Geant4—a simulation toolkit

Models, methods and technologies Maria Grazia Pia INFN Genova, Italy [email protected] M. Batič1,2, P. Saracco1, G. Weidenspointner3 1INFN Sezione di Genova, Italy 2Jozef Stefan Institute, Ljubljana, Slovenia 3Max-Planck-Institut für extraterrestrische Physik and Halbleiterlabor, Germany CAARI 2012 Maria Grazia Pia, INFN Genova Fort Worth, Texas, USA 5-10 August 2012 1 S. Agostinelli et al. Geant4: a simulation toolkit NIM A, vol. 506, no. 3, pp. 250-303, 2003 3137 citations Most cited CERN publication in Web of Science℠ (excluding Rev. Part. Properties) Nuclear Science & Technology Instruments & Instrumentation J. P. Biersack and L. G. Haggmark, 3627 citations A Monte-Carlo computer-program for the transport of energetic ions in amorphous targets NIM, vol.174, no. 1-2, pp. 257-269, 1980 L. R. Doolittle, 2071 citations Algorithms for the rapid simulation of Rutherford backscattering spectra NIM B, vol. 9, no. 3, pp. 344-351, 1985 Source: Thomson-Reuters, Web of Science℠ (7 August 2012) Maria Grazia Pia, INFN Genova 2 Monte Carlo simulation in literature Fraction of published papers Fraction of IEEE TNS papers mentioning Monte Carlo or simulation 0.6 0.5 0.4 0.3 0.2 Same trend in NIM 0.1 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 0.0 Year M. G. Pia, T. Basaglia, Z.W. Bell, P.V. Dressendorfer The butterfly effect: correlations between modeling in nuclear-particle physics and socioeconomic factors Maria Grazia Pia, INFN Genova IEEE NSS 2010 Conf. Rec. 3 Courtesy K. Amako et al., KEK Courtesy CMS Collaboration Object oriented toolkit for the simulation of particle interactions with matter Born from the requirements of large scale HEP experiments Courtesy ATLAS Collaboration Courtesy H. Araujo and A. Howard, IC London Widely used in: § Space science and astrophysics § Medical physics, nuclear medicine § Radiation protection § Accelerator physics § Pest control, food irradiation § Humanitarian projects, security § etc. § Technology transfer to industry, hospitals… IST and INFN Genova Courtesy GATE Collaboration ZEPLIN III Courtesy R. Nartallo et al.,ESA Courtesy Borexino Maria Grazia Pia, INFN Genova 4 Boolean operations Courtesy of ATLAS Collaboration Geometry " Role ATLAS – detailed detector description ~5.2 M volume objects ~110 K volume types – efficient navigation across volumes " One can do fancy things with Geant4 geometry… 1 GeV proton in the earth’s geomagnetic field ISS e.m. fields Courtesy Laurent Desorgher, University of Bern Courtesy T. Ersmark, KTH Stockholm Maria Grazia Pia, INFN Genova RADMON, CERN 5 Physics Electromagnetic § electrons and positrons § photons (including optical photons) § muons § charged hadrons § ions Multiple scattering Bremsstrahlung Ionisation Annihilation Photoelectric effect Compton scattering Rayleigh scattering γ conversion Synchrotron radiation Transition radiation Cherenkov Refraction Reflection Absorption Scintillation Fluorescence Auger emission Hadronic §Theory-driven § Parameterized § Data-driven Maria Grazia Pia, INFN Genova 6 Geant4 Visualization Readout Geant4 architecture! Persistency Interfaces Run Event Open source code Freely available Tracking Source code, libraries, data files, documentation Digis Hits Processes http://cern.ch/geant4 Track Particle Geometry Graphic Reps Developed and managed by an international team of scientists Material Global Maria Grazia Pia, INFN Genova Intercoms Open to contributions from the scientific community 7 PIXE in Geant4 (and other codes) " Specialized codes for PIXE analysis (and simulation) – GeoPIXE, GUPIX, PIXAN, PixeKLM, Sapix,WinAxil, Wits-HEX etc. – DATTPIXE, DT2, LibCPIXE, VIBA-Lab1/2 etc. – Concerned with experimental observables relevant to material analysis " Impact ionisation by electrons – EGS*, Geant4, Penelope – Cross sections available for all shells " " " " Geant4 R&D with