42 const int max_iter = 50;
44 double e_previous_step = 0.0;
45 const double tolerance = 5.e-4;
47 for (iter = 0; iter < max_iter; iter++) {
50 if (std::abs(
e_ - e_previous_step) < tolerance) {
53 const double gamma_inv = std::sqrt(1.0 -
v_.
sqr());
57 auto T_mub_mus = eos.
solve_eos(
e_, gamma_inv * nb_, gamma_inv *
ns_);
70 if (iter == max_iter) {
71 std::cout <<
"Warning from solver: max iterations exceeded." 72 <<
" Accuracy: " << std::abs(
e_ - e_previous_step)
73 <<
" is less than tolerance " << tolerance << std::endl;
91 return out <<
"T[mu,0]: " << node.
Tmu0() <<
", nb: " << node.
nb()
92 <<
", ns: " << node.
ns() <<
", v: " << node.
v()
93 <<
", e: " << node.
e() <<
", p: " << node.
p()
94 <<
", T: " << node.
T() <<
", mub: " << node.
mub()
95 <<
", mus: " << node.
mus();
99 const std::array<int, 3> n_cells,
100 const std::array<double, 3> origin,
101 bool periodicity,
double e_critical,
102 double t_start,
double delta_t,
104 : eos_typelist_(list_eos_particles()),
105 N_sorts_(eos_typelist_.size()),
111 lat_ = make_unique<RectangularLattice<ThermLatticeNode>>(
112 lat_sizes, n_cells, origin, periodicity, upd);
113 const std::array<double, 3> abc =
lat_->cell_sizes();
122 bool ignore_cells_under_treshold) {
126 for (
auto &node : *
lat_) {
130 if (!ignore_cells_under_treshold ||
131 node.Tmu0().x0() + std::abs(node.Tmu0().x1()) +
132 std::abs(node.Tmu0().x2()) + std::abs(node.Tmu0().x3()) >=
134 node.compute_rest_frame_quantities(
eos_);
143 +0.5 *
lat_->cell_sizes()[0]),
145 +0.5 *
lat_->cell_sizes()[1]),
147 +0.5 *
lat_->cell_sizes()[2]));
151 ParticleList &plist,
const FourVector required_total_momentum) {
152 const auto &log = logger<LogArea::GrandcanThermalizer>();
156 log.info(
"Required 4-momentum: ", required_total_momentum);
157 log.info(
"Sampled 4-momentum: ", conserved.momentum());
159 (required_total_momentum.
threevec() - conserved.momentum().threevec()) /
161 log.info(
"Adjusting momenta by ", mom_to_add);
162 for (
auto &particle : plist) {
163 particle.set_4momentum(particle.type().mass(),
164 particle.momentum().threevec() + mom_to_add);
169 const ThreeVector beta_CM_generated = conserved.momentum().velocity();
173 double E_expected = required_total_momentum.
abs();
174 for (
auto &particle : plist) {
175 particle.boost_momentum(beta_CM_generated);
176 E += particle.momentum().x0();
180 double a, a_min, a_max, er;
181 const int max_iter = 50;
183 if (E_expected >= E) {
191 a = 0.5 * (a_min + a_max);
193 for (
const auto &particle : plist) {
194 const double p2 = particle.momentum().threevec().sqr();
195 const double E2 = particle.momentum().x0() * particle.momentum().x0();
196 E += std::sqrt(E2 + a * (a + 2.0) * p2);
204 log.debug(
"Iteration ", iter,
": a = ", a,
", Δ = ", er);
206 }
while (std::abs(er) > tolerance && iter < max_iter);
208 log.info(
"Renormalizing momenta by factor 1+a, a = ", a);
209 for (
auto &particle : plist) {
210 particle.set_4momentum(particle.type().mass(),
211 (1 + a) * particle.momentum().threevec());
212 particle.boost_momentum(-beta_CM_required);
223 for (
size_t i_type = 0; (i_type < N_sorts_) && (N_to_sample > 0); i_type++) {
224 if (
get_class(i_type) != particle_class) {
248 const double gamma = 1.0 / std::sqrt(1.0 - cell.
