In this section, keywords for the all electron calculation are explained. These keywords discussed here are important for all calculations including the generation of pseudopotentials and pseudoatomic basis functions, since both the generations of pseudopotentials and pseudo-atomic orbitals are based on the all electron calculation. The list of keywords and some comment for the all electron calculation are as follows:
Choose GGA, LDA, or LDA-VWN
Give the total number of electrons. It is also possible to give the number of electrons corresponding to not only a neutral atom, but also a positive or negative charged atom.
Set grid.xmin ((a.u.)=exp(grid.xmin)), where is the minimum radius from which radial differential equations are solved toward a distant. An appropriate value for grid.xmin is -7.0 from H to Kr, and -10.0 for heavier atoms.
Set grid.xmin ((a.u.)=exp(grid.xmax)), where is the maximum radius from which radial differential equations are solved toward the origin. An appropriate value for grid.xmin is 2.5 to 4.0, but could depend on whether there are delocalized states or not.
Set the number of grids to solve radial differential equations. A larger number of grids gives a higher degree of accuracy, while the computational time increases. An appropriate value for grid.num is 3000 to 12000. For heavier atoms, the use of a larger number of grids is better to achieve a reliable calculation.
It is possible to change the number of grids for in output files by the keyword, grid.num.output, although the actual calculation is performed using grid.num. Usually, we use 500 for it.
Set the maximum number of SCF iteration.
Choose a method for charge mixing. Either simple, GR-Pulay, or Pulay is available. In most cases, the simple mixing scheme is enough to achieve a sufficient convergence.
Set the minimum mixing weight.
Set the maximum mixing weight.
Set previous SCF steps for charge mixing in the GR-Pulay or Pulay method.
Set a SCF iteration number from which the GR-Pulay or Pulay starts.
Set scf.criterion. At least 1.0e-10 for the keyword should be chosen for a convergent calculation.