Status of Si DIowa PFA by Usha Mallik
Status of Si. D/Iowa PFA by Usha Mallik The University of Iowa for Ron Cassell, Mat Charles, TJ Kim, Christoph Pahl Linear Collider Workshop of the Americas, Albuquerque, NM, Sept 29 -Oct 3, 2009 1
The Detector (Si. D 02) 2
The Particle Flow Algorithm Goal: To obtain dijet mass resolution M/M < 3 -4% (Z width) E(cm)/E(cm) < 3 -4% for e+e- qq (q=u, d, s) Resolution for PFA : = EM neu. had conf Attempt to minimize conf in the PFA In calorimetry /E 1/ E… … but in a PFA the confusion increases with E At high energies leakage is also important Generally PFA ~ between E and E 3
Overview at LOI (April 2009) e+e- qq (q=u, d, s) at Ecm = 100 Ge. V qq 100 e+e- Z (qq) Z ( ) at Ecm = 500 Ge. V ZZ rms 90 For qqbar events, E 1 = E 2 = Ecm/2 and E 1 = E 2 = Ecm / 2 M 212 = 2 E 1 E 2(1 - cos 12) and M 12/ M 12 = Ecm/Ecm 4
In December 2008 5
Energy dependence 6
Leakage study at 500 Ge. V and 1 Te. V Marty Breidenbach helped produce a Si. D 02 -like detector with 6 HCAL Ron Cassell generated the events and produced the files for 1 Te. V, 500 Ge. V, 200 Ge. V • Change Steel for Cu for absorber • Increase to 54 layers from 40 layers in HCAL • 1. 7 more material in HCAL • No gap between HCAL and Muon endcap (instead of 10 cm) Compare sid 02 with sid 02 -Cu at various energies by looking at: • # hits in Muon detector (indicates punch through, a measure of leakage) • Energy resolution 7
Punch-through muon hits Si. D 02 -Cu Si. D 02 8
Resolution study (Si. D 02 -Cu comparison) real tracking Si. D 02 -Cu Si. D 02 9
Lessons learned • Leakage is present in substantial amount • Not the whole story at all • Confusion clearly important at 500 Ge. V, dominant at 1 Te. V • Back to the drawing board • Anatomy of the events 10
p(left) = 105 Ge. V, p(right) = 97 Ge. V Angle < 1 degree, connected `seeds’ p = 7. 7 Ge. V, blue piece picked up by cone (2 nd pass) Refined. Cheat. Cluster e+e- qq at 500 Ge. V Refined. Cluster - sharedhits reconstructed 11
has a low energy 12 Ge. V neutral hadron and several photons present in the ECAL; interaction of charged hadron p (orange) = 119 Ge. V, E/p match, enough hits (green) = 17 Ge. V reconstructed Refined. Cheat. Cluster e+e- qq at 500 Ge. V Refined. Cluster - sharedhits 12
Refined. Cluster - sharedhits In 97 Ge. V track-cluster `cone’ gives high score to the stub and is connected; 105 Ge. V can not access the stub reconstructed ECAL Monte Carlo Backscatter p(orange) = 97 Ge. V p(blue) = 105 Ge. V e+e- qq at 500 Ge. V 13
e+e- qq at 500 Ge. V p (green) = 40. 8 Ge. V, p (blue) = 2. 7 Ge. V Higher score by cone to green cluster seed, blue has implied cluster connected to seed 2. 7 Ge. V reconstructed ECAL 14
Algorithm modifications/additions • Cone algorithm is too aggressive! • Mostly the cone algorithm picks up MIP-like pieces • Use reconstructed shower information (not only stubs) • Use directional information • Low-momentum tracks steal pieces from high-momentum showers • Iteration starts with lowest momentum track and assigns clusters • Keep clusters available for others tracks even if assigned • Use geometry information (proximity) to adjudicate cluster assignments between tracks • Misc: • Can Barrel Muon be used as a backing calorimeter, for merged high p tracks ? • Backscattering ? 15
Conclusion • Much better understanding of weak points of algorithm • Hitting our stride in aftermath of LOI • Christoph Pahl joined the effort, can now afford an FTE • Clear path to improve pattern recognition • Lots of work to do! 16
Leakage study (Si. D 02 -Cu comparison) cheat tracking 17
- Slides: 17