I have two cases where,
(1). if we do structure optimization (ligand adsorption on a 2D-surface) without vdW's correction and followed by single point calculation when vdW's correction is activated will give me ground energy that include vdW correction.
(2). In the start activate vdW's correction and let system fully relax. In this case ground energy is not very different but final geometry is.
I calculated these two cases using SIESTA and included GRIMM's correction, I am wondering to know which configuration is more accurate ? In first case, it easily converged and time saving than case. (2) which i think is more realistic.
Yes, they do. One thing you can do is to relax with out dispersion. Then reoptimize with dispersion.
Your second way is correct and the structure from this calculation will be more accurate and reliable. As Bharat Medasani said, first you can fully relax the system without vdw dispersion corrections and then again do full relaxation calculation with dispersion correction. In method 1, you do single point calculation with dispersion; this is not correct because the system is not relaxed.
As Karthikeyen mentioned, the best way to deal with the situation is to optimized the system without vdW correction. Once the optimization is over, take the relaxed geometry and reoptimized it using vdW correction. Its a computationally cost effective approach. But, your first option is not acceptable due to lack of accuracy. However, if you have some experimental evidences for that particular system, in some cases option 1 can give reliable outcomes.
Greatly thankful for quick and helpful answer.
Can you please comment on the difference in final geometries in both cases, if i'm not wrong dispersion correction if just numerical addition of something ? How the geometry changes ?
The change of geometry depends on the chemistries, but to my understanding, Grimme's corrections is empirical technique that modifies the energies and hence the energy landscape. So the ground state configuration changes with/without the dispersion correction. I forgot the exact details about the dispersion correction to comment in a more detailed way. You can publish bot geometries so people can pick what they want.
vdW correction is an empirical term. Without going to the detail in theory, the simple thing is that, if you pick the geometry from your first run and put it for next run with vdW correction your calculations will be less expensive. As you have already optimized the geometry once, that significantly reduce the computational cost.
If you're simply using the Grimme correction to account for the vdW interactions it will add an insignificant amount of CPU-time compared to your overall DFT calculation. So I do not understand why you would like to firstly optimize the geometry without the vdW correction. This might even give you a geometry that is further away from the "true" geometry than your starting geometry (if you're good at guessing starting geometries). Unless you want to compare the geometries with and without vdW interactions, perform the geometry optimization with vdW interactions from the very start. In general, this will require less geometry optimization steps than if you first did the optimization without vdW, thereby saving you some time..
performing Grimme correction from the very start is very expensive way to perform simulation. i think single point calculation of optimized geometry will add the the dispersion correction to the final energy which is an effective way to perform simulation. need guidance if i am wrong
Shabeer Ahmad Mian In my experience correction should be added from the start as I have observed two different optimized geometries with and without correction. And scientific feeling is, later geometry is closer to the true geometry.
- Badges
- Science topic
- Similar topics
- Filing