Questions on increase accuracy of Bader charges

[gaoxuew]

I am trying to get the Bader charges of atoms in a Cu(110) surface. The structure of the slab is in the attached 'surface.png'. c axis is perpendicular to the surfaces of the slab. After Bader charge analysis using the outputs from VASP calcualtions, I got the Bader charges on the atoms as seen in the attached 'bader.png'. My question are: 1) How to increase the accuracy of the Bader charges? For example, the bader charges of atoms in the same layer of the slab should be very close. But in my case, the distribution of the Bader charges on each layer is relatively larege.
2) If this is just numerical errors, is there any suggestions on setting VASP input flags to improve the results?

I really appreciate for your suggestions.

[graeme]

Try using the "-b weight" flag for higher accuracy. This method, developed in the Trinkle group, divides cells on the boundary between atoms with a weight based upon linear interpolation.

Increase the grid density also improves accuracy for all methods.