urgent! how dose ATK deal with charged molecule in molecule_configuration

[hzkvictory]

When an extra electron is added to molecule, the molecule get charged.
In Siesta, when extra electron is added, the default cell is filled with jellium-like opposite charge for compensation, so the total box is neutral. In order to remove the interaction of charged default cells, a Madelung correction term is applied to the energy to make it converge much faster with cell size.
( for molecules or atoms, this is done only if the cell is SC, FCC or BCC).
However, in the forum I haven't found how ATK remove the effect of extra compensate electron on repeated default cells in calculation ,
I don't know whether ATK use the same method as Siesta ?
In addition, how is the default cell for molecule in molecule_configuration like, especially when molecule is being charged (like SC)?

urgent, please reply soon as possible.

[kstokbro]

In ATK we use the correct procedure for charged systems, which for molecules  is to apply multipole boundary conditions.
I.e. for a molecule systems we calculate all the multipoles of the charge density and this sets up the boundary conditions of the cell.
This, requires that you select MultigridSolver for the Poisson solver and select MultipoleBoundaryConditions, which is the default for molecules.

[hzkvictory]

Thank you very much, sir.
I am very interested in the related theory, could you tell me the references of this method ?

[kstokbro]

Check:
http://www.quantumwise.com/documents/manuals/latest/ReferenceManual/index.html/chap.appendix.html#sect1.appendix.poisson

The multipole expansion to solve poissons equation is a standard used in many QuantumChemistry code which use a real space poisson solver. It seems that  the teory is so basic that I could not find any references to it.

The basic idea is the following. You have a real space poisson solver, for such a solver you need to specify boundary conditons.
For a molecule, you can expand the long range electro-static potential in charge multipoles around some center. In this way we ensure that our solutions has the correct assymptotic behaviour. The accuracy of the method depends on how many moments you include (we use up to l=2) and that the box contains all the charge density.

Thus, the method does not need any compensation charge, and is therefore very accurate, you may check that we get very good values for ionization energies with ATK, see forinstance
http://www.quantumwise.com/documents/tutorials/latest/BenzeneSET/index.html/chap.isolated.html

The only drawback of this method is that the real space poisson solver is slower than the fft method. However, we are working on improving the speed of our real space poisson solver.

[GJK]

Hi,

I have seen through the link http://www.quantumwise.com/documents/tutorials/latest/BenzeneSET/index.html/chap.isolated.html given in this post for calculating the Electron affinity and Ionization energy of benzene molecule using Total energy concept for Molecular configuration of Benzene.

My question is 
Can we use the same calculation procedure for calculating the Electron affinity and Ionization energy of graphene which is optimized in bulk configuration.

Note:As already discussed by Dr.Anders that the total energy of bulk and molecule configuration will not differ so much if enough large vacuum is provided at the bulk configuration.I have verified his disussion in my studies and did not find any appreciable change in total energy.

Regards
GJK

[kstokbro]

In this case you cannot use the multipole boundary conditions, but you should use periodic in the 2-d plane and dirichlet perpendicular to the plane and include sufficient vacuum.
Since graphene is infinite the unit cell representing it will not be charged when you add an electron and the ionization and affinities are given directly from the bandstructure of the neutral system, i.e. the top of the valence band and bottom of conduction band. Graphene is a semi-metal i.e. no bandgap.