Vibrational Frequency Calculation: Imaginary Frequencies from Optimized Minimum structures (H2O)

[manoj_wijesingha]

Hi,
I need to get the vibrational frequency calculation for the Metal doped Graphene. After the calculations, there are four imaginary frequencies, and some imaginary frequencies are higher (1.028868 cm-1, 1.334421 cm-1, 3.595799 cm-1, 125.990998 cm-1).

Then I wanted to run for H2O molecule as a test frequency calculation. The results are the same as the Metal doped Graphene. I could run the vibrational frequency calculation for H2O molecule after the relax calculation. But there are three imaginary frequencies for this H2O molecule. I don’t understand why is showing the imaginary frequencies for the optimized minimum structure.
Could you please explain me and figure out and what should I do for the frequency calculation.

I have attached my files of H2O.
Thank you.

INCAR:
System = FREQ
ISTART = 0
ICHARG = 2
PREC=Accurate
EDIFF=1E-6

ENMAX = 450
ISMEAR = 0
SIGMA = 0.05
LREAL=Auto

IBRION = 5
ISIF = 2
NSW = 1
EDIFFG = -0.01
ALGO=Fast
NFREE = 2
POTIM=0.005

KPOINTS:
KPoints-point
0
Monkhorst Pack
3 3 1
0 0 0

POSCAR
H2O _2
1.0000000000000000
8.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 8.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 8.0000000000000000
O H
1 2
Direct
-0.0000340711360937 0.0000000000000000 0.0000000000000000
0.0745271855680487 0.9040070284914102 0.0000000000000000
0.0745271855680487 0.0959929715085898 0.0000000000000000


Imaginary frequencies of first frequency calculation (H2O):
5.875007 cm-1, 148.058529 cm-1, 202.491147 cm-1


Then I increased EDIFF to 1E-8. After that the imaginary frequency (H2O) are;
3.087036 cm-1, 53.908671 cm-1, 77.342731 cm-1

[fabien_tran1]

Hi,

Could you also provide the OUTCAR of your calculation and mention which PAW dataset (POTCAR) you used? Have you used Phonopy for calculating the vibrational frequencies?

[manoj_wijesingha]

Hi,
Thank you so much and really appreciate your response. I have attached the OUTCAR files and POTCAR files. I used potpaw_PBE.54 dataset for the POTCAR. I did not use Phonopy for the frequency calculation.
Could you please assist me for the calculation of vibrational frequency.

Thank you.

Files.zip

[fabien_tran1]

For the moment, I can mention two things:

1) In principle, the calculation of phonon frequencies with finite differences should be done with a supercell that is large enough to describe properly long-wavelength phonons. With a single cell, only the frequencies at the GAMMA point may be correct, while at other points the values may be (wrongly) imaginary. See, e.g., https://iopscience.iop.org/article/10.1 ... 075/ac78b3

2) The calculation of phonon frequencies requires rather accurate settings. As you observed, there is a rather large difference between EDIFF=1E-6 and 1E-8.

[manoj_wijesingha]

Hi,
Thank you for your informative response. So, do you suggest Phonophy for the vibrational frequency calculations and is this better for the large super cell (5x5x1) calculation?
Thank you.

[fabien_tran1]

Not necessarily. For the moment, the important point is to consider a supercell. See also Phonons_from_finite_differences.

[manoj_wijesingha]

Hi,
Thank you for your guidance. If the supercell is large, how we can calculate the correct frequencies and how we can identify the structures are minima or transition structures? Could you please explain what should I do for the frequency calculation for the super cell 5x5x1.
This test H2O molecule is very simple calculation but it gives imaginary frequencies. Could you please suggest for me the process to compute frequency calculations to identify the minima and the transition states.

Thank you.

[fabien_tran1]

The topic has been moved to the forum "From Users for Users". Actually, now this is more about learning (from literature and documentation) than solving technical problems (.e.g., bugs).

[nidhi_tiwari]

Dear manoj_wijesingha,

I have also obtained similar result (three imaginary values out of six) in the frequency calculation of a single water molecule in a 7.34X7.34X20 Angstrom supercell, 4X4X1 Gamma KPOINTS, and EDIFF=1E-6.

I am really curious to know why an optimized molecule could show imaginary frequencies? I have attached the files of my calculation. Is there any mistake? If not, how can we rely of imaginary frequency values to distinguish transition states?

[fabien_tran1]

Among the 9 printed frequency modes of an isolated water molecule, only the first three are physically meaningful, see the following:
tutorials/latest/molecules/part3/#molecules-e10
https://www.chem.purdue.edu/jmol/vibs/h2o.html

Besides, the proper way to do a calculation on an isolated molecule is to choose only one k-point and to ensure that the cell is large enough to avoid interaction between the periodic images of the molecule.

[nidhi_tiwari]

Thank you for the answer.

[alex]

Hello, Fabian is right.

I might add some remarks if consider doing thermodynamics with partition sums caculated from these frequency analysis:
you need to do projection of the rotational and translational modes of your system (in case its gas phase, non-linear: 3N-6 degrees of freedom should ring a bell).
Some reading with the proper math: Wilson, JR., J. C. Decius, Paul C. Cross: Molecular Vibrations.

Or just use phonopy, it should also do the trick.

Best regards,

alex