barbara mascialinogeant4 workshopcatania, october 4-9 2004 electromagnetic physics validation...

Barbara Mascialino Geant4 Workshop Catania, October 4-9 2004

Electromagnetic physicsvalidation

Katsuya Amako,Susanna Guatelli, Vladimir Ivanchenko, Michel Maire, Barbara

Mascialino, Koichi Murakami, Sandra Parlati, Andreas Pfeiffer, Maria Grazia Pia, Takashi

Sasaki, Lazslo Urban

Geant4 Workshop Catania, October 4th-9th 2004


The project• The project is based on a geographically spread

collaboration:

INFN Genova

INFN Gran Sasso

Standard Group

KEK THANKS TO KOICHI MURAKAMI, TAKASHI

SASAKI, KATSUYA AMAKO FOR THE VERY FRUITFUL

COLLABORATION!

Preliminary results were presented at last Geant4 Workshop and at IEEE-NSS in Portland. Now the project has reached a mature state.


Aim of the project

• Project for the validation of all Geant4 electromagnetic models against established references

• The project s made-up by two parts:

PHYSICAL TEST

GOODNESS-OF-FIT TESTING

test50

Goodness-of-Fit statistical toolkit

Quantitative statistical comparisons allow:- an evaluation of Geant4 physics goodness- how the specific models behave in the same experimental condition

POSSIBILITY OF CHOOSING THE MOST APPROPRIATE MODEL

Chi-squared stability study


Photon Attenuation Coefficient

Photon Cross Sections (attenuation coefficients with only one process activated)

Electrons CSDA range and Stopping Power (no multiple scattering, no energy fluctuations)

Protons CSDA range and Stopping Power (no multiple scattering, no energy fluctuations)

Alpha particles CSDA range and Stopping Power

(no multiple scattering, no energy fluctuations)

First phase: validation against the NIST database

Elements: Be, Al, Si, Fe, Ge, Ag, Cs, Au, Pb, U Energy range: 1 keV – 100 GeV

Testing activity has been automatised (thanks to Sandra Parlati and Koichi Murakami)


Photons: attenuation coefficient

χ2/ν stability study

Be Z dependency?


Photon attenuation coefficient: statistical results

χ2 /ν p-value χ2 /ν p-value χ2 /ν p-value

Be 1.26 0.16 0.08 1 1.01 0.45

Al 0.31 1 0.07 1 0.34 1

Si 0.32 1 0.10 1 0.56 0.97

Fe 0.11 1 0.08 1 0.20 1

Ge 0.15 1 0.05 1 0.28 1

Ag 0.22 1 0.15 1 0.45 1

Cs 0.19 1 0.37 1 0.81 0.75

Au 0.08 1 0.05 1 0.71 0.86

Pb 0.13 1 0.08 1 0.52 0.99

U 0.04 1 0.06 1 0.75 0.93

NIST – XCOMLowE Livermore

NIST – XCOMLowE Penelope

NIST – XCOMStandard


Photons: photoelectric cross section


Be Cs

Z dependency?


Photon photoelectric cross section: statistical results


Be 0.05 1 0.05 1 0.82 0.63

Al 0.05 1 0.03 1 0.27 1

Si 0.07 1 0.02 1 0.17 1

Fe 0.09 1 0.05 1 0.11 1

Ge 0.21 1 0.08 1 0.07 1

Ag 0.07 1 0.05 1 0.07 1

Cs 0.52 0.94 0.21 1 1.18 0.27

Au 0.14 1 0.15 1 0.57 0.93

Pb 0.25 1 0.35 1 0.51 0.96

U 0.35 1 0.37 0.99 0.42 0.99





Photons: Compton cross section

The 1keV deviation effect is evident in both LowE Penelope and Standard packages

χ2/ν = with the

1 keV point

without the 1 keV point

LowE Penelope

17.30 0.46

Standard 9.0 1.65

As an example, let us consider Ag:


Photon Compton cross section: statistical results


Be 0.12 1 0.07 1 0.53 0.85

Al 0.07 1 0.18 1 0.49 0.86

Si 1.61 0.09 0.15 1 0.60 0.77

Fe 0.30 1 0.69 0.75 0.27 0.99

Ge 0.13 1 1.68 0.07 0.38 0.96

Ag 0.09 1 0.65 0.79 0.46 0.92

Cs 0.14 1 0.36 0.97 0.62 0.72

Au 0.16 1 0.41 0.94 1.04 0.41

Pb 0.23 1 1.31 0.21 1.18 0.29

U 0.11 1 0.13 1 0.08 1





Compton cross sections χ2/ν stability study

(without the E=1 keV point)

