PEPSetWhichEigenpairs#

Specifies which portion of the spectrum is to be sought.

Synopsis#

#include "slepcpep.h" 
PetscErrorCode PEPSetWhichEigenpairs(PEP pep,PEPWhich which)

Logically Collective

Input Parameters#

pep - the polynomial eigensolver context
which - the portion of the spectrum to be sought

Options Database Keys#

-pep_largest_magnitude - Sets largest eigenvalues in magnitude
-pep_smallest_magnitude - Sets smallest eigenvalues in magnitude
-pep_largest_real - Sets largest real parts
-pep_smallest_real - Sets smallest real parts
-pep_largest_imaginary - Sets largest imaginary parts
-pep_smallest_imaginary - Sets smallest imaginary parts
-pep_target_magnitude - Sets eigenvalues closest to target
-pep_target_real - Sets real parts closest to target
-pep_target_imaginary - Sets imaginary parts closest to target
-pep_all - Sets all eigenvalues in an interval or region

Notes#

The parameter ‘which’ can have one of these values

PEP_LARGEST_MAGNITUDE - largest eigenvalues in magnitude (default)
PEP_SMALLEST_MAGNITUDE - smallest eigenvalues in magnitude
PEP_LARGEST_REAL - largest real parts
PEP_SMALLEST_REAL - smallest real parts
PEP_LARGEST_IMAGINARY - largest imaginary parts
PEP_SMALLEST_IMAGINARY - smallest imaginary parts
PEP_TARGET_MAGNITUDE - eigenvalues closest to the target (in magnitude)
PEP_TARGET_REAL - eigenvalues with real part closest to target
PEP_TARGET_IMAGINARY - eigenvalues with imaginary part closest to target
PEP_ALL - all eigenvalues contained in a given interval or region
PEP_WHICH_USER - user defined ordering set with PEPSetEigenvalueComparison()

Not all eigensolvers implemented in PEP account for all the possible values stated above. If SLEPc is compiled for real numbers PEP_LARGEST_IMAGINARY and PEP_SMALLEST_IMAGINARY use the absolute value of the imaginary part for eigenvalue selection.

The target is a scalar value provided with PEPSetTarget().

The criterion PEP_TARGET_IMAGINARY is available only in case PETSc and SLEPc have been built with complex scalars.

PEP_ALL is intended for use in combination with an interval (see PEPSetInterval()), when all eigenvalues within the interval are requested, and also for computing all eigenvalues in a region with the CISS solver. In both cases, the number of eigenvalues is unknown, so the nev parameter has a different sense, see PEPSetDimensions().