v().
sqr());
249 const double N_this_cell =
257 for (
int i = 0; i <
mult_int_[type_index]; i++) {
260 double partial_sum = 0.0;
261 int index_only_thermalized = -1;
262 while (partial_sum < r) {
263 index_only_thermalized++;
264 partial_sum +=
N_in_cells_[index_only_thermalized];
266 const int cell_index = cells_to_sample_[index_only_thermalized];
279 particle.
set_4momentum(m, phitheta.threevec() * momentum_radial);
283 plist.push_back(particle);
288 double time,
int ntest) {
289 const auto &log = logger<LogArea::GrandcanThermalizer>();
294 const double gamma = 1.0 / std::sqrt(1.0 - cell.
v().
sqr());
295 for (
size_t i = 0; i <
N_sorts_; i++) {
304 for (
size_t i = 0; i <
N_sorts_; i++) {
315 const auto Nbar_antibar = bessel_sampler_B.sample();
322 for (
size_t i = 0; i <
N_sorts_; i++) {
326 std::pair<int, int> NS_antiS;
332 NS_antiS = bessel_sampler_S.
sample();
334 NS_antiS = std::make_pair(
337 if (NS_antiS.first - NS_antiS.second !=
347 for (
size_t i = 0; i <
N_sorts_; i++) {
351 std::pair<int, int> NC_antiC;
356 conserved_initial.
charge() - ch_sampled);
357 NC_antiC = bessel_sampler_C.
sample();
359 NC_antiC = std::make_pair(
362 if (NC_antiC.first - NC_antiC.second !=
363 conserved_initial.
charge() - ch_sampled) {
374 for (
size_t itype = 0; itype <
N_sorts_; itype++) {
378 const double e_init = conserved_initial.
momentum().
x0();
381 e_tot += particle.momentum().x0();
383 if (std::abs(e_tot - e_init) > 0.01 * e_init) {
384 log.debug(
"Rejecting: energy ", e_tot,
" too far from e_init = ", e_init);
394 int S_plus = 0, S_minus = 0, B_plus = 0, B_minus = 0, E_plus = 0, E_minus = 0;
396 auto condition1 = [](int, int, int) {
return true; };
398 while (conserved_initial.
momentum().
x0() > energy ||
409 auto condition2 = [](
int S, int, int) {
return (S < 0); };
411 while (S_plus + S_minus > conserved_initial.
strangeness()) {
415 if (S_plus + S_minus + s_part >= conserved_initial.
strangeness()) {
422 auto condition3 = [](
int S, int, int) {
return (S == 0); };
427 while (conserved_remaining.
momentum().
x0() > energy ||
438 auto condition4 = [](
int S,
int B, int) {
return (S == 0) && (B < 0); };
440 while (B_plus + B_minus > conserved_remaining.
baryon_number()) {
443 if (B_plus + B_minus + bar >= conserved_remaining.
baryon_number()) {
450 auto condition5 = [](
int S,
int B, int) {
return (S == 0) && (B == 0); };
451 conserved_remaining = conserved_initial - QuantumNumbers(
sampled_list_);
454 while (conserved_remaining.
momentum().
x0() > energy ||
455 E_plus < conserved_remaining.
charge()) {
465 auto condition6 = [](
int S,
int B,
int C) {
466 return (S == 0) && (B == 0) && (C < 0);
469 while (E_plus + E_minus > conserved_remaining.
charge()) {
472 if (E_plus + E_minus + charge >= conserved_remaining.
charge()) {
479 auto condition7 = [](
int S,
int B,
int C) {
480 return (S == 0) && (B == 0) && (C == 0);
482 conserved_remaining = conserved_initial - QuantumNumbers(
sampled_list_);
485 while (conserved_remaining.
momentum().
x0() > energy) {
494 const auto &log = logger<LogArea::GrandcanThermalizer>();
495 log.info(
"Starting forced thermalization, time ", time,
" fm/c");
502 for (
auto &particle : particles) {
503 const bool is_on_lattice =
504 lat_->value_at(particle.position().threevec(), node);
505 if (is_on_lattice && node.
e() >
e_crit_) {
519 log.info(
"Removed ",
to_remove_.size(),
" particles.");
527 const size_t lattice_total_cells =
lat_->size();
528 for (
size_t i = 0; i < lattice_total_cells; i++) {
533 log.info(
"Number of cells in the thermalization region = ",
547 throw std::invalid_argument(
548 "This thermalization algorithm is" 549 " not yet implemented");
565 struct to_average on_lattice = {0.0, 0.0, 0.0, 0.0, 0.0};
566 struct to_average in_therm_reg = {0.0, 0.0, 0.0, 0.0, 0.0};
567 double e_sum_on_lattice = 0.0, e_sum_in_therm_reg = 0.0;
568 int node_counter = 0;
569 for (
const auto &node : *
lat_) {
570 const double e = node.e();
571 on_lattice.T += node.T() * e;
572 on_lattice.mub += node.mub() * e;
573 on_lattice.mus += node.mus() * e;
574 on_lattice.nb += node.nb() * e;
575 on_lattice.ns += node.ns() * e;
576 e_sum_on_lattice += e;
578 in_therm_reg.T += node.T() * e;
579 in_therm_reg.mub += node.mub() * e;
580 in_therm_reg.mus += node.mus() * e;
581 in_therm_reg.nb += node.nb() * e;
582 in_therm_reg.ns += node.ns() * e;
583 e_sum_in_therm_reg += e;
588 on_lattice.T /= e_sum_on_lattice;
589 on_lattice.mub /= e_sum_on_lattice;
590 on_lattice.mus /= e_sum_on_lattice;
591 on_lattice.nb /= e_sum_on_lattice;
592 on_lattice.ns /= e_sum_on_lattice;
595 in_therm_reg.T /= e_sum_in_therm_reg;
596 in_therm_reg.mub /= e_sum_in_therm_reg;
597 in_therm_reg.mus /= e_sum_in_therm_reg;
598 in_therm_reg.nb /= e_sum_in_therm_reg;
599 in_therm_reg.ns /= e_sum_in_therm_reg;
602 std::cout <<
"Current time [fm/c]: " << clock.
current_time() << std::endl;
603 std::cout <<
"Averages on the lattice - T[GeV], mub[GeV], mus[GeV], " 604 <<
"nb[fm^-3], ns[fm^-3]: " << on_lattice.T <<
" " << on_lattice.mub
605 <<
" " << on_lattice.mus <<
" " << on_lattice.nb <<
" " 606 << on_lattice.ns << std::endl;
607 std::cout <<
"Averages in therm. region - T[GeV], mub[GeV], mus[GeV], " 608 <<
"nb[fm^-3], ns[fm^-3]: " << in_therm_reg.T <<
" " 609 << in_therm_reg.mub <<
" " << in_therm_reg.mus <<
" " 610 << in_therm_reg.nb <<
" " << in_therm_reg.ns << std::endl;
611 std::cout <<
"Volume with e > e_crit [fm^3]: " <<
cell_volume_ * node_counter
int charge() const
The charge of the particle.
const double e_crit_
Critical energy density above which cells are thermalized.
void add_particle(const ParticleData &p, double factor)
Add particle contribution to Tmu0, nb and ns May look like unused at first glance, but it is actually used by update_lattice, where the node type of the lattice is templated.
A class to pre-calculate and store parameters relevant for density calculation.
PdgCode pdgcode() const
Get the pdgcode of the particle.
ThermalizationAlgorithm
Defines the algorithm used for the forced thermalization.
HadronClass get_class(size_t typelist_index) const
Defines the class of hadrons by quantum numbers.
