ABINIT Discussion Forums

I have trying to calculate phonon frequencies for SnO2 (Tin Dioxide, rutile structure, tetragonal) but I do not obtain "good" frecuencies compared to experimental and also compared to reference " DFT study of the electronic, vibrational, and optical properties of SnO2" (P.D. Borges, L.M.R. Scolfaro, H.W. Leite Alves, E.F. da Silva jr.; Theor Chem Acc Vol. 126, pp. 39-44 (2010)) wich also calculates phonon frequencies using ABINIT. For example: lowest phonon frequency is 141 cm-1 while it should be 104 between 87 and 123 (experimental) or close to 83 cm-1 (calculated at LDA from ABINIT in ref. mentioned). I am confused to what to do to improve the results to get closer to at least the mentioned results in the reference. I have tried yet to increase ecut from 50 to 55 and to 60 Hartree without futher improvements. I attach the input file for everyone can see if I am doing something wrong or can give me suggestions. I will apreciate all the help given.
Cheers, Sergio Ferrari

Hi!

1] Are your ecut and ngkpt converged wrt phonon frequencies?

2] Are your unit cell and atomic positions relaxed?

3] Did you use the same parameters (ecut, ngkpt,...) for cell relaxation and response function? What value for Pressure do you get in your response function output?

Best regards,

Igor

Hi Igor,
Thanks for your answer, let me respond to your questions:

ilukacevic wrote:1] Are your ecut and ngkpt converged wrt phonon frequencies?

As I said in my post, increasing ecut gave me the same result; so at least respect ecut ecut had converged wrt phonon frequencies. About ngkpt, the number of k-points used was the typical for this unit cell, and I have use that set for relaxing the atomic positions and the cell previously. The variable ngkpt was converged at leat for relaxing purpouses, but I admit I did not make a study of convergence of ngkpt wrt phonon frequencies; ALTOUGH I have recently made a run with a higher ngkpt (ngkpt 6 6 12) with almost the same results: If there the problem is a question of rising ecut and/or ngkpt I wonder how much futher I have to increase them to get a more proper result

ilukacevic wrote:2] Are your unit cell and atomic positions relaxed?

As I said in the previous answer to you question, I have made a relaxation of internal coordinates and cell ( tolmxf 1.0e-8) previous to the phonon frequencies calculation (with ngkpt 5 5 10, ecut 50) and used the results for the phonon calculations.

ilukacevic wrote:3] Did you use the same parameters (ecut, ngkpt,...) for cell relaxation and response function? What value for Pressure do you get in your response function output?

Yes, of course the parameters for cell/internal coordinates relaxation were the same use for responde function. They were ngkpt 5 5 10, ecut 50; and the pressure that I got was "Pressure= 3.2606E-06 GPa" what seems low enought to consider that the structre is relaxed.

Please if you have any have any other suggestions they will be welcome!

Pushing futher the ecut and ngkpt, I just made a run that first optimize the internal coordinates, then later cell constants and internal coordinates (toldmxf 1.e-8) with ngkpt 6 6 12, and ecut 70 Ha. Pressure obtained was "Pressure= 4.6927E-07 GPa". Phonon frequencies: (optical modes, cm-1)
1.420420E+02
1.662215E+02
2.853423E+02
2.888886E+02
3.021415E+02
4.218785E+02
4.336116E+02
4.616864E+02
4.966128E+02
4.966130E+02
5.981731E+02
6.396225E+02
6.711675E+02
7.741852E+02
7.999499E+02

still fail to reproduce the experimental results or the results reported in the reference mentioned ((P.D. Borges, L.M.R. Scolfaro, H.W. Leite Alves, E.F. da Silva jr.; Theor Chem Acc Vol. 126, pp. 39-44 (2010)) wich also claims to use ABINIT for the calculation of phonon frequencies. I suppose I can continue increasing ecut or/and ngkpt; but I still fear that I will go on having this problems and there may be something that it is escaping to my sight. Please help me!

Hi!

Maybe the problem is in the pseudopotentials used? It could be that they used different number of valence electrons in their pseudo's. They don't specify in the paper which ones they used.

Best regards,

Igor

Hi

A few questions/comments:

1. You have 18 frequencies, have you tried to calculate the phonon frequencies with only the 3-atom primitive unit cell ?
2. Are you using PAW dataset or norm conserving pseudopotentials?
3. Are your first three frequencies zero?
4. Go in the output file, and search for the decomposition of the second derivative of the total energy (in PAW: search for the term "Fourteen", in NCPP: search for the term "Thirteen"). At the end of the decomposition, there should be "2DTE variational" and "2DTE non-var". These numbers should be similar (up to 10^-3). Check if they are.
5. Do not use more cpus than kpoints in your calculations. Especially, at Gamma, there are few kpoints because of symmetries, so be sure your number of cpus is smaller than the reduced number of kpoints.