134 by wanghonghx


									Geant4: an update

An overview of Geant4’s
recent developments

 John Apostolakis, CERN
 Makoto Asai, SLAC
         for the Geant4 collaboration
1. Brief introduction to Geant4
2. Physics highlights
      Modeling
      validation
3. New capabilities
      Detector description and collision detection
4. Some current Developments
      In progress
      Planned for 2003
24th March 2003                                       2
    Part 1

Toolkit structure
GEANT 4 introduction

Detector simulation tool-kit for HEP
     offering alternatives, allowing for tailoring
Software Engineering and OO technology
     provide the method for building, maintaining it.
Requirements from HEP & other domains:
     LHC, heavy ions, CP violation, cosmic rays
     medical and space science applications
World-wide collaboration
     RD44 1994-1998
     MoU 1999-today

24th March 2003                                          4
Geant4 Overview
Extensive & transparent physics models
     electromagnetic, hadronic, optical, decay, …
Powerful structure and kernel
     tracking, stacks, geometry, hits, …
     visualization, GUI, persistency.
Efficiency enhancing techniques
     Framework for fast simulation (shower parameterization)
     Variance reduction / event biasing

24th March 2003                                          5
    Part 2

Physics Highlights
Physics Development Highlights
Geant4 releases Dec 2001-today included
     New EM processes
          And improvements to existing processes
     New theoretical hadronic models
          In particular for the cascade energy range
     The release of „tailored‟ physics lists
          For different hadronics use cases.
     Numerous physics improvements
          Including, for example
          Charge state for recoils
          Improved X-sections for e-Nuclear, with hard scattering

24th March 2003                                                      7
  Significant developments in
  EM (std) in 2002
 Multiple scattering (L. Urban)
     Angular distributions (see next slides)
 Ultra relativistic energies (H. Burkardt, S. Kelner, R. Kokoulin)
   g m m process
 Ionization for Generic Ions (V. Ivanchenko)

 New model of Transition radiation (V. Grichine)
     for TR detectors
 Redesign of few processes
     prototype model approach for Ionization and Bremsstrahlung (V.

24th March 2003                                                       8
Multiple                     15.7 MeV electrons on gold foil

Small differences
  between G4 & G3
  observed below 1 MeV
- Results competitive
  versus data in G4 3.2
- Differences traced to
  Multiple Scattering

MS modeling improved
  in Geant4 4.0 & 5.0

Examples of
   comparisons to data    Geant4 4.0
                          (Dec 2001)
   Thanks to L. Urban                           Angle (deg)
  24th March 2003                                              9
 modeling of
   in Geant4 5.0

Modeling &
  comparisons:       Geant4 5.0
                     (Dec 2002)
          L. Urban

                                  Angle (deg)
  24th March 2003                               10
 Hadronic physics:
  models, processes and ‘lists’
Five level implementation Components can be assembled in an
framework                   optimized way for each use case.
Variety of models and         Element
   for each energy regime,   particle
    particle type, material
    alternatives with different
    strengths and CPU                           Pre-compound

    requirements.                               model

 Illustrative example of assembling                CHIPS
  models into an inelastic process for set
  of particles
   Uses levels 1 & 2 of framework                        Parame-
 24th March 2003                                                    11
Models: Cascade energy range
   Parameterized process (1997)
   Chiral Invariant Phase Space decay,“CHIPS”                         M Kosov,
         For g-Nucleus, p capture, string-‟backend‟              P Degtyarenko,
             First release Dec 2001 in Geant4 4.0                   JP Wellisch
             Refinements and extension in 2002
   Bertini cascade (Dec 2002, Geant4 5.0)                             A Heikinen
         Re-engineered from HETC by HIP                               N Stepanov
             See the presentation of A Heikinen                             JPW
   Binary cascade model (Frankfurt, CERN)                               G Folger
         First release for nucleon induced interactions (in G4 5.0)         JPW
   Extensive verification suite
      See the presentation by D. Wright, V. Ivantchenko
   For further details,
      see the next presentation (J.P. Wellisch)

24th March 2003                                                               12
Tailored Physics ‘lists’
  Created and distribute “educated guess” physics lists
      That correspond to the major use cases of Geant4 involving
       hadronic physics,
      to use directly, and as a starting point for users to modify,
           facilitate the specialization of those parts of hadronic physics lists
            that vary between use cases.
  First released in September 2002
      Using physics models of Geant4 4.1.
  Revised with experience of comparisons with data
           updated with physics models of Geant4 5.0 in March 2003
  Find them on the G4 hadronic physics web pages

