Ciao Giacomo,
No matter how many (movable) layers you include, if the relaxed
structure is a real energy minimum, any small displacement of the
coordinates you allowed to move in the first place, must increase
the energy. Agree ?. Thus you cannot have negative (i.e. imaginary)
frequencies.
Perhaps you're not aware of the structure of the *.FC file, let me
explain. Each (movable) atom is given a small displacement (0.05 Bohr
or so) along the two senses of every cartesian axes, +x,-x,... ,+z,-z.
For each of these steps the forces over the whole box are recorded.
Namely, each movable atom gets a block of 6N lines and 3 columns in
the *.FC file.
My previous comment referred to checking the forces, say, for +dx and
-dx, they should be opposite or close to.
Now, your FC calculation must have been for a 132 atoms block, though
you allowed to move only, say, the first 44. Your FC file contains then
6*44*132 lines. However, for the Vibra utility, the forces on atoms
45 and on must be whiped out from the FC file. Namely, you cheat Vibra
by telling your cell contains only 44 atoms, and give the respective
coordinates of course. Thus your new FC file must contain 6*44*44 lines.
IS THIS WHAT YOU'RE DOING ?. I bet not, so your "disastrous" result ;-).
Best,
Roberto
On Fri, 7 Sep 2012, Giacomo Giorgi wrote:
Ciao Roberto, Hi Alexander,
thanks for the kind advices.
Actually, my geometry optimizations where initially performed on a slab where
only the bottom layer was kept frozen.
Then on such optimized structure I have frozen also the second layer,
considering that if I want to study the mechanism of anchoring only the top
layer (24 atoms) and the 2 molecules (As I told I am trying to compare my
results with those of an experimental spectrum obtained at high concentration
of adsorbate) anchored on top of it will contribute to the IR spectrum.
Then I have reduced even more the convergence on the forces. Let's say from
0.05 eV/Ang to 0.01 eV/Ang. I did the FC calculation getting these "glorious"
results... :(
I wonder if the Asymmetric nature of the slab can play a role in my results.
In other words I was wondering if for example the negative (Really a too
Large number) Frequencies can be due to the vibrational modes of surface
atoms (the top layer for which I am also calculating the Frequencies) on top
of which the organic molecules are anchored.
In other words, should I consider only the Frequencies of the adsorbed
molecule? Or am I free to include atoms of the surface?
Roberto, regarding your advice to check if for
opposite displacements I am getting opposite forces, I wonder if you refer to
a Symmetric slab with molecules anchored on both top and bottom layer.
Thanks a lot!
Very best regards,
Giacomo
On Thu, 6 Sep 2012, R.C.Pasianot wrote:
Ciao Giacomo,
I know nothing about Born charges, but did FC calculations a number
of times: 67 negative frequencies out of 132 means your calculation
is garbage. FC calculations are quite delicate in the sense that you
need, 1st very well relaxed structures, and 2nd very well converged
forces in the FC calculation step.
At 1st glance, the only thing I wonder regarding your data file is
if those 400 Ry of MeshCutoff are large enough to eliminate the
egg box effect for a box having an edge of 40 Ang along z axis.
This might be affecting force precision while evaluating the FC
matrix. You might tell if something along these speculations is
going on by looking at the FC file itself; watch e.g. if for
opposite displacements you are getting opposite forces, check that
the largest force occurs on the atom being displaced, that the
total force for the whole bov is a good zero. etc.
Good luck,
Roberto
On Thu, 6 Sep 2012, Giacomo Giorgi wrote:
Dear All,
I am doing for the first times the calculation of Frequency with
SIESTA. My system is an asymmetric slab (2 bottom layers are frozen) and
on top of it I anchored an organic molecule.
In total 92 atoms. But freezing the first two layers I need to know the
frequencies of only 44 atoms, i.e. 44*3 =132 Eigenvectors.
Anyway, by checking both the .bands and the .vectors files I observe that
the frequencies are highly underestimated with respect to the experimental
ones (my last calculated frequency, the 132nd, is at 1649.101 cm-1, while
experimentally a net visible peak is at 2900 cm-1)
Moreover the first 67 (!!) of 132 Eigenvectors are characterized by a
negative Frequency.
Below my Vibra .fdf
Any comment and hint is extremely appreciated.
Additionally, as previously posted (but no answer) even if I set
BornCharge True
in the fdf, no .BC file was generated thus no IR Spectrum... :(
Many thanks,
Giacomo
SystemName 101
SystemLabel 101
################# options for Vibra ####################
Eigenvectors .true.
SuperCell_1 0
SuperCell_2 0
SuperCell_3 0
BandLinesScale ReciprocalLatticeVectors
%block BandLines
1 0.000 0.000 0.000 \Gamma
%endblock BandLines
# Kpoints
KPointMPSampling F
%block PolarizationGrids
6 3 3
4 8 4
5 5 10
%endblock PolarizationGrids
BornCharge True
########################################################
MD.TypeOfRun FC # Type of dynamics:
MD.FCfirst 49
MD.FClast 92
MD.FCdispl 0.02 ang
NumberOfAtoms 44
NumberOfSpecies 4
%block ChemicalSpeciesLabel
1 22 Ti
2 8 O
3 6 C
4 1 H
%endblock ChemicalSpeciesLabel
PAO.BasisSize DZP
XC.functional LDA
XC.authors CA
ElectronicTemperature 1500.0 K
LatticeConstant 1. Ang
%block LatticeVectors
7.569 0 0
0 10.2394 0
0 0 40
%endblock LatticeVectors
MeshCutoff 400.0 Ry
MaxSCFIterations 200
DM.MixingWeight 0.008
DM.NumberPulay 5
DM.Tolerance 1.d-4
ParallelOverK .true.
SolutionMethod diagon
WriteDM .true.
DM.UseSaveDM .true.
PAO.EnergyShift 0.01 Ry
MD.MaxForceTol 0.01 eV/Ang
WriteForces true
WriteCoorCerius true
WriteCoorXmol true
WriteMullikenPop 1
WriteMDXmol true
UseSaveData .true.
%block GeometryConstraints
position from 1 to 48 # 2 layers
%endblock GeometryConstraints
AtomicCoordinatesFormat NotScaledCartesianAng
%block AtomicCoordinatesAndAtomicSpecies
....
....
%endblock AtomicCoordinatesAndAtomicSpecies
