SCF convergence problem when using dipole correction

[li bingxin]

Dear CP2K experts,

Recently I am doing a single-point energy calculation of Mg (0001) slab with numerous explicit water molecules. The structure of the system is obtained from a certain step of a long-time MD calculation (I have attached one structure here). The basic idea of this calculation is to obtain the Hartree potential of the whole system and calculate the work function of the slab