[Compaq]

I am slightly confused about how I should explain that one method performs differently when two basis sets of different size are used. For example, consider calculating the potential energy surface for diargon at different internuclear separations. Using B3LYP/aug-cc-pVDZ and /aug-cc-pVTZ, I get pretty much the exact same results. How can I explain this properly?

On an intuitive level, I want to say that the double zeta basis set lets the wave function vary sufficiently to reach the accuracy limit. The wave function converges toward the same energies at all distances for both basis sets, so apparently the double zeta basis is sufficiently large for the given system.

But if someone said "why" to my explanations, I would not know what to say. I feel I lack the insight. How can i obtain insight? How should I study computational chemistry most efficiently? I want to learn and be good at this, but I am not sure how.

This question became some sort of a ramble, but I appreciate any guidance if you have any.

[Irlanur]

I don't think there is too much to understand here. Basis sets and exchange-correlation functionals are a bloody mess. Of course sometimes there are things which make total sense, e.g. if you need polarisable basis sets or dispersion corrected functionals, but in general, this is trial and error (at the moment).

[Compaq]

If the method was MP2 instead of DFT, would the situation be different then? That is, more "understanding" available?

[pm133]

I am slightly confused about how I should explain that one method performs differently when two basis sets of different size are used. For example, consider calculating the potential energy surface for diargon at different internuclear separations. Using B3LYP/aug-cc-pVDZ and /aug-cc-pVTZ, I get pretty much the exact same results. How can I explain this properly?

On an intuitive level, I want to say that the double zeta basis set lets the wave function vary sufficiently to reach the accuracy limit. The wave function converges toward the same energies at all distances for both basis sets, so apparently the double zeta basis is sufficiently large for the given system.

But if someone said "why" to my explanations, I would not know what to say. I feel I lack the insight. How can i obtain insight? How should I study computational chemistry most efficiently? I want to learn and be good at this, but I am not sure how.

This question became some sort of a ramble, but I appreciate any guidance if you have any.

I don't think there is too much to understand here. Basis sets and exchange-correlation functionals are a bloody mess. Of course sometimes there are things which make total sense, e.g. if you need polarisable basis sets or dispersion corrected functionals, but in general, this is trial and error (at the moment).

If the method was MP2 instead of DFT, would the situation be different then? That is, more "understanding" available?

Let's start with the second quote above. There is a HUGE amount to understand here. The basis set aug-cc-pvdz has 4 seperate parts to it. The "aug", the "cc", the "p" and the "vdz". In order to explain what is going on and why you are seeing such quick convergence you need to be able to understand what these 4 parts are doing. Start taking parts out and repeat the process. Try various combinations as well. This will tell you what the model you are using predicts the important parts of the interaction between the atoms to be. You will almost certainly find that at least some of your permutations will result in a slower convergence. That will be telling you something.

I notice you are not using dispersion corrections. You should include some. You will also need to consider basis set superposition error corrections. These two are the bare minimum if you are trying to calculate reasonable results with B3LYP where there are weak interactions involved. B3LYP has these two well known problems. They counter balance each other in some cases allowing "spectacular" results. For this reason B3LYP became mainstream. However it has become clear that these two effects don't cancel each other out and in those circumstances your results can be useless.

Finally, you may well see differences with MP2 and to understand that, you need to understand the difference between what DFT and MP2 are doing. It's important to bear in mind that MP2 is not variational.

I hope I've shown that in contradiction to the second quote there is a HUGE amount to be understood before you can answer your question.

Good luck and welcome to the game of computational chemistry. It's great fun but realising it's not a "draw and click" technique (not even with DFT) is usually a shock.

[Irlanur]

I hope I've shown that in contradiction to the second quote there is a HUGE amount to be understood before you can answer your question.

We obviously have a very different meaning of "understanding".

I am out of here.