EIC code displayed by LXR master/​include/​celeritas/​em/​distribution/​EnergyLossUrbanDistribution.hh	Source navigation Diff markup Identifier search General search
	Version: master test Revert   Change

File indexing completed on 2025-02-22 10:31:16

 //----------------------------------*-C++-*----------------------------------//
 // Copyright 2022-2024 UT-Battelle, LLC, and other Celeritas developers.
 // See the top-level COPYRIGHT file for details.
 // SPDX-License-Identifier: (Apache-2.0 OR MIT)
 //---------------------------------------------------------------------------//
 //! \file celeritas/em/distribution/EnergyLossUrbanDistribution.hh
 //---------------------------------------------------------------------------//
 #pragma once
 
 #include <cmath>
 
 #include "corecel/Assert.hh"
 #include "corecel/Macros.hh"
 #include "corecel/math/Algorithms.hh"
 #include "celeritas/em/data/FluctuationData.hh"
 #include "celeritas/random/distribution/PoissonDistribution.hh"
 #include "celeritas/random/distribution/UniformRealDistribution.hh"
 
 #include "EnergyLossGaussianDistribution.hh"
 #include "EnergyLossHelper.hh"
 
 namespace celeritas
 {
 //---------------------------------------------------------------------------//
 /*!
  * Sample from the Urban model of energy loss fluctuations in thin layers.
  *
  * The Urban model is used to compute the energy loss fluctuation when \f$
  * \kappa \f$ is small. It assumes atoms have only two energy levels with
  * binding energies \f$ E_1 \f$ and \f$ E_2 \f$ and that the particle-atom
  * interaction will either be an excitation with energy loss \f$ E_1 \f$ or \f$
  * E_2 \f$ or an ionization with energy loss proportional to \f$ 1 / E^2 \f$.
  * The number of collisions for each interaction type has a Poisson
  * distribution with mean proportional to the interaction cross section. The
  * energy loss in a step will be the sum of the energy loss contributions from
  * excitation and ionization.
  *
  * When the number of ionizations is larger than a given threshold, a fast
  * sampling method can be used instead. The possible energy loss interval is
  * divided into two parts: a lower range in which the number of collisions is
  * large and the energy loss can be sampled from a Gaussian distribution, and
  * an upper range in which the energy loss is sampled for each collision.
  *
  * See section 7.3.2 of the Geant4 Physics Reference Manual and GEANT3 PHYS332
  * section 2.4 for details.
  */
 class EnergyLossUrbanDistribution
 {
   public:
     //!@{
     //! \name Type aliases
     using FluctuationRef = NativeCRef<FluctuationData>;
     using Energy = units::MevEnergy;
     using Mass = units::MevMass;
     //!@}
 
   public:
     // Construct from particle properties
     inline CELER_FUNCTION
     EnergyLossUrbanDistribution(FluctuationRef const& shared,
                                 MaterialTrackView const& cur_mat,
                                 Energy unscaled_mean_loss,
                                 Energy max_energy,
                                 Mass two_mebsgs,
                                 real_type beta_sq);
 
     // Construct from helper-calculated data
     explicit inline CELER_FUNCTION
     EnergyLossUrbanDistribution(EnergyLossHelper const& helper);
 
     // Sample energy loss according to the distribution
     template<class Generator>
     inline CELER_FUNCTION Energy operator()(Generator& rng);
 
   private:
     //// TYPES ////
 
     using Real2 = Array<real_type, 2>;
 
     //// DATA ////
 
     real_type max_energy_;
     real_type loss_scaling_;
     Real2 binding_energy_;
     Real2 xs_exc_;
     real_type xs_ion_;
 
     //// CONSTANTS ////
 
     //! Relative contribution of ionization to energy loss
     static CELER_CONSTEXPR_FUNCTION real_type rate() { return 0.56; }
 
     //! Number of collisions above which to use faster sampling from Gaussian
     static CELER_CONSTEXPR_FUNCTION size_type max_collisions() { return 8; }
 
     //! Threshold number of excitations used in width correction
     static CELER_CONSTEXPR_FUNCTION real_type exc_thresh() { return 42; }
 
     //! Ionization energy [MeV]
     static CELER_CONSTEXPR_FUNCTION real_type ionization_energy()
     {
         return value_as<Energy>(EnergyLossHelper::ionization_energy());
     }
 
     //! Energy point below which FWHM scaling doesn't change [MeV]
     static CELER_CONSTEXPR_FUNCTION real_type fwhm_min_energy()
     {
         return 1e-3;  // 1 keV
     }
 
     //// HELPER FUNCTIONS ////
 
     template<class Engine>
     CELER_FUNCTION real_type sample_excitation_loss(Engine& rng);
 
     template<class Engine>
     CELER_FUNCTION real_type sample_ionization_loss(Engine& rng);
 
     template<class Engine>
     static CELER_FUNCTION real_type sample_fast_urban(real_type mean,
                                                       real_type stddev,
                                                       Engine& rng);
 };
 
 //---------------------------------------------------------------------------//
 // INLINE DEFINITIONS
 //---------------------------------------------------------------------------//
 /*!
  * Construct from distribution parameters.
  */
 CELER_FUNCTION EnergyLossUrbanDistribution::EnergyLossUrbanDistribution(
     FluctuationRef const& shared,
     MaterialTrackView const& cur_mat,
     Energy unscaled_mean_loss,
     Energy max_energy,
     Mass two_mebsgs,
     real_type beta_sq)
     : max_energy_(max_energy.value())
 {
     CELER_EXPECT(unscaled_mean_loss > zero_quantity());
     CELER_EXPECT(two_mebsgs > zero_quantity());
     CELER_EXPECT(beta_sq > 0);
 
     // Width correction: the FWHM of the energy loss distribution in thin
     // layers is in most cases too small; this width correction is used to get
     // more accurate FWHM values while keeping the mean loss the same by
     // rescaling the energy levels and number of excitations. The width
     // correction algorithm is discussed (though not in much detail) in PRM
     // section 7.3.3
     loss_scaling_
         = real_type(0.5)
               * min(this->fwhm_min_energy() / max_energy_, real_type(1))
           + real_type(1);
     real_type const mean_loss = unscaled_mean_loss.value() / loss_scaling_;
 
     // Material-dependent data
     CELER_ASSERT(cur_mat.material_id() < shared.urban.size());
     UrbanFluctuationParameters const& params
         = shared.urban[cur_mat.material_id()];
     binding_energy_ = params.binding_energy;
     xs_exc_ = {0, 0};
 
     // Calculate the excitation macroscopic cross sections and apply the width
     // correction
     auto const& mat = cur_mat.make_material_view();
     if (max_energy_ > value_as<Energy>(mat.mean_excitation_energy()))
     {
         // Common term in the numerator and denominator of PRM Eq. 7.10
         // two_mebsgs = 2 * m_e c^2 * beta^2 * gamma^2
         real_type const w = std::log(value_as<units::MevMass>(two_mebsgs))
                             - beta_sq;
         real_type const w_0
             = value_as<units::LogMevEnergy>(mat.log_mean_excitation_energy());
         if (w > w_0)
         {
             if (w > params.log_binding_energy[1])
             {
                 real_type const c = mean_loss * (1 - this->rate()) / (w - w_0);
                 for (int i : range(2))
                 {
                     // Excitation macroscopic cross section (PRM Eq. 7.10)
                     xs_exc_[i] = c * params.oscillator_strength[i]
                                  * (w - params.log_binding_energy[i])
                                  / params.binding_energy[i];
                 }
             }
             else
             {
                 xs_exc_[0] = mean_loss * (1 - this->rate())
                              / params.binding_energy[0];
             }
 
             // Scale the binding energy and macroscopic cross section (i.e.,
             // the mean number of excitations)
             real_type scaling = 4;
             if (xs_exc_[0] < this->exc_thresh())
             {
                 scaling = real_type(0.5)
                           + (scaling - real_type(0.5))
                                 * std::sqrt(xs_exc_[0] / this->exc_thresh());
             }
             binding_energy_[0] *= scaling;
             xs_exc_[0] /= scaling;
         }
     }
 
     // Calculate the ionization macroscopic cross section (PRM Eq. 7.11)
     constexpr real_type e_0 = EnergyLossUrbanDistribution::ionization_energy();
     xs_ion_ = mean_loss * (max_energy_ - e_0)
               / (max_energy_ * e_0 * std::log(max_energy_ / e_0));
     if (xs_exc_[0] + xs_exc_[1] > 0)
     {
         // The contribution from excitation is nonzero, so scale the ionization
         // cross section accordingly
         xs_ion_ *= this->rate();
     }
 }
 
 //---------------------------------------------------------------------------//
 /*!
  * Construct from helper-calculated data.
  */
 CELER_FUNCTION EnergyLossUrbanDistribution::EnergyLossUrbanDistribution(
     EnergyLossHelper const& helper)
     : EnergyLossUrbanDistribution(helper.shared(),
                                   helper.material(),
                                   helper.mean_loss(),
                                   helper.max_energy(),
                                   helper.two_mebsgs(),
                                   helper.beta_sq())
 {
 }
 
 //---------------------------------------------------------------------------//
 /*!
  * Sample energy loss according to the distribution.
  */
 template<class Generator>
 CELER_FUNCTION auto
 EnergyLossUrbanDistribution::operator()(Generator& rng) -> Energy
 {
     // Calculate actual energy loss from the loss contributions from excitation
     // and ionization
     real_type result = this->sample_excitation_loss(rng)
                        + this->sample_ionization_loss(rng);
 
     CELER_ENSURE(result >= 0);
     return Energy{loss_scaling_ * result};
 }
 
 //---------------------------------------------------------------------------//
 /*!
  * Calculate the energy loss contribution from excitation for the Urban model.
  */
 template<class Engine>
 CELER_FUNCTION real_type
 EnergyLossUrbanDistribution::sample_excitation_loss(Engine& rng)
 {
     real_type result = 0;
 
