Binding Energy in Hybrid system

[SKM]

Hi All,
seen some of the posts regarding Binding energy Calculation, but i could not get my doubt clarified, so posting here w.r.t my system

" I have a Hybrid structure A.B [A=say perovskite ; B=charge carrier material]"
I wish to find the binding energy between these A and B when both are combined.

What i did earlier on some other similar type system, was:
Step-1: Geometry/energy optimised the combined system [A.B]
Step-2: Geometry / energy optimised the [A] alone
Step-3: Geometry/energy optimised the {B] alone
all steps used same INCAR file.
from the OSZICAR files of these three systems, i did the Binding Energy as

Binding Energy = F[A.B] - F[A] - F

Now my queries are :
1. Is my above calculation correct? (irrespective of whther its LDA, GGA etc, as long as i used same method for all three)?
1a). If not, what should have been done?
1b). Even if its correct, is there any simple way to calculate to reduce 3 steps of calculations? (or alternate? )
Why am I asking this 1b) alternate way is because, i thought i can get binding energy from charge density difference calculations. i.e. to plot charge density difference we do Optimise system [A.B] first, and from that Optimised system separate POSCAR files of each [A] and and run with NSW=0 and write CHGCAR files, right? So my point is from these three calculations also, we can pickup the values of F=xxxxx from their OSZICAR files and use the same formula above and calculate the Binding energy, with that be OK?
Thanks
Regards

[martin.schlipf]

In general, your approach is correct, but you need to be careful with the specifics. E.g., the systems might converge differently w.r.t. the energy cutoff, so it is not sufficient to just use one value for both. Other considerations are the system size convergence for the combined system, lattice mismatch at the interface, rearrangements (e.g. interdiffusion), ... Keep in mind to converge w.r.t. to the target quantity (binding energy).

As for a way to combine the calculations. I don't see how this would be easily possible. Typically it doesn't matter anyway, because the time to calculate the combined system dominates the one needed for individual systems due to the much larger unit cell.

[SKM]

Thank you Martin, for the reply. I have some more queries based on your reply and otherwise as well.

1. From your reply

E.g., the systems might converge differently w.r.t. the energy cutoff, so it is not sufficient to just use one value for both.

So, you mean even if i use same INCAR file, (i.e. same ENCUT value for these three steps) is not enough to estimate binding energy?

2. Then again the other considerations you mentioned

Other considerations are the system size convergence for the combined system, lattice mismatch at the interface, rearrangements (e.g. interdiffusion), ... Keep in mind to converge w.r.t. to the target quantity (binding energy).

what do you mean by the above
(i) system size convergence for the combined system,
(ii) Lattice mismatch-this one i am aware of and calculated for each hybrid system, and of course i tried to design the lowest possible mismatch
(iii) rearrangements, and
(iv) target quantity (what do you mean by converge w.r.t this BE???)

How do you take care and consider these parameters to assess Binding energy?

3. Additional Query is:

considering my original post and the Step-1, after this Step-1 (i.e. of A.B system), the CONTCAR is written. As per the VASP Manual (dated Oct 29, 2018) Section 5.8, after each ionic step CONTCAR is written. Assume that the last Ionic step took, say 35 electronic steps to converge to EDIFF and thereafter its written CONTCAR (assume that at this stage EDIFFG also fulfilled and thus reached 'required accuracy'). i also referred to Sec 7.6.2 of the manual, wherein its mentioned

Never take the energy obtained at the end of a relaxation run, if you allow for cell
shape relaxations (the final basis set might not be isotropic).

As i have NOT DONE volume and cell relaxation in my example, so this Sec 7.6.2 doesn't apply to my example.

Now, if i take this CONTCAR and re-run one static calculation with NSW=0, and IBRION= -1, Why it is taking more than 35 electronic steps to converge?

My point here is, the Ionic positions that were converged in optimisation step, were the ones just prior to achieve the electronic convergence and not the finally written CONTCAR, which was written after electronic convergence.

Is my above point is correct? if not, can anyone, correct me and also direct me to any specific part of manual or any reference.

sorry for long description.

Regards

[SKM]

Hi

Thanks for the reply. but i am not very clear about the suggestions you made.

Meanwhile i have another issue: (in continuation of this calculation)
consider my original query. Now the issue is :

after relaxing the system A.B.
i run the two parts static calculations (after separating the two parts after relaxation -i.e. POSCAR-A and POSCAR-B) and run static calculations.

the combined A.B system took 63 SCF steps
system Part-A, took 100 SCF steps
and System Part-B, is not converged even after 340 SCF steps.

used same INCAR files for all 3 steps with IBRION = -1 NSW=0 and added the NGX/Y/Z and NGXF/XY/XZ explicitly taken from previous OUTCAR file.

Can anyone help in suggestion?

Regards

[martin.schlipf]

So, you mean even if i use same INCAR file, (i.e. same ENCUT value for these three steps) is not enough to estimate binding energy?

