| 1 | //
|
|---|
| 2 | // ********************************************************************
|
|---|
| 3 | // * License and Disclaimer *
|
|---|
| 4 | // * *
|
|---|
| 5 | // * The Geant4 software is copyright of the Copyright Holders of *
|
|---|
| 6 | // * the Geant4 Collaboration. It is provided under the terms and *
|
|---|
| 7 | // * conditions of the Geant4 Software License, included in the file *
|
|---|
| 8 | // * LICENSE and available at http://cern.ch/geant4/license . These *
|
|---|
| 9 | // * include a list of copyright holders. *
|
|---|
| 10 | // * *
|
|---|
| 11 | // * Neither the authors of this software system, nor their employing *
|
|---|
| 12 | // * institutes,nor the agencies providing financial support for this *
|
|---|
| 13 | // * work make any representation or warranty, express or implied, *
|
|---|
| 14 | // * regarding this software system or assume any liability for its *
|
|---|
| 15 | // * use. Please see the license in the file LICENSE and URL above *
|
|---|
| 16 | // * for the full disclaimer and the limitation of liability. *
|
|---|
| 17 | // * *
|
|---|
| 18 | // * This code implementation is the result of the scientific and *
|
|---|
| 19 | // * technical work of the GEANT4 collaboration. *
|
|---|
| 20 | // * By using, copying, modifying or distributing the software (or *
|
|---|
| 21 | // * any work based on the software) you agree to acknowledge its *
|
|---|
| 22 | // * use in resulting scientific publications, and indicate your *
|
|---|
| 23 | // * acceptance of all terms of the Geant4 Software license. *
|
|---|
| 24 | // ********************************************************************
|
|---|
| 25 | //
|
|---|
| 26 | // $Id: G4LivermoreComptonModel.cc,v 1.7 2009/06/10 13:32:36 mantero Exp $
|
|---|
| 27 | // GEANT4 tag $Name: geant4-09-03-cand-01 $
|
|---|
| 28 | //
|
|---|
| 29 | //
|
|---|
| 30 | // Author: Sebastien Inserti
|
|---|
| 31 | // 30 October 2008
|
|---|
| 32 | //
|
|---|
| 33 | // History:
|
|---|
| 34 | // --------
|
|---|
| 35 | // 18 Apr 2009 V Ivanchenko Cleanup initialisation and generation of secondaries:
|
|---|
| 36 | // - apply internal high-energy limit only in constructor
|
|---|
| 37 | // - do not apply low-energy limit (default is 0)
|
|---|
| 38 | // - remove GetMeanFreePath method and table
|
|---|
| 39 | // - added protection against numerical problem in energy sampling
|
|---|
| 40 | // - use G4ElementSelector
|
|---|
| 41 |
|
|---|
| 42 | #include "G4LivermoreComptonModel.hh"
|
|---|
| 43 |
|
|---|
| 44 | //....oooOO0OOooo........oooOO0OOooo........oooOO0OOooo........oooOO0OOooo....
|
|---|
| 45 |
|
|---|
| 46 | using namespace std;
|
|---|
| 47 |
|
|---|
| 48 | //....oooOO0OOooo........oooOO0OOooo........oooOO0OOooo........oooOO0OOooo....
