We should interface also to the Jacobi-Davidson solver. It is also supported in Spectra so I think it will be relatively easy.
https://spectralib.org/doc/classspectra_1_1davidsonsymeigssolver
Jacobi-Davidson should be faster that implicitly-restarted Lanczos when a small part of the (extreme) spectrum is needed.
We may also wish to support shift-invert mode for some cases, where the spectral lines of interest may be the near-zero lines.
We should interface also to the Jacobi-Davidson solver. It is also supported in
Spectraso I think it will be relatively easy.https://spectralib.org/doc/classspectra_1_1davidsonsymeigssolver
Jacobi-Davidson should be faster that implicitly-restarted Lanczos when a small part of the (extreme) spectrum is needed.
We may also wish to support shift-invert mode for some cases, where the spectral lines of interest may be the near-zero lines.