1. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 1/25 Data Challenge II Kickoff Workshop Simulation and Reconstruction Overview Tracy Usher (representing a lot of people – mis- statements are mine, credit is all theirs…)

2. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 2/25 Goals of this Talk Overview of Simulation/Reconstruction: –Introduce the simulation and reconstruction to those unfamiliar with it Highlight changes/improvements since DC-1 –Attempting to focus on those simulation and reconstruction features which are new for DC-2 Though most recent changes are foremost in memory… Provide background for next talk: –Hopefully, this serves as a useful introduction to Bill Atwood’s talk which will describe the performance of the reconstruction, again highlighting what is new from DC-1

3. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 3/25 Outline Framework –Gaudi environment Simulation –Overview of the flow of simulation –A look at the LAT Geometry as used in simulation and reconstruction –Simulating events with Geant4 –Digitization Reconstruction –Overview and flow of the reconstruction –Calorimeter Reconstruction Clustering/Moments Analysis Energy correction methods MIP Finder –Tracker Reconstruction The basic steps –ACD Recon –“Analysis” and Output

4. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 4/25 Sim/Recon Framework: Gaudi Gaudi –An experiment independent framework providing the necessary interfaces and services for HEP event data processing applications –Used by several LHC experiments Sim/Recon interacts with Gaudi through: –Gaudi Algorithms Algorithms define the reconstruction tasks and control the sequence in which they are performed –Gaudi Services Provide “global” information across events (e.g. geometry) –Gaudi Tools Implementation of specific reconstruction tasks Interact with Algorithms through abstract interfaces to allow interchangeability –Gaudi Transient Data Store (TDS) Globally accessible repository containing the results of each reconstruction task Provides communication between the stages of reconstruction Gaudi allows us to configure event processing “on the fly” –Job Options files define event processing and Input/Output for a given job

5. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 5/25 Simulation Overview Goal of simulation is to produce “raw data”, representing the response of the LAT to a given event, which can be fed to the reconstruction Logically breaks into four steps: –Event generation – “source particle” –Simulation of LAT response to the generated event –Digitization: the conversion of simulated LAT response to “raw data” –Simulation of Trigger

6. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 6/25 Flow of Simulation Gaudi Algorithm Sructure for DC-2 New Event Event Generator (FluxAlg) (PointingAlg) Event Simulation (G4Generator) Output Data to TDS (or PDS) LAT Geometry (G4GeometrySvc) (also accessed by the reconstruction) Sources, Orbit Orientation, Etc. (Many improvements and additions from DC1 - See Toby’s Talk later) Digitization (ACDDigiAlg) (CalDigiAlg) (TkrDigiAlg) Trigger / Filter (TriggerAlg) (OnboardFilter) Geant4 Event/Detector Simulation Convert simulated detector response to “raw data” (calibration effects enter here) Output of digis given to Trigger and Onboard Filter Events MUST pass Trigger in order to get reconstructed Onboard Filter results recorded, events not rejected

7. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 7/25 LAT Geometry Simulation and Reconstruction

8. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 8/25 LAT Geometry Simulation and Reconstruction

9. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 9/25 LAT Geometry Simulation and Reconstruction Main Point: –Most detailed LAT geometry to date including all features which would significantly impact DC-2 results From active devices To surrounding support structures Including electronics, wrapping, … etc. Large number of updates to geometry since DC-1 –Primarily CAL and ACD –For complete list see the xmlGeoDbs change log at: http://www-glast.stanford.edu/cgi-bin/viewcvs/xmlGeoDbs/ChangeLog?rev=1.413&view=markup Still working on improvements for DC-3: –e.g. Curved ACD tiles One other important point: –The reconstruction, through the propagator, uses the same geometry as the simulation

10. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 10/25 Overview of Geant4 Simulation Geant4 determines the evolution of the event –G4 given “source” which is propagated through the LAT –G4 output to TDS: McParticles – particles produced during trackinig McPositionHits – positions and energy deposited in Silicon and ACD McIntegratingHits – energy deposited in “sub-cubes” of CsI crystals Passed on to Digitization stage For DC-2 we are using G4 v6.2.2 –“Vanilla” G4 for everything… –However, we continue to experience “problems” with the multiple scattering Native G4 v6.2.2 Multiple Scattering “too small” Continue to use formalism circa G4 v3 –Have recently upgraded to G4 v7 where this problem is claimed to be fixed (finally)… Incident Gamma Conversion Point Incident Gamma Conversion Point Propagation of primary e + and e - through LAT McPositionHits in Active Silicon of Tracker Incident Gamma Conversion Point Propagation of primary e + and e - through LAT McPositionHits in Active Silicon of Tracker Production and Tracking of Secondaries McIntegratngHits produced in the Calorimeter Crystals

11. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 11/25 Digitization Digitization: Convert Monte Carlo information to detector response –Simulates the response of the detector to the simulated event TKR –Input: McPositionHits – position/energy in active silicon –Output: hit strips and ToT (Time over Threshold) –New for DC-2: Implementation of “bad” (hot/dead) strips Implementation of cable controller buffer limit –Truncation of data in “big” events (64 strips/plane end, 128 strips/cable) Improved ToT response CAL –Input: McIntegratingHits – Energy deposited in cubes subdividing the Calorimeter crystals –Output: ADC counts for both diodes at both ends –New for DC-2 (but mostly in well over 1 year ago, past time horizon…): New code throughout –Improved/updated xTal response –Calibrations (DC-2 uses “ideal” though) ACD –Input: McPositionHits – position/energy in ACD Tiles/ribbons –Output: PMT ADC counts per end Many updates / improvements from DC-1

12. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 12/25 Reconstruction Overview Goal of the reconstruction: –Determine the incident gamma direction and energy –Provide tools for suppressing background Follow a two pass approach for Cal and Tracker –Best energy from Calorimeter needs tracking information –Best tracks from Tracker needs best energy reconstruction –Solve by breaking into two passes Background suppression –ACD Recon –MIP Finder Analysis step at end of reconstruction –Create output ntuple for detailed performance studies –Event Classification –Create FT1 output for analysis NOTE: –Will not discuss Calibrations Critical piece of “real” reconstruction But (mostly) not using calibrations for DC2

13. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 13/25 Flow of Reconstruction Gaudi Algorithm Sructure: Envisioned for DC-2 “Raw data” from TDS CalRecon Pass I TkrRecon Pass I CalRecon Pass II TkrRecon Pass II AcdRecon Event “Analysis” Root tree/ntuple output to PDS CalXtalRecAlg CalClusterAlg CalMipFinderAlg CalEventEnergyAlg (parametric only) TkrClusterAlg TkrFilterAlg (inactive for DC2) TkrFindAlg TkrTrackFitAlg TkrVertexAlg CalEventEnergyAlg TkrTrackFitAlg TkrVertexAlg AcdReconAlg meritAlg (output ntuple) Key Existing Algorithms New Algorithms Not “Recon” ClassifyAlg (Classification Analysis) FT1Alg (FT1 tuple)

14. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 14/25 Calorimeter Recon CalRecon extensively restructured for DC-2 –Break into logical pieces –Follow TkrRecon organization paradigm Consists of a number of tasks handled by Gaudi Algorithms: –CalXtalRecAlg (Recon First Pass Only) Takes in raw data (CalDigis – diode readouts) and produces calibrated output data (CalXtalRecData – energy, position) for input to rest of CalRecon. –CalClusterAlg (Recon First Pass Only) Associates CalXtalRecData objects into a “Cal Cluster” Performs a “Moments Analysis” to determine principal axis, etc. –CalMipFinderAlg (Recon First Pass Only) Search for MIP signature(s) in CalXtalRecData –CalEventEnergyAlg (Recon First and Second Pass) Run various energy correction algorithms

15. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 15/25 Calorimeter Recon Clustering and Moments Analysis CalXtalRecAlg: Convert diode ADC counts to energy, position in crystals: - Select range/diode - Pedestal subtraction - Correct for Light attenuation - etc. Use “ideal” response – no calibrations Much work here – all “new” code from DC-1 (but nearly 2 years ago now!) 1 GeV Gamma Energy Flow Monte Carlo Version of event “McIntegratingHits” “McPositionHits” 1 GeV Gamma Energy Flow CalXtalRecAlg Version of event “McPositionHits” “hit” crystals 1 GeV Gamma Energy Flow CalXtalRecAlg Version of event “hit” crystals Red Box size proportional to total energy deposited, longitudinal position from readout asymmetry at each end of crystal 1 GeV Gamma Energy Flow CalClusterAlg Version of event Energy centroid for each layer Cluster Centroid from Moments Analysis Cluster Axis from Moments Analysis CalClusterAlg: Associate hit crystals to form a single cluster Determine cluster centroid and axis via a Moments Analysis Provide “raw” calorimeter energy - simple sum of crystal energies New “code” from DC-1 New “TDS” output from DC-1 Much work here as well

16. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 16/25 Calorimeter Reconstruction Energy Correction Algorithms The Problem: –Large phase space of Glast Energy range from 20 MeV to 300 GeV Large FoV: ~2.5 str –Tracker above Up to 1.5 radiation lengths BEFORE the Cal –Gaps between towers Crystals ~187 mm, gaps ~35 mm –Cal is “only” 8.6 radiation lengths deep –etc. Corrections: –Energy Deposit in the Calorimeter e.g. Wallet Card Shower Profile correction –a and b are parameters »b scales the radiation length »a sets the location of the energy centroid t –Detector / Geometry (see next slide) Huge Amount of Effort in this area –Cannot do justice to this work in these few slides…

17. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 17/25 2 GeV Photon Cal E = 950 MeV Conversion Bottom of Cal Calorimeter Reconstruction Energy Correction Algorithms Correct for: –Leakage out the back Cal is 8.6 rl deep Correct for: –Leakage out the back Cal is 8.6 rl deep –Gaps between towers 2 GeV photon Cal E = 980 GeV conversion Bottom of Cal 2 GeV Photon Cal E = 200 MeV Conversion Gap between towers Correct for: –Leakage out the back Cal is 8.6 rl deep –Gaps between towers –Edge of Calorimeter 2 GeV Photon Cal E = 200 MeV Edge of Cal 100 MeV Photon Cal E = 60 MeV Conversion Positron stops in Tracker Electron ranges in Calorimeter Correct for: –Leakage out the back Cal is 8.6 rl deep –Gaps between towers –Edge of Calorimeter –Loss in the Tracker

18. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 18/25 Calorimeter Reconstruction Energy Correction Algorithms - continued Maximum Likelihood Estimator –Replaces the previous (DC-1) Last Layer Leakage method with Maximum Likelihood estimator –In addition, method extended to the low energy range to correct energy taking into account energy lost in tracker Extends method to provide correction in range from 50 MeV to 300 GeV –Also, angular acceptance extended from around 30 o to 50 o –See Pol D’Avezac’s note: http://polywww.in2p3.fr/glast/CalLikelihood.pdf http://polywww.in2p3.fr/glast/CalLikelihood.pdf Full Profile Shower Fit –Extension of previous 1D method (not used in DC-1) to a full 3D shower profile fit –Finds energy correction for events above ~1 GeV –Angular acceptance extended to 90 o –See Philippe Bruel’s note: http://polywww.in2p3.fr/~bruel/CalFullProfile.ps http://polywww.in2p3.fr/~bruel/CalFullProfile.ps Parametric Method –Parameterized shower profile model –Provides corrected energy in range 20 MeV to 300 GeV and over the full field of view –See Bill Atwood’s talk: http://www-glast.slac.stanford.edu/software/AnaGroup/Atwood-EnergyRecon-2May2005.ppt http://www-glast.slac.stanford.edu/software/AnaGroup/Atwood-EnergyRecon-2May2005.ppt The following talk will compare these methods

19. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 19/25 Calorimeter Recon MIP Finder The MIP finder analyzes the calorimeter data looking for a Minimum Ionizing Particle signature –Goal is to eliminate a large source of charged particle backgrounds which enter from the sides and bottom of the Calorimeter The Method in two bullets or less: –Select CalXtalRecData objects in Calorimeter Take those crystals with energy deposits consistent with a MIP energy deposit Look for a straight line trajectory through these crystals Does Full fit of MIP track – returns a number of powerful background rejection variables –Can then reject as background those events with clear MIP tracks Completely new for DC-2 See: http://www-glast.slac.stanford.edu/software/AnaGroup/Piron_ana2005-08-08.pdf

20. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 20/25 Current Status: –Blue Lines - Protons –Red Lines - Gammas –Selection Cuts: Bottom right shows efficiency to identify MIPs (or misidentify Gammas) (Plots from Sylvain Guiriec) Calorimeter Recon MIP Finder http://www-glast.slac.stanford.edu/software/DataChallenges/DC2/MarchWorkshop/Presentations/Ana05-08-29-1.ppt Pre-cuts : TkrNumTracks > 0 && CalCsIRLn > 4 && CalEnergyRaw > 5 && CalTotalCorr 0 Cuts : CalMipErm / CalMipArcLen < 1 && CalMipChi2 < 30 && CalMipEcor between 8 and 25 MeV && CalMipEcorRms < 5 && dirErr < 0.7

21. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 21/25 1 GeV Gamma Conversion point e + and e - trajectories 1 GeV Gamma Conversion point Pattern Track trajectories 1 GeV Gamma Conversion point Fit Track Trajectories 1 GeV Gamma Conversion point Vector resulting from vertexing of fit track pair for this event 1 GeV Gamma Conversion point Vector resulting from vertexing of fit track pair for this event The Illustrated Tracker Reconstruction Tkr Clustering: Associate adjacent hit strips to form strip clusters (corrected for hot and dead strips) Pattern Recognition: Associate strips to form candidate tracks (Combinatoric pattern recognition using Kalman Filter to accept/reject candidate hits) Track Fit: Take candidate tracks and track energy estimate and fit using a Kalman Filter Vertexing: Combine fit tracks to form common vertex point - If one track, vertex is head of track - If two tracks don’t “combine” well, keep as separate vertices

22. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 22/25 Tracker Reconstruction What is new for DC-2? –Pretty much everything… Algorithm Structure remains same –Two pass reconstruction format same as for DC-1 But… Underlying code extensively reworked –Combo Pat Rec Combinatoric loops restructed and optimized to improve performance Uses same KF as track fit for hit selection Much improved performance –Track Fit New Kalman Filter Track fit parameters (energy loss, particle type, etc.) controlled from Job Options file –Vertexing New scheme for deciding whether to pair tracks or not Completely new TDS structure

23. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 23/25 ACD Reconstruction “Active Distance” outside tile boundary no tile hit inside tile boundary [cm] 1 GeV Muon Reconstructed Track Struck ACD Tile Projection of Track back to nearest ACD Tile ACD: Primary Background Suppression Tool Project Track to plane of struck tile Calculate distance to nearest edge 3D calculation if outside tile 2D calculation if inside tile Sign Positive if track projection inside the tile Negative if track projection outside the tile Reject if inside the tile For complete details see Eric Charle’s IA Workshop talk on Monday February 27

24. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 24/25 Post Reconstruction Processing The Final Steps meritAlg –Controls the creation of the merit output ntuple Used, for example, in various event classification analyses ClassifyAlg –Runs the classification analysis: Energy Selection PSF Tail suppression Background Rejection –For DC-2 this is the Insightful Miner Analysis provided by Bill Atwood (see next talk) Future versions will allow alternate analyses as well (e.g. Toby Burnett’s Random Forests analysis) FT1Alg –Produce the Fits Level 1 output Will be discussed in more detail later

25. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 25/25 Main Points for DC-2 Very Detailed LAT Geometry –Really getting to the very fine details at this point Most of what is missing is already in the pipeline for DC-3 Very Detailed Geant4 simulation –Based on G4 v6.2.2 Multi-pass Reconstruction Algorithms –New algorithms to improve energy resolution and background suppression –Many improvements to existing algorithms –Resulting in improved energy and psf resolution and background rejection Enhanced post reconstruction processing –Perform Event Classification as part of reconstruction

26. Science Analysis Software DC-2 Kickoff Workshop March 1-3, 2005 Tracy Usher 26/25 Summary Since DC-1 TkrRecon, CalRecon and AcdRecon have undergone extensive restructuring with many improvements… –Substantial updates to existing code –Major overhaul of the algorithm output –Implementation of many new tools along full reconstruction chain Resulting in major improvements –Background rejection –PSF Tail suppression –Energy Resolution –Reconstruction efficiency –Event classification See next talk for the results!