Dear Ziba,it seems that you are randomly trying different options.Try to be
more systematic.MeshCutoff of 870 Ry seems to be an overkill -but certainly not
an error; if your tests show that you need as much as that, so be it.Did you
have a look at your structures with visualisation tools - is everything as it
should be, no obvious structure error?If your imaginary frequency is genuine
and reproducible,I'd look at the eigenvector of this mode, give a small
displacementto atoms along this eigenvector, and make the relaxation again. The
total energy ought to be lower than in your initial starting point;this is the
idea of an imaginary frequency: you displace the atomswithin the corresponding
mode, and the energy goes down.Hence you were not at the equilibrium; try
better.No way around.Best regardsAndrei Postnikov
- Ziba Torkashvand a écrit :
>
Thank you Professor Postnikov,Here I will provide two examples of my workI have
a hexagonal structure containing 18 atoms1. I have substituted one atom with a
gaseous one and relaxed it up to 0.000794 eV/A and 0.00458798 kBarthen I
calculated the phonon and fortunately, there is no negative frequency while the
force for the first step is approximated as 0.008174 eV/A
2. I have substituted two atoms with two gaseous ones and relaxed it up to
0.000549 eV/A and 0.00290065 kBarthen I calculated the phonon and
unfortunately, there are frequencies as -206.30292747012965 while the force for
the first step is approximated as 0.015862 eV/A
For both cases, I have used same inputs asFor relaxation:MeshCutoff870 Ry
DM.MixingWeight 0.02
DM.NumberPulay 3
%block kgrid_Monkhorst_Pack
6 0 0 0.0
0 6 0 0.0
0 0 1 0.0
%endblock kgrid_Monkhorst_Pack
For phonon: MeshCutoff750 Ry
DM.MixingWeight 0.02
DM.NumberPulay 8
%block kgrid_Monkhorst_Pack
3 0 0 0.0
0 3 0 0.0
0 0 1 0.0
%endblock kgrid_Monkhorst_Pack
I repeated phonon calculation for the second case with %block
kgrid_Monkhorst_Pack
6 0 0 0.0
0 6 0 0.0
0 0 1 0.0
%endblock kgrid_Monkhorst_Pack
and the resulting force for the first step is 0.013928 eV/A
I'm confused about what is happening in these two similar examples.
--
SIESTA is supported by the Spanish Research Agency (AEI) and by the European
H2020 MaX Centre of Excellence (http://www.max-centre.eu/)