Hello Murat,
The additional functions improve the calculation, in general trends. As higher basis, for example 6-31++G(d,p), more reliable should be the results. First of all is take into account the magnitude of you system and the time you want to spend in each calculus. For this you must take into account:
-Double or triple zeta (6-31G or 6-311G)
-Which kind of system, metals, group P compounds, organic systems.
-Which kind of interactions, are you studying a reaction for example, there are two fragments in the same calculus? If it is yes, then you must take into account BSSE corrections.
-With the fragments there are two options, use the Counterpoise correction in gaussian if you use small basis sets or use as large basis set as you can (think that decreases the error). For example, with 6-31++G(d,p) you should use BSSE correction but with 6-311++G(2df,2p) the error should be enough small.
In summary, the increase of basis set is a relation between ACCURACY-TIME COST. Even so, this is a "qualitative" and general explication. The experience will made to select directly the best basis set for each system or THE BEST CASE, I recommend you read the book: Cramer J. C., Essentials of Computational Chemistry . In this kind of literature there are a section dedicated to explain in detail the basis sets.
I hope it helps you
Hello Murat,
The additional functions improve the calculation, in general trends. As higher basis, for example 6-31++G(d,p), more reliable should be the results. First of all is take into account the magnitude of you system and the time you want to spend in each calculus. For this you must take into account:
-Double or triple zeta (6-31G or 6-311G)
-Which kind of system, metals, group P compounds, organic systems.
-Which kind of interactions, are you studying a reaction for example, there are two fragments in the same calculus? If it is yes, then you must take into account BSSE corrections.
-With the fragments there are two options, use the Counterpoise correction in gaussian if you use small basis sets or use as large basis set as you can (think that decreases the error). For example, with 6-31++G(d,p) you should use BSSE correction but with 6-311++G(2df,2p) the error should be enough small.
In summary, the increase of basis set is a relation between ACCURACY-TIME COST. Even so, this is a "qualitative" and general explication. The experience will made to select directly the best basis set for each system or THE BEST CASE, I recommend you read the book: Cramer J. C., Essentials of Computational Chemistry . In this kind of literature there are a section dedicated to explain in detail the basis sets.
I hope it helps you
Joaquim Mª Rius Bartra's answer is good.
According to your choice, it should be B3LYP/6-31G(d,p)
Hi Joaquim Mª Rius Bartra, that was a great answer. I do have a question though with regards to the basis set and BSSE. I've never been able to get Gaussian to use an implicit solvent model and counterpoise at the same time. Why does this not seem to be possible? Therefore, to reduce the error as much as possible without using counterpoise I used a basis set of 6-311++G(2df,2p).
Hello Anthony,
About run Counterpoise and implicit solvent I have never tried. Maybe the reason is technical, it is possible that Counterpoise is not prepared to add salvation.
About the basis set to reduce the error I think that it is enough. Even so, it is good always to validate your calculus and see that really, the error is small. For example, you can run the Counterpoise with this basis set or try the next or before level of this basis and compare. Then you will be able to see the magnitude of the BSSE error and if it is enough small.
Thanks for that information. I think you might be right, thinking back I do believe you get an error if the SCRF and counterpoise keywords are used at the same time. I've just never been sure why this was the case.
Hi Murat
Adding polarization and diffuse function to the basis set enhances you result and it depends on your model compound.
Although what Mr. Joaquim Mª Rius Bartra stated above I think that'll answer your most of the questions quite right.
Hello Murat;
The additional functions as higher basis (6-31+G* / 6-311+G*) will constitute a good compromise between accuracy and computational costs.
I'd like to add that typically you need diffuse functions when you have a system where you will face long-range interactions, i.e. hydrogen bonds, ligand fields, or, if you compute electronically excited states, when you're dealing with Rydberg states. In the short-range domain at typical chemical bond lengths you do not need to add diffuse functions as they will not yield higher accuracy in this case. As you are working with solvent environment you will definitely need long-range interactions, hence diffuse functions are something you should apply for accuracy of your results.
Polarization functions serve another purpose. With these you can model the correct shape of your atomic orbitals which then become molecular orbitals within the typically applied LCAO approach. As an example, you can get SP3 hybrid orbitals. The larger your basis set gets, the less important it is to add polarization functions as the larger basis sets typically contain higher angular momentum wave functions. As long as your not going to apply quadruple zeta basis sets or even larger ones, polarization functions are a very good way to increase accuracy and, at the same time, save computer time.
Finally, I'd like to recommend a trick to avoid the counterpoise correction. Simply run a calculation with the fragments being seperated by several Angstroms (at least 10 A, better more). Check if there is still some overlap of the basis functions between the fragments. It is important to apply the same basis set, method and environment as in the original computation. This yields proper relative energies of the fragments without the need for a counterpoise correction.
Calcualation of basis-set-sensitive properties as electric (hyper)polarizability require the use of prepared, property oriented basis sets. Polarization and diffuse GTF play a major here. You will find a gew pertinent remarks in some textbooks. Follow closely related papers for more useful instructions.
Good luck in your efforts
GM
Dear All,
Above comments have answered the question, however, I would also like to add few more points. See, there is no any general rule to add polarisation and diffuse function. Rather, it is the compromisation between the accuracy and the computational cost. Even though, its better to use as large basis set coupled with diffuse and polarisation functions.
Polarisation function provide sufficient mathematical flexibility to adequately describe the wave function (allow basis functions not to be centered on atoms). This flexibility is almost always added in the form of basis functions corresponding to one quantum number of higher angular momentum than the valence orbitals. E.g., Adding d functions to the nitrogen basis set causes HF theory to predict correctly a pyramidal minimum for ammonia molecule.
Regarding Diffuse function, it is eseential for anions, highly excited electronic states, and loose supermolecular complexes. When a basis set does not have the flexibility necessary to allow a weakly bound electron to localize far from the remaining density, significant errors in energies and other molecular properties can occur. For combating that, one needs to use dissuse functions.
I hope it will help you to find answer to your question
Dear Murat,
Inclusion of polarization and diffused functions depends on the property one is aiming at and the system one is dealing with.
Hello:
Inclusion of polarization and diffuse functions depends on the electrical property, The additional of higher higher basis sets like 6-31++G(d,p), more reliable should be the results.
- Badges
- Science topic