Can cluster molecule models with vacuum on a,b and c sides be run with ISIF=2?

[SKM]

Hi
I have an arrangement of organic molecules (2 or more numbers) arranged in a specific manner (pi--pi stack, Z-arrangement, cross-linked) with some spacing between molecules to start with. As I want them to be like clusters of molecules, I introduce a vacuum in all 3 directions of the crystal model. A sample model is attached. In such cases, I usually optimise the models with ISIF=2, so that the cell volume will not be changed or minimised during the optimisation, to retain the cluster mode. At the same time, I have a doubt that if i don't give ISIF=3, the atomic movement may restricted, compared with that of ISIF=2 case.
Is my concern has a point? if so, how do i optimise such models with minimum tags in INCAR?

Regards

[andreas.singraber]

Hello!

With ISIF = 2 you are doing the right thing, you just need to make sure that there is enough vacuum such that interactions with periodic images do not occur (or at least are at an acceptable level). Also, with ISIF = 2 there is no restriction to the atoms' movement, every atom can move in any direction. Actually, there is no need to compute the stress tensor at all, so ISIF = 0 would be the best choice because it saves some computation time.

In contrast, if you would choose ISIF = 3 it is unclear how the additional cell degrees of freedom will propagate. Assuming there is enough vacuum around the isolated molecules the stress tensor should be zero. Then, either the cell parameters stay static or due to rounding errors start to move in random directions. Since there is no restoring force (the stress tensor is still zero) this movement can continue (and may even accelerate) until numbers overflow or the cell comes close to the molecules. Both cases are not desirable, the latter even alters your simulation setup because the molecules will not be isolated any more. Hence, avoid ISIF = 3 and other options freeing cell degrees of freedom for isolated molecules.

All the best,
Andreas Singraber

[SKM]

Thank you for the clear reply. it helps my understanding further.

Regards

[SKM]

Thank you.
But i got a further doubt, about ISIF=0, as you suggested.

but VASP Wiki says, this setting for Molecular Dynamics. But my case is purely the optimisation of geometry.

Regards

[andreas.singraber]

Hello!

The arguments I presented apply to geometry optimization as well. Let's quickly outline all ISIF options we discussed:

• ISIF = 0: The stress tensor will not be computed, all atoms can move in any direction, cell shape and volume is not modified.
• ISIF = 2: The stress tensor will be computed, all atoms can move in any direction, cell shape and volume is not modified.
• ISIF = 3: The stress tensor will be computed, all atoms can move in any direction, cell shape and volume can be modified.

As you can see neither ISIF = 0 nor ISIF = 2 will result in any changes to the cell shape and volume. Hence, it is up to you whether you want to compute the stress tensor or not. I suggested ISIF = 0 to save some time but most likely there will not be a noticeable difference. Maybe it is better to keep it at ISIF = 2 so you can check the stress tensor.

Anyway, also in geometry optimization it make no sense for molecules in vacuum to free the cell degrees of freedom with ISIF = 3. Simply, because the only "optimal" cell for a molecule in vacuum would be an infinite cell. However, we cannot do this in VASP so we have to find a compromise between two extremes:

• a very large cell with lots of vacuum: this is closer to the "perfect" molecule in vacuum but comes at the cost of more plane waves which slows down the calculation considerably.
• a smaller cell, where molecules start to "see" their periodic images: this may be faster but what we simulate is then not an isolated molecule any more.

You will have to find an acceptable amount of vacuum for your particular system. However, finding this compromise is not something that can be achieved with ISIF = 3. Instead, we you will need to use ISIF = 2 and adapt the cell manually in the POSCAR file.

All the best,
Andreas Singraber

[SKM]

Perfect explanation, for this.
Thank you.

Regards