Running the same molecule using different quantum chemistry (QM) programs does not produce the same results. This post summarizes the crucial point that computational chemists should consider before running any QM calculation.
It can be said that all QM programs use different algorithms and different default settings (check out their manual), such as the SCF algorithm, numerical integration algorithm, number of points in space for energy, and geometry calculations (grid).
First, you have to check the geometry specification (initial structure) in the input files to ensure that the programs are dealing with the same molecule. The symbol of the element, number, and coordinate of the atoms must be correct. Others are the total charge of the molecule, spin multiplicity, and basis set. The programs must use the same unit for geometry. Most QM programs use Angstroms, whereas the others use atomic units (A.U.) by default.
Caution must be taken in setting basis sets too. There are two types of basis functions (basis sets) for d orbitals that are used in QM programs: the spherical d-function (5d function) and the Cartesian d-function (6d function). Using different d-functions yields different basis numbers. Some program uses spherical d-function as the default, but others use Cartesian d-function. You have to check the number of basis functions!
Then you should also check whether the computed energies are the same. Go to the tail of the output file, where you can find the total and nuclear repulsion energies. The same molecule must have the same nuclear repulsion energies. If the nuclear repulsion energies are different, you have to check the clue above again.
Another thing is the default setting in each program, such as the cut-off level for the energy and gradient in self-consistent field (SCF) calculations. You should (have to) set the cut-off or threshold of different programs with the same criteria.
Moreover, for density functional theory (DFT) calculation, you should know how many grids (in the default setting) the program uses. When you increase the number of grids, this will help you to reach the minima/global point correctly. However, the calculation will also take a longer time. Also, the recipe of exchange-correlation (XC) functional that each program uses is different. This also gives the difference in energy. You have to check out what kind of exchange and correlation programs employ.
Rangsiman Ketkaew