Transcript Geant4 status and applications John Apostolakis, CERN for Geant4 collaboration
Geant4 status and applications
John Apostolakis, CERN for Geant4 collaboration
Contents
Geant4’s kernel EM Physics processes Comparisons with data Hadronic framework, models Utilization of Geant4 Status and plans Geant4 4.0 to be released Friday 14 th 12th December 2001 2
G
EANT
4
Detector simulation tool-kit for HEP offers alternatives, allows 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 12th December 2001 3
Geant4 Collaboration
12th December 2001 Collaborators also from non member institutions, including Budker Inst. of Physics IHEP Protvino MEPHI Moscow Pittsburg University 4
Geant4 Capabilities
Powerful structure and kernel tracking, stacks, geometry, hits, … Extensive & transparent physics models electromagnetic, hadronic, … Framework for fast simulation Additional capabilities/interfaces persistency, visualization , ...
12th December 2001 5
Geant4 kernel
Creates & manages runs, events , tracks a run is configuration of geometry, physics & event generator.
Allows particles to be prioritized 3 default stacks, postponement.
easily at no cost Tracking is general (unique) physics lists can be tailored for different use cases Enables the creation of user-defined hits able to handle pile-up.
Versatile volumes and navigator for Geometry 12th December 2001 6
Geant4 geometry: what it does
Describes a Detector Hierarchy of volumes Many volumes repeat Volume & sub-tree Up to millions of volumes for LHC era Import detectors from CAD systems Navigates in Detector Locates a point Computes a step Linear intersection Field propagation 12th December 2001 7
Propagating in a field
Charged particles follow paths that approximate their curved trajectories in an electromagnetic field.
It is possible to tailor the accuracy of the splitting linear segments, of the curve into the accuracy of each intersection of a volume boundary, trading accuracy against performance .
These can be set to different values for a single volume or for a hierarchy.
12th December 2001 8
Electromagnetic processes
Initial goal (RD44) was to create EM physics at least at equivalent to Geant-3 New concepts were pursued: Distinction between production threshold and tracking cut Expressing production cuts in terms of range instead of energy Creating an effective range from the distance to the nearest boundary 12th December 2001 9
Electromagnetic processes
Processes new to Geant4 Multiple Scattering : new model includes lateral displacement no path length restriction New process: Transition radiation from physics models Photo-Absorption Ionisation model Alternative energy loss Referred to as PAI model 12th December 2001 10
Electromagnetic physics
Gammas: 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, Photons: Cerenkov, Rayleigh, Reflection, Refraction, Absorption, Scintillation High energy muons and lepton-hadron interactions Alternative implementation (“low energy”) for applications that need to go below 1 KeV Down to 250eV (e+/ g ), O(0.1) m m for hadrons 12th December 2001 11
Shower profile
1 GeV electron in H 2 O G4, Data G3 Good agreement seen with the data 12th December 2001 12
Cuts
, in kernel & user hook
No tracking cut particles are tracked down to zero energy, range Coherent “production cuts” for secondaries validity range of models fully exploited kernel can enforce consistent production cuts yet processes can ask to override when they need to.
treatment of boundary effects (Figures) Cuts in range rather than Energy Geant3 used cuts in Energy - inefficient significant gain in results quality vs CPU usage User can cut in Energy, track length, TOF .. 12th December 2001 14
5 cm Pb, CO 2 , Pb, CO 2
G
EANT
4
Cut: 2 mm Pb 2.5 MeV CO 2 55keV
GEANT3
12th December 2001 15
Changing cuts
Results much more stable in variation of cuts even track length Also see shower profiles for different cuts (next slide) between 10mm and 50 microns 12th December 2001 16
PbWO 4 5 GeV e 12th December 2001 17
Secondaries produced only if they could escape
Lead CO 2 Lead CO 2 Range < safety Secondaries CANNOT leave Pb: NOT produced Range > safety Secondaries could leave Pb: produced 12th December 2001 18
Sampling calorimeter
Sampling calorimeter visible energy tests all EM processes for e-, e+ and photon Data from Sicapo Col. NIM A332 (85-90) 1993 12th December 2001 19
Alternatives for Energy Loss
‘Standard’ differential Extended down to 1 KeV Creates more secondaries near volume borders Photo-absoption Ionisation model increased precision suited for gases/thin absorbers Utilised in ATLAS TRT detector ‘Low-energy’ down to 250 eV with shell effects, … Integral Energy Loss processes Current implementation withdrawn for now (maintenance/uses) 12th December 2001 20
Multiple scattering model
A new model for multiple scattering based on the Lewis theory is implemented since public b release in 1998. It randomizes momentum direction and displacement of a track.
Step length, time of flight, and energy loss along the step are affected, and It does not constrain the step length.
12th December 2001 21
Multiple scattering
Examples of comparisons: 15.7 MeV e on gold foil (figure this page) Thanks to L. Urban 12th December 2001 Angle (deg) 22
Multiple Scattering: One new development
12th December 2001 Thanks to L.Urban
Included in Geant4 4.0
23
Low energy EM processes
Photons, electrons down to 250 eV Xsec from EADL libr.
Thanks to M.G. Pia, P. Nieminen, ..
Hadron EM interactions See 5-001 for an overview of Geant4 Low Energy EM Physics 12th December 2001 E(KeV) Photon transmission through 1 mm Pb, showing shell effects 24
Comparison projects
Established join projects for comparing experiment or test-beam data . Results of EM comparisons: 2000-2001. Geant4 with Collaboration with experiments ATLAS (projects with data of test beams of 4 calorimeters) BaBar (with experiment data for tracker, drift chamber) Following slides are taken from the presentations of the experiments at conferences & workshops For latest example: see presentation of D. Barberis at LC workshop, CERN 15 Nov 2001.
12th December 2001 25
12th December 2001 26
12th December 2001 Thanks to R. Manzini, P. Loch, Atlas FCAL 27
Atlas FCal1 electron Energy resolution
12th December 2001 Thanks to Rachid Manzini & Peter Loch & Atlas FCAL 29
12th December 2001
A 50 GeV e
-
in the Atlas EM barrel
calorimeter 30
Liquid Argon EM Calorimetry
muons in LAr EM Barrel Thanks to G. Parrour, K. Kordas Atlas LAr 12th December 2001 31
Liquid Argon EM Calorimetry
electrons in LAr EM Barrel Linearity of response: - within 0.5% Thanks to Gaston Parrour, Kostas Kordas, Atlas LAr 12th December 2001 32
Liquid Argon EM Calorimetry
• Geant4 describes better than Geant3 energy deposits as measured with muon test beam data – agreement G4 test beam is within 1% • Geant4 (as well as Geant3) describes well the linearity of electron response • Geant4 predicts larger energy resolution for electrons that Geant3. Direct comparison with test beam data still in progress 33 12th December 2001
Liq Argon Hadronic Calorimetry
pion shower in LAr Hadronic End Cap electron energy resolution in LAr Hadronic End Cap 12th December 2001 34
Liquid Argon Hadronic Calorimetry
pion energy resolution in LAr Hadronic End Cap • Electrons: • Geant4 predicts less visible energy in LAr than Geant3 (~3%) and more energy in absorber (~0.1%). Total energy is the same • energy resolution well reproduced by Geant3: Geant4 gives too good resolution • Pions: • first results of simulation with Geant4 look reasonable • more detailed comparisons with test beam data in progress • open questions being discussed with Geant4 people 12th December 2001 35
Tile Calorimeter: electrons
=0.35 in Tilecal module 0 • Electron energy resolution somewhat too good: sampling term 16% instead of 24% (was the same for Geant-3) Electron energy resolution 12th December 2001 • Visible energy vs impact point has the correct shape but amplitude of variations and energy dependence do not match test beam data Visible energy vs impact position for 20 GeV and 100 GeV electrons 36
Observations on the comparisons
Geant4 results are more stable than Geant3 Many results of Geant4 versus data reasonable agreement seen in many observables, issues/differences exist that need to be resolved. Invaluable feedback obtained, with improvements developed and planned.
The good spirit of the collaboration has allowed us to resolve many issues (THANKS).
12th December 2001 37
Hadronic processes
Five level implementation framework allows models to be used in combination at different levels Solving the mix and match problem in the framework Variety of models and cross-sections for each energy regime, particle type, material alternatives with different strengths and computing resource requirements Components can be assembled in an optimized way for each use case.
A simple example is illustrated in figure (next page) 12th December 2001 38
Assembling processes
Element particle Pre-compound model CHIPS Illustrative example of assembling models into an inelastic process for set of particles Uses levels 1 & 2 of framework QGSM Parame terized 12th December 2001 39
Hadronic processes
Each hadronic process may have one or more cross section data sets and final state production models associated with it. Each one has its own applicability.
We define “data set” and “model” broadly A “data set” is an object that encapsulates methods and data for calculating total cross sections.
A “model” is an object that encapsulates methods and data for calculating final states.
12th December 2001 40
Hadronic processes at rest
At Rest processes pion absorption kaon absorption neutron capture antiproton annihilation antineutron annihilation mu capture At Rest processes may generate secondaries after some time interval.
12th December 2001 41
Hadronic processes in flight
Four types of processes Elastic scattering Inelastic scattering Fission Capture Examples Parameterization driven models originally based on GHEISHA with many improvements Data driven models based on ENDF/B-VI Theory driven models for inelastic scattering 12th December 2001 42
Modeling approaches
1. Data driven approach Neutrons from numerous evaluated data libraries down to thermal energies, up to 20 MeV Isotope production (see next slide) Induced Fission & Capture (H.Fesefeldt) used above 20 MeV Photon-evaporation, radioactive decay, etc.
2. Parameterized models Gheisha + fixes + new parameterizations (H.F, TRIUMF) 12th December 2001 43
Modeling approaches (cont.)
3. Theoretical models , from low E to high E Pre-Compound Model + Evaporation Phase Cascades, CHIPS and QMD models String models Excitation, fragmentation, hadronisation models Interface to event generator(s) In future 12th December 2001 44
Pre-Compound Model & Evaporation Phase
Traditional pre-equilibrium model as good as existing ones Evaporation: Weisskopf-Ewing model Fermi breakup model Model for fission Multi-fragmentation model (Bondorf) Photon Evaporation only missing Internal Conversion (suppressed by more than 10^4, funding expected) Future: 2nd Pre-Compound, from HETC re-eng. (in 2002) 12th December 2001 45
Cascade energy range
Parameterized Chiral Invariant Phase Space decay, “ CHIPS ” 1st implementation now (Jefferson Lab.) collaboration milestone 2001 (in release) Bertini cascade (from HETC) collaboration milestone for 2001 (missed) Kinetic model (INFN, Frankfurt): future Further future: Relativistic QMD (Frankfurt), rewrite of INUCL code (Helsinki) 12th December 2001 46
String models
FTF string model, derived from Fritiof but no Rutherford scattering Quark Gluon String model (~ Dual Parton) for proton, neutron, p, K + /K induced reaction string decay as in J ETSET following Kaidalov’s formulation using FTF algorithm for energy transfer in case of single diffraction (~6% cases) future: K 0 , g , anti-nucleon induced reactions 12th December 2001 47
Future additions
Quark molecular dynamics model (Frankfurt) Nucleus-Nucleus via QMD (Frankfurt) for light nuclei using pre-compound and cascades ablation/abration model Parton cascade (ansatz of K. Geiger) ‘Re-use’ of Pythia7 for hadron-nuclear & hadron-hadron interactions 12th December 2001 48
Some of the Improved Hadronic Physics 1999-2001
Neutron & proton induced isotope production models up to 100 MeV (J.P. Wellisch) Multi-fragmentation and pre-compound redesign & refinement (V. Lara) Additional string model (J.P.Wellisch) for proton, neutron, p, K+/K- induced reactions Special cross-section classes for neutron, proton, and ion induced reactions (D. Axen, M. Laidlaw, J.P.W.) Retuning of High Energy Models (H. Fesefeldt) | (JPW) Doppler broadening of neutron X-section on the fly 12th December 2001 49
Isotope Production
Isotopes produced by neutrons on Lead 208 Small dots: evaluated data Circles with error bars: Geant4 latest model included since Geant4 1.0
J.P.Wellisch
50 12th December 2001
CHIPS
New physics model/event generator From ~100 MeV to ~10 GeV Applications to date: Pbar annihilation at rest Pi capture Gamma-nucleus interactions Intranuclear transport after high energy interaction Schedule for release end 2001 12th December 2001 51
Pion capture on
12
C nucleus
12th December 2001 52
Thanks to M. Kossov, J.P. Wellisch, P.V. Degtaryenko
Example gamma nuclear reactions
For more on Hadronic Physics in Geant4, see the presentation in Track 5 by J.P. Wellisch, 5-004 “Hadronic Shower Models in Geant4: Validation strategy and Results” 53 12th December 2001
Atlas Tile Calorimeter: pions
First Results of parameterised models in GEANT4 similar to GHEISHA in GEANT3
E(GeV) 12th December 2001 Thanks to Atlas TileCal 54
Parameterization/Fast Simulation
Fast Simulation Manager Framework for parameterization Takes over from detailed simulation can return to detailed simulation (eg cracks) Can trigger on particle, volume, ..
Parallel geometrical description User must create his/her own (for now) This parameterisation scheme utilised For fast simulation of TR, PAI 12th December 2001 55
Visualization
Much functionality is implemented Several drivers: OpenGL ,
VRML,
Open Inventor , Opacs,
DAWN renderer (G4)
Also choice of User Interfaces: Terminal (text) or GUI: Momo (G4), OPACS Editors for geometry, EM physics code generation 12th December 2001 56
Object Persistency: Hits & other
To store hits, use object persistency Abstract interface ODBMS solution via RD45 (Objectivity) Tracker-type and calorimeter-type hits Saw minimal performance & storage overhead Minimal modifications G4 kernel untouched Also store: Trajectories , Runs, Events, Geometry 12th December 2001 57
Interface to external tools
Through abstract interfaces
No dependence
Minimize coupling of components
Example: AIDA & Analysis Tools AIDA
Courtesy of A. Pfeiffer, CERN
Lizard 12th December 2001 Java Analysis Studio Similar approach: • • • • graphics (G)UI persistency etc.
58
Variance reduction
Geant4 has one event biasing option for the transportation of “low energy” neutrons.
The formulae are derived from MARS95.
Producing small number of secondaries weighted to represent a large number of particles.
• This option is now available for neutrons of E < 5 GeV.
It is possible to use other methods now, but in user code.
New general purpose built-in methods are being created: Importance biasing: splitting/roulette.
Leading particle ‘filters’ X section enhancement Prototype implementations and applications are simple to create, given the flexibility of the toolkit ‘Friendly-user’ use will be available soon – contact us We intend to release them in 2002 – exact time to be determined.
12th December 2001 59
Use of Geant4
Three years from the first Geant4 public release, major experiments has already started to use Geant4 intensively: BaBar: Full migration from G3 based simulator HARP: Running today with G4 as only simulation Atlas : G4 validation for all detector test beams 2000/1 CMS : Simulation of full detector based on G4 in 2001 Usage of Geant4 has expanded to fields other than HEP.
Accelerator applications (T9 Harp, Muon fact.) Space and medical applications 12th December 2001 60
BaBar full simulator ‘BOGUS’
B aBar O bject-oriented G eant4-based U nified S imulator Over 100 million events generated (by mid-Nov.) Robust - Crash rate less than 1 per million events No significant memory leaks Performance as good as BBSIM (Geant3) Comparisons with experimental data and BBSIM, were undertaken, using the full reconstruction chain.
Decision was taken to move to Geant4 and to utilize it exclusively for 2001 data.
Full production with BOGUS is ongoing: plans to simulate 200 million events in 2001, 500 million by early 2002.
12th December 2001 61
12th December 2001 62
12th December 2001 63
Transition Radiation Tracker
support and alignment plate radiator matrix straws carbon fiber shell zoom of module end plate Thanks to Atlas TRT 12th December 2001 64
TRT: Energy Loss in Straws
300 GeV muons 20 GeV pions 20 GeV electrons
12th December 2001 Energy loss measured in ATLAS test beam compared to Geant-3 and to Geant-4 simulations (PAI model) including effects of detector and electronics (K.A.Assamagan): • spectra match reasonably for different particles and energies • some more checks needed for electrons 65 Thanks to Atlas TRT
Cosmic rays, jovian electrons
X-Ray Surveys of Asteroids and Moons
Solar X-rays, e, p
Courtesy SOHO EIT
Induced X-ray line emission: indicator of target composition (~100 m m surface layer) Geant3.21
ITS3.0, EGS4 Geant4 C, N, O line emissions included
ESA Space Environment & Effects Analysis Section
12th December 2001 66
Superficial brachytherapy Leipzig applicators
1,2 1,0 0,8 0,6 0,4 0,2 0,0 0
Brachytherapy
at the Natl. Inst. for Cancer Research (IST-Genova)
10 20
Experimental validation
30 40
Simulation Nucletron
50 -35 -40 -45 -50 -55 -60 0 -40 -5 -10 -15 -20 -25 -30 -30 -20 -10 0 10 20 30 40 80% 60% 40% 20% 10%
12th December 2001
Performance
Geometry navigation Geant4 automatically optimizes the user’s geometrical description. And it provides faster navigation than optimized Geant3 descriptions.
EM Physics computing performance goals For the same physics performance, we seek speed at the level of Geant3 or better.
Keeping physics performance constant, optimise the speed that maintains the performance.
In two current speed benchmarks (thin silicon & simplified sampling calorimeter) these goals are achieved.
12th December 2001 68
Performance
In experimental setups Atlas EM Barrel: comparable performance Atlas FCAL: Geant4 is 3x faster than Geant3 BaBar: comparable performance.
Our goal is that Geant4 is at least as fast as Geant3 in almost every case when its power and features are well exploited.
And, where required, ‘new’ techniques including shower parameterisation can be used to obtain large speedups (in acceptable trade-off with accuracy) 12th December 2001 69
The next release of Geant4
The next release of Geant4 is 4.0, a major release on December 14 th has been scheduled to include New theoretical hadronic models CHIPS for gamma-Nucleus, p capture and intranuclear transport Cascade+Pre-compound re-engineered from HETC (not ready) Ability to reduce initialisation time By saving/retrieving physics processes’ table A field can now be set to any volume or volume tree Overriding a potential global ‘default’ field • Note that, for now, all fields must be addressed in global coordinates Ability to set different Cuts for different regions In a first implementation (for urgent clients) 12th December 2001 70
Next release
New UI option: ‘GAG’ as plug-in version for JAS enabling user to control the execution of Geant4, while taking histograms In Visualization Completion of the ‘new’ visualization commands.
The first release of the XML ==> DAWN file converter Analysis - Example utilising new AIDA interfaces, version 2.2 That includes support for tuple, cloud of points & plotter. In addition all current fixes, and numerous improvements in functionality and physics modelling are also included.
12th December 2001 71
Geant4 architecture
Domain decomposition Software Engineering plays a fundamental role in Geant4 User Requirements • • • formally collected systematically updated PSS-05 standard 12th December 2001 hierarchical structure of • • • spiral iterative approach Software Process regular assessments and improvements monitored following the ISO 15504 model sub-domains Uni-directional • • Object Oriented methods • • OOAD use of CASE tools essential for distributed parallel development contribute to the transparency of physics flow of dependencies • • • • • commercial tools code inspections Quality Assurance automatic checks of coding guidelines testing procedures at unit and integration level dedicated testing team Use of Standards • de jure and de facto 72
Some future plans/ directions
To facilitate the specialization of those parts of physics lists that vary between use cases. To create and distribute “educated guess” physics lists corresponding to the major use cases of Geant4 involving hadronic physics, As an aid and starting point for users.
To enable the use of different models for EM physics (eg Ionisation) in one application likely through the use of ‘models’ etc 12th December 2001 73
http://cern.ch/geant4/
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. 12th December 2001 74
THE END
Note that it was not possible to give appropriate credit here to all those who have contributed to the Geant4 toolkit, the accompanying tools, the topics presented, all the applications featured.
Or to express the appreciation to Users, especially the donors of feedback
THANKS TO ALL
12th December 2001 75