krylov_bem_solver_new

the Krylov charge density solver is a solver using Krylov subspace methods to solve the linear charge densities equation. It offers the choice of GMRES and BiCGStab and requires to specify a matrix vector multiplication method to convert charge densities into potentials on the source elements. As of today only the fast multipole matrix is available. Optionally a preconditioner can be added. For details on the matrix and preconditioners see the next examples.

XML element: <krylov_bem_solver_new>
C++ type: KKrylovChargeDensitySolver
Also available as: krylov_charge_density_solver
Child elements

Element

Description

fast_multipole_matrix

The fast multipole matrix uses the KATRIN FFTM method to compute the multiplication of the charge density problem matrix with a vector.

fftm_multiplication

The fast multipole matrix uses the KATRIN FFTM method to compute the multiplication of the charge density problem matrix with a vector.

krylov_preconditioner

the preconditioner is a inner Krylov solver as right preconditioner for Krylov solvers.

Attributes

Attribute

Type

Description

intermediate_save_interval

unsigned int

iterations_between_restarts

unsigned int

specify the maximuum number of iterations before the algorithm will restart. This means using the last solution but discarding the constructed Krylov subspace. For more information on the theory of restarted gmres or bicgstab, google it.

max_iterations

unsigned int

specify the maximum number of iterations after which the solver will finish regardless of convergence

show_plot

bool

works with vtk only: display the convergence curve in a graph

solver_name

string

specify the solver type by name. gmres and bicgstab are available.

time_check_interval

unsigned int

time_limit_in_seconds

double

tolerance

double

specify the desired reduction of the start residual to define convergence. This defaults to 1e-8, meaning that the algorithm will stop when the norm of the residual has fallen to 1e-8 times its value in the first iteration (the norm of the vector describing the desired potential distribution).

use_display

bool

display progress in the form of iteration number and residual on the console

use_timer

bool

Example:

<krylov_charge_density_solver
    solver_name="gmres"
    tolerance="1e-8"
    max_iterations="500"
    iterations_between_restarts="500"
    use_display="true"
    show_plot="true"
>
    <fast_multipole_matrix/>
</krylov_charge_density_solver>

Used as: