Geant4 New Features Joseph Perl (SLAC/SCCS) G4NAMU AAPM Minneapolis 22 July 2007

Geant4 New Features - Joseph Perl (SLAC/SCCS)2 Contents Why 8.3 right before 9.0? What was new in release 8.3 and 9.0 Current best practices for medical physics applications Highlights form recent Geant4 External Review Features to come in the next year Extra slides: –What was new in releases 8.0, 8.1, 8.2

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)3 Why 8.3 right before 9.0 Geant4 would not normally have two releases so close to one another –8.3 was released 4 May 20007, –9.0 was release 29 June 2007 Release 8.3 was a special case, driven by the need to fix one specific issue –the Hadronic shower shape –in time for the LHC experiments. These experiments did not want to have an other significant changes from the code they had already extensively tested, release 8.2. –Because they are a major provider of Geant4 funding, they get what they want.

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)4 New in Release 8.3 Revised Hadronic modeling, resulting in longer shower shapes: –the FTF model now has improved treatment of diffraction. –As a result the longitudinal shower component now starts later and ends later, which goes in the direction of the data. –The addition of quasi-elastic scattering (G4QuasiElasticChannel) to the QGS model is also expected to improve shower shapes by reducing the deep inelastic component.

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)5 New in Release 9.0 of interest to Medical Physics Kernel / Geometry –Parallel navigation, first implementation –Revision of biasing/scoring processes –Example RE02, Nested Parameterization

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)6 Parallel navigation In previous versions, we had several somewhat ad-hoc ways of overlaying a parallel world: –Ghost volume for shower parameterization assigned to G4GlobalFastSimulationManager –Readout geometry assigned to G4VSensitiveDetector –Importance field geometry for geometry importance biasing assigned to importance biasing process –Scoring geometry assigned to scoring process Release 9.0 merges all of these into a common parallel world scheme: –Readout geometry for sensitive detector will be kept for backward compatibility. –Other current “ parallel world schemes ” become obsolete.

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)7 Parallel navigation Occasionally, it is not straightforward to define sensitivity, importance or envelope to be assigned to volumes in the mass geometry. –Typically a geometry built machinery by CAD, GDML, DICOM, etc. has this difficulty. New parallel navigation functionality allows the user to define more than one world simultaneously. –New G4Transportation process sees all worlds simultaneously. –A step is limited not only by the boundary of the mass geometry but also by the boundaries of parallel geometries. –Materials, production thresholds and EM field are used only from the mass geometry. –In a parallel world, the user can define volumes in arbitrary manner with sensitivity, regions with shower parameterization, and/or importance field for biasing. Volumes in different worlds may overlap.

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)8 New exampleN07 Mass geometry –sandwich of rectangular absorbers and scintillators Parallel scoring geometry –Cylindrical layers

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)9 Current Best Practices for simulating DICOM-like 3D voxels with material parameterization These are the suggestions from Makoto Asai: –There is no silver bullet. You can try some/all of these options combined. Huge number of cells –If 3D parameterized volume with material parameterization is used, Huge memory size in case of 3D optimization –Instead use the new Nested Parameterization –Example in geant4/examples/RunAndEvent/RE02 Capability has been in Geant4 since release 8.1. Example was added at Release 8.2.

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)10 Material map –Though number of materials may be small, each voxel must at least have a pointer to a material, hence huge map of these pointers. –Split voxel geometry into reasonable number of regions, assign a dedicated stack to each region. For example 5*5*5 = 125 regions. –Load material map (from file on disk) only for one region. If a track reaches to the boundary of the region you are currently simulating, suspend the track. –Simulate all the tracks in one region. Once a region becomes empty, load material map for another region and simulate all tracks in that region. –Note that some tracks may come back to a region you have already simulated. Event biasing –In particular, geometrical importance biasing and secondary particle splitting. –You must validate results of your biasing options with full simulation. Shower parameterization –Instead of having a full EM shower, you may want to consider the shower parameterization in particular for the core part of the shower. Dedicated navigator –For a regular geometry, consider implementing a dedicated navigator that is specialized for regular pattern of voxels. –Geant4 will eventually provide such a navigator option ready-made (comments in a moment) Parallelization –Allocate good number of CPUs … More hints for large numbers of voxels

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)11 General Hints for Speeding Up Geant4 We are trying to improve the speed of Geant4 –But, since it is a general-purpose toolkit for which we give the user great control, it is possible for the user to make the simulation unnecessarily slow. For general applications –Check methods which are invoked frequently: UserSteppingAction() ProcessHits() ComputeTransformation() GetField() etc. –In such methods: avoid string manipulation, file access cout unnecessary object instantiation or deletion unnecessary massive polynomial calculations such as sin(), cos(), log(), exp().

