Shielding tensors on dummy atoms

[vaclavh]

Dear all,
I need to calculate NMR shielding tensors on dummy atoms (no nucleus and no basis functions, serve only as passive observers). Is there a way to do this in ATK?
- I found that the plane wave calculator can be used to calculate the shielding tensors but it does not seem to support the ghost atoms.
- Is there any way around such as defining my own atoms with no charge and no basis functions?

Thanks for help
Vaclav

[filipr]

The NMR method uses the GIPAW approach which again is based on the PAW formalism for treating the atomic cores. The PAW formalism is basically a way to use a mixed basis of plane waves and localized orbitals. In GIPAW the shielding tensors are calculated by projecting on the localized basis, which is described by the PAW dataset/pseudopotential. Maybe it would be possible to create a custom  PAW dataset for your "ghost" atoms where you set the charge to zero, but I honestly don't know if this will work as there are many assumptions in the code and it wasn't designed with this in mind (in PW you don't need ghost atoms as the basis spans all space). It may work, it may run but give wrong results or it may straight up crash - I don't know and I even wrote the code.

You are welcome to try it out, though. I can give you some hints as to how to do it, if you wish, but first consider if you want to try this very experimental approach.

[vaclavh]

Dear Filip,
thanks a lot for your help.

Perhaps I should better clarify what is the intention of the calculation. I would like to calculate the Nucleus Independent Chemical Shifts (NICS) which is a measure of induced ring currents in the material. My background is mostly calculations of large molecules in Gaussian, where this can be easily calculated using GIAO NMR. The dummy atoms serve as probes for measuring the induced magnetic fields - the shielding tensors on these atoms basically describe the induced fields without the influence of the nucleus, that's why they should have no charge. They should have no basis because otherwise, in the vicinity of other atoms (which is my case), there could be an unrealistic interaction with the dummy atoms.

Now the question is how to do this in ATK.
- One option would be to do it in the LCAO method, where the ghost atoms are defined and calculate the shielding tensors manually in the script. One problem is how to define the ghost atoms without the basis fcs, another one is that the calculation of the shielding tensor might be quite a challenge.
-Another one would be to use the GIPAW method and define the ghost atom with zero charge and no (or very small) localized basis
Let me know what you think is the best approach and I am definitely willing to try it out.
Thanks, Vaclav

[filipr]

Dear Vaclav,

NMR chemical shifts aren't implemented for the LCAO method in QuantumATK, only the plane-wave based GIPAW method is implemented. As I mentioned above, in this framework it isn't so easy to define "ghost atoms", as you don't have the arbitrary local basis sets as one has in LCAO and Gaussian based codes. I think that it is in theory possible to create a zero charged PAW dataset/pseudopotential and use that to create "ghost atoms" on which one can then calculate the chemical shifts. I am unsure whether the implementation currently supports this as I wrote above.

Even if it is possible to create a "ghost atom" PAW dataset and it works, there are some limitations to the approach, e.g. that the ghost atoms can't be placed too close to other atoms as the PAW augmentation spheres are not allowed to overlap.

[vaclavh]

Dear Filip,
as you wrote, the basis set is a combination of planar waves and localized atom-centered orbitals. Perhaps if one could use a zero-charge atom and manually set the localized part of the basis to a very small spherical function, that would be a way to go. But I see that this might be challenging and no guarantee for success.

Let me think about it for a while, maybe I'll come up with an alternative solution.

Thanks for your help! V.