|
|---|
| 49 |
|
|---|
| 50 | G4LivermoreComptonModel::G4LivermoreComptonModel(const G4ParticleDefinition*,
|
|---|
| 51 | const G4String& nam)
|
|---|
| 52 | :G4VEmModel(nam),isInitialised(false),meanFreePathTable(0),
|
|---|
| 53 | scatterFunctionData(0),crossSectionHandler(0)
|
|---|
| 54 | {
|
|---|
| 55 | lowEnergyLimit = 250 * eV;
|
|---|
| 56 | highEnergyLimit = 100 * GeV;
|
|---|
| 57 | // SetLowEnergyLimit(lowEnergyLimit);
|
|---|
| 58 | SetHighEnergyLimit(highEnergyLimit);
|
|---|
| 59 |
|
|---|
| 60 | verboseLevel=0 ;
|
|---|
| 61 | // Verbosity scale:
|
|---|
| 62 | // 0 = nothing
|
|---|
| 63 | // 1 = warning for energy non-conservation
|
|---|
| 64 | // 2 = details of energy budget
|
|---|
| 65 | // 3 = calculation of cross sections, file openings, sampling of atoms
|
|---|
| 66 | // 4 = entering in methods
|
|---|
| 67 |
|
|---|
| 68 | if( verboseLevel>0 ) {
|
|---|
| 69 | G4cout << "Livermore Compton model is constructed " << G4endl
|
|---|
| 70 | << "Energy range: "
|
|---|
| 71 | << lowEnergyLimit / eV << " eV - "
|
|---|
| 72 | << highEnergyLimit / GeV << " GeV"
|
|---|
| 73 | << G4endl;
|
|---|
| 74 | }
|
|---|
| 75 | }
|
|---|
| 76 |
|
|---|
| 77 | //....oooOO0OOooo........oooOO0OOooo........oooOO0OOooo........oooOO0OOooo....
|
|---|
| 78 |
|
|---|
| 79 | G4LivermoreComptonModel::~G4LivermoreComptonModel()
|
|---|
| 80 | {
|
|---|
| 81 | if (crossSectionHandler) delete crossSectionHandler;
|
|---|
| 82 | if (scatterFunctionData) delete scatterFunctionData;
|
|---|
| 83 | }
|
|---|
| 84 |
|
|---|
| 85 | //....oooOO0OOooo........oooOO0OOooo........oooOO0OOooo........oooOO0OOooo....
|
|---|
| 86 |
|
|---|
| 87 | void G4LivermoreComptonModel::Initialise(const G4ParticleDefinition* particle,
|
|---|
| 88 | const G4DataVector& cuts)
|
|---|
| 89 | {
|
|---|
| 90 | if (verboseLevel > 3)
|
|---|
| 91 | G4cout << "Calling G4LivermoreComptonModel::Initialise()" << G4endl;
|
|---|
| 92 |
|
|---|
| 93 | if (crossSectionHandler)
|
|---|
| 94 | {
|
|---|
| 95 | crossSectionHandler->Clear();
|
|---|
| 96 | delete crossSectionHandler;
|
|---|
| 97 | }
|
|---|
| 98 |
|
|---|
| 99 | // Reading of data files - all materials are read
|
|---|
| 100 |
|
|---|
| 101 | crossSectionHandler = new G4CrossSectionHandler;
|
|---|
| 102 | crossSectionHandler->Clear();
|
|---|
| 103 | G4String crossSectionFile = "comp/ce-cs-";
|
|---|
| 104 | crossSectionHandler->LoadData(crossSectionFile);
|
|---|
| 105 |
|
|---|
| 106 | G4VDataSetAlgorithm* scatterInterpolation = new G4LogLogInterpolation;
|
|---|
| 107 | G4String scatterFile = "comp/ce-sf-";
|
|---|
| 108 | scatterFunctionData = new G4CompositeEMDataSet(scatterInterpolation, 1., 1.);
|
|---|
| 109 | scatterFunctionData->LoadData(scatterFile);
|
|---|
| 110 |
|
|---|
| 111 | // For Doppler broadening
|
|---|
| 112 | shellData.SetOccupancyData();
|
|---|
| 113 | G4String file = "/doppler/shell-doppler";
|
|---|
| 114 | shellData.LoadData(file);
|
|---|
| 115 |
|
|---|
| 116 | if (verboseLevel > 2)
|
|---|
| 117 | G4cout << "Loaded cross section files for Livermore Compton model" << G4endl;
|
|---|
| 118 |
|
|---|
| 119 | InitialiseElementSelectors(particle,cuts);
|
|---|
| 120 |
|
|---|
| 121 | if( verboseLevel>0 ) {
|
|---|
| 122 | G4cout << "Livermore Compton model is initialized " << G4endl
|
|---|
| 123 | << "Energy range: "
|
|---|
| 124 | << LowEnergyLimit() / eV << " eV - "
|
|---|
| 125 | << HighEnergyLimit() / GeV << " GeV"
|
|---|
| 126 | << G4endl;
|
|---|
| 127 | }
|
|---|
| 128 | //
|
|---|
| 129 | if(isInitialised) return;
|
|---|
| 130 | fParticleChange = GetParticleChangeForGamma();
|
|---|
| 131 | isInitialised = true;
|
|---|
| 132 | }
|
|---|
| 133 |
|
|---|
| 134 | //....oooOO0OOooo........oooOO0OOooo........oooOO0OOooo........oooOO0OOooo....
|
|---|
| 135 |
|
|---|
| 136 | G4double G4LivermoreComptonModel::ComputeCrossSectionPerAtom(
|
|---|
| 137 | const G4ParticleDefinition*,
|
|---|
| 138 | G4double GammaEnergy,
|
|---|
| 139 | G4double Z, G4double,
|
|---|
| 140 | G4double, G4double)
|
|---|
| 141 | {
|
|---|
| 142 | if (verboseLevel > 3)
|
|---|
| 143 | G4cout << "Calling ComputeCrossSectionPerAtom() of G4LivermoreComptonModel" << G4endl;
|
|---|
| 144 |
|
|---|
| 145 | if (GammaEnergy < lowEnergyLimit || GammaEnergy > highEnergyLimit) return 0.0;
|
|---|
| 146 |
|
|---|
| 147 | G4double cs = crossSectionHandler->FindValue(G4int(Z), GammaEnergy);
|
|---|
| 148 | return cs;
|
|---|
| 149 | }
|
|---|
| 150 |
|
|---|
| 151 | //....oooOO0OOooo........oooOO0OOooo........oooOO0OOooo........oooOO0OOooo....
|
|---|
| 152 |
|
|---|
| 153 | void G4LivermoreComptonModel::SampleSecondaries(std::vector<G4DynamicParticle*>* fvect,
|
|---|
| 154 | const G4MaterialCutsCouple* couple,
|
|---|
| 155 | const G4DynamicParticle* aDynamicGamma,
|
|---|
| 156 | G4double, G4double)
|
|---|
| 157 | {
|
|---|
| 158 |
|
|---|
| 159 | // The scattered gamma energy is sampled according to Klein - Nishina formula.
|
|---|
| 160 | // then accepted or rejected depending on the Scattering Function multiplied
|
|---|
| 161 | // by factor from Klein - Nishina formula.
|
|---|
| 162 | // Expression of the angular distribution as Klein Nishina
|
|---|
| 163 | // angular and energy distribution and Scattering fuctions is taken from
|
|---|
| 164 | // D. E. Cullen "A simple model of photon transport" Nucl. Instr. Meth.
|
|---|
| 165 | // Phys. Res. B 101 (1995). Method of sampling with form factors is different
|
|---|
| 166 | // data are interpolated while in the article they are fitted.
|
|---|
| 167 | // Reference to the article is from J. Stepanek New Photon, Positron
|
|---|
| 168 | // and Electron Interaction Data for GEANT in Energy Range from 1 eV to 10
|
|---|
| 169 | // TeV (draft).
|
|---|
| 170 | // The random number techniques of Butcher & Messel are used
|
|---|
| 171 | // (Nucl Phys 20(1960),15).
|
|---|
| 172 |
|
|---|
| 173 | G4double photonEnergy0 = aDynamicGamma->GetKineticEnergy();
|
|---|
| 174 |
|
|---|
| 175 | if (verboseLevel > 3) {
|
|---|
| 176 | G4cout << "G4LivermoreComptonModel::SampleSecondaries() E(MeV)= "
|
|---|
| 177 | << photonEnergy0/MeV << " in " << couple->GetMaterial()->GetName()
|
|---|
| 178 | << G4endl;
|
|---|
| 179 | }
|
|---|
| 180 |
|
|---|
| 181 | // low-energy gamma is absorpted by this process
|
|---|
| 182 | if (photonEnergy0 <= lowEnergyLimit)
|
|---|
| 183 | {
|
|---|
| 184 | fParticleChange->ProposeTrackStatus(fStopAndKill);
|
|---|
| 185 | fParticleChange->SetProposedKineticEnergy(0.);
|
|---|
| 186 | fParticleChange->ProposeLocalEnergyDeposit(photonEnergy0);
|
|---|
| 187 | return ;
|
|---|
| 188 | }
|
|---|
| 189 |
|
|---|
| 190 | G4double e0m = photonEnergy0 / electron_mass_c2 ;
|
|---|
| 191 | G4ParticleMomentum photonDirection0 = aDynamicGamma->GetMomentumDirection();
|
|---|
| 192 |
|
|---|
| 193 | // Select randomly one element in the current material
|
|---|
| 194 | // G4int Z = crossSectionHandler->SelectRandomAtom(couple,photonEnergy0);
|
|---|
| 195 | const G4ParticleDefinition* particle = aDynamicGamma->GetDefinition();
|
|---|
| 196 | const G4Element* elm = SelectRandomAtom(couple,particle,photonEnergy0);
|
|---|
| 197 | G4int Z = (G4int)elm->GetZ();
|
|---|
| 198 |
|
|---|
| 199 | G4double epsilon0 = 1. / (1. + 2. * e0m);
|
|---|
| 200 | G4double epsilon0Sq = epsilon0 * epsilon0;
|
|---|
| 201 | G4double alpha1 = -std::log(epsilon0);
|
|---|
| 202 | G4double alpha2 = 0.5 * (1. - epsilon0Sq);
|
|---|
| 203 |
|
|---|
| 204 | G4double wlPhoton = h_Planck*c_light/photonEnergy0;
|
|---|
| 205 |
|
|---|
| 206 | // Sample the energy of the scattered photon
|
|---|
| 207 | G4double epsilon;
|
|---|
| 208 | G4double epsilonSq;
|
|---|
| 209 | G4double oneCosT;
|
|---|
| 210 | G4double sinT2;
|
|---|
| 211 | G4double gReject;
|
|---|
| 212 |
|
|---|
| 213 | do
|
|---|
| 214 | {
|
|---|
| 215 | if ( alpha1/(alpha1+alpha2) > G4UniformRand())
|
|---|
| 216 | {
|
|---|
| 217 | // std::pow(epsilon0,G4UniformRand())
|
|---|
| 218 | epsilon = std::exp(-alpha1 * G4UniformRand());
|
|---|
| 219 | epsilonSq = epsilon * epsilon;
|
|---|
| 220 | }
|
|---|
| 221 | else
|
|---|
| 222 | {
|
|---|
| 223 | epsilonSq = epsilon0Sq + (1. - epsilon0Sq) * G4UniformRand();
|
|---|
| 224 | epsilon = std::sqrt(epsilonSq);
|
|---|
| 225 | }
|
|---|
| 226 |
|
|---|
| 227 | oneCosT = (1. - epsilon) / ( epsilon * e0m);
|
|---|
| 228 | sinT2 = oneCosT * (2. - oneCosT);
|
|---|
| 229 | G4double x = std::sqrt(oneCosT/2.) / (wlPhoton/cm);
|
|---|
| 230 | G4double scatteringFunction = scatterFunctionData->FindValue(x,Z-1);
|
|---|
| 231 | gReject = (1. - epsilon * sinT2 / (1. + epsilonSq)) * scatteringFunction;
|
|---|
| 232 |
|
|---|
| 233 | } while(gReject < G4UniformRand()*Z);
|
|---|
| 234 |
|
|---|
| 235 | G4double cosTheta = 1. - oneCosT;
|
|---|
| 236 | G4double sinTheta = std::sqrt (sinT2);
|
|---|
| 237 | G4double phi = twopi * G4UniformRand() ;
|
|---|
| 238 | G4double dirx = sinTheta * std::cos(phi);
|
|---|
| 239 | G4double diry = sinTheta * std::sin(phi);
|
|---|
| 240 | G4double dirz = cosTheta ;
|
|---|
| 241 |
|
|---|
| 242 | // Doppler broadening - Method based on:
|
|---|
| 243 | // Y. Namito, S. Ban and H. Hirayama,
|
|---|
| 244 | // "Implementation of the Doppler Broadening of a Compton-Scattered Photon
|
|---|
| 245 | // into the EGS4 Code", NIM A 349, pp. 489-494, 1994
|
|---|
| 246 |
|
|---|
| 247 | // Maximum number of sampling iterations
|
|---|
| 248 | G4int maxDopplerIterations = 1000;
|
|---|
| 249 | G4double bindingE = 0.;
|
|---|
| 250 | G4double photonEoriginal = epsilon * photonEnergy0;
|
|---|
| 251 | G4double photonE = -1.;
|
|---|
| 252 | G4int iteration = 0;
|
|---|
| 253 | G4double eMax = photonEnergy0;
|
|---|
| 254 | do
|
|---|
| 255 | {
|
|---|
| 256 | iteration++;
|
|---|
| 257 | // Select shell based on shell occupancy
|
|---|
| 258 | G4int shell = shellData.SelectRandomShell(Z);
|
|---|
| 259 | bindingE = shellData.BindingEnergy(Z,shell);
|
|---|
| 260 |
|
|---|
| 261 | eMax = photonEnergy0 - bindingE;
|
|---|
| 262 |
|
|---|
| 263 | // Randomly sample bound electron momentum
|
|---|
| 264 | // (memento: the data set is in Atomic Units)
|
|---|
| 265 | G4double pSample = profileData.RandomSelectMomentum(Z,shell);
|
|---|
| 266 | // Rescale from atomic units
|
|---|
| 267 | G4double pDoppler = pSample * fine_structure_const;
|
|---|
| 268 | G4double pDoppler2 = pDoppler * pDoppler;
|
|---|
| 269 | G4double var2 = 1. + oneCosT * e0m;
|
|---|
| 270 | G4double var3 = var2*var2 - pDoppler2;
|
|---|
| 271 | G4double var4 = var2 - pDoppler2 * cosTheta;
|
|---|
| 272 | G4double var = var4*var4 - var3 + pDoppler2 * var3;
|
|---|
| 273 | if (var > 0.)
|
|---|
| 274 | {
|
|---|
| 275 | G4double varSqrt = std::sqrt(var);
|
|---|
| 276 | G4double scale = photonEnergy0 / var3;
|
|---|
| 277 | // Random select either root
|
|---|
| 278 | if (G4UniformRand() < 0.5) photonE = (var4 - varSqrt) * scale;
|
|---|
| 279 | else photonE = (var4 + varSqrt) * scale;
|
|---|
| 280 | }
|
|---|
| 281 | else
|
|---|
| 282 | {
|
|---|
| 283 | photonE = -1.;
|
|---|
| 284 | }
|
|---|
| 285 | } while ( iteration <= maxDopplerIterations &&
|
|---|
| 286 | (photonE < 0. || photonE > eMax || photonE < eMax*G4UniformRand()) );
|
|---|
| 287 |
|
|---|
| 288 | // End of recalculation of photon energy with Doppler broadening
|
|---|
| 289 | // Revert to original if maximum number of iterations threshold has been reached
|
|---|
| 290 |
|
|---|
| 291 | if (iteration >= maxDopplerIterations)
|
|---|
| 292 | {
|
|---|
| 293 | photonE = photonEoriginal;
|
|---|
| 294 | bindingE = 0.;
|
|---|
| 295 | }
|
|---|
| 296 |
|
|---|
| 297 | // Update G4VParticleChange for the scattered photon
|
|---|
| 298 |
|
|---|
| 299 | G4ThreeVector photonDirection1(dirx,diry,dirz);
|
|---|
| 300 | photonDirection1.rotateUz(photonDirection0);
|
|---|
| 301 | fParticleChange->ProposeMomentumDirection(photonDirection1) ;
|
|---|
| 302 |
|
|---|
| 303 | G4double photonEnergy1 = photonE;
|
|---|
| 304 |
|
|---|
| 305 | if (photonEnergy1 > 0.)