GrandCanThermalizer(const std::array< double, 3 > lat_sizes, const std::array< int, 3 > n_cells, const std::array< double, 3 > origin, bool periodicity, double e_critical, double t_start, double delta_t, ThermalizationAlgorithm algo)
Default constructor for the GranCanThermalizer to allocate the lattice.
const size_t N_sorts_
Number of different species to be sampled.
The ThreeVector class represents a physical three-vector with the components .
double p() const
Get pressure in the rest frame.
void sample_multinomial(HadronClass particle_class, int N)
The sample_multinomial function samples integer numbers n_i distributed according to the multinomial ...
constexpr double really_small
Numerical error tolerance.
ThermLatticeNode()
Default constructor of thermal quantities on the lattice returning thermodynamic quantities in comput...
std::array< double, 3 > solve_eos(double e, double nb, double ns, std::array< double, 3 > initial_approximation)
Compute temperature and chemical potentials given energy-, net baryon-, net strangeness density and a...
int baryon_number() const
static double net_strange_density(double T, double mub, double mus)
Compute net strangeness density.
Class to handle the equation of state (EoS) of the hadron gas, consisting of all hadrons included int...
FourVector Tmu0_
Four-momentum flow of the cell.
void thermalize(const Particles &particles, double time, int ntest)
Main thermalize function, that chooses the algorithm to follow (BF or mode sampling).
int baryon_number() const
double mub() const
Get the net baryon chemical potential.
double ns() const
Get net strangeness density of the cell in the computational frame.
void update_lattice(RectangularLattice< T > *lat, const LatticeUpdate update, const DensityType dens_type, const DensityParameters &par, const Particles &particles, const bool compute_gradient=false)
Updates the contents on the lattice.
int baryon_number() const
double abs() const
calculate the lorentz invariant absolute value
void thermalize_BF_algo(QuantumNumbers &conserved_initial, double time, int ntest)
Samples particles according to the BF algorithm by making use of the.
void from_table(EosTable::table_element &res, double e, double nb) const
Get the element of eos table.
double e() const
Get energy density in the rest frame.
ThreeVector threevec() const
LatticeUpdate
Enumerator option for lattice updates.
std::pair< int, int > sample()
Sample two numbers from given Poissonians with a fixed difference.
double mus() const
Get the net strangeness chemical potential.
double T() const
Get the temperature.
bool is_tabulated() const
Create an EoS table or not?
void print_statistics(const Clock &clock) const
Generates standard output with information about the thermodynamic properties of the lattice...
const ThermalizationAlgorithm algorithm_
Algorithm to choose for sampling of particles obeying conservation laws.
void set_formation_time(const double &form_time)
Set the absolute formation time.
void add_values(const ParticleData &p)
Add the quantum numbers of a single particle to the collection.
void set_rest_frame_quantities(double T0, double mub0, double mus0, const ThreeVector v0)
Set all the rest frame quantities to some values, this is useful for testing.
void update_thermalizer_lattice(const Particles &particles, const DensityParameters &par, bool ignore_cells_under_threshold=true)
Compute all the thermodynamical quantities on the lattice from particles.
ThreeVector v_
Velocity of the rest frame.
double p_
Pressure in the rest frame.
static double net_baryon_density(double T, double mub, double mus)
Compute net baryon density.
double ns_
Net strangeness density of the cell in the computational frame.
double mub_
Net baryon chemical potential.
ParticleList sampled_list_
Newly generated particles by thermalizer.
ThreeVector velocity() const
Get the velocity (3-vector divided by zero component).
ThreeVector v() const
Get 3-velocity of the rest frame.
void set_4momentum(const FourVector &momentum_vector)
Set the particle's 4-momentum directly.
Non-strange mesons (S = 0) with electric cherge Q < 0.
A container for storing conserved values.
void compute_rest_frame_quantities(HadronGasEos &eos)
Temperature, chemical potentials and rest frame velocity are calculated given the hadron gas equation...
double sample_momenta_from_thermal(const double temperature, const double mass)
Samples a momentum from the Maxwell-Boltzmann (thermal) distribution in a faster way, given by Scott Pratt.
void compute_N_in_cells_mode_algo(F &&condition)
Computes average number of particles in each cell for the mode algorithm.
std::vector< size_t > cells_to_sample_
Cells above critical energy density.
Mesons with strangeness S < 0.
The intention of this class is to efficiently sample from the Bessel distribution ...
double N_total_in_cells_
Total number of particles over all cells in thermalization region.
void thermalize_mode_algo(QuantumNumbers &conserved_initial, double time)
Samples particles to the according to the mode algorithm.
const ParticleType & type() const
Get the type of the particle.
Define the data structure for one element of the table.
void renormalize_momenta(ParticleList &plist, const FourVector required_total_momentum)
Changes energy and momenta of the particles in plist to match the required_total_momentum.
Clock tracks the time in the simulation.
Neutral non-strange mesons.
Non-strange mesons (S = 0) with electric cherge Q > 0.
ParticleData sample_in_random_cell_mode_algo(const double time, F &&condition)
Samples one particle and the species, cell, momentum and coordinate are chosen from the corresponding...
std::vector< double > mult_sort_
Real number multiplicity for each particle type.
ParticleList to_remove_
Particles to be removed after this thermalization step.
FourVector momentum() const
Mesons with strangeness S > 0.
std::vector< double > N_in_cells_
Number of particles to be sampled in one cell.
int binomial(const int N, const T &p)
Returns a binomially distributed random number.
double nb_
Net baryon density of the cell in the computational frame.
static double partial_density(const ParticleType &ptype, double T, double mub, double mus)
Compute partial density of one hadron sort.
void set_4position(const FourVector &pos)
Set the particle's 4-position directly.
double e_
Energy density in the rest frame.
The ThermLatticeNode class is intended to compute thermodynamical quantities in a cell given a set of...
HadronClass
Specifier to classify the different hadron species according to their quantum numbers.
std::vector< int > mult_int_
Integer multiplicity for each particle type.
double mus_
Net strangeness chemical potential.
int poisson(const T &lam)
Returns a Poisson distributed random number.
ThreeVector uniform_in_cell() const
void sample_in_random_cell_BF_algo(ParticleList &plist, const double time, size_t type_index)
The total number of particles of species type_index is defined by mult_int_ array that is returned by...
double current_time() const
std::array< double, 7 > mult_classes_
The different hadron species according to the enum defined in.
double nb() const
Get net baryon density of the cell in the computational frame.
Angles provides a common interface for generating directions: i.e., two angles that should be interpr...
std::unique_ptr< RectangularLattice< ThermLatticeNode > > lat_
The lattice on which the thermodynamic quantities are calculated.
void distribute_isotropically()
Populate the object with a new direction.
The Particles class abstracts the storage and manipulation of particles.
DensityType
Allows to choose which kind of density to calculate.
std::ostream & operator<<(std::ostream &out, const ActionPtr &action)
Convenience: dereferences the ActionPtr to Action.
double mult_class(const HadronClass cl) const
The FourVector class holds relevant values in Minkowski spacetime with (+, −, −, −) metric signature.
double cell_volume_
Volume of a single cell, necessary to convert thermal densities to actual particle numbers...
HadronGasEos eos_
Hadron gas equation of state.
ParticleData contains the dynamic information of a certain particle.
void boost_momentum(const ThreeVector &v)
Apply a Lorentz-boost to only the momentum.
const ParticleTypePtrList eos_typelist_
List of particle types from which equation of state is computed.
static double pressure(double T, double mub, double mus)
Compute pressure .
const FourVector & momentum() const
Get the particle's 4-momentum.
static double energy_density(double T, double mub, double mus)
Compute energy density.
FourVector Tmu0() const
Get Four-momentum flow of the cell.