KERNEL_TRUNCATION for charged system

[jess_white]

Hi,

I am currently trying to do charged calculations as I have a CO3^2- molecule adsorbing onto a diopside surface. To do the charge I initially did dipoles but then saw that this would be a better method (you will see lots of # in the INCAR of things I tried).

What I am finding is that with the calculations they are very crazy when I consider the KERNEL_TRUNCATION for both the adsorbed surface and the clean surface. When I commented out the KERNEL in the INCAR for the surface it goes back to normal so I know its something to do with the KERNEL_TRUNCATION addition.

I have done the following:
- removed the dispersion forces
- increased vacuum as I know that was a previous issue
- have ISPIN = 1 and ISPIN = 2 (both gained same results)

I even changed the surface normal to see if that would do anything but still got the crazy results. Any help would be greatly appreciated!

[henrique_miranda]

Hi Jess,

Could you explain a little bit more what you mean by crazy results

I am not familiar with this particluar system, what you are trying to calculate and what results you are expecting.
Do you mean the total energies, forces, dipole moments?
What are the differences between when you use Coulomb truncation and not? Could you share the input and output files for both runs with and without the coulomb truncation?

Sharing the input and output files for the other tests you did would also be help me understand the issue.

[jess_white]

Hi Henrique,

Thanks for your response!
So what I mean by crazy results is that I expect to get a total energy of around -1000 eV for the system but when I add in the KERNEL_TRUNCATION it goes to -10000 eV. I have attached them and I hope they load properly. I am trying to calculate carbonate adsorption on the surface without the addition of Na to balance the charge - as CO3 has a 2- charge so I changed the NELECT to account for 2 more electrons.

The no_kernel does have solvation which I then removed when doing the charged calculations as I was trying to ensure to get the best results as possible. Please let me know if you require anything else.

Thanks so much!

[jess_white]

Don't need to post this but letting you know that I am currently making like a table to make it easier to compare the results and to understand the problem

[henrique_miranda]

I see I see. That is crazy indeed

I am looking at both output files the one without and the one with the kernel and it seems that the calculations are not completely consistent:

1. in one of then you set LSOL and in the other one not.
2. the POSCAR is not exactly the same
3. in one case ISPIN=2 and in the other one not.

To really pin down the issue, it would be helpful if you could repeat the exact same calculation just changing the INCAR tag related to coulomb truncation.
For a faster test I would suggest also to deactivate the ionic relaxation IBRION and perform a single SCF loop.

[jess_white]

diopside no kernel = -1171.64
diopside with kernel = -14235.54
diopside surface = -1142.03
diopside surface with kernel = -11923.99

Head of the POSCAR (could the issue be the z axis?)
1.00000000000000
13.3273639679000002 0.0000000000000000 0.0000000000000000
2.1764672898000001 10.3851944947000003 0.0000000000000000
1.2158538284000000 2.5382592667999999 26.3527136722000002
Ca_sv/a4736583 Mg/a90315392b5 Si/79d9987ad87 O/818f92134a0a C/253f7b50bb8d
16 16 32 99 1

forsterite no kernel = -1547.36
forsterite with kernel = -1578.95
forsterite surface = -1522.51
surface with kernel = -1522.51

Head of POSCAR
1.00000000000000
11.9974079132000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 9.5581750870000004 0.0000000000000000
0.0000000000000000 0.0000000000000000 31.9194335938000009
Mg Si O C
64 32 131 1

lizaradite no kernel = -959.90
lizaradite with kernel = -959.90
lizardaite surface = -905.43
lizaradite surface with kernel = -904.07

Head of POSCAR
1.00000000000000
10.6950040887847049 0.0000000000000000 0.0000000000000000
-5.3475020444931891 9.2621452346127953 0.0000000000000000
0.0000000000000009 0.0000000000000015 24.6682361192188395
Si Mg O H
16 24 72 32

talc no kernel = -1187.07
talc with kernel = -1556.44
talc surface = -1166.80
talc surface with kernel = -1466.36

Head of POSCAR
1.00000000000000
9.1764230728000005 0.0000000000000000 0.0000000000000000
0.0045896612000000 10.6283588035999994 0.0000000000000000
-0.1092733047000000 0.4700083260000000 28.3786530472999985
Mg Si O H C
24 32 99 16 1

wollastonite no kernel = -1759.03
wollastonite with kernel = -6826.99
wollastonite surface = -1731.48
wollastonite surface with kernel = -6682.72

Head of POSCAR
1.00000000000000
15.5493106842000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 14.7846975326999992 0.0000000000000000
-1.3854849006000001 0.0000000000000000 26.6618732573000017
Ca Si O C
48 48 147 1

carbonate molecule with kernel = -28.60

I should also mention that I have done the calculations by adding a Na to counter balance the charge and the adsorption energies were between -0.5eV and -1eV.

Hope this gives enough information

As the attachments are quite large I will be adding them one by one

[jess_white]

Here are some of the diopside results

[jess_white]

more attachments

[jess_white]

more attachments

[jess_white]

more attachments

[jess_white]

attachments

[jess_white]

more attachments

[jess_white]

more attachments

[jess_white]

Sorry I must of pressed submit before it loaded. Really sorry I had to put the attachments like that. They were too big to combine in any other way

[henrique_miranda]

Ok thanks for sharing the files but now I am a bit overwhelmed by information.
I will try to help but let's first look at the first set of files (diopside):

surf: diopside_surf.zip
1_kernel : diopside_1_kernel_co3.zip
1 : diopside_1_co3.zip

between 1_kernel and 1 only the kernel truncation tags change.
Between surf and 1 it seems that your intention is to only change the POSCAR but GGA=RE and IVDW=11 is also changed. Is that a mistake?
You also change NELECT but that is fine, I understand you want to add a charged molecule and compare the energies.

To compute the adsorption energy, if I understand correctly, you want:
E_ads = total energy from the surface with molecule - the energy of the molecule - the energy of the surface

The current problem you are facing is that the total energy of the surface with the molecule is changing quite a lot if you use coulomb truncation or not.
I think the correct procedure is to use couloumb truncation also in the calculation of the energy of the surface and use the same amount of vaccum in both calculations.

In fact, I would suggest that you start with the smallest vaccum 26.35A and converge the adsorption energy w.r.t vaccum and not the total energy of each individual calculation.
I doubt you need to use such a large amount of vaccum as 76A.

Let me know if this helps!