NICADD LCD Simulations

1
NICADD LCD Simulations
Current Process Overview Status Report Future
Plans
Guilherme Lima Jeremy McCormick
2
Current Simulation Pipeline
Event Generation
JAS, Pandora-Pythia
stdHEP
Detector Simulation
Gismo, LCDG4
SIO
SIO File Server
SIO File (remote)
Analysis
JAS
Note Each arrow represents a probable file
transfer.
3
Event Generation

• Single Particles
• diagnostic particles with JAS SingPartGen class
• helpful for debugging and results certification
• stdHEP MC particles, typically electrons, pions
  or muons
• E 2, 5, 10, 15, 20, 50 GeV fixed theta/phi or
  randomized within boundaries
• Complex Events
• pandora-pythia (TAULA, Pandora, Pythia)
• processes include E E-, Gamma Gamma, E- E-,
  E-() Gamma, Gamma E-, etc.
• E E- event generations with constrained decays
  and radiations
• produceOnly() and onlyDecay() used to constrain
  decays 20 different settings allStates,
  leptonsOnly, quarksOnly, c/b/eOnly, mu/tauOnly,
  WW/ZZOnly, etc.

4
Detector Simulation

• Gismo
• C program, maintained _at_ SLAC
• stdHEP input with SIO output
• uses lcdparms for detector geometry
• no longer in development, difficult to install,
  old and unsupported, physics model lacking slow
• Geant4
• same I/O format and detector geometry as Gismo
• Last main revision LCDG4 reL01 still in beta
  phase.
• reL01 fork for non-projective cells in HCal
• LCD-Mokka still Seg Faults.
• Mokka is available with variety of
  colliders-detectors, but none match LCDG4s input
  geometry.

5
Analysis
JAS

• only good option for SIO data analysis from
  LCDG4 and Gismo
• visualization with histograms, plots, graphs
• files remotely accessible by SIO-server
• wired display, MC Tree, MC Table and LCD Event
  allow graphical and tabular data display
• new version includes AIDA integration
• browseable database of SIO files _at_ nicadd

6
Simulation Requests
Argonne 95,000 events, LCDG4 (Gismo also ?
190,000) 4 single particles (pi, photon,
neutron, Z) 2 complex decays (rho ? pi gamma,
Sigma ? n pi) SDJan03 XML Dhiman 60,000 events
(30,000 2), LCDG4/Gismo 6 E E- events with
constrained decays event files _at_
rowlf/var/ftp/pub/lcdprj/stdhep-files/dhiman_ SD
Jan03 XML Vishnu 30,000 events in NP text file
format single chpi cells2,4,6,9,12
GeV5,10,15,20,30,50 SDMar01 NP XML
Completion
65,000
-
24,000
7
LCDG4 Status

• Non-projective Geometry
• additional tag to detector dtd specify length
  or width
• simulates rectangular cells, adjusted slightly
  upward for integers at each layer
• event-file format
• EVENT this_event_num
• ECAL num_hits
• ecal_hits
• HCAL num_hits
• hcal_hits
• hits layer, phi, theta, x, y, z, edep_cell,
  edep_absorber
• ToDo List
• SD modifications offset, possibly different
  shapes (hexagons), mod by layer
• NP cells in ecal
• SIO output for NP
• test GDML and STEP input

8
LCDG4-related Problems

• Extending Difficult
• limited by XML format
• Class structure is not well-designed.
• Sensitive detector and geometry code requires an
  overhaul.
• Codebase
• no support for non-projective readout in main
  code trunk
• lack of code documentation
• some recoding and misuse of Geant4 functionality
• coding style, including indentation and variable
  naming, is inconsistent
• Unused code is still compiled.
• Overall Problems
• Poor class design creates major redundencies and
  inefficiencies.
• lacks a coherent long-term design plan and path
• checkout and build procedure requires much
  manual configuration