24th March 2003                                                                  13
Use cases of Physics Lists
   HEP calorimetry.                    low energy dosimetric
   HEP trackers.                        applications
   'Average' HEP collider detector     high energy production targets
   Low energy dosimetric                  e.g. 400GeV protons on C or Be
    applications                        medium energy production
    with neutrons                        targets
   low energy nucleon                     e.g. 15-50 GeV p on light targets
    penetration shielding               LHC neutron fluxes
   linear collider neutron fluxes      Air shower applications
   high energy penetration             low background experiments
   medical and other life-saving
    neutron applications

                  Contributors: http://cern.ch/geant4/organisation/
24th March 2003      working_groups.html#wg.Had                           14
Physics lists for calorimetry
 LHEP is the fastest for CPU                QGSC, is similar but uses
      uses the LEP and HEP                   CHIPS for fragmentation
       parameterized models for                 The CHiral Invariant Phase-
       inelastic scattering.                     Space decay (CHIPS)
      uses theory-driven modeling           FTFP replaces instead the
        for reactions of ps, Ks, and          string
        nucleons.                               with a diffractive string
     It employs                                  excitation
      Quark Gluon String Model                    similar to that in FRITJOF,
          for the 'punch-through'                  and the Lund fragmentation
           interactions of the projectile           functions.
      A Pre-equilibrium decay model
          with an extensive evaporation
           phase to model the nucleus
           'after the punch'.

24th March 2003                                                                15
Comparison projects
Joint efforts for comparing Geant4 with
 experiment & test-beam data.
     Results of EM comparisons: „peak‟ between 2000-
     Hadronic comparisons: 2002-ongoing.
Collaboration with experiments
     ATLAS (projects with data of numerous test beams)
     BaBar (with data for tracker, drift chamber)
Many results have been presented at
 conferences & workshops, eg Calor 2002.
     And at regular LHC experiment-Geant4 physics
      comparisons meetings

24th March 2003                                           16
Mix and match
  problem seen in
Problem disappears,
  as expected, when
  utilising theoretical
    Eg Quark-Gluon String
      Model + CHIPS

For latest results please                First results from April/May 2002
   see the presentations of
   JP Wellisch & Atlas

                              Thanks to Atlas HEC and J.P. Wellisch
   24th March 2003                                                     18
                                 As presented at Geant4 Workshop, 30th September 2002
                                                            Thanks to P. Loch, Atlas
   Geant4 Hadronic Signals in ATLAS Calorimeters
 Calorimeter pion response:

                                             e/π signal ratio
                                                                    HEC Pions
    after discovery of “mix-and-match”
problem (transition from low energy to high
energy char-ged pion models) in the
deposited energy from energy loss of
charged particles in pion showers in the HEC
(G4 4.0, early 2002): fixes suggested by
H.P. Wellisch (LHEP, new energy thresholds
in model transition + code changes) and QGS
model tested;

     e/π signal ration in HEC and TileCal still                     LHEP
not well reproduced by Geant4 QGS or
LHEP - but better than with GCalor in
                                             e/π signal ratio


   energy dependence in HEC in QGS
smoother, “discontinuities” between ~20
GeV and ~80 GeV gone;
  24th March 2003                                                Pion energy [GeV]   19
                               Word highlights: JA, March 2003
     Part 3
  New capabilities

Detector description
Other Development highlights
Detector description
     New ways to create geometries
     Tools to detect incorrect geometry definitions
     A different field for any volume (or volume tree)
          Overriding a global field
Ability to reduce initialisation time
     By saving/retrieving physics processes‟ table
Variance reduction / event biasing
     Importance: biasing by geometry
     Leading particle biasing

24th March 2003                                           21
Improvements in Geometry
 Reflection of volume hierarchies                     I Hrivnacova
     Eg to create endcap geometry                          G Cosmo
                                                          V Grichine
 Improved voxelisation for performant navigation
     3-D for parameterized volumes
                                                             G Cosmo
          Now equal performance to „placed‟ volume
     Option to avoid voxelizing some volumes

 „Illegal‟ geometries detected & rejected                   G Cosmo
     E.g. incompatible daughters (placed & parameterized)

 XML binding: GDML 1.0 released                         R Chytracek
          Specification & Implementation
                  • Refinements currently on „hold‟.

