The CMS Simulation Software

1
The CMS Simulation Software

• Julia Yarba, Fermilab
• (for the CMS Collaboration)
• IEEE/NSS 2006
• Oct 30 Nov 2 2006
• San Diego, California (USA)

2
Overview

• Though in operation for a number of years, its a
  live system goals, requirements, tools evolve
  throughout the lifetime of the experiment
• Based on Geant4 (7.1 in transition to 8.1)
• physics processes electro-magnetic and hadronic
  interactions
• tools for detector geometry and sensitive element
  response
• interfaces for tuning and monitoring particle
  tracking
• New CMS offline framework and Event Data Model
• Manages application control at run time
• Relies on the concept of event processing module
  (EDProducer)
• Interface to common tools (generators, mag.field,
  MC truth)
• Ensures provenance tracking and event immutability

Julia Yarba, FNAL IEEE/NSS 2006
3
Simulation Software CMS Solution
CMSSW the new framework - ties pieces together
Application control
Object browsing
Event generation PYTHIA, Particle Gun,
User Actions
HepMC
Visualization
Mixing Module
User Actions
User Actions
User Actions
Simulation Geant4 (FAMOSGFlash)
SimHit Data File (Hit level information, linked
to MC truth)
Validation Suite
Geometry Detector Description Database (XML
C) Sensitive Volumes Interface
Digitization subsystem-specific packages
Digi data file (Data-like, linked to MC truth)
Misalignment Simulation (under development)
ROOT based persistency format
Reconstruction
Julia Yarba, FNAL IEEE/NSS 2006
4
Interface to Geant4

• Core application framework-based Event Data
  Producer customized RunManager as interface
  between Geant4 and CMS Event Data Model
• Geometry record is available to either simulation
  or reconstruction via the framework EventSetup
• uses XML-based Detector Description machinery,
  configurable at run time via a hierarchy of XML
  files converts DD solids and materials to Geant4
  counterparts
• Sensitive detectors associated with geometrical
  volumes through XML configuration files at run
  time
• Magnetic field based on dedicated geometry of
  magnetic volumes provided by independent
  subsystem via EventSetup field selection,
  propagation tuning configurable at run time

Julia Yarba, FNAL IEEE/NSS 2006
5
Interface to Geant4 (cont.)

• Variety of lists (LHEP, QGSP/QGSP_EMV, QGSC,
  FTFP,) for modeling physics processes run-time
  selection of physics list and production cuts,
  activation/tailoring of individual processes
• Variety of Physics event generators (particle
  guns, Pythia, Herwig,) generator information
  stored in HepMC format and interfaced to G4Event
• User actions allow access to Geant4 objects at
  any stage (run, event, track, step) used for
  tuning, diagnostics, custom bookkeeping
• Monte Carlo truth record with decay/interaction
  history of the generators particles and selected
  tracks from Geant4 simulation

Julia Yarba, FNAL IEEE/NSS 2006
6
The CMS Detector
Different subsystems have different simulation
requirements ? Region based optimization

• 22 m long, 15 m in diameter
• Over a million geometrical volumes
• Many complex shapes

Julia Yarba, FNAL IEEE/NSS 2006
7
Central Detector Subsystems
Simulated tracks and hits in CMS Tracker

• Tracker (talk by F.Ambroglini)
• Critical region, due to its own physics
• significance, and the effect on the overall
• simulation accuracy
• Detailed geometry description of active and
• passive volumes
• Extensive validation single particle,
  minbias,
• physics channels

Simulated tracks and energy deposition in CMS
ECAL

• Electromagnetic calorimeter (talk by
  F.Cossutti)
• Resolution dominated by effects not included in
• shower simulation
• Excellent agreement between test beam and
• Geant4 simulation
• Highly sensitive to changes in radiation,
  showering
• in tracker

Julia Yarba, FNAL IEEE/NSS 2006
8
Central Detector Subsystems (cont.)

• Hadronic calorimeter
• Test beam data available (2002-2004,
• various HCAL modules, preceded
• by ECAL prototype e, p, and µ beams)
• Agreement with simulation within large
• systematic uncertainties of the data
• Sensitive to accuracy of the Geant4
• modeling of hadronic showers results
• depend on the choice of physics list

HCAL response linearity vs energy
H(150)?ZZ?4µ

• Muon system
• Comparison G3/G4 with single ? 10-104GeV
• Improvements in Geant4 µ brem., ee- prod.,
• µ-nuclear interactions, multiple scattering
• Test beam setup
• 2 chambers with or without iron slab, to study
• effect of ? showers in passive material

Julia Yarba, FNAL IEEE/NSS 2006
9
Forward Subsystems

• Essential for diffractive and Heavy Ion programs
• CASTOR calorimeter
• Totem telescopes
• Zero Degree Calorimeter
• Simulation to study energy
• resolution, leakage analysis
• of test beam data (Nov.05)
• Other components
• Roman Pots
• Luminosity Monitor

Totem Telescopes at 7.5 lt z lt 13.5m
Castor Calorimeter at 14.37 m (5.3 lthlt 6.7)
Julia Yarba, FNAL IEEE/NSS 2006
10
Event Mixing and Digitization

• In-time pileup
• LHC will produce 3 (low lum.) or 25 (high
  lum.)
• minbias interactions/crossing, on top of the
  trigger event
• Out-of-time pileup
• Coming from bunch crossings before/after the
  trigger event
• Pileup events simulated separately from the
  physics events merge of simulation outputs at
  hit level (reuse)
• Performed by a dedicated module, in a separate
  step
• Followed by simulation of the electronic readouts
  (Digis)
• Dedicated Digi module for each subsystem
  (separate steps)

Julia Yarba, FNAL IEEE/NSS 2006
11
Software Validation (talk by X.Huang)

• Validation of physics processes modeling, via
  dedicated test beam setup simulation compared vs
  test beam data feedback to Geant4

• Software Validation Suite, to ensure simulation
  (or other) software reliability,
  release-to-release, when changing Geant4 version,
  etc

• Proved very useful, in particular in recent
  tests of Geant4.8.1-based version

Julia Yarba, FNAL IEEE/NSS 2006
12
Performance and Production

• With the new/upgraded software over 60 millions
  events simulated by the production team since
  July 2006 (CSA06)
• Failure rate 1/104
• Expected to improve as we switch to Geant4.8.1
• Speed (3.6GHz CPU, Geant4.8.1 with QGSP_EMV
  physics list, interactive testing) very
  preliminary
• Minimum bias events 37 seconds per event
• H ? eeµµ 197 seconds per event
• CMS strategy
• equal number of simulated and real events (
  1.5x109/year)
• Aim to achieve this with a mixture of full and
  fast simulation

Julia Yarba, FNAL IEEE/NSS 2006
13
Summary and Outlook

• In CMS, the Geant4-based Object Oriented
  simulation has been successfully implemented
• Ported to the new framework, widely validated
  (feedback reported to Geant4), used for physics
  and detector studies
• Proven to be robust, powerful, maintainable
• Capable to fulfill emerging requirements
• Further developments
• Finalize geometry updates (ECAL,)
• Include shower parametrization (GFlash, hadronic)
• Performance improvement (Geant4.8.1, local
  magnetic field stepper)

Julia Yarba, FNAL IEEE/NSS 2006