The point is you need to converge it. Just choosing the same value is not enough.

Then again the other considerations you mentioned

Other considerations are the system size convergence for the combined system, lattice mismatch at the interface, rearrangements (e.g. interdiffusion), ... Keep in mind to converge w.r.t. to the target quantity (binding energy).

what do you mean by the above
(i) system size convergence for the combined system,
(ii) Lattice mismatch-this one i am aware of and calculated for each hybrid system, and of course i tried to design the lowest possible mismatch
(iii) rearrangements, and
(iv) target quantity (what do you mean by converge w.r.t this BE???)

In your case the target quantity is the BE. I just wanted to emphasize that it is not enough to converge, e.g., the band gap, and then expect that the same cutoffs will work for the BE as well. You should perform your convergence tests on the quantity you are interested in. (Though it might be helpful to conduct preliminary tests on simpler systems).
How do you take care and consider these parameters to assess binding energy? If you create an interface between two materials they can interdiffuse so it might not be as simple as just stacking them on top of each other. Here it depends a lot on the experimental setup that you want to compare to. Also you need to consider the size of the interface. Unless you know specifically how many layers of each material are used, you typically want to have thick enough layers to recover some bulk-like property in the center of the layers. Hence you need to increase the size until that is converged.

considering my original post and the Step-1, after this Step-1 (i.e. of A.B system), the CONTCAR is written. As per the VASP Manual (dated Oct 29, 2018) Section 5.8, after each ionic step CONTCAR is written. Assume that the last Ionic step took, say 35 electronic steps to converge to EDIFF and thereafter its written CONTCAR (assume that at this stage EDIFFG also fulfilled and thus reached 'required accuracy'). i also referred to Sec 7.6.2 of the manual, wherein its mentioned

Never take the energy obtained at the end of a relaxation run, if you allow for cell
shape relaxations (the final basis set might not be isotropic).

As i have NOT DONE volume and cell relaxation in my example, so this Sec 7.6.2 doesn't apply to my example.

Now, if i take this CONTCAR and re-run one static calculation with NSW=0, and IBRION= -1, Why it is taking more than 35 electronic steps to converge?

Did you reuse the WAVECAR from the previous calculation? If not then Vasp needs to reoptimize the wave functions.
Also note that the Vasp manual is no longer maintained. All information within should have been moved to our wiki.

My point here is, the Ionic positions that were converged in optimisation step, were the ones just prior to achieve the electronic convergence and not the finally written CONTCAR, which was written after electronic convergence.

There are two optimizations, the electronic one and the ionic one. The WAVECAR contains the optimized wave functions after the electronic minimization, the CONTCAR the optimized ion positions after the ionic relaxation.

[SKM]

Thank you for the instant reply.

before i seek more clarification on some points on your reply, i found that WAVECAR is not written even if i use LWAVE = .TRUE. in INCAR, after electronic steps are finished but the final accuracy is not achieved. Thats why i could not use ISTART=1, in continuation of optimisation.

I think (if i am right) once the previous electronic convergence happened within NELM, the WAVECAR file is used for start of next ionic step, and the previous WAVECAR is erased.

If this is correct, can you suggest what is needed to save WAVECAR file, so the i can continue the calculation with ISTRAT=1 and having WAVECAR file.

Regards

[martin.schlipf]

The WAVECAR file should be written if the code exits normally including cases where the target accuracy was not reached.

[SKM]

Thanks Martin for the reply.

i faced another issue now and the WAVECAR file if written could have saved the time to continuation run.

I have a structure having 552 atoms and submitted for geometry optimisation.
Gamma centred KPOINTS is [4 4 1]
Used IALGO=38, ISMEAR= -5, LWAVE = .TRUE.
NBANDS works out to be 1248
the issue is, even one ionic step not finished in 15 hours, and 94 electronic steps done and there is no WAVECAR file written as its not completed even a single ionic step.

Is there a way to get WAVECAR file quickly at least once with different settings, and then use the WAVECAR for re-run with suitable and improved settings/tags in INCAR?

Regards

[martin.schlipf]

You can always just run a smaller number of steps NELM. Just keep in mind not to do relaxation steps, if you electronic structure is not converged.

[SKM]

Dear Martin,

sorry i did not get what you mentioned

Just keep in mind not to do relaxation steps, if you electronic structure is not converged.

1. What do you mean 'not to do relaxation'? do you mean NSW=0?
2. As within NELM itself the electronic steps not converged to the EDIFF value, so it means not converged. So, no WAVECAR is written.

Regards

[martin.schlipf]

Yes, if your calculation is not converged the forces are not accurate so a relaxation step may lead you in a completely wrong direction. So first find a setting how to obtain the electronic ground state and then switch on relaxation by setting NSW or IBRION.

The WAVECAR should be written even if EDIFF is not reached. If you find that it is not, either you have LWAVE = .false. in your INCAR file, or your time runs out before the writing finishes, or the file would be so big that the OS doesn't allow its creation.