On Wednesday, November 2, 2016 at 9:48:38 AM UTC-4, Steven G. Johnson wrote: > > > > On Wednesday, November 2, 2016 at 6:43:48 AM UTC-4, Alejandro Castellanos > wrote: >> >> I use this command: *l, v = eigs(M,nev=15,which = :SM ,ritzvec=true)* >> > > How are you constructing the matrix M? > > Apparently eigs thinks that your matrix is symmetric positive-definite > (SPD), because it is trying to use a Cholesky factorization. Schrodinger > operators are indefinite unless the potential V is >= 0, so this won't > work. (If you're calling cholfact, don't.) >
Oh, nevermind, it is using an LDTL factorization, which is appropriate for symmetric-indefinite operators. Seems like it might be a bug in the eigs function — it shouldn't have a problem with singular matrices.
