# Question concerning example 5 - phonon dispersion of MgO

Hi all,
I just tried my first GULP calc - example 5 - to calculate the phonon dispersion of MgO in gamma-X direction. The program seems to run but the output is unreasonable - there are 10 dispersion branches with the incorrect TO and LO mode frequencies at gamma point. There should only be 6 dispersion branches (effectively 4 because TO and TA are doubly degenerate) and I would think that this should be independent of model parameters. I am running on MAC-OS Sierra 10.12.6.

My question is - were the parameters in example5.gin carefully chosen to give results consistent with literature? I am an experimentalist and do not have a good sense yet of the reasonableness of model parameters.

David

Hi David,

The parameters for MgO in the example may not be the absolutely best (though this is hard to define) but are pretty typical of a shell model for this system and so the results won’t change much with other parameters. Of course it has the limitations of a dipole-only shell model (e.g. no Cauchy violation etc). As to the phonon dispersion, it’s important to check that you are sampling the direction you think you are. GULP works with the primitive cell for the calculation which is generated from the F centred cell to maintain the same orientation. If you are specifying the dispersion in terms of the F centred cell you will need the “kfull” keyword to switch from the primitive to centred cell Brillouin zone.
Regards,
Julian

Hi Julian,

I am also quite interested in Example 5, and when I run it, I also find 10 dispersion branches. Could you tell me how to modify the input file to examine the dispersion along the Gamma-X direction?

Connor

Hi Julian,
thanks for the quick reply. I now have TWO questions about the Buckingham parameters in example5.gin.

Q1. According to the help file for “buckingham” there are no defaults . How does the program know which are the energy units?

Q2. According to the help file format line there are 8 possible inputs, that is:
atom1 atom2 A rho C rmax <3*flags>

but I only see 7 inputs specified in example5.gin

buck
O shel Mg core 1295.553402 0.300000 0.00000 0.000 8.000
buck
O shel O shel 22764.000000 0.149000 27.88000 0.000 12.000

How do I know which parameters are specified?

Thank-you,
David

there should be an in the 8 inputs , that is
atom1 atom2 A rho C rmax <3*flags>
but only 7 inputs are specified

buck
O shel Mg core 1295.553402 0.300000 0.00000 0.000 8.000
buck
O shel O shel 22764.000000 0.149000 27.88000 0.000 12.000

for some reason, the rmin" which is before “rmax” in the list disappearing when I hit reply.

Hi David,
Q1: The energy units of GULP should always be eV unless you specify an option/sub-option to change this to something else.
Q2: The flags only are required for a fitting run and so the input would be:
atom1 atom2 A rho C rmin rmax (i.e. 7 inputs)
The inclusion of rmin is optional and would default to 0 (i.e. the normal sensible value) if not specified.
The output of the run should tell you what parameters have been read and their units - see line 184 in the output for example5 or thereabouts.
Regards,
Julian

Hi Connor,
Just add “kfull” as a keyword and the direction you are interested for the FCC cell & I hope this should work.
Cheers
Julian

NB: Example5 doesn’t use symmetry and so I should have added that to get a difference between kfull and not you would specify the structure as:

cell
4.212000 4.212000 4.212000 90.000000 90.000000 90.000000
fractional
Mg core 0.000000 0.000000 0.000000
O core 0.500000 0.500000 0.500000
O shel 0.500000 0.500000 0.500000
space
225

With symmetry you can then switch between the primitive and F centred cell.

Thanks again Julian for your responses. However the problem of 10 different \omega values for a given k vector (which Connor also observes) has nothing to do with the correct direction in k-space. Unless I am mistaken, there should only be maximum 6 different \omega values at any point in the Brillouin zone because MgO has a 2 atom basis.

Is it possible that there is some control for convergence within GULP?

Hi David,
This is where the “NB” I added to the reply to Connor’s post comes. Example 5 is using an explicit F-centred cell with 8 atoms in it and so there are a potential 24 modes (including translation/acoustic). If you want to work with the primitive 2 atom cell then it’s important to use the symmetry-based crystal structure input given in the post with the space group. Then you’ll only have 2 atoms and a maximum of 6 modes.
Regards,
Julian

Hi Julian,

yes I see that. If I add “eigen” to the calculation to calculate the modes, I see that you get 24 possible modes. The problem is that the incorrect number are degenerate.

David

Hi David,
I’ve compared the symmetry-adapted run with kfull specified as a keyword against the curve plotted for [001] in <I> Phonon Dispersion Relations Database (for example) and both have 4 branches with the same shape & so I think everything in GULP is fine.
Regards,
Julian

Yes it does. Thanks! That is exactly the help I needed.
Best Regards,
David