| Re: q: charges? basis sets/potentials validation? ( No.1 )|
- Date: 2006/11/15 20:37
- Name: Vasilii Artyukhov
- "2) is the calculation with OpenMX "size consistent"? I.e., can one calculate reagent A, then reagent B, then its product AB as clusters, separately, and take E(AB)-E(A)-E(B) as the reaction energy?"
To get good accuracy of bond energies, you must ensure that you are using exactly the same bases, pseudopotentials, unit cell size, and the integration grid (scf.energycutoff). Since you're trying to use automated determination of the cell size, you get different cells (and hence, different grids as well) for A, B, and AB. Therefore, you should manually set the same cell size for all three calculations. This will lead to some error cancellation and improve your results.
Still, your calculations will suffer from the so-called basis set superposition error. Since you are using different basis sets for the particles (even if these consist of same functions but centered at different points), the error cancellation when subtracting energies is not that precise. To correct this, you should either use dummy atoms to ensure that you have the same bases for all calculations, or use other types of basis sets such as plane waves (which are not available in OpenMX or other O(N) codes). Of course, the same holds for Gaussian bases.
| Re: q: charges? basis sets/potentials validation? ( No.2 )|
- Date: 2006/11/18 22:59
- Name: Grigory
- Thank you!