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)12 General Hints for Speeding Up Geant4 For relatively complex geometry or high energy applications –Kill unnecessary secondary particles as soon as possible –Utilize G4Region for regional cut-offs, user limits –For geometry, consider replica rather than parameterized volume as much as possible. Also consider nested parameterization. –Do not keep too many trajectories. For relatively simple geometry or low energy applications –Do not store the random number engine status for each event.

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)13 External Review An external review of Geant4 was held at CERN this Spring

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)14 External Review - Executive Summary

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)15 What New Features can you Expect in Next Year (of interest to Medical Physics Users) Interactive commands to set up scorers –Beta by end of this year –Will be done by the run/event group Penelope Physics –Makoto is now starting work on an interface from Geant4 to the Penelope Fortran code so that Geant4 users can access Penelope physics straight from the Penelope release –Will be distributed from the Penelope web site Dedicated navigator for voxel geometry –Paganetti pioneered this a long time ago –Geant4 has been promising to offer this as a built-in part of the code –We are very aware that the wait for this has been too long Speed improvements –Fermilab group has recently joined the Geant4 collaboration –Specific part of their scope-of-work is to speed up the code software engineers, rather than physicists, taking a look at the code

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)16 Extra Slides What was new in releases 8.0, 8.1, 8.2

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)17 Geant general picture 1.New capabilities –EM (‘standard’) : new MSC model implementation –New design for particle definitions, no longer ‘static’ –Concrete scoring and filter classes –Geometry overlap check at construction 2. Improvements and fixes –to existing hadronic physics modeling & models –in physics process implementations 3.Migrations, updates –Support for CLHEP 2.0.X series –Keeping compatibility with 1.9.X series –Usage of replacing obsolete

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)18 Geant general picture 1.Adoption of a new Geant4 Software License 2.New capabilities –Updated particle definitions to match PDG-2005 –New ‘material scanner’ utility –New ‘tessellated’ solids for interface with CAD systems –ß-release of very low-energy processes in water 3.Numerous improvements in EM “standard” physics –Simulation of back-scattering improved –Tail of angular distribution now material dependent –Visible energy in sampling calorimeters less sensitive to production threshold 3.Fixes in hadronic physics and reviewed physics-lists –New combined elastic-scattering models –Extended use of Binary and Bertini cascade models

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)19 EM Physics New stepping algorithm for multiple- scattering process (since 8.0) –To improve behavior of low energy particles –Significantly reduced cut dependency –More precise visible energy in sampling calorimeters –Limits step size for particles –Extra CPU cost when using same value of production thresholds –Mechanism provided to deactivate step limitation Refinements to multiple scattering process (version 8.1) –Improved back-scattering –Improved scattering on very thin layers –Tail of angular distribution now material dependent –Even less sensitive to production thresholds in sampling calorimeters Reintroduced option to create secondaries below the cut near geometrical boundaries (sub-cutoff) Improved simulation of ionization for ions –Data from NIST databases used

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)20 Physics … more Synchrotron radiation (version 8.1) New process version released: new methods for angular distributions; enhanced to work with local fields and set polarization Transition radiation (version 8.1) New algorithm for XTR angular distribution Multiple scattering does not use table –Needed to ensure repeatability Refined PAI (Photon-Absorption-Ionisation) model Low energy EM (version 8.1) New models for photoelectric angular distribution ß-release of low energy processes in water down to eV scale (Geant4- DNA)

22 July 2007Geant4 New Features - Joseph Perl (SLAC/SCCS)21 More … Data sets –New data set for high precision neutron processes G4NDL 3.9 (release 8.1) –New data set for low-energy EM processes G4EMLOW 4.0 (release 8.1) Examples –New extended examples: TestEm0 - uses G4EmCalculator to print x-sections and stopping power TestEm16 - simulation of synchrotron radiation Hadr01 - simulation of proton/ion beam interaction in water Parameterization/gflash - usage of gflash for shower parameterization Persistency/P01 - object persistency through Reflex –New advanced examples: Microbeam - cellular irradiation beam line at CENBG facility, France Raredecay_calorimetry - photonuclear reactions for photon inefficiency Radiation_monitor - LHC radiation monitoring detectors –Many improvements and fixes …