Can someone give me a bit more information about what mathematical method Mathematica is using, (preconditioner, filtering-restarting, deflation) in the Lanczos implementation when I input something like this:
{eig, eivec} =
Eigensystem[
SparseArray[(Heff + ConjugateTranspose[Heff])/2], -1,
Method -> {"Arnoldi", "StartingVector" -> startvec,
"Tolerance" -> tolerance, "MaxIterations" -> iterations}]`
I want to always compute single eigenvalue with the smallest magnitude.
Arnoldimethod uses the ARPACK library. – ilian Jan 07 '16 at 15:36