From: bauer@lbm.mw.tu-muenchen.de (Sebastian Bauer) Newsgroups: sci.math.num-analysis Subject: Re: Eigen values of Ax=kBx Date: 16 Sep 1995 13:45:27 GMT In article <43cemr$qjc@reznor.larc.nasa.gov>, "C.J. Reddy" writes: |> I am looking for any public domain mathematical software to compute |> eigen values of the following equation |> |> Ax=kBx |> |> where A and B are sparse, complex symmetric matrices and k's are |> the eigen values to be found. |> |> Till now I have come across routines that would compute eigen values |> of matrix of Ax=0 and not in the above form.(eg QMRPACK etc). |> |> I would appreciate, any help in this direction. |> |> C.J.Reddy Hi, I guess with complex symmetric you mean complex Hermitian. LAPACK contains the FORTRAN routines (simple driver routines) CHEGV/CHPGV (single precision) and ZHEGV/ZHPGV (double precision), which solve the generalized eigenvalue problem Ax=kBx for Hermitian matrices A and B whereby B has to be positive definite. The LAPACK package is available on all NETLIB-servers. In case you need theses routines in C, they are available in the package CLAPACK also from NETLIB. I took a short look into the GAMS search tree for numerical software (http://gams.cam.nist.gov/) into the Class D4b3 (complex Hermitian generalized matrix eigenvalue problems - see http://gams.cam.nist.gov/cgi-bin/gams-serve/class/D4b3.html), and it didn't show any routines for sparse complex systems. Netlib containes besides LAPACK also the package LANZ, which can solve generalized sparse eigenvalue problems, but I think, it can't handle complex matrices. So I think, it may be quite difficult for you to get exactly what you want without programming it yourself. Sebastian -- Sebastian Bauer Institute B for Mechanics, Technical University of Munich