Geant4: an update
John Apostolakis, CERNMakoto Asai, SLAC
for the Geant4 collaboration
Version 0.5.4 23rd March 2003, 16:30
24th March 2003 2
Outline
Brief introduction to Geant4Physics highlights
Modeling validation
New capabilities Detector description and collision
detectionSome of the Developments
In progress Planned for 2003
24th March 2003 3
GEANT 4
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:
LHC heavy ions, CP violation, cosmic rays medical and space science applications
World-wide collaboration
24th March 2003 4
Geant4 Capabilities
Extensive & transparent physics models electromagnetic, hadronic, optical, decay, …
Powerful structure and kernel tracking, stacks, geometry, hits, …
Interfaces visualization, GUI, persistency.
Efficiency enhancing techniques Framework for fast simulation (shower
parameterization) Variance reduction / event biasing
24th March 2003 5
Physics Development Highlights
Geant4 releases Dec 2001-today included New theoretical hadronic models
CHIPS for -Nucleus, capture, ..Cascade models: Bertini (HETC) and Binary
New EM processes to pair new TR models
Numerous physics improvementsIncluding, for exampleCharge state for recoilsImproved X-sections for e-Nuclear, with hard
scattering
24th March 2003 8
Multiple scattering
Examples of comparisons: 15.7 MeV e-
on gold foil
(figure this page)
Modelling & comparisons:
L. UrbanAngle (deg)
24th March 2003 9
Multiple scattering
Angle (deg)
Examples of comparisons: 15.7 MeV e-
on gold foil
Modelling & comparisons:
L. Urban
24th March 2003 10
Significant developments in EM (std) in 2002 Multiple scattering (L. Urban)
Ultra relativistic energies (H. Burkardt, S. Kelner, R. Kokoulin)
Ionization for Generic Ions (V. Ivanchenko)
Models of Transition radiation detectors (V. Grichine)
Redesign of few processes: prototype model approach for Ionization and Bremsstrahlung (V. Ivanchenko)
24th March 2003 11
Comparison projects
Joint efforts for comparing Geant4 with experiment & test-beam data. Results of EM comparisons: 2000-2002. Hadronic comparisons: 2002-ongoing.
Collaboration with experiments ATLAS (projects with data of test beams of 4 calorimeters) BaBar (with experiment data for tracker, drift chamber)
Many results have been presented (by the experiments) at conferences & workshops eg Calor 2002. Presentations at quarterly LHC experiment-Geant4 physics
comparisons meetings, eg 4th March 2003.
24th March 2003 12
Hadronic physics: models, processes and ‘lists’
Illustrative example of assembling models into an inelastic process for set of particles Uses levels 1 & 2 of framework
CHIPS
QGSM Parame-terized
Ene
rgy
Element
particle
Pre-compoundmodel
Five level implementation frameworkVariety of models and cross-sections
for each energy regime, particle type, materialalternatives with different strengths and CPU requirements.
Components can be assembled in an optimized way for each use case.
Z
24th March 2003 13
Models: Cascade energy range
Parameterized process (1997)Chiral Invariant Phase Space decay,“CHIPS”
Development 2000-2001 (M Kosov, P Degtyarenko, JP Wellisch) Refinements and extension in 2002
Bertini cascade (Dec 2002, Geant4 5.0) Re-engineered from HETC,
See the presentation of A Heikinen
Binary cascade model (Frankfurt, CERN) First release for nucleon induced interactions (in Geant4
5.0) Extensive verification suite
See the presentation by D. Wright For other details,
see the next presentation (J.P. Wellisch)
24th March 2003 14
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 Latest:
updated with physics models of Geant4 5.0 in March 2003 Find them on the G4 hadronic physics web pages
http://cmsdoc.cern.ch/~hpw/GHAD/HomePage
24th March 2003 15
Use cases of Physics Lists Calorimetry Tracker detectors Average HEP detector High energy physics
calorimetry. High energy physics
trackers. 'Average' HEP collider
detector Low energy dosimetric
applicationswith neutrons
low energy nucleon penetration shielding
linear collider neutron fluxes high energy penetration
shielding
medical and other life-saving neutron applications
low energy dosimetric applications
high energy production targetse.g. 400GeV protons on C or Be
medium energy production targetse.g. 15-50 GeV p on light
targets LHC neutron fluxes Air shower applications low background experiments
Contributors: http://cern.ch/geant4/organisation/ working_groups.html#wg.Had
24th March 2003 17
Resolution
Original (org) results from Calor 2002 presentation,
(March 2002).
