly observe CPV in B-system KEKB PEP-II ! βγ = 0 . 28 Belle II √ s = 10 . 58 GeV B 0 ∆ z ≈ c β γ τ B ≈ 126 µ m e + Υ (4 S ) e − Υ (4 S ) E ( e + ) = 4 GeV E ( e − ) = 7 GeV B 0 B -meson lifetime h b ¯ h b ¯ h ¯ h b ¯ b i b i bd i d i E ( e + ) = 3 . 1 GeV BaBar E ( e − ) = 9 GeV βγ = 0 . 56 E ( e + ) = 3 . 5 GeV E ( e − ) = 8 GeV Belle βγ = 0 . 42 Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 8
What is the di ff erence between Belle and Belle II? 50:1
What is the di ff erence between Belle and Belle II? 50:1 Expected data set increase and ~ increase in inst. Luminosity 50 1 = LHCb Upgrade BaBar and Belle = BaBar and Belle Belle II = LHCb today CLEO As significant for us as the energy increase from 7/8 TeV to 13 TeV at the LHC
KEKB → SuperKEKB L (without crab) present KEKB Nano-beam scheme: Half crossing angle: � � Squeeze vertical beam spot to 50 nm Hourglass condition: 1 � m � y* >~ L= � x / � � SuperKEKB 100 � m 22 mrad 5mm crossing angle ~50nm 1 � m 100 � m 5mm 83 mrad crossing angle 13 15 12 Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 11
final focussing magnets April 11, 2017
Belle → Belle II Electrons (7 GeV) Positrons (4 GeV) Increased luminosity comes at a price: much larger beam backgrounds Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 13
Belle → Belle II Electromagnetic Calorimeter: KL and Muon detection system Thallium activated Caesium Iodide scintillation crystals RPC based Particle identification Vertex detectors Time-of-propagation counter, Aerogel Cherenkov ring detector 2 layers of Pixel (DEPFET) + 4 layers of strips (DSSD) KL & Muon Central drift chamber: Cherenkov / TOP Gas mixture of Helium and Ethan (C 2 H 6 ) Drift chamber Vertex Calorimeter Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 14
Belle → Belle II Electromagnetic Calorimeter: KL and Muon detection system Thallium activated Caesium Iodide scintillation crystals RPC based Particle identification Vertex detectors Time-of-propagation counter, Aerogel Cherenkov ring detector 2 layers of Pixel (DEPFET) + 4 layers of strips (DSSD) Central drift chamber: Gas mixture of Helium and Ethan (C 2 H 6 ) Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 15
Tracks
Central drift chamber: 14336 sense wires Gas mixture of Helium and Ethan (C 2 H 6 ) 56 layers Aim: Convert unmarked Hits … Cells of sense wires and field wires r drift ∝ t drift Illustrations: Oliver Frost, Sarah Neuhaus
Central drift chamber: 14336 sense wires Gas mixture of Helium and Ethan (C 2 H 6 ) 56 layers Beam background e + e − … into charged particle trajectories Physics collision Illustrations: Oliver Frost, Sarah Neuhaus
Original question by E. Paoloni
VXD Online and Offline Tracking Vertex detector (VXD) Sector-Map To reduce combinatorics: Group hits into sectors that will contain Einteilung des SVD in Sektoren; a likely neighbouring hit Sektoren durch die “vernünftige” Spur in großer Simulation geht, sind “Freunde”; Done in 3D Used in track reconstruction 37 Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 20
z-Vertex Trigger • Typical B-Meson trigger requires 3 tracks (at least one in each hemisphere) • A lot of interesting low-multiplicity events are missed 3 tracks or more e − e + − → τ − τ + 2 track requirement 40 % 20 % > 2 2 > 2 N backward 1 2 1 0 0 N forward Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 21
z-Vertex Trigger • Typical B-Meson trigger requires 3 tracks (at least one in each hemisphere) • A lot of interesting low-multiplicity events are missed 3 tracks or more e − e + − → τ − τ + 2 track requirement 40 % 20 % > 2 2 > 2 N backward 1 2 1 0 0 N forward Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 22
Z distribution # of events / 5 mm Interesting 1000 background Belle physics s c 800 collisions i s y h p 600 e + e − 400 z 200 0 -40 -30 -20 -10 0 10 20 30 40 z (cm) Use FPGA based L1 trigger with neural network to “learn” z direction from drift chamber input (TS) Track Segment • position and drift time of TS priority wires Input: Output: ut z estimate ≈ 15 mm • 2D track estimates ( p T , ϕ ) Plots: Sebastian Skambraks
Neural Networks
Neural Networks and Lepton colliders • Fairly clean environment (even with beam background) and no pile-up Simuliertes Beispiel- Ereignis Elektron- Positron- Beschleuniger 6 http://www.nikhef.nl/~i93/img/Event6_top.png • Allows use of multivariate methods to implement a “Full Event Interpretation” Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 25
Vollständige Ereignisinterpretation Rekombinations-Effizienz O(1%) nach … ● Rekonstruktions-Effizienz ○ “angemessener” Reinheit ○ … erfordert bereits Betrachtung von Allows one to reconstruct the missing ~ p B tag = − ~ > 10,000 Zerfallsketten p B sig Four-momentum on the signal side 52
Vollständige Ereignisinterpretation Rekombinations-Effizienz O(1%) nach … ● Rekonstruktions-Effizienz ○ “angemessener” Reinheit ○ … erfordert bereits Betrachtung von > 10,000 Zerfallsketten 52 B 0 , B + Reconstruct O( 1000-10000 ) of hadronic and semileptonic modes, achieves an e ffi ciency of about O(1%)
Physics
The big flavour questions and one anomaly B-Factory measurement candy bowl η 0.5 m / m ∆ ∆ s d sin(2 ) β Current CKM fit 0.4 contours hold 68.3% 0.3 0.2 Exclusive |V / V | cb ub Inclusive |V / V | ub cb Direct γ 0.1 Future and |V / V | γ ub cb at 1 σ , 3 σ , 5 σ 0 0 0.1 0.2 0.3 0.4 0.5 ρ Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 29
The big flavour questions and one anomaly τ − − W − V cb ν ¯ ¯ V qb ν τ b B-Factory measurement candy bowl u q c u R(D*) BaBar, PRL109,101802(2012) 0.5 η 2 0.5 = 1.0 contours ∆ χ m / m ∆ ∆ Belle, PRD92,072014(2015) s d LHCb, PRL115,111803(2015) sin(2 ) β SM Predictions Belle, PRD94,072007(2016) 0.45 Current CKM fit Belle, PRL118,211801(2017) R(D)=0.300(8) HPQCD (2015) 0.4 LHCb, FPCP2017 R(D)=0.299(11) FNAL/MILC (2015) Average R(D*)=0.252(3) S. Fajfer et al. (2012) 0.4 contours hold 68.3% 0.3 0.35 4 σ 0.3 2 σ 0.2 Exclusive |V / V | cb ub Inclusive |V / V | ub cb 0.25 Direct HFLAV γ 0.1 FPCP 2017 Future and |V / V | γ 0.2 ub cb 2 P( χ ) = 71.6% at 1 σ , 3 σ , 5 σ 0.2 0.3 0.4 0.5 0.6 0 R(D) 0 0.1 0.2 0.3 0.4 0.5 ρ Florian Bernlochner Belle II @ KIT : From Tracks to Neural Networks to Physics 30
Recommend
More recommend
