#include <StrangImplicitSpectralEM.H>

Inheritance diagram for StrangImplicitSpectralEM:

Public Member Functions
	StrangImplicitSpectralEM ()=default

	~StrangImplicitSpectralEM () override=default

	StrangImplicitSpectralEM (const StrangImplicitSpectralEM &)=delete

StrangImplicitSpectralEM &	operator= (const StrangImplicitSpectralEM &)=delete

	StrangImplicitSpectralEM (StrangImplicitSpectralEM &&)=delete

StrangImplicitSpectralEM &	operator= (StrangImplicitSpectralEM &&)=delete

void	Define (WarpX *a_WarpX) override
	Read user-provided parameters that control the implicit solver. Allocate internal arrays for intermediate field values needed by the solver.

void	PrintParameters () const override

void	OneStep (amrex::Real start_time, amrex::Real a_dt, int a_step) override
	Advance fields and particles by one time step using the specified implicit algorithm.

void	ComputeRHS (WarpXSolverVec &a_RHS, const WarpXSolverVec &a_E, amrex::Real start_time, int a_nl_iter, bool a_from_jacobian) override
	Computes the RHS of the equation corresponding to the specified implicit algorithm. The discrete equations corresponding to numerical integration of ODEs are often written in the form U = b + RHS(U), where U is the variable to be solved for (e.g., the solution at the next time step), b is a constant (i.e., the solution from the previous time step), and RHS(U) is the right-hand-side of the equation. Iterative solvers, such as Picard and Newton, and higher-order Runge-Kutta methods, need to compute RHS(U) multiple times for different values of U. Thus, a routine that returns this value is needed. e.g., Ebar - E^n = cvac^20.5dt(curl(Bbar) - mu0Jbar(Ebar,Bbar)) Here, U = Ebar, b = E^n, and the expression on the right is RHS(U).

amrex::Real	GetThetaForPC () const override

Public Member Functions inherited from ImplicitSolver
	ImplicitSolver ()=default

virtual	~ImplicitSolver ()=default

	ImplicitSolver (const ImplicitSolver &)=delete

ImplicitSolver &	operator= (const ImplicitSolver &)=delete

	ImplicitSolver (ImplicitSolver &&)=delete

ImplicitSolver &	operator= (ImplicitSolver &&)=delete

bool	IsDefined () const

void	CreateParticleAttributes () const

bool	DoParticleSuborbits ()

int	numAMRLevels () const

const amrex::Geometry &	GetGeometry (int) const

const amrex::Array< FieldBoundaryType, 3 > &	GetFieldBoundaryLo () const

const amrex::Array< FieldBoundaryType, 3 > &	GetFieldBoundaryHi () const

amrex::Array< amrex::LinOpBCType, 3 >	GetLinOpBCLo () const

amrex::Array< amrex::LinOpBCType, 3 >	GetLinOpBCHi () const

const amrex::Vector< amrex::Array< amrex::MultiFab , 3 > >	GetMassMatricesCoeff () const
	Return pointer to MultiFab array for mass matrix.

void	ComputeJfromMassMatrices (const bool a_J_from_MM_only)

Private Member Functions
void	UpdateWarpXFields (WarpXSolverVec const &a_E, amrex::Real half_time)
	Update the E and B fields owned by WarpX.

void	FinishFieldUpdate (amrex::Real a_new_time)
	Nonlinear solver is for the time-centered values of E. After the solver, need to use m_E and m_Eold to compute E^{n+1}.

Private Attributes
WarpXSolverVec	m_E
	Solver vectors to be used in the nonlinear solver to solve for the electric field E. The main logic for determining which variables should be WarpXSolverVec type is that it must have the same size and have the same centering of the data as the variable being solved for, which is E here. For example, if using a Yee grid then a container for curlB could be a WarpXSovlerVec, but magnetic field B should not be.

WarpXSolverVec	m_Eold

amrex::Vector< std::array< std::unique_ptr< amrex::MultiFab >, 3 > >	m_Bold
	B is a derived variable from E. Need to save Bold to update B during the iterative nonlinear solve for E. Bold is owned here, but only used by WarpX. It is not used directly by the nonlinear solver, nor is it the same size as the solver vector (size E), and so it should not be WarpXSolverVec type.

Additional Inherited Members
Protected Member Functions inherited from ImplicitSolver
void	parseNonlinearSolverParams (const amrex::ParmParse &pp)
	parse nonlinear solver parameters (if one is used)

void	InitializeMassMatrices ()
	Initialize the Mass Matrices used for plasma response in nonlinear Newton solver.

void	PrintBaseImplicitSolverParameters () const

void	PreRHSOp (const amrex::Real a_cur_time, const int a_nl_iter, const bool a_from_jacobian)
	Perform operations needed before computing the Right Hand Side.

void	SyncMassMatricesPCAndApplyBCs ()
	Communicate Mass Matrices used for PC and apply boundary conditions.

void	SetMassMatricesForPC (const amrex::Real a_theta_dt)
	Scale mass matrices used for PC by c^2mu0theta*dt and add 1 to diagonal terms.

void	CumulateJ ()
	Add J from particles included in mass matrices at start of each Newton step.

void	SaveE ()
	Save E at start of each Newton step.

amrex::Array< amrex::LinOpBCType, 3 >	convertFieldBCToLinOpBC (const amrex::Array< FieldBoundaryType, 3 > &) const
	Convert from WarpX FieldBoundaryType to amrex::LinOpBCType.

Protected Attributes inherited from ImplicitSolver
WarpX *	m_WarpX
	Pointer back to main WarpX class.

bool	m_is_defined = false

int	m_num_amr_levels = 1
	Number of AMR levels.

amrex::Real	m_dt = 0.0
	Time step.

amrex::Real	m_theta = 0.5
	Time-biasing parameter for fields used on RHS to advance system.

NonlinearSolverType	m_nlsolver_type
	Nonlinear solver type and object.

std::unique_ptr< NonlinearSolver< WarpXSolverVec, ImplicitSolver > >	m_nlsolver

amrex::ParticleReal	m_particle_tolerance = amrex::ParticleReal(1.0e-10)
	tolerance used by the iterative method used to obtain a self-consistent update of the particle positions and velocities for given E and B on the grid

int	m_max_particle_iterations = 21
	maximum iterations for the iterative method used to obtain a self-consistent update of the particle positions and velocities for given E and B on the grid

bool	m_particle_suborbits = false
	whether to use suborbits for particles that fail to converge in m_max_particle_iterations iterations

bool	m_print_unconverged_particle_details = false
	whether the details of unconverged particles are printed out during the particle evolve loops

bool	m_skip_particle_picard_init = false
	whether to skip the full Picard update of particles on the initial newton step

bool	m_use_mass_matrices = false
	Whether to use mass matrices for the implicit solver.

bool	m_use_mass_matrices_jacobian = false
	Whether to use mass matrices to compute J during linear stage of JFNK.

bool	m_use_mass_matrices_pc = false
	Whether to use mass matrices in the preconditioner.

amrex::IntVect	m_ncomp_xx = amrex::IntVect{0}
	Direction-dependent component number of mass matrices elements.

amrex::IntVect	m_ncomp_xy = amrex::IntVect{0}

amrex::IntVect	m_ncomp_xz = amrex::IntVect{0}

amrex::IntVect	m_ncomp_yx = amrex::IntVect{0}

amrex::IntVect	m_ncomp_yy = amrex::IntVect{0}

amrex::IntVect	m_ncomp_yz = amrex::IntVect{0}

amrex::IntVect	m_ncomp_zx = amrex::IntVect{0}

amrex::IntVect	m_ncomp_zy = amrex::IntVect{0}

amrex::IntVect	m_ncomp_zz = amrex::IntVect{0}

amrex::Vector< amrex::Array< amrex::MultiFab *, 3 > >	m_mmpc_mfarrvec
	Array of multifab pointers to mass matrix preconditioner.

Constructor & Destructor Documentation

◆ StrangImplicitSpectralEM() [1/3]

StrangImplicitSpectralEM::StrangImplicitSpectralEM ( )

default

◆ ~StrangImplicitSpectralEM()

StrangImplicitSpectralEM::~StrangImplicitSpectralEM ( )

overridedefault

◆ StrangImplicitSpectralEM() [2/3]

StrangImplicitSpectralEM::StrangImplicitSpectralEM ( const StrangImplicitSpectralEM & )

delete

◆ StrangImplicitSpectralEM() [3/3]

StrangImplicitSpectralEM::StrangImplicitSpectralEM ( StrangImplicitSpectralEM && )

delete

Member Function Documentation

◆ ComputeRHS()

void StrangImplicitSpectralEM::ComputeRHS	(	WarpXSolverVec &	a_RHS,
		const WarpXSolverVec &	a_E,
		amrex::Real	a_time,
		int	a_nl_iter,
		bool	a_from_jacobian )

overridevirtual

Computes the RHS of the equation corresponding to the specified implicit algorithm. The discrete equations corresponding to numerical integration of ODEs are often written in the form U = b + RHS(U), where U is the variable to be solved for (e.g., the solution at the next time step), b is a constant (i.e., the solution from the previous time step), and RHS(U) is the right-hand-side of the equation. Iterative solvers, such as Picard and Newton, and higher-order Runge-Kutta methods, need to compute RHS(U) multiple times for different values of U. Thus, a routine that returns this value is needed. e.g., Ebar - E^n = cvac^2*0.5*dt*(curl(Bbar) - mu0*Jbar(Ebar,Bbar)) Here, U = Ebar, b = E^n, and the expression on the right is RHS(U).

Implements ImplicitSolver.

◆ Define()

void StrangImplicitSpectralEM::Define ( WarpX * a_WarpX )

overridevirtual

Read user-provided parameters that control the implicit solver. Allocate internal arrays for intermediate field values needed by the solver.

Implements ImplicitSolver.

◆ FinishFieldUpdate()

void StrangImplicitSpectralEM::FinishFieldUpdate ( amrex::Real a_new_time )

private

Nonlinear solver is for the time-centered values of E. After the solver, need to use m_E and m_Eold to compute E^{n+1}.

◆ GetThetaForPC()

amrex::Real StrangImplicitSpectralEM::GetThetaForPC ( ) const

inlinenodiscardoverridevirtual

Implements ImplicitSolver.

◆ OneStep()

void StrangImplicitSpectralEM::OneStep	(	amrex::Real	a_time,
		amrex::Real	a_dt,
		int	a_step )

overridevirtual

Advance fields and particles by one time step using the specified implicit algorithm.

Implements ImplicitSolver.

◆ operator=() [1/2]

StrangImplicitSpectralEM & StrangImplicitSpectralEM::operator= ( const StrangImplicitSpectralEM & )

delete

◆ operator=() [2/2]

StrangImplicitSpectralEM & StrangImplicitSpectralEM::operator= ( StrangImplicitSpectralEM && )

delete

◆ PrintParameters()

void StrangImplicitSpectralEM::PrintParameters ( ) const

overridevirtual

Implements ImplicitSolver.

◆ UpdateWarpXFields()

void StrangImplicitSpectralEM::UpdateWarpXFields	(	WarpXSolverVec const &	a_E,
		amrex::Real	half_time )

private

Update the E and B fields owned by WarpX.

Member Data Documentation

◆ m_Bold

amrex::Vector<std::array< std::unique_ptr<amrex::MultiFab>, 3 > > StrangImplicitSpectralEM::m_Bold

private

B is a derived variable from E. Need to save Bold to update B during the iterative nonlinear solve for E. Bold is owned here, but only used by WarpX. It is not used directly by the nonlinear solver, nor is it the same size as the solver vector (size E), and so it should not be WarpXSolverVec type.

◆ m_E

WarpXSolverVec StrangImplicitSpectralEM::m_E

private

Solver vectors to be used in the nonlinear solver to solve for the electric field E. The main logic for determining which variables should be WarpXSolverVec type is that it must have the same size and have the same centering of the data as the variable being solved for, which is E here. For example, if using a Yee grid then a container for curlB could be a WarpXSovlerVec, but magnetic field B should not be.

◆ m_Eold

WarpXSolverVec StrangImplicitSpectralEM::m_Eold

private

The documentation for this class was generated from the following files:

/home/docs/checkouts/readthedocs.org/user_builds/warpx/checkouts/6270/Source/FieldSolver/ImplicitSolvers/StrangImplicitSpectralEM.H
/home/docs/checkouts/readthedocs.org/user_builds/warpx/checkouts/6270/Source/FieldSolver/ImplicitSolvers/StrangImplicitSpectralEM.cpp

Public Member Functions

Private Member Functions

Private Attributes

Additional Inherited Members

Constructor & Destructor Documentation

◆ StrangImplicitSpectralEM() [1/3]

◆ ~StrangImplicitSpectralEM()

◆ StrangImplicitSpectralEM() [2/3]

◆ StrangImplicitSpectralEM() [3/3]

Member Function Documentation

◆ ComputeRHS()

◆ Define()

◆ FinishFieldUpdate()

◆ GetThetaForPC()

◆ OneStep()

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ PrintParameters()

◆ UpdateWarpXFields()

Member Data Documentation

◆ m_Bold

◆ m_E

◆ m_Eold