Open symbols from additional physics lists JPW, May 2002, using geant4 4.0-patch2
(released: end Feb 2002).
Thanks to Atlas HEC and J.P. Wellisch
Status of May 2002
24th March 2003 18
BaBar
Geant4 based simulation since 2001 production.
More than 109 events (through Oct 2002)
Used Geant4 3.1+fixes, own transport.
24th March 2003 20
Other Development highlights
Various improvements Ability to reduce initialisation time
By saving/retrieving physics processes’ table A different field for any volume (or volume tree)
Overriding a potential global ‘default’ field Additional ways to create geometries Detect and debug incorrect geometry definitions
Variance reduction Importance: biasing by geometry
In real or ’ghost’ geometry• Eg enabling simulation of shielding applications with
improved time efficiency by large factors Leading particle biasing
• a-la MARS 95, for En<5GeV
24th March 2003 21
Improvements in Geometry Reflection of volume hierarchies
Eg to create endcap geometry
Improved voxelisation for performant navigation 3-D for parameterized volumes
Now equal performance to ‘placed’ volume Option to avoid voxelising some volumes
‘Illegal’ geometries detected & rejected E.g. incompatible daughters (placed & parameterised)
XML binding: GDML 1.0 released Specification Implementation
• Refinements currently on ‘hold’.
G Cosmo
R Chytracek
I Hrivnacova G Cosmo
V Grichine
Debugging geometriesDebugging 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• By S. Tanaka, released circa 1997
New built-in run-time commands to run verification tests by its own navigation
Using solids and a grid of rays • Created DC Williams; released in 4.0
New example with full tracking / navigationUses the full hierarchy & the Navigator
• Created by M Liendl; released in 5.0 Thanks to S. Tanaka
24th March 2003 23
Variance reduction Geant4 has had an event biasing option for the transportation of
“low energy” neutrons. The formulae are derived from MARS95. New, 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 created.
Importance biasing: Splitting/Russian roulette (first released in G4 4.1, June 2002). Importance values can be associated to a ‘mass’ geometry or
to a ghost geometry. Varied options in driving MC ‘history’ and scoring tallies; No changes to the kernel were required, due to the flexibility of the
toolkit.
M Dressel
24th March 2003 24
CPU Performance Our 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 issue 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
Geometry, hitsNew
“DTREE”: hierarchy display HEPREP and Wired driver
Other Current Drivers OpenGL
Popular, hardware speed VRML
Portable, lower detail DAWN renderer
High quality Postscript Also from others, eg
IGUANA (for CMS simulation)
DAWN rendererThanks to S. Tanaka
Iguana, thanks to L.Tuura, I. Osborne
Visualization
24th March 2003 26
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 27
Cuts per region (precis) Geant4 has had a unique production threshold (‘cut’)
expressed in length (range of secondary). For all volumes Possibly different for each particle.
This promotes Good consistency in energy deposition less use of CPU in dense materials (compared to a common
Energy) Yet appropriate length scales can vary greatly between
different areas of a large detector Eg a vertex detector (5 m) and a muon detector (2.5 cm). Having a unique (low) cut can create a performance penalty.
New functionality, enabling the tuning of production thresholds at the level of a sub-
detector. Created Region (or sub-detector)
24th March 2003 28
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 material.
Requests from ATLAS, BABAR, CMS, LHCb, …, to allow several cuts; Globally or per particle;
24th March 2003 29
Motivation for several cuts
Having a unique cut can be the source of performance penalties;
Part of the detector with lower cut needs fixes the cut for the all simulation; Can be far too low than necessary in
other parts;Silicon vertex detector: a few 10 m;Hadronic calorimeter: 1 cm;
Other parts being geometrically far, to.
24th March 2003 31
Region: example & properties Introduce the concept of
« region »: Set of geometry volumes,
typically of a sub-system;• Eg: barrel + end-caps of the
calorimeter; Or any group of volumes;
A cut in range is associated to a region;
a different range cut for each particle is allowed in a region .
Typical Uses barrel + end-caps of the
calorimeter can be a region; “Deep” areas of support
structures can be a region.
Region B
RegionB
Region A
Region B
Region B
Region C
c
24th March 2003 33
Cuts per region status
Design and implementation have been made without severe design revision of the existing
GEANT4; First implementation available in latest release
(Feb)
Comparable run-time performanceToday 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 34
In progress (also)
The refinement of the design of EM physics processes through the use of ‘models’. To enable the specialization of key features; To enable the easy use of different models
for a single process (e.g. Ionization) in one application.
Additional variance reduction techniques Filter for enhancing processes in hadronic
interactions.
24th March 2003 35
Scheduled 2003 Development: highlights
Cuts per region Additional physics processes/models
Hadronic: induced binary cascade model, .. EM: refinements to multiple scattering; “models”.
Refinements, including Improvement to recoil in elastic scattering Improved X-sections for pions.
Variance reduction Physics process enhancement Refinements to importance biasing. Leading particle
24th March 2003 36
Geant4 Collaboration
Collaborators also from non-member institutions, including
Budker Inst. of PhysicsIHEP Protvino
MEPHI Moscow
Lebedev
24th March 2003 37
Review and Releases
Review October 2002 Report available at http://cern.ch/geant4
Developments available in releases Every two months Latest release (February)
Included cuts per region Upcoming releases
Planned: Geant4 5.1 release end-April Geant4 5.2 scheduled for end-June Release timeframe(s) adjusted for customer needs
Workplan User & Experiment Requirements and Requests Next major release Geant4 6.0 scheduled for December.
24th March 2003 38
SummaryResults 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.
Geant4 is in production use today in running HEP experiments (BaBar, HARP)
Geant4 is evolving With requirements from LHC exper., BaBar and
numerous other experiments and application domains.Refinements & development are ongoing.
http://cern.ch/geant4/
THE END
Slides after this are backup slides, deleted from the presentation
24th March 2003 40
Support: new & continued
Documentation 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 41
Testing and QA 2002/3
Establishment of ‘statistical testing’ suite Automated comparison of physics
quantities Against ‘standard’ data (eg NIST)In ‘test-beam’ applicationsIncluding ‘regression testing’.For details see
Establishing a benchmark suite for computing performance.
24th March 2003 42
Geant4 Capabilities
Extensive & transparent physics models electromagnetic, hadronic, optical, decay, …
Powerful structure and kernel tracking, stacks, geometry, hits, …
Interfaces visualization, GUI, persistency.
Efficiency enhancing techniques Framework for fast simulation (shower
parameterization) Variance reduction / event biasing
24th March 2003 43
Summary
Geant4 is in production use today in running HEP experiments (BaBar, HARP)
Results of comparing Geant4 versus data are growing month by month, have provided important ‘yardsticks’.
Geant4 has demonstrated important strengths: stability of results, flexibility, transparency.
Refinements & development are ongoing.
24th March 2003 44
Testing tools
Improvements of tools for automation of testing, followup Tinderbox Bonsai LXR for use in testing followup
24th March 2003 45
Examples of improvements
Fixes and improvements in Geant4 release 4.1 (June 2002)
Geometry Fix for voxelisation of reflected volumes Fix for exit normal angle Fix for problem in very small step in field
EM Improvements in Multiple Scattering, Ionisation, ..
Hadronics Fix for energy conservation in parametrised
models. Fix for small peak at =0 in parametrised models.