     // Mean and variance for fast sampling from Gaussian
     real_type mean = 0;
     real_type variance = 0;
 
     for (int i : range(2))
     {
         if (xs_exc_[i] > this->max_collisions())
         {
             // When the number of collisions is large, use faster approach
             // of sampling from a Gaussian
             mean += xs_exc_[i] * binding_energy_[i];
             variance += xs_exc_[i] * ipow<2>(binding_energy_[i]);
         }
         else if (xs_exc_[i] > 0)
         {
             // The loss due to excitation is \f$ \Delta E_{exc} = n_1 E_1 + n_2
             // E_2 \f$, where the number of collisions \f$ n_i \f$ is sampled
             // from a Poisson distribution with mean \f$ \Sigma_i \f$
             unsigned int n = PoissonDistribution<real_type>(xs_exc_[i])(rng);
             if (n > 0)
             {
                 UniformRealDistribution<real_type> sample_fraction(n - 1,
                                                                    n + 1);
                 result += sample_fraction(rng) * binding_energy_[i];
             }
         }
     }
     if (variance > 0)
     {
         // Sample excitation energy loss contribution from a Gaussian
         result += this->sample_fast_urban(mean, std::sqrt(variance), rng);
     }
     return result;
 }
 
 //---------------------------------------------------------------------------//
 /*!
  * Calculate the energy loss contribution from ionization for the Urban model.
  */
 template<class Engine>
 CELER_FUNCTION real_type
 EnergyLossUrbanDistribution::sample_ionization_loss(Engine& rng)
 {
     real_type result = 0;
 
     constexpr real_type e_0 = EnergyLossUrbanDistribution::ionization_energy();
     real_type const energy_ratio = max_energy_ / e_0;
 
     // Parameter that determines the upper limit of the energy interval in
     // which the fast simulation is used
     real_type alpha = 1;
 
     // Mean number of collisions in the fast simulation interval
     real_type mean_num_coll = 0;
 
     if (xs_ion_ > this->max_collisions())
     {
         // When the number of collisions is large, fast sampling from a
         // Gaussian is used in the lower portion of the energy loss interval.
         // See PHYS332 section 2.4: Fast simulation for \f$ n_3 \ge 16 \f$
 
         // Calculate the maximum value of \f$ \alpha \f$ (Eq. 25)
         alpha = (xs_ion_ + this->max_collisions()) * energy_ratio
                 / (this->max_collisions() * energy_ratio + xs_ion_);
 
         // Mean energy loss for a single collision of this type (Eq. 14)
         real_type const mean_loss_coll = alpha * std::log(alpha) / (alpha - 1);
 
         // Mean number of collisions of this type (Eq. 16)
         mean_num_coll = xs_ion_ * energy_ratio * (alpha - 1)
                         / ((energy_ratio - 1) * alpha);
 
         // Mean and standard deviation of the total energy loss (Eqs. 18-19)
         real_type const mean = mean_num_coll * mean_loss_coll * e_0;
         real_type const stddev
             = e_0 * std::sqrt(xs_ion_ * (alpha - ipow<2>(mean_loss_coll)));
 
         // Sample energy loss from a Gaussian distribution
         result += this->sample_fast_urban(mean, stddev, rng);
     }
     if (xs_ion_ > 0 && energy_ratio > alpha)
     {
         // Sample number of ionizations from a Poisson distribution with mean
         // \f$ n_3 - n_A \f$, where \f$ n_3 \f$ is the number of ionizations
         // and \f$ n_A \f$ is the number of ionizations in the energy interval
         // in which the fast sampling from a Gaussian is used
         PoissonDistribution<real_type> sample_num_ioni(xs_ion_ - mean_num_coll);
 
         // Add the contribution from ionizations in the energy interval in
         // which the energy loss is sampled for each collision (Eq. 20)
         UniformRealDistribution<real_type> sample_fraction(
             alpha / energy_ratio, 1);
         for (auto n = sample_num_ioni(rng); n > 0; --n)
         {
             result += alpha * e_0 / sample_fraction(rng);
         }
     }
     return result;
 }
 
 //---------------------------------------------------------------------------//
 /*!
  * Fast sampling of energy loss in thin absorbers from a Gaussian distribution.
  */
 template<class Engine>
 CELER_FUNCTION real_type EnergyLossUrbanDistribution::sample_fast_urban(
     real_type mean, real_type stddev, Engine& rng)
 {
     if (stddev <= 4 * mean)
     {
         EnergyLossGaussianDistribution sample_eloss(Energy{mean},
                                                     Energy{stddev});
         return value_as<Energy>(sample_eloss(rng));
     }
     else
     {
         UniformRealDistribution<real_type> sample(0, 2 * mean);
         return sample(rng);
     }
 }
 
 //---------------------------------------------------------------------------//
 }  // namespace celeritas

This page was automatically generated by the 2.3.7 LXR engine. The LXR team