[PROBLEM] canonical_configuration and small positive frequencies

[flokno]

When creating samples via canonical_configurations and input dispersions which have very small but positive frequencies, the resulting displacements can be very large. This comes from the fact that TDEP treats frequencies larger than lo_freqtol (=1e-9) as positive: https://github.com/tdep-developers/tdep/blob/main/src/libolle/type_forceconstant_secondorder_aux.f90#L439

If the phonon dispersion is close to zero at a commensurate point, a large displacement will therefore be created.

One possible way out would be to check that the resulting amplitudes are not too big or something, or throw an error

[OrbitalC]

A solution I have seen for this kind of problem is to set all frequencies lower than some threshold to some positive frequency, a small one but large enough to have sensible displacements. Maybe the positive frequency is something that can be inferred from the rest of the spectra ?
Compared to ignoring the low/imaginary frequencies, this allow to still excite the system on the specific directions of the mode, which helps for the next iteration.
Since at convergence everything should be positive anyway, this have no impact on the final results.
However, I think that throwing a warning when this happens is a good idea.

[mjv500]

It's a little dangerous for systems with low frequency modes, as the sampling amplitudes will never be correct if the cutoff is higher than the soft mode frequency. As you converge this threshold should be reduced somehow The physical thing which limits the actual amplitude of the displacements are higher order phi. We could imagine a better estimation formula based on the phi2,3,4 at a given iteration... but then we're not Gaussian and variational (gasp). M.

…

On Fri, Aug 2, 2024 at 5:18 PM Aloïs Castellano ***@***.***> wrote: A solution I have seen for this kind of problem is to set all frequencies lower than some threshold to some positive frequency, a small one but large enough to have sensible displacements. Maybe the positive frequency is something that can be inferred from the rest of the spectra ? Compared to ignoring the low/imaginary frequencies, this allow to still excite the system on the specific directions of the mode, which helps for the next iteration. Since at convergence everything should be positive anyway, this have no impact on the final results. However, I think that throwing a warning when this happens is a good idea. — Reply to this email directly, view it on GitHub <#82 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABWOT32V6JJ4ODO3LQYZVDTZPOPODAVCNFSM6AAAAABL2PCPH6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENRVGYZTAMRZGM> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Professor Matthieu J Verstraete Fellow, American Physical Society Chair, Steering Committee, European Theoretical Spectroscopy Facility www.etsf.eu Alumnus Fellow, Young Academy of Europe yacadeuro.org Institute for Theoretical Physics (ITP) Department of Physics Buys Ballot Gebouw/Building, Princetonplein 5, office University of Utrecht, 3584 CC Utrecht ITP Secretariat: +31 30 253 5928 E-mail: ***@***.*** Group web page (Liège): http://www.nanomat.ulg.ac.be/ Nanomat lab, Q-Mat center, Université de Liège Département de Physique, Bat. B5a, 4/50 Allée du 6 août, 19 B-4000 Sart Tilman, Liège Belgium Phone : +32 4 366 90 17 European Theoretical Spectroscopy Facility (ETSF) Mail : ***@***.*** ***@***.*** ***@***.***