|
|---|
| 306 | {
|
|---|
| 307 | fParticleChange->SetProposedKineticEnergy(photonEnergy1) ;
|
|---|
| 308 | }
|
|---|
| 309 | else
|
|---|
| 310 | {
|
|---|
| 311 | photonEnergy1 = 0.;
|
|---|
| 312 | fParticleChange->SetProposedKineticEnergy(0.) ;
|
|---|
| 313 | fParticleChange->ProposeTrackStatus(fStopAndKill);
|
|---|
| 314 | }
|
|---|
| 315 |
|
|---|
| 316 | // Kinematics of the scattered electron
|
|---|
| 317 | G4double eKineticEnergy = photonEnergy0 - photonEnergy1 - bindingE;
|
|---|
| 318 |
|
|---|
| 319 | // protection against negative final energy: no e- is created
|
|---|
| 320 | if(eKineticEnergy < 0.0) {
|
|---|
| 321 | fParticleChange->ProposeLocalEnergyDeposit(photonEnergy0 - photonEnergy1);
|
|---|
| 322 | return;
|
|---|
| 323 | }
|
|---|
| 324 | G4double eTotalEnergy = eKineticEnergy + electron_mass_c2;
|
|---|
| 325 |
|
|---|
| 326 | G4double electronE = photonEnergy0 * (1. - epsilon) + electron_mass_c2;
|
|---|
| 327 | G4double electronP2 = electronE*electronE - electron_mass_c2*electron_mass_c2;
|
|---|
| 328 | G4double sinThetaE = -1.;
|
|---|
| 329 | G4double cosThetaE = 0.;
|
|---|
| 330 | if (electronP2 > 0.)
|
|---|
| 331 | {
|
|---|
| 332 | cosThetaE = (eTotalEnergy + photonEnergy1 )* (1. - epsilon) / std::sqrt(electronP2);
|
|---|
| 333 | sinThetaE = -1. * sqrt(1. - cosThetaE * cosThetaE);
|
|---|
| 334 | }
|
|---|
| 335 |
|
|---|
| 336 | G4double eDirX = sinThetaE * std::cos(phi);
|
|---|
| 337 | G4double eDirY = sinThetaE * std::sin(phi);
|
|---|
| 338 | G4double eDirZ = cosThetaE;
|
|---|
| 339 |
|
|---|
| 340 | G4ThreeVector eDirection(eDirX,eDirY,eDirZ);
|
|---|
| 341 | eDirection.rotateUz(photonDirection0);
|
|---|
| 342 |
|
|---|
| 343 | // SI - The range test has been removed wrt original G4LowEnergyCompton class
|
|---|
| 344 |
|
|---|
| 345 | fParticleChange->ProposeLocalEnergyDeposit(bindingE);
|
|---|
| 346 |
|
|---|
| 347 | G4DynamicParticle* dp = new G4DynamicParticle (G4Electron::Electron(),
|
|---|
| 348 | eDirection,eKineticEnergy) ;
|
|---|
| 349 | fvect->push_back(dp);
|
|---|
| 350 | }
|
|---|
| 351 |
|
|---|