χ2/ν stability studySi

Pb

Ge

Au


Photons: pair production cross section

Beryllium


Be(not compatible with the NIST)

deviations



Photon pair production cross section: statistical results


Be 0.05 1 8.07 <0.001 7.01 <0.001

Al 0.07 1 0.83 0.61 1.07 0.39

Si 0.09 1 0.76 0.69 0.93 0.52

Fe 0.07 1 0.32 0.99 0.30 0.99

Ge 0.04 1 0.25 1 0.18 1

Ag 0.06 1 0.15 1 0.23 1

Cs 0.24 1 0.54 0.89 0.30 0.99

Au 0.08 1 0.09 1 0.23 1

Pb 0.11 1 0.08 1 0.22 1

U 0.09 1 1.18 1 0.26 0.99




Rem

ovin

g th

e 1

keV

poi

nt


Photons: Rayleigh cross section

deviations



Ge

PbU

Au

Fe

Si

Al


Photon Rayleigh cross section: statistical results

χ2 /ν p-value χ2 /ν p-value

Be 0.27 0.99 0.17 1

Al 1.15 0.97 24.86 <0.001

Si 0.68 0.77 25.08 <0.001

Fe 0.27 1 11.08 <0.001

Ge 17.97 <0.001 1.06 0.39

Ag 1.10 0.36 1.74 0.08

Au 1.90 0.05 18.99 <0.001

Pb 8.40 <0.001 22.10 <0.001

U 11.31 <0.001 28.19 <0.001



Tes

t re

sults

are

not

con

sist

ent


• The disagreement between NIST reference data and data coming from the recent library EPDL97 (provided by Lawrence Livermore National Laboratory) within the range of energies between 1 keV and 1 MeV has been already underlined and discussed in a recent paper by Zaidi*.

• In his paper Zaidi concluded that EPDL97 is the most up-dated, complete and consistent data library available at the moment.

For these features, it should be considered as a standard.

Critical discussion of this result

* Zaidi H., 2000, Comparative evaluation of photon cross section libraries for materials of interest in PET Monte Carlo simulation IEEE Transaction on Nuclear Science 47 2722-35


Electrons: stopping powerχ2/ν stability study

NIST – ESTARLowE Livermore

NIST – ESTARLowE Penelope

NIST – ESTARStandard

χ2/ν = -0.032 + 0.0074 Z R2=0.995 p<0.0001

χ2/ν = -0.032 + 0.0074 Z R2=0.995 p<0.0001

χ2/ν = -0.046 + 0.0073 Z R2=0.989 p<0.0001

The three models are equivalent

Strange effect (as a function of Z)

BESTFIT

BESTFIT

BESTFIT


Electrons stopping power: statistical results


Be 0.03 1 0.02 1 0.02 1

Al 0.06 1 0.05 1 0.06 1

Si 0.07 1 0.06 1 0.06 1

Fe 0.15 1 0.14 1 0.13 1

Ge 0.18 1 0.17 1 0.15 1

Cs 0.36 1 0.35 1 0.31 1

Au 0.57 0.96 0.56 0.96 0.54 0.97

Pb 0.58 0.95 0.57 0.96 0.56 0.97

U 0.65 0.91 0.64 0.92 0.64 0.92





Electrons: CSDA range


Ag(to be explained)

The three models are equivalent



Be 0.21 1 0.24 1 0.17 1

Al 0.05 1 0.05 1 0.04 1

Si 0.04 1 0.04 1 0.04 1

Fe 0.02 1 0.02 1 0.02 1

Ge 0.02 1 0.01 1 0.02 1

Ag 0.10 1 1.06 0.39 0.02 1

Cs 0.009 1 0.006 1 0.008 1

Au 0.005 1 0.004 1 0.005 1

Pb 0.005 1 0.004 1 0.004 1

U 0.004 1 0.003 1 0.003 1




Electrons CSDA range: statistical results


• LowE Ziegler 85• LowE Ziegler 2000• ICRU

• Standard

At low energies: free electron gas modelAt middle energies (~ MeV): parametrisationsAt high energies: Bethe Bloch

Statistical comparison cannot lead to a real physics validation, but we can only compare two different models (NIST – Ziegler)

Protons and alpha particles

NIST database


Protons: stopping power

χ2/ν stability studyLowE ICRU StandardLowE Ziegler 85lOWe Ziegler 2000NIST - PSTAR


Protons stopping power: statistical results

χ2 /ν p χ2 /ν p χ2 /ν p χ2 /ν p

Be 0.008 1 0.41 1 0.66 0.94 0.01 1

Al 0.02 1 0.05 1 0.06 1 0.02 1

Si 0.008 1 0.41 1 0.81 0.79 0.01 1

Fe 0.02 1 0.03 1 0.03 1 0.02 1

Ge 0.12 1 0.53 1 0.72 0.89 0.11 1

Ag 0.06 1 0.15 1 0.10 1 0.04 1

Au 0.08 1 0.27 1 0.17 1 0.09 1

Pb 0.06 1 0.12 1 0.17 0.21 0.05 1

U 0.10 1 1.28 0.12 1.28 0.12 0.09 1

NIST – PSTARLowE ICRU49

NIST – PSTARLowE Ziegler85

NIST – PSTARStandard



Protons: CSDA range


LowE ICRU StandardLowE Ziegler 85LowE Ziegler 2000NIST - PSTAR


Protons CSDA range: statistical results

χ2 /ν p χ2 /ν p χ2 /ν p χ2 /ν p

Be 0.005 1 0.11 1 1.03 0.41 0.47 1

Al 0.03 1 0.01 1 0.01 1 0.27 1

Si 0.02 1 0.29 1 0.32 0.99 0.23 1

Fe 0.06 1 0.20 1 0.18 1 0.32 1

Ge 0.13 1 0.26 1 0.27 1 0.24 1

Ag 0.08 1 0.34 1 0.15 1 0.15 1

Au 0.09 1 0.91 0.62 0.47 1 0.14 1

Pb 0.09 1 0.21 1 0.17 0.99 0.12 1

U 0.11 1 1.23 0.16 1.18 0.21 0.13 1

NIST – PSTARLowE ICRU49


NIST – PSTARStandard



Alpha particles: stopping powerWORK IN PROGRESS

LowE ICRU StandardLowE Ziegler 77NIST - ASTAR


Alpha particles: CSDA rangeWORK IN PROGRESS

LowE ICRU StandardLowE Ziegler 77NIST - ASTAR


Statistical comparisons (II)

Concerning alpha particles, this is the second iteration of production and analysis since last July.

This because thanks to the quantitative analysis we could detect a conceptual flaw in physics tables treatment for both protons and alpha particles.

Systematic data analysis allowed to improve the physical models.


• Low Energy Livermore is the most compatible with the NIST reference (Rayleigh scattering is a special case)

• Low Energy Penelope is quite compatible with NIST reference except for some problems exhibited in Compton scattering and pair production cross sections

• Standard electrons are compatible with NIST, photons are quite compatible, but exhibit some problems

SUMMARY: photons and electrons


• While NIST represents an established reference for photon and electron processes, the reference for protons and alpha processes in controversial at least in the lower energy ranges.

• Two reference data compilations ICRU/NIST and Ziegler.

• Quantitative comparisons available for all NIST quantities for protons and alpha particles.

SUMMARY: protons and alpha particles


Conclusions

• Validation of all Geant4 Electromagnetic models against the NIST database

• Quantitative statistical analysis on all the comparisons

• Fully automated testing system (thanks to Sandra Parlati and Koichi Murakami)

• Objective comparison among Geant4 models (with respect to the NIST reference)

• Mature project and results will be presented at IEEE-NSS – paper submitted for publication next month


Future perspectives

• Final states

• angular distributions and spectra

• The first results will be shown and discussed in the parallel section Physics Book introductory talk by Susanna Guatelli

barbara mascialinogeant4 workshopcatania, october 4-9 2004 electromagnetic physics validation...

Documents

pvalue be0

pvalue be1

koichi murakami slide

nist deviations

project project

photon compton cross

statistical results

kev point lowe penelope