PIXE started in 2001 Major step forward in 2008 Further refinements in progress Open to evolution: OO technology + physics Maria Grazia Pia, INFN Genova 8 eROSITA extended Roentgen Survey with an Imaging Telescope Array Systematic detection and analysis of all obscured accreting Black Holes in nearby galaxies and of ten thousand galaxy clusters On-board the Russian "Spectrum-Roentgen-Gamma" (SRG) satellite to be launched from Baikonur in 2013 and placed in an L2 orbit Maria Grazia Pia, INFN Genova 9 PIXE Particle Induced X-ray Emission Particle = proton, ion (electron) The physics of PIXE is simple The simulation of PIXE is NOT simple interplay of continuous and discrete transport methods Maria Grazia Pia, INFN Genova 10 PIXE simulation recipe Ingredients " Proton (ion) ionisation cross sections for individual shells Specific to PIXE " " Radiative and non-radiative transition probabilities Atomic binding energies (X-ray energies) Instructions " Create a vacancy in the shell occupancy due to ionisation " A touch of spice De-excite the atom through fluorescence and Auger emission Infrared divergence of the cross section for δ-ray emission Monte Carlo codes neutralize the spice by introducing “cuts” (i.e. a threshold for δ-ray production) Below the threshold: ionisation treated as continuous energy loss Above the threshold: emission of an electron PIXE is and intrinsically discrete process, which is intimately associated with a continuous-discrete one One of the ingredients (ionisation cross sections) Note Maria Grazia Pia, INFN Genova is available in limited quantity (for a few inner shells only) 11 Condensed-discrete ionisation Depending on the δ-ray production cut, the “total” cross section may be smaller than the K shell ionisation cross section L K ECPSSR (ISICS) Si ionisation by protons Maria Grazia Pia, INFN Genova Can we deal with PIXE consistently? " What can we do with existing instruments? " Longer term R&D on transport methods 12 Determination of a vacancy Relative probability of ionising a shell w.r.t. the other shells But we do not know the total ionisation cross section (without introducing a dependency on cuts) We can calculate ionisation cross sections only for K, L, M shells OK for light elements OK for relative estimates Overestimate inner shell vacancies for heavier elements In principle, the issue would be solved if we could calculate ionisation cross sections for all shells Maria Grazia Pia, INFN Genova 13 Domain decomposition No dependence of PIXE component on δ-ray production threshold Dependency on “cut” only in client ionisation process Reuse in other context Use Geant4 Atomic Relaxation component Data-driven model Performance! Maria Grazia Pia, INFN Genova Order of magnitude faster than analytical calculation 14 Proton, K shell " " " " " " " Proton, L shell " " " " " " PWBA ECPSSR ECPSSR United Atom Miyagawa et al. Orlic et al. Sow et al. Proton, M shell " " " Ionisation cross sections PWBA ECPSSR ECPSSR Hartree-Slater ECPSSR United Atom ECPSSR relativistic Paul & Sacher Kahoul et al. PWBA ECPSSR ECPSSR United Atom Maria Grazia Pia, INFN Genova " Theoretical " Empirical Released as PIXE 2010 data library Thanks to RSICC! α, K shell " " " " " PWBA ECPSSR ECPSSR Hartree-Slater ECPSSR United Atom Paul & Bolik α, L and M shell " " " PWBA ECPSSR ECPSSR United Atom Theoretical cross sections tabulated by ISICS (thanks to S. Cipolla!) Subject to rigorous experimental validation 15 Experimental validation, K shell Experimental data from Paul & Sacher compilation Ta F Ionisation cross sections X-ray production cross sections Maria Grazia Pia, INFN Genova 16 K shell, protons χ2 test: p-values One test case per element, for each σ model Maria Grazia Pia, INFN Genova 17 χ2 test: % test cases with p-value > 0.05 ECPSSR ECPSSR rel. ECPSSR-HS ECPSSR-UA Paul-Sacher All data 67 77 74 68 Kahoul 71 46 70 48 Exclude higher energy data 69 75 86 69 Contingency tables: α = 0.05 ECPSSR with Hartree-Slater correction best at low energy All data K shell χ2 test χ2 test Conclusions based on objective, quantitative analysis Maria Grazia Pia, INFN Genova Exclude high E ECPSSR OK High energy: scarce measurements to really test ECPSSR rel. Pass Fail p-value Fisher p-value Pearson χ2 p-value Yates χ2 Pass Fail p-value Fisher p-value Pearson χ2 p-value Yates χ2 χ2 test Pass Fail p-value Fisher p-value Pearson χ2 p-value Yates χ2 ECPSSR ECPSSR-HS 44 50 21 15 0.327 0.240 0.327 ECPSSR ECPSSR-HS 44 55 20 9 0.034 0.020 0.035 Kahoul et al. ECPSSR-HS 30 50 35 15 0.001 < 0.001 18 0.001 Same validation process for L shell Experimental data L1 Z=67 Compilations by Sokhi & Crumpton Orlic et al. Sow & Tang L2 Z=79 Maria Grazia Pia, INFN Genova L3 Z=47 19 Contingency tables: α = 0.05 L shell χ2 test χ2 test: % with p-value > 0.05 L1 L2 L3 ECPSSR 64 68 89 ECPSSR UA 64 79 89 Miyagawa et al. 26 78 85 Orlic et al. 41 70 56 Sow et al. 56 85 81 ECPSSR - United Atom best compatibility with experimental data ECPSSR OK Orlic et al.: worst compatibility with experimental data Maria Grazia Pia, INFN Genova ECPSSR-UA ECPSSR Pass 65 62 Fail 19 22 p-value Fisher 0.720 p-value Pearson χ2 0.590 p-value Yates χ2 0.719 χ2 test ECPSSR-UA Miyagawa Pass 65 51 Fail 19 30 p-value Fisher 0.060 2 p-value Pearson χ 0.043 p-value Yates χ2 0.063 χ2 test Pass Fail p-value Fisher p-value Pearson χ2 p-value Yates χ2 χ2 test Pass Fail p-value Fisher p-value Pearson χ2 p-value Yates χ2 ECPSSR-UA 65 19 0.005 0.003 0.005 ECPSSR-UA 65 19 0.717 0.620 0.754 Orlic 45 36 Sow 60 21 20 M shell The same analysis will be performed once adequate experimental data are available… Collaboration with experimental groups is welcome Maria Grazia Pia, INFN Genova 21 K, L, M shell ionisation PIXE 2010 Cross section data library (p, α) Publicly distributed by RSICC and NEA http://www-rsicc.ornl.gov/codes/dlc/dlc2/dlc-246.html Theoretical calculations (ISICS) • PWBA • ECPSSR • ECPSSR – United Atom • ECPSSR – Hartree-Slater • ECPSSR – relativistic Fitted, K shell • Paul & Sacher (p) • Kahoul et al. (p) • Paul & Bolik (α) Maria Grazia Pia, INFN Genova http://www.oecd-nea.org/tools/abstract/detail/nea-1868/ Fitted, L shell • Miyagawa et al. • Sow et al. • Orlic et al. 22 Which ECPSSR calculator? W. Brandt and G. Lapicki, “Energy-loss effect in inner-shell Coulomb ionization by heavy charged particles,” Phys. Rev. A, vol. 23, pp. 1717–1729, 1981 " ISICS (C) (C++) – Z. Liu and S. J. Cipolla, “ISICS: A program for calculating K-, L- and M-shell cross sections from ECPSSR theory using a personal computer,” Comp. Phys. Comm., vol. 97, no. 3, pp. 315–330, 1996 – and following updates " ERCS08 (FORTRAN) – V. Horvat, “ERCS08: A FORTRAN program equipped with a Windows graphics user interface that calculates ECPSSR cross sections for the removal of atomic electrons,” Comp. Phys. Comm., vol. 180, no. 6, pp. 995–1003, 2009. " Z. Šmit’s code KIO/LIO (FORTRAN) – Z.Šmit,“K-shell ECPSSR cross sections for analytical applications,” NIM. B, vol. 36, no. 3, pp. 254–258, 1989 – Code kindly provided by the author Relative comparison to experimental data Maria Grazia Pia, INFN Genova 23 Quantification of relative accuracy 200 1800 ERCS08 KIO w.r.t. ISICS2011 1400 1000 800 150 125 100 75 600 50 400 25 200 0 0 -0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 50 K shell Al 40 30 20 10 0 -2 10 10 -1 1 Energy (MeV) Maria Grazia Pia, INFN Genova 0.08 -0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 0.1 (σISICS - σERCS08,LIO) / σISICS (!ISICS - !ERCS08,KIO) / !ISICS Cross section (kb) Entries 1200 L3 shell ERCS08 LIO w.r.t. ISICS2011 175 K shell Entries 1600 10 " χ2 test of compatibility with experiment over individual elements, for each calculator " Pass/fail: α = 0.05 " Categorical analysis to compare