Modified Ta PAW potential to calculate accurate EFGs

[knlausch]

Dear VASP community,

I am Ian Yap Chang Jie, a PhD student from the Civil Engineering department of the University of Duisburg Essen and in collaboration with different theoretical and experimental groups. I work in nuclear spectroscopy (Time differential perturbed angular correlation or TDPAC) measuring electric field gradients (See wiki/index.php/Electric_Field_Gradient). In particular, I am working on rutile TiO2 with Tantalum (Ta)-induced substitutional defects. In addition to our experimental work, we have been using CP-PAW, developed by Peter Blöchl, to calculate EFGs of the material. However, we are currently considering switching to VASP, because it offers the HSE06 exchange functional in addition to the PAW method.

As EFG calculations require high-quality PAW sets, we must include the semi-core electrons of titanium and tantalum. We already had a look at the list of available PAW potentials under wiki/index.php/Available_PAW_potentials and found that for Ti, we could use the Ti_Sv potential, which has 12 valence electrons corresponding to 3s2 3p6 3d2 4s2 configuration. However, for Ta, we could only find the Ta_pv potential, which has a valency of 11, which I believe corresponds to 5p6 5d3 6s2. In our studies using CP-PAW, we found that the 5s2 and 4f14 states must be included in the valence states to generate an accurate EFG, which replicates our experimental Vzz values. So, we would need to increase the total number of valence electrons to 26-27 (depending on the charge of the supercell that we have generated).
Hence, I would like to ask if there exists a Ta PAW potential that we could use for this purpose? And if so, how do we activate it?

Thank you for considering my question, and I hope to hear from your soon.

Best Regards,
Ian Yap Chang Jie

[andreas.singraber]

Dear Ian Yap Chang Jie,

I brought up your request in today's meeting. As you already found out we currently do not supply a matching PAW potential for Ta. However, we will try to generate one and do some quick tests. Then we can send it to you for further testing and application. I will report back soon with an update, please stay tuned...

All the best,
Andreas Singraber

[knlausch]

Dear Andreas Singraber,

thank you very much for bringing up our request and generating a modified potential for Ta! I'm looking forward to receiving the potential so that we can do some test calculations and compare them to our current data.

Thank you again and best regards,
Ian Yap Chang Jie

[andreas.singraber]

Hello again!

I already sent you an email with the modified Ta potential. It would be great to know whether it worked as desired. If there are any further questions just continue this topic, thank you!

All the best,
Andreas Singraber

[knlausch]

Dear Andreas Singraber,

thank you very much! We received the potential and will run our tests as soon as possible. We will update the post once we have the results or if anything else comes up.

Best regards,
Ian Yap Chang Jie

[CrimsonFirebird]

Good evening Andreas Singraber,

I apologize for the lack of update till now, as my group has been trying to acquire our own license to run in our own capacity.

Just a question. I understand that from the over completeness problem, the Ta(TiO2) system will not run when the normal configuration is used. This is seen from the CP-PAW program where under the electronic minimalization using the damped newton dynamics, we would need to reduce the timestep to a low value to allow for the wavefunctions to properly run. How do we do that for the enhanced Ta case?

If that helps, here is our INCAR file that is successful for the ionic minimalization of the pure TiO2 cell using HSE06 (Ti_sv and O PAW potentials):

SYSTEM = (Ta)TiO2 optimization
PREC = Accurate
ISTART = 2
ICHARG = 2

ENCUT = 520
ISPIN = 1
NELMIN = 5
NELM = 100
ALGO = Normal
EDIFF = 0.000000001

IBRION = 2
ISIF = 3
NSW = 100
EDIFFG = -0.000001
POTIM = 0.5

ISMEAR = -5 ! Tetrahedron method with Bloechl corrections
SIGMA = 0.01

LHFCALC = T
HFSCREEN = 0.2
AEXX = 0.25
GGA = PE

LORBIT = 11

LPLANE = .TRUE.
NCORE = 4
LSCALU = .FALSE.
NSIM = 4

LEFG=.TRUE.

The error code that is returned is:


| Orbital orthonormalization failed in the inversion of matrix |
| LAPACK: Routine ZPOTRF failed! kpoint: 1 spin: 1 |
| |
| Possible solutions for this issue are: |
| - Please check whether the atoms are too close to each other. |
| - Decreasing POTIM might help during relaxations or MDs. |
| - Deleting CHGCAR or WAVECAR if incompatible with the structure. |
| - Try calculating without symmetrizing (ISYM = 0 or ISYM = -1). |
| - As last resort: using a different PAW for some of the atoms. |

I tried decreasing POTIM to 0.01 under IBRION = 2, including ISYM=0 and disabling the whole ionic relaxation altogether (which indicates electronic divergence).

Thank you for reading my email, I hope to hear from you soon.

Have a great Merry Christmas and a Happy New Year.

Best Regards,
Ian Yap Chang Jie

[CrimsonFirebird]

Apologies for my tardiness, I have just realized that in the process of constructing my supercell, I did not set the fractional coordinates properly. I divide the actual coordinates by the original lattice constant and not the supercell lattice constant.

Now the program runs nicely with the current INCAR file.

Best Regards,
Ian Yap Chang Jie