Good day everyone,
I have collected few numeric data from Gaussian output, but may I know the error value for Gaussian data? Is that standard or I should calculate it depend on my data?
It depends what do you mean with error.
If you run same calculation once again with same sofware the difference in final energies will be somewhere in 6-8th digit after the coma, its just "noise" and play no role in any practical use.
Theoretically, running same calculation with same method, basis set, integration grid, cutoffs, etc but with another software SHOULD give you same result (with noise difference). In reality its not often the case and thats why to compare TOTAL (absolute) energies from different software is almost always very bad idea. Relative energies however are normally nicely comparable, already when you have same method and basis set (same integration grid will make difference even smaller, but its not an easy task to make integration grid absolutely identical. In my experience relative energies should not differ more than 0.1-0.5 kcal/mol.
Finally the difference of obtained results from higher-level methods or from experimental values may be huge and depends on used method and system of interest. Usage of inappropriate methods may give you qualitatively wrong pictures, hundreds of kcal/mol is not the limit.
If the question is "how many digits should be used" - in calculation of relative energies - use all what Gaussian gives to you. If you want to show absolute energies (in a.u.) in supplementary - normally 5 digits after the coma should be enough. Relative energies in paper - depends on method, basis, task, molecule. For DFT reaction barriers or energy of conformers, strian and so on I would put 1, maximum 2 digits after the coma in kcal/mol, also following option is possible: 123.4(5) kcal/mol. You should keep in mind that "chemical accuracy" is around 1 kcal/mol, so providing many digits after the coma is not making your result more correct. Its the same case like writing that melting point of Hafnium is 2233,314159 C.
Error of what? What are you calculating? Energies, frequencies, optimized coordinates, thermodynamic data, etc?
Dmitry Sharapa
A detailed, comprehensive, and nice explanation; I was enjoyed!
Saeed
- Badges
- Science topic