Dear Erich,
As I am sure you are aware, if x is an eigenvector corresponding to an
eigenvalue lambda, then alpha*x, where alpha is a non-zero scalar, is
also an eigenvector corresponding to lambda. So, in particular we can
choose alpha = +1 or -1. The LAPACK symmetric eigenvalue routines
return orthonormal eigenvectors, so that in particular norm(x) = 1,
where norm is the 2-norm (Euclidean length), but do not choose any other
scaling to further particularise x.
Best wishes,
Sven Hammarling.
________________________________________________________________________
This e-mail has been scanned for all viruses by Star. The
service is powered by MessageLabs. For more information on a proactive
anti-virus service working around the clock, around the globe, visit:
http://www.star.net.uk
________________________________________________________________________
|