poisson.F90 File Reference

#include "petsc_version_defs.h"
#include <petsc/finclude/petsc.h>

Include dependency graph for poisson.F90:

Data Types
module	f90moduleinterfaces
	Explicit interfaces to somve PETSc function used by the FSA solver. More...
interface	f90moduleinterfaces::pcsetapplicationcontext
interface	f90moduleinterfaces::pcgetapplicationcontext
interface	f90moduleinterfaces::matcreateshell
interface	f90moduleinterfaces::matshellsetcontext
interface	f90moduleinterfaces::matshellgetcontext

Functions/Subroutines
subroutine	init_poisson (grid, psn, iflag_dummy)
subroutine	pcshermanmorrisonapply (pc, xin, yout, ierr)
	Applies a Sherman-Morrison preconditioner. More...
subroutine	create_2field_solver (grid, bd, solver, solver_data)
	Sets up a 2x2 block solver for the Poisson equation. The first row is the Poisson equation, the second row is a constraint equation for the flux-surface averaged potential. More...
subroutine	make_12mat (grid, Bmat, BIdentMat, FSAMass_Te, solver, bd, scale, xgc_petsc)
subroutine	make_21mat (grid, Cmat, Cio, Bmat, CConstBCVec, bd, xgc_petsc, petsc_xgc_bd)
subroutine	solve_axisymmetric_poisson (iflag)
	High level Poisson solver routine for n=0, offers choice of outer iterative solver, two-step solver, and FSA solver. More...
subroutine	solve_turb_poisson (iflag, ipc)
	High level Poisson solver routine for |n|>0 Currently, only one implementation is available. More...
subroutine	set_rhs_poisson (grid, psn, solver, iflag, is_turb, is_two_step)
	Routine to set and process the rhs arrays for both axisymmetric and turbulent poisson solvers. Also sets boundary data for XGCA. More...
subroutine	save_rhs_poisson (grid, psn, solver, iflag, ipc, do_solve_poisson)
	Save the charge density for time averaging of the Poisson equation. More...
subroutine	spectral_solver_decompose (grid, field_inout, field_ntor)
	Decompose field into spectral components treated by the spectral solver and the rest treated by the regular solver. More...
integer function	get_ntor_num_from_ntor_real (ntor_real, nphi, wedge_n)
	Calculates the numerical toroidal mode number for a given real toroidal mode number. More...
subroutine	spectral_solver_reassemble (grid, field_inout, field_ntor)
	Reassemble the full field from the spectral components treated by the spectral solver, and the rest treated by the regular solver. More...
subroutine	set_petsc_rhs_vec_axisymmetric_poisson (grid, psn, solver)
	Places rhs and (optionally) boundary condition data into a PETSc vector. More...
subroutine	set_initial_guess_axisymmetric_poisson (grid, psn)
	Sets initial guess for the iterative axisymmetric poisson solver using the simple00 solver. The initial guess is stored in psnpot0m. More...
subroutine	solve_axisymmetric_poisson_iter (grid, psn)
	Poisson solver that uses an iterative method to solve for the axisymmetric electrostatic potential: The Poisson equation for the axisymmetric mode is \[ -\nabla_\perp \cdot \xi \nabla_\perp \overline{\phi}_{k+1} + \frac{e n_0/T_{e,0}} \overline{\phi}_{k+1} = e \left( \langle \overline{\delta n_i} \rangle_g - \overline{\delta n_{e,NA}} \right) + \frac{e n_0/T_{e,0}} \langle \phi_{k} \rangle_{fs} \] where \(k\) is the iteration index, \(xi\) is the electric susceptibility, \(e\) is the elementary charge, \(n_0\) and \(T_0\) are the background density and temperature, \(\langle \overline{\delta n_i} \rangle_g\) is the gyroaveraged ion (gyrocenter) charge density, \(\overline{\delta n_{e,NA}}\) is the non-adiabatic electron charge density, and \(\langle\dots\rangle_{fs}\) is the the flux-surface average. \(\overline{\dots}\) is the toroidal average. More...
subroutine	axisym_mat_mult (A, X, Y, ierr)
	Subroutine used to apply the shell matrix representing the left hand side of the axisymmetric Poisson equation. More...
subroutine	axisym_pc_apply (pc, X, Y, ierr)
	Subroutine used to apply the shell preconditioner used in the iterative solution of the axisymmetric Poisson equation. More...
subroutine	solve_axisymmetric_poisson_two_step_fsa (grid, psn)
	Simple two-step 2D solver driver for the axisymmetric potential and one-step solver ("FSA-solver") with constraint equation for the flux-surface averaged potential. For the two-step solver, the RHS is the flux-surface averaged charge density, the LHS consists only of the polarization operator. The result of the 2D solve is flux-surface averaged; i.e., the result is \(\langle\phi\rangle\) The Poisson equation for the flux-surface averaged mode is \[ -\nabla_\perp \cdot \xi \nabla_\perp \overline{\phi} = e \left\langle \left( \langle \overline{\delta n_i} \rangle_g - \overline{\delta n_{e,NA}} \right) \right\rangle_{fs} \] where \(xi\) is the electric susceptibility, \(e\) is the elementary charge, \(n_0\) is the background density, \(\langle \overline{\delta n_i} \rangle_g\) is the gyroaveraged ion (gyrocenter) charge density, \(\overline{\delta n_{e,NA}}\) is the non-adiabatic electron charge density, and \(\langle\dots\rangle_{fs}\) is the the flux-surface average. \(\overline{\dots}\) is the toroidal average. More...
subroutine	post_process_axisymmetric_poisson (grid, psn)
	Routine to post process the solution from the axisymmetric poisson solver. More...
subroutine	solve_turb_poisson_default (grid, psn)
	Routine to solve the non-axisymmetric Poisson equation: \[ -\nabla_\perp \cdot \xi \nabla_\perp \delta\phi + \frac{e n_0/T_{e,0}} \delta\phi = e \left( \langle \delta n_i \rangle_g - \overline{\delta n_{e,NA}} \right) \] . More...
subroutine	post_process_turb_poisson (grid, psn)
	Routine to post process the solution from the non-axisymmetric Poisson equation. More...

Function/Subroutine Documentation

subroutine axisym_mat_mult	(		A,
			X,
			Y,
			ierr
	)

Subroutine used to apply the shell matrix representing the left hand side of the axisymmetric Poisson equation.

Parameters

[in]	A	Shell matrix which multiplies X, Petsc Mat
[in]	X	Vector to be multiplied, Petsc Vec
[out]	Y	Vector to store product A*X, Petsc Vec
[out]	ierr	PETSc Error code

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine axisym_pc_apply	(		pc,
			X,
			Y,
			ierr
	)

Subroutine used to apply the shell preconditioner used in the iterative solution of the axisymmetric Poisson equation.

Parameters

[in]	pc	Shell preconditioner which is applied to residual X, Petsc PC
[in]	X	Residual vector, Petsc Vec
[out]	Y	Preconditioned residual vector, Petsc Vec
[out]	ierr	PETSc Error code

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine create_2field_solver	(	type(grid_type)	grid,
		type(boundary2_type)	bd,
		type(xgc_solver)	solver,
		type(solver_init_data), intent(in)	solver_data
	)

Sets up a 2x2 block solver for the Poisson equation. The first row is the Poisson equation, the second row is a constraint equation for the flux-surface averaged potential.

Parameters

[in]	grid	XGC grid object, type(grid_type)
[in]	bd	Object containing the indices of the solver boundary vertices in the XGC mesh.
[in,out]	solver	XGC solver object into which to put the new solver, type(xgc_solver)
[in]	solver_data	Data needed for settug up the ssolver, type(solver_init_data)

Here is the call graph for this function:

Here is the caller graph for this function:

integer function get_ntor_num_from_ntor_real	(	integer, intent(in)	ntor_real,
		integer, intent(in)	nphi,
		integer, intent(in)	wedge_n
	)

Calculates the numerical toroidal mode number for a given real toroidal mode number.

Parameters

[in]	ntor_real	Real toroidal mode number, integer
[in]	nphi	Number of toroidal planes per wedge, integer
[in]	wedge_n	Number of wedges per toroidal circuit, integer

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine init_poisson	(	type(grid_type)	grid,
		type(psn_type)	psn,
		integer	iflag_dummy
	)

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine make_12mat	(	type(grid_type)	grid,
		intent(inout)	Bmat,
			BIdentMat,
		intent(in)	FSAMass_Te,
		type(xgc_solver)	solver,
		type(boundary2_type), intent(in)	bd,
		real (8), intent(in)	scale,
		dimension(grid%nnode), intent(in)	xgc_petsc
	)

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine make_21mat	(	type(grid_type)	grid,
			Cmat,
			Cio,
		intent(in)	Bmat,
			CConstBCVec,
		type(boundary2_type), intent(in)	bd,
		dimension(grid%nnode), intent(in)	xgc_petsc,
		dimension(grid%nnode), intent(in)	petsc_xgc_bd
	)

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine pcshermanmorrisonapply	(		pc,
			xin,
			yout,
			ierr
	)

Applies a Sherman-Morrison preconditioner.

