Snapshots from the Geant4 Review J. Apostolakis

Snapshots from the Geant4 Review

• J. Apostolakis

• PH/SFT Group Meeting

• 11 May 2007

Overview

• The Review

• Mandate, members
• The presentations
• User presentations
• Geant4 presentations
• Closeout

• Sampling from the presentations

• Users
• Geant4
• Key recommendations

Geant4 Review Mandate

• The mandate of the Geant4 Review Committee in 2007 is to investigate the physics precision, computational speed and general usability of the Geant4 software for the major existing and upcoming Geant4 use cases in its various application domains, and to issue recommendations for future software improvements and developments. In performing this task, the Commitee may also address auxiliary issues that it considers relevant.

• You can find the agenda from

• http://cern.ch/geant4/collaboration/Geant4-Review2007.html

Panel members

• Review Panel Members

• Irène Buvat, INSERM France

• Fabio Cossuti, INFN Trieste

• Frank Gaede, DESY, Hamburg

• Nobu Katayama, KEK, Tsukuba

• Andreas Morsch, CERN, Geneva

• Robert Reed, Vanderbilt University, Nashville TN

• Sayed Rokni, SLAC, Mountain View CA

• Lembit Sihver, Chalmers University, Sweden

• Charles Young, SLAC (chair)

• You can find the agenda from

• http://cern.ch/geant4/collaboration/Geant4-Review2007.html

User Presentations

• Input from user communities

• LHC: Physics, CPU performance, usability
• ILC and other HEP experiments
• Space (radiation effects on electronics)
• Medical
• Shielding (requirements)

• Several speakers only had a few days to collect information and assemble the talks

This presentation

• Covers only select topics

• Provides excerpts of some user talks

• I have picked a few topics, trying to

• Emphasize HEP aspects different from the ones we see regularly at LCG Physics Validation meetings
• Give a flavour of Geant4’s use and requirements from outside HEP

• Also selected

• Small sample of Geant4 results
• Key reviewers’ observations and recommendations

User topics selected

• HEP: ILC

• Requirements
• Challenges and opportunities

• Space

• Electronics in space
• the system, strengths, needs

• Medical

• Outline:

• ILC physics focus

• hadronic showers data and simulation

• requests from other HEP experiments

• some considerations

• New Physics:

• - rare processes/limited statistics

• Many final states involving heavy bosons (Z,W,H):

• e+e-  WW  , e+e-  ZZ 

• Hadronic decay of W and Z

• - branching ratio ~70%
• - result in two hadronic jets

• Requires excellent

CALICE: from MC to reality

Event display

Hadron event

Deep Analysis of hadronic shower

Getting ready to compare data/MC

Models comparison

The ILC request to G4

• The ILC community has to design, optimize and compare concepts of the next generation of HEP detectors

• The most controversial part for simulations is the hadronic sector

• The energy of interest for ILC hadronic physics is lower than that of LHC (mean E~5-10 GeV, @ 500 GeV s)

• In this energy range hadronic models in G4 are sometimes inadequate or outdated

• CALICE can offer to G4 the possibility of active collaboration in comparing and tuning all models with unique precision and new benchmark parameters

• For our community it is convenient to work within one framework but it is vital to have available all intra-nuclear cascade models (from FLUKA and MCNP)

• A clear set of expert-recommended parameters is needed to ease the user task

The rest of the non-LHC community

• F. Gaede and C. Young collected answers to a poll among HEP experiments:

• here reported only the comments concerning additional physics request to G4

• MINOS / NOVA

• Try to migrate from G3 to G4, problems with geometry, no phys request yet

• NUMI (beamline simulation)

• Default G3+FLUKA, started with G4, request to use FLUKA

• MiniBooNE

• G4 only for neutrino flux prediction  p-Be interaction (100MeV – 10 GeV)

• Request broader range of models, in particular FLUKA

• BaBar

• Finished validation of G4.7, improvements compared to G4.6, proceed to G4.8

• no additional physics requests

• Also other experiments have replied which do not use G4:

• Belle, STAR and PHENIX (RHIC), OPERA, K2K/T2K (because of no FLUKA),

• KEK/E391, J-PARK/E14 (not yet, planed if FLUKA)

Additional Remarks

• Comments/suggestions collected:

• Request of improvement of default parameter sets

• possibility: multiple recommended sets for different users

•  not all uses are MC experts, choice from experts is normally the best

• Request to improve geometry: more flexible definition and debug (GUI), possibility to interface to existing programs (CAD)

• Both BaBar and Bell are interested to know how good are hadronic interactions in G4 for B and super-B factories

• Offers of cooperation from the users:

• Some users have offered to collaborate with GEANT4 on areas where they have particular expertize and interest, e.g. polarized EM processes (DESY Zeuthen), hadronic showers model validation (CALICE) Is GEANT4 interested in such collaboration? How should users proceed? 

Geant4 & Radiation Effects in Electronics The RADSAFE Strategy

• Robert A. Weller

• Institute for Space & Defense Electronics

• Vanderbilt University, Nashville, TN, USA

• Vanderbilt: R. A. Reed M. H. Mendenhall

• B. D. Sierawski K. M. Warren R. D. Schrimpf

• L. W. Massengill A. F. Witulski D. R. Ball

• J. A. Pellish C. L. Howe A. D. Tipton

• M. L. Alles A. L. Sternberg B. E. Templeton

• SLAC: M. Asai T. Koi D. H. Wright

• NASA Goddard: M. A. Xapsos K. A. LaBel

• NASA Marshall: J. H. Adams J. W. Watts

• Sponsoring Agencies: NASA, DTRA, AFOSR, AEDC

Overview

• Application: First-principles simulation ofradiation effects in semiconductors in terrestrial and space systems.

• • Motivation: Maximize efficiency and minimize cost of design and testing for arbitrary radiation environments.

• Approach: RADSAFE - A strategy for integrating commercial and custom software systems.

• Geant4 application: MRED, the radiation event generator.

Why [RADSAFE] use[s] Geant4 (Instead of X, Y, or Z)

• Comprehensive physics.

• Range of available processes, classes.

• Quality/size of the developer community.

• Continuing developer/user interaction.

• The toolkit philosophy. (It’s not a program!)

• Total user control over input and output.

• Available source code.

• Platform independence.

• C++ language. (Not Fortran, Java, etc.)

• Our (unique?) requirements described above.

Priorities for Semiconductor Electronics Applications

• Physics, tracking down to nm scale (at least nothing, i.e. cuts, breaks).

• The best possible dE/dx for all particles, including ions especially, at all energies, few keV to 100+ GeV.

• a) Correct fluctuation.
• b) Discrete/continuous partitioning.
• c) Microstructure of individual tracks, including e-h production (research area).

• Coulomb scattering of nuclei for computing displacement damage. (VU is donating ScreenedScattering class.)

• Must have accurate mass, energy, angle distributions of nuclear reaction fragments.

• a) Heaviest fragments are most important.
• b) Goal: threshold-100 GeV/u; Interim goal 5-10 MeV/u to 20 GeV/u.

• …

Profiling of GATE-Geant4 performance

Current VRT Implementations

• Variance Reduction Techniques (VRTs)

• Importance sampling
• Photon splitting
• Russian roulette
• Weight window sampling
• Weight roulette
• Scoring

• A number of VRT implementations are already available in Geant4, but

• Some of the Geant4 classes can be optimized more,
• Further classes implementing VRTs could be developed

Suggested actions regarding VRTs

• Collaboration between

• GATE VRTs workgroup
• Geant4 VRTs workgroup

• Aim of collaboration

• Determination of the existing G4 classes need to be optimized
• Determination of further G4 classes possibly needed to be implemented
• Implementation of GATE-specific classes within the Geant4 framework
• Publication of a GATE-specific patch for Geant4

• Suggested VRTs for Geant4 follow:

• Implementation of flags (probably at GATE) for each one of the following VRTs
• => VRTs should be activated or deactivated by the user

• Contact: Nicolas Karakatsanis : [e-mail]

Using G4 for medical applications – Further suggestions

• The documentation on the facrange (facgeom and related) parameters seems important for medical application users

• “Wiki” documentation

• cooperative community-build documentation in a “wiki” format
• interesting way to improve the documentation, particularly in the developer section
• “Wiki” technology is now well known and mature
• allows any user to modify and improve the on-line documentation
• Our experience shows that
• is very efficient and
• it also encourages information exchange in the community

Beta Ray Point Source Distributions Using GEANT4 – Comparative Study

• L. Maigne, C.O. Thiam

• Laboratoire de Physique Corpusculaire, 24 avenue des Landais, 63177 AUBIERE cedex

Beta Ray Point Source Distributions Using GEANT4 – Comparative Study

• Comparison of G4 Versions (1MeV, 2MeV)

Beta Ray Point Source Distributions Using GEANT4 – Comparative Study

• Conclusion

• High discrepancies between G4.5 and G4.6, G4.6 and G4.7 =>
• Possible cause: Multiple scattering implementation?

Beta Ray Point Source Distributions Using GEANT4 – Comparative Study

• Comparison of Geant4 with other MC calculations ( kinetic energy = 3MeV, 4MeV )

Beta Ray Point Source Distributions Using GEANT4 – Comparative Study

• Conclusion

• High discrepancies between Geant4.8 and other MC packages (around 10%)
• Possible reason: still undefined

Dosimetric characteristics of an I125 Brachytherapy Source Using Geant4 – A comparative study

• Comparison of G4 Standard / Low Energy

Themes

• Computing Performance is vital

• in many HEP, Space, Medical applications

• More validation of physics models

• Demonstrating physics performance

• Larger choice of physics models

• Ion-ion, hadronics (1

• Simpler documentation

• Physics lists choice
• How to improve application performance

And the Geant4 presentations

• Twice as many as the user presentations

• … choose a few …

Simulation of the Fano cavity setup

• Sabine Elles, Vladimir Ivanchenko,

• Michel Maire, Laszlo Urban

• March 2007

Fano Cavity: Geant4 v 6.2 results

Effects of all improvements (8.2ref3)

TARC Fluence Binary cascade

• 2.5 GeV/c p

• Simulated with ‘QGSP_BIC_HP’ physics list

• Binary Cascade
• HP ‘High Precision’ Neutrons

• Red: cylinder

• Yellow: sphere

• Black: Full 4/cos

Ratios of fluence

• Ratio of simulated 4/cos versus Data

• Two-sets of data

• Dominated by systematic errors of experiment

n-C total cross-section

CPU Performance – with hadronics

From the Reviewers

• (draft form – closeout session)

• 20 April 2007

• JA Note: I have summarised and simplified – the original slides are available in the agenda

EM Physics

• Observations

• heavy ion stopping models not fully available.

• Documentation hard to find

• …

• Recommendations

• Provide guidance on selection between the [std vs low-energy] models for specific species, energies, etc.

• Integrate models into a single package for EM physics, …

• Encourage the rapid integration of ICRU 73 heavy ion stopping

• Setup a validation webpage for EM physics

• Provide detail references for validation papers for various models
• Comparisons to data etc…

• Provide guidance about trade off between accuracy and speed for various range cuts

Hadronic Physics

• the committee notes that there is a strong request to use the FLUKA hadronic model with Geant4.

• a serious effort has been undertaken by the Geant4 collaboration and the LCG Physics Validation group to test and validate the Geant4 hadronic transport models.

G4 Hadronic Physics

• comparison to test beam data has pointed to important shortcomings of the G4 hadronic models

• Note ongoing improvement. Concern on not seeing ‘a detailed planning (milestones, manpower) for an improvement of the hadronic package’.

• ‘put in place a set of simple hadronic benchmarks which allow to identify such very basic problems like disagreement with well-known shower shapes’.

Computing Performance

• Observations

• CPU optimization is critical for LHC and other heavy users (eg medical and space).

• Multi-prong approach is needed.

• Optimization of GEANT4 toolkit.
• Trade-off between CPU time and simulation detail. This is a user choice. - yet help from GEANT4 is crucial
• Variance reduction techniques (VRT) can cut CPU needs greatly. A number of VRT are available in GEANT4.

CPU Performance Recommendations

• Systematically track code performance, for each part, model.

• Perform systematic checks on

• Memory use
• Architecture, compiler/options
• Multi core CPUs/advanced instructions

• Encourage users to do same and provide feedback.

• Extend the computing professionals to review/optimize other parts.

• Provide a plan regarding the expected performance in the next few years.

• Create a document on performance optimization guides.

• Provide a simple mechanism for users to turn off “irrelevant” processes for a given region.

Documentation: Observations

• GEANT4 has provided among others

• Application Developers Guide.
• Installation Guide, ..

• No clear recommendation about the validity ranges for the many physics models (e.g. the two EM models) and the different Physics Lists.

• Few references to validation and benchmarking papers on GEANT4 web.

• Limited information about connection between the physics models and their implementation.

Documentation: Recommendations

• Provide clear recommendations when the different packages and Physics Lists should be used and their validity ranges.

• Document the limitations, and validity and applicability ranges of various hadronic/EM models.

• When models overlap in validity range, document the tradeoffs.

• Describe the connection between physics models and how they are implemented.

• Provide greater details in release notes.

• …

Validation: Recommendations

• Create a common validation procedure, possibly automated, to be run at every release.

• Define a procedure that quantifies the validation results, and make them easily available to users.

• For example,

• Timing checks, reporting key results in release notes.
• Benchmarking against other MC transport codes.
• “Recruit” additional beta testers from different user communities.
• Provide references of validation and benchmarking on the web
• …