9
Overall Issues

• General Problems
• manpower and expertise, many simultaneous
  demands and needs, unclear priorities
• lack of in-house design docs or Road Map
• un-automated simulation process means lots of
  time running and babysitting requests
• Geometry
• XML format limits customization only cylinders
  and barrels.
• idealized structures with limited utility beyond
  a certain LCD design phase
• SD is always projective.
• Simulator Zoo
• LCDG4 local capability only (nicadd) with no
  certification study.
• Mokka tested only on pre-defined
  detectorsnon-working Mokka geometry drivers for
  XML input LCD-Mokka from Ties has a Seg Fault in
  XML library.
• Non-projective fork of LCDG4 difficult to
  reintegrate into main trunk.
• Pipeline
• perl and bash scripts plus manual command-line
  work (Ex. - sftp, ftp or scp for files.)
• Job submission, event generation and detector
  simulation have little to no User Interface
  outside of the command-line.
• Analysis tools dont always accommodate geometry
  modifications.

10
Short-term Goals (2-4 weeks)

• Prepare correct, non-projective geometry output
  in Hcal and provide analysis tools for the data.
• Author report on NP findings including JAS
  histograms and plots.
• Event requests from Argonne and Dhiman for
  Cornell plus Vishnus NP.
• Debugged LCDG4 running on the Fermilab cluster
• LCDG4 and LCD-Mokka mutual results certification
  for identical detector geometries (requires
  working LCD-Mokka first).
• Complete CVS reorganization.
• Clean up LCDG4 code remove unused code
  document standardize on indentation and variable
  naming conventions.
• Study Geant4 toolkit to understand class
  structures, application scope and built-in
  capabilities.
• Extend non-projective geometry to EMCal.
• Experiment with STEP and GDML as geometry input
  formats.

11
Midterm Plans (4-6 Months)

• Gismo to G4
• Choose one option
• 1) overhaul LCDG4, especially geometry and SD
• 2) drop LCDG4 for another package (Mokka)
• 3) G4 design document and start a new project
  based on first principles
• Event Generation
• UI and server processing for event generation
  automation
• pandora_pythia Java library
• Analysis
• analysis classes for SIO files based on
  non-projective geometry formats
• General
• establish collaborative effort with CERN and
  SLAC for LCD simulation tool development with
  Geant4
• design better distribution and packaging system
  for NICADD LCD software and results files
• plan develop tools pipeline for LCD computing
  in all 3 phases

12
Long-Term Musings (1-2 years)

• Ideal G4 Package
• Flexible geometry input from GDML, STEP or SQL.
• XML-based application config file for
  flexibility
• utilise Geant4 eventgen and visualization
  capabilities
• flexibility in input and output formats and
  locations (remote would be nice)
• simulator engine that can be run remotely via
  server commands (XML-RPC/CORBA/SOAP)
• local computing cluster for pipeline automation
• Simulation Pipeline
• catalog and distribution system for software and
  simulation or event generation result files
• Event Generation, Detector Simulation and
  Analysis tools must be powerful, flexible and
  configurable.
• Would you build an LCD without blueprints? No.
  Similarly, producing good software usually
  requires a design phase.

13
Future System Architecture
14
References

• Pandora-Pythia
• http//www-sldnt.slac.stanford.edu/nld/new/Docs/
  Generators/PANDORA_PYTHIA.htm
• JAS http//www-sldnt.slac.stanford.edu/jas/
• Geant4 http//wwwasd.web.cern.ch/wwwasd/geant4/g
  eant4.html
• stdHEP http//cepa.fnal.gov/CPD/stdhep/
• SIO (now within LCIO) http//www-it.desy.de/phys
  ics/projects/simsoft/lcio/
• Mokka http//polype.in2p3.fr/geant4/tesla/www/mo
  kka/mokka.html
• Gismo http//www-sldnt.slac.stanford.edu/nld/new
  /Docs/Gismo/GismoInstall.htm
• LCDG4 http//nicadd.niu.edu/simulations/software
  /index.html
• fbsnghttp//www-isd.fnal.gov/fbsng/
• Mokka LCDG4 Geometries
• http//nicadd.niu.edu/jeremy/mokka-lcdg4.html
• LCD Simulation at NICADD
• http//nicadd.niu.edu/dhiman/talks/lcdg4_slac03
  _030521.pdf
• Mokka/LCDG4 Comparison http//www-conf.slac.stan
  ford.edu/lcsimworkshop/talks/mokka.pdf
• GDML http//gdml.web.cern.ch/gdml/