Parameters

[in,out]	pc	PETCs PC object (manages preconditioners)
	xin	???, Vec
	yout	???, Vec
[out]	ierr	PETc error code

Here is the caller graph for this function:

subroutine post_process_axisymmetric_poisson	(	type(grid_type)	grid,
		type(psn_type)	psn
	)

Routine to post process the solution from the axisymmetric poisson solver.

Parameters

[in]	grid	XGC grid data object
[in,out]	psn	XGC field data object

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine post_process_turb_poisson	(	type(grid_type), intent(in)	grid,
		type(psn_type), intent(inout)	psn
	)

Routine to post process the solution from the non-axisymmetric Poisson equation.

Parameters

[in]	grid	XGC grid data object
[in,out]	psn	XGC field data object

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine save_rhs_poisson	(	type(grid_type), intent(in)	grid,
		type(psn_type), intent(inout)	psn,
		type(xgc_solver), intent(inout)	solver,
		integer, intent(in)	iflag,
		integer, intent(in)	ipc,
		logical, intent(out)	do_solve_poisson
	)

Save the charge density for time averaging of the Poisson equation.

Parameters

[in]	grid	XGC grid data structure, type(grid_type)
[in,out]	psn	XGC field data structure, type(psn_type)
[in,out]	solver	XGC solver object, type(xgc_solver)
[in]	iflag	Adiabatic solve indicator, integer
[in]	ipc	RK2 stage index, integer
[out]	do_solve_poisson	Whether to solve Ampere's law after saving RHS, logical

Here is the caller graph for this function:

subroutine set_initial_guess_axisymmetric_poisson	(	type(grid_type), intent(in)	grid,
		type(psn_type), intent(inout)	psn
	)

Sets initial guess for the iterative axisymmetric poisson solver using the simple00 solver. The initial guess is stored in psnpot0m.

Parameters

[in]	grid	Solver grid data, type(grid_type)
[in,out]	psn	Field data; the pot0m is stored there, type(psn_type)

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine set_petsc_rhs_vec_axisymmetric_poisson	(	type(grid_type)	grid,
		type(psn_type)	psn,
		type(xgc_solver)	solver
	)

Places rhs and (optionally) boundary condition data into a PETSc vector.

Parameters

[in]	grid	Solver grid data, type(grid_type)
[in,out]	psn	Field data; the rhs XGC data is stored there, type(psn_type)
[in,out]	solver	XGC solver object; the rhs PETSc vector is stored there, type(xgc_solver)

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine set_rhs_poisson	(	type(grid_type)	grid,
		type(psn_type)	psn,
		type(xgc_solver)	solver,
		integer, intent(in)	iflag,
		logical, intent(in)	is_turb,
		logical, intent(in)	is_two_step
	)

Routine to set and process the rhs arrays for both axisymmetric and turbulent poisson solvers. Also sets boundary data for XGCA.

Parameters

[in]	grid	XGC grid data object
[in,out]	psn	XGC field data object
[in,out]	solver	XGC solver object, type(xgc_solver)
[in]	iflag	Flag to indicate whether adiabatic elec or full eq. is solved, integer
[in]	is_turb	If .true. set RHS for turbulence solve, otherwise for axisymmetric solve
[in]	is_two_step	Whether this is for setting the RHS for the (n=0,|m|>0) step of the two-step solver

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine solve_axisymmetric_poisson

(

integer

iflag

)

High level Poisson solver routine for n=0, offers choice of outer iterative solver, two-step solver, and FSA solver.

Parameters

[in]

iflag

Flag to indicate whether adiabatic elec or full eq. is solved, integer

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine solve_axisymmetric_poisson_iter	(	type(grid_type), intent(in)	grid,
		type(psn_type), intent(inout)	psn
	)

Poisson solver that uses an iterative method to solve for the axisymmetric electrostatic potential: The Poisson equation for the axisymmetric mode is

\[ -\nabla_\perp \cdot \xi \nabla_\perp \overline{\phi}_{k+1} + \frac{e n_0/T_{e,0}} \overline{\phi}_{k+1} = e \left( \langle \overline{\delta n_i} \rangle_g - \overline{\delta n_{e,NA}} \right) + \frac{e n_0/T_{e,0}} \langle \phi_{k} \rangle_{fs} \]

where \(k\) is the iteration index, \(xi\) is the electric susceptibility, \(e\) is the elementary charge, \(n_0\) and \(T_0\) are the background density and temperature, \(\langle \overline{\delta n_i} \rangle_g\) is the gyroaveraged ion (gyrocenter) charge density, \(\overline{\delta n_{e,NA}}\) is the non-adiabatic electron charge density, and \(\langle\dots\rangle_{fs}\) is the the flux-surface average. \(\overline{\dots}\) is the toroidal average.

The RHS and LHS of the axisymmetric equation can be Fourier-filtered (poloidal modes). The LHS of the non-axisymmetric equation can be Fourier-filtered in the toroidal and poloidal direction. The solver in both cases is a linear finite-element solver on an unstructured triangular grid.

Parameters

[in]	grid	Solver grid data, type(grid_type)
[in,out]	psn	Field data; the potential is stored there, type(psn_type)

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine solve_axisymmetric_poisson_two_step_fsa	(	type(grid_type), intent(in)	grid,
		type(psn_type), intent(inout)	psn
	)

Simple two-step 2D solver driver for the axisymmetric potential and one-step solver ("FSA-solver") with constraint equation for the flux-surface averaged potential. For the two-step solver, the RHS is the flux-surface averaged charge density, the LHS consists only of the polarization operator. The result of the 2D solve is flux-surface averaged; i.e., the result is \(\langle\phi\rangle\) The Poisson equation for the flux-surface averaged mode is

\[ -\nabla_\perp \cdot \xi \nabla_\perp \overline{\phi} = e \left\langle \left( \langle \overline{\delta n_i} \rangle_g - \overline{\delta n_{e,NA}} \right) \right\rangle_{fs} \]

where \(xi\) is the electric susceptibility, \(e\) is the elementary charge, \(n_0\) is the background density, \(\langle \overline{\delta n_i} \rangle_g\) is the gyroaveraged ion (gyrocenter) charge density, \(\overline{\delta n_{e,NA}}\) is the non-adiabatic electron charge density, and \(\langle\dots\rangle_{fs}\) is the the flux-surface average. \(\overline{\dots}\) is the toroidal average.

NOTE: The two-step solver does not support non-zero Dirichlet boundary conditions!

The FSA solver solves the same equation as the outer-iterative solver (solve_axisymmetric_poisson_iter), but with a twist. The flux-surface averaged potential is treated as an independent field \(\lambda\) in the Poisson equation. In order to make the equation consistent, a second (constraint) equation is added that enforces \(\lambda=\langle\phi\rangle_{fs}\). So the FSA solver solves a 2x2 block system.

Parameters

[in]	grid	Solver grid data, type(grid_type)
[in,out]	psn	Field data; the potential is stored there, type(psn_type)

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine solve_turb_poisson	(	integer	iflag,
		integer	ipc
	)

High level Poisson solver routine for |n|>0 Currently, only one implementation is available.

Parameters

[in]	grid	XGC grid data object
[in,out]	psn	XGC field data object
[in]	iflag	Flag to indicate whether adiabatic elec or full eq. is solved, integer
[in]	ipc	RK2 stage index, integer

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine solve_turb_poisson_default	(	type(grid_type), intent(in)	grid,
		type(psn_type), intent(inout)	psn
	)

Routine to solve the non-axisymmetric Poisson equation:

\[ -\nabla_\perp \cdot \xi \nabla_\perp \delta\phi + \frac{e n_0/T_{e,0}} \delta\phi = e \left( \langle \delta n_i \rangle_g - \overline{\delta n_{e,NA}} \right) \]

.

Parameters

[in]	grid	XGC grid data object
[in,out]	psn	XGC field data object

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine spectral_solver_decompose	(	type(grid_type), intent(in)	grid,
		real(kind=8), dimension(grid%nnode), intent(inout)	field_inout,
		real(kind=8), dimension(grid%nnode,2), intent(out)	field_ntor
	)

Decompose field into spectral components treated by the spectral solver and the rest treated by the regular solver.

Parameters

[in]	grid	XGC grid object, type(grid_type)
[in,out]	field_inout	Input data, real(8)
[out]	field_ntor	Output

Here is the call graph for this function:

Here is the caller graph for this function:

subroutine spectral_solver_reassemble	(	type(grid_type), intent(in)	grid,
		real(kind=8), dimension(grid%nnode), intent(inout)	field_inout,
		real(kind=8), dimension(grid%nnode,2), intent(inout)	field_ntor
	)

Reassemble the full field from the spectral components treated by the spectral solver, and the rest treated by the regular solver.

Parameters

[in]	grid	XGC grid object, type(grid_type)
[in,out]	field_inout	Input data, real(8)
[in,out]	field_ntor	Output

Here is the call graph for this function:

Here is the caller graph for this function: