Mumpster
http://mumpster.org/

Positive-indefinite rank deficient matrix
http://mumpster.org/viewtopic.php?f=8&t=1808
Page 1 of 1

Author:  michoski [ Tue Jul 05, 2016 9:04 am ]
Post subject:  Positive-indefinite rank deficient matrix

Hi all,

I am solving a matrix problem that corresponds to doubly periodic boundary
condition in a FEM representation. It leads to a positive-indefinite rank deficient matrix,
the null space dimension is one, and it's a vector of some constant C, I think. As far as I am thinking, the system can be written Ax=B, but since the boundaries, there is a null vector y, such that A(x+y) = B, for some Ay=0. I would like to solve the system under the global constraint that y (or, the effect of y on the solution x) vanishes. I am not sure how to make this happen in MUMPS. I think in petsc something like KSPSetNullSpace() might work...?

Right now I set:

MatMumpsSetIcntl (F, 7, 2);

MatMumpsSetIcntl (F, 24, 1);

MatMumpsSetCntl (F, 1, 0.1);

MatMumpsSetCntl (F,4, 1.E-6);

But the solution I get back is still off by a constant. Is there a way of determining a unique solution?

Thanks,
Craig

Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/