compatibility with experiment – Fisher’s exact test, χ2 test 24 χ2 test Statistical analysis 350 K shell 300 χ2 test ISICS 2011 Pass 51 K shell Fail 15 p-value Fisher 0551 2 p-value Pearson χ 0.426 p-value Yates χ2 0.550 Entries 250 200 150 100 50 0 -5 -4 -3 -2 -1 0 1 2 3 4 5 (!calculated - !experimental) / error Number of standard deviations w.r.t. experiment 60 L3 shell 50 40 Entries ISICS 2011 ERCS08 Pass 51 51 K shell Fail 15 15 p-value Fisher 1 p-value Pearson χ2 1 2 p-value Yates χ 0.835 KIO 47 19 Statistically equivalent accuracy χ2 test ISICS 2011 ERCS08 Pass 66 64 L shell Fail 18 20 p-value Fisher 1 p-value Pearson χ2 1 p-value Yates χ2 0.835 30 χ2 test 20 10 0 -5 -4 -3 -2 -1 0 1 2 (!calculated - !experimental) / error Maria Grazia Pia, INFN Genova 3 4 5 ISICS 2011 Pass 66 L shell Fail 18 p-value Fisher 0.161 2 p-value Pearson χ 0.219 2 p-value Yates χ 0.160 LIO 58 26 25 ISICS reengineering ISICS reengineered into a C++ class Can be used as a component in other systems Generated improvements to ISICS Windows version Maria Grazia Pia, INFN Genova 26 Geant4 atomic relaxation Geant4 simulation based on EADL Validation of X-ray energies against Des Lattes et al. experimental review compilation 2007 Maria Grazia Pia, INFN Genova 27 Atomic Binding Energies • • • • • • • 2011 Bearden and Burr Carlson EADL Sevier (1979) used by GUPIX Table of Isotopes, 7th edition (1978): Shirley + Uppsala Group Table of Isotopes, 8th edition(1996): Larkins Williams (X-ray Data Booklet, CRC Handbook) 10 600 Note the different scale! 7.5 500 400 2.5 Difference (eV) Difference (eV) 5 0 -2.5 Carlson ToI 1996 ToI 1978 Sevier -5 -7.5 -10 0 10 20 30 Maria Grazia Pia, INFN Genova 40 50 60 Atomic number 70 80 300 EADL 200 100 0 90 100 -100 0 10 20 30 40 50 60 Atomic number 70 80 90 100 28 Do they make any difference? p impact ionisation σECPSSR X-ray energies 0.1 1 % difference w.r.t. DesLattes et al 0.75 Relative difference (%) Relative difference KL2transition 0.5 0.25 0 -0.25 -0.5 EADL Carlson ToI 1996 ToI 1978 Williams Sevier -0.75 -1 -1.25 -1.5 C, K shell 0.05 0 -0.05 Carlson EADL -0.1 -0.15 differerence w.r.t. using Bearden & Burr -0.2 -2 10 10 20 30 40 50 60 70 80 90 10 -1 100 1 10 E (MeV) Atomic number Z t-test for no mean difference, α = 0.05 χ2 test ToI 1996 EADL Pass 44 16 Fail 4 32 p-value Fisher <0.0001 2 p-value MariaYates GraziaχPia, INFN<0.0001 Genova statistically significant EADL inaccuracy χ2 test w.r.t. experiment, α = 0.05 χ2 test Bearden-Burr EADL Pass 51 39 Fail 16 28 p-value Fisher p-value Pearson χ2 p-value Yates χ2 0.042 0.027 0.043 29 Radiative transition probabilities 2009 Hartree-Slater (EADL) vs. Hartree-Fock calculation methods χ2 test % test cases with p-value > 0.05 χ2 test Comparison with experimental data Hartree-Slater Hartree-Fock Pass 8 16 Fail 9 1 p-value Fisher 0.007 p-value Yates χ2 0.003 Transition probabilities calculated with Hartree-Fock methods are significantly more accurate Maria Grazia Pia, INFN Genova 30 3 cm Cu shield eROSITA shielding optimization Cu+ 1 mm Al shield eROSITA CCD detector @ L2 Cu+Al+1 mm B4C shield http://www.ge.infn.it/geant4/physics/pixe/erosita.html Maria Grazia Pia, INFN Genova Geant4 advanced example 31 empirical 1st cycle α Geant4Abdelouahed ECPSSR Geant4pii Z=29 empiricalp p Z=29 p GUPIX Maria Grazia Pia, INFN Genova 32 In my end is my beginning… T.S. Eliot, East Coker Geant4 provides functionality for PIXE simulation Rigorous, quantitative validation of all the physics components specific to PIXE simulation The software is adequate for experimental applications concerned with relative effects of experimentally relevant spectra Successfully used for eROSITA shielding optimisation Conceptual issues of condensed-discrete transport are being investigated Collaboration is welcome! http://cern.ch/geant4/ Maria Grazia Pia, INFN Genova http://www.ge.infn.it/geant4/ 33