24th March 2003                                                    22
Debugging geometries
 It is easy to create overlapping volumes
       a volume that protrudes from its mother,
       2+ volumes that intersect in common mother
    During tracking Geant4 does not check for
     malformed geometries
 The problem of detecting „significant‟
  overlaps is now addressed by
    DAVID that intersects volumes directly
        ( Uses graphical representations )
           •   Created by S. Tanaka, released ca 1997
    New commands to run verification tests
           •   Created by DC Williams; released in 4.0
    New example with full tracking / navigation
           •   Created by M Liendl; released in 5.0

                                                         Thanks to S. Tanaka
 Variance reduction
 Geant4 had leading particle biasing option for “low energy” neutrons.
     Now redesigned and improved, implementation in Geant4 4.1.
     It was possible to use other methods, but only in user code.
 Now new general purpose built-in methods have been released
     Further refinements & methods are under development.
 Importance biasing:                                                           M Dressel
     Splitting/Russian roulette (first released in G4 4.1, June 2002).
     Importance values can be associated to a volume
         In the „mass‟ geometry or in a dedicated „parallel‟ geometry.
     Enabling simulation of shielding applications with improved time efficiency
      by large factors
         Varied options in driving MC „history‟ and scoring tallies
         No changes to the kernel were required, due to the flexibility of the toolkit.
 Leading particle biasing
     a-la MARS 95, for En<5GeV                                                 N.Kanaya

 24th March 2003                                                                           24
CPU Performance
 Our first simple benchmarks:
     Geometry faster, EM shower setups: competitive
 Performance in experimental setups (with Geant4 releases
  2 and 3) was comparable to Geant3
     few counterexamples, including BTeV ECAL.
 New performance issues arose with Geant4 4.0
     and were addressed (in the patches & release 4.1)
 Difficult cases remain, including
     Some setups of EM showers and field propagation, factor ~ 2x
 Collecting a set of benchmarks
     To follow computing performance regularly
 Goal is that Geant4 is at least as fast as Geant3 in almost
  all cases
     When its power is used.
24th March 2003                                                      25
                                        DAWN renderer
Visualization                           Thanks to S.

Geometry, hits
   “DTREE”: hierarchy display
    HEPREP driver for WIRED
Other Current Drivers
   DAWN Renderer
   Also from others, eg
         IGUANA (for CMS simulation)
 24th March 2003   Iguana, thanks to L.Tuura, I. Osborne   26
     Part 4
Current development

In Progress 2003 (highlights)
Cuts per region
     See next slides
Improvements of multiple scattering
     in straggling, backscattering
Additional refinements of physics lists
     Continuous updates
Design iteration of EM (std) processes
     With benefits in tailoring, maintenance
Further extension and automation of testing
     Statistical testing: „benchmarks‟ and test-beams

24th March 2003                                          28
Cuts in Geant4 (to date)
 Geant4 has had a unique production threshold („cut‟)
  expressed in length (range of secondary).
     For all volumes
     Possibly different for each particle.
 This promotes
     Clear criteria for locality of energy deposition
     better use of CPU – less „wasted‟ in dense materials
 Yet appropriate length scales can vary greatly between
  different areas of a large detector
   Eg a vertex detector (5 mm) and a muon detector (2.5 cm).
     Having a unique (low) cut can create a performance penalty.
 So the part of the detector with the lowest cut need
  fixed the cut for all the simulation.

24th March 2003                                                     29
Motivation for several cuts
Having a unique cut enforced a choice
     Sacrificing accuracy of energy deposition
     Accepting a performance penalty
Lifting the uniqueness of cuts
     Requested from LHC experiments & BaBar
     Implemented by introducing geometrical
          And enabling the choice of thresholds in a region.

24th March 2003                                             30
Region & its properties
 Introduce the concept of
  « region »:
      Set of geometry volumes,                               Region B
       typically of a sub-system;
                  • Eg: barrel + end-caps of the
      Or any group of volumes;
 A cut in range is associated to a                 Region               Region
  region;                                                     Region C
          a different range cut for each             B                    B
           particle is allowed in a region .
 Typical Uses
      barrel + end-caps of the
       calorimeter can be a region;                           Region B
      “Deep” areas of support                     Region A
       structures can be a region.

24th March 2003                                                              31
Cuts per region status
Design and implementation have been made
     without severe design revision of the existing GEANT4;
     First implementation available in latest b release
Comparable run-time performance
          Today a penalty within 5% is seen, due to redundant checks
           included for verification purposes
„Full release‟ will be in Geant4 5.1 (end April)
     With further refinements, tests, validation.

24th March 2003                                                    32
MS in
Multiple scattering:


 Backscattering

 Straggling

 Transmitted

 24th March 2003       33
                     MS: straggling
                           Fit to data

Lateral straggling
of 2.5 MeV protons
After mylar foils

                                  Geant4 5.1 (April 2003)

   24th March 2003                                          35
In progress (also)
The refinement of the design of EM
 physics processes through the use of
     To enable the specialization of key features;
     To enable the easy use of different models
      for a single process (e.g. Ionization) in one
Additional variance reduction techniques
     Filter for enhancing processes in hadronic
24th March 2003                                       36
Some further 2003
development highlights
Additions to physics processes/models
     p induced binary cascade model, ..
     EM-std implementation with “model” approach.
      Refinements, including
          Improvement to recoil in elastic scattering
          Improved X-sections for pions.

Revisions of the „tailored‟ physics lists
     Incorporating results of validation
Variance reduction
     Physics process enhancement
     Leading particle biasing
     Plus refinements to importance biasing
24th March 2003                                          37
Review and Releases
d Review October 2002
      Report available at http://cern.ch/geant4
 Developments available in b releases
      Every two months
      Latest b release (February)
          Included cuts per region
 Upcoming releases
      Next minor release is Geant4 5.1 planned for end-April
          Incorporating cut per region, developments in progress.
          Release timeframe selected to aid in CMS production.
      „Scheduled‟ release Geant4 5.2 for end-June
          Further refinements, developments
      2003 work items & planned release contents to be available soon
          Started from User & Experiment Requirements and Requests
          Next major release Geant4 6.0 is scheduled for December 2003.

24th March 2003                                                            38

Results of comparing Geant4 versus data,
   Have & are providing excellent „yardsticks‟ of EM perf.
   Are testing the hadronics well, with increasing coverage
Geant4 has demonstrated important strengths:
   stability of results, flexibility, transparency.
   it is in production use today in running HEP
    experiments (BaBar, HARP)
Geant4 is evolving
   With the feedback from LHC exper., BaBar and
    numerous other experiments and application domains.
Refinements & development are ongoing.
24th March 2003                                           39

   Thanks to all
After the END …

  Slides after this are backups,
  not part of the presentation.

              v0.8 24th March 2003, 18:40 GMT
  Electromagnetic physics
     Gamma-conversion, Compton scattering, Photo-electric effect
 Leptons(e, m), charged hadrons, ions
     Energy loss (Ionisation, Bremstrahlung) or PAI model energy loss,
      Multiple scattering, Transition radiation, Synchrotron radiation,
     Cerenkov, Rayleigh, Reflection, Refraction, Absorption, Scintillation

 High energy m
 Alternative implementation
     „Standard‟ for applications that do not need to go below 1 KeV
     „Low Energy‟: down to 250eV (e+/g), O(0.1) mm for hadrons
24th March 2003          specialized HEP applications                 42
1 GeV
   in H2O

 24th March 2003   43
Support: new & continued

        Revisions of the user and reference guides
             After assessments of overall structure & detailed
        LXR for code reference
             see http://geant4www.triumf.ca/lxr/
  New tool for collecting requirements

  Continued Support
        of users‟ questions, problems
             HyperNews, Problem reporting system, email.
        of comparisons with data
             By wide variety of users, in HEP, space, medical phys., ..
24th March 2003                                                            45
Testing and QA 2002/3
Establishment of „statistical testing‟ suite
     Automated comparison of physics quantities
          Against „standard‟ data (eg NIST)
          In „test-beam‟ applications
          Including „regression testing‟.
          For details see
Establishing a benchmark suite for
 computing performance.

24th March 2003                                    46
  Examples of improvements
Fixes and improvements in Geant4 release 4.1
  (June 2002)
     Fix for voxelisation of reflected volumes
     Fix for exit normal angle
     Fix for problem in very small step in field
     Improvements in Multiple Scattering, Ionisation, ..
     Fix for energy conservation in parametrised models.
     Fix for small peak at f=0 in parametrised models.

24th March 2003                                             47
                  Geant4 based
                  More than
                    109 events
                    Oct 2002)
                  Used Geant4
24th March 2003            48
       Cuts/Region Introduction
    A Cut here is a « production threshold »;
        Only for physics processes that have infra-red divergence
        Not tracking cut; (which does not exist in Geant4)
    GEANT4 up to now allows a unique cut in range;
        One cut in range for each particle;
              By default is the same cut for all particles;
        Consistency of the physics simulated:
              A volume with dense material will not «dominate» the
               simulation time at the expense of sensitive volumes with light
    Requests from ATLAS, BABAR, CMS, LHCb, …, to allow
     several cuts;
        Globally or per particle;

24th March 2003                                                                 51
Geant4 Collaboration

                       Collaborators also from non-
 Lebedev               member institutions, including
                         Budker Inst. of Physics
                             IHEP Protvino
                            MEPHI Moscow

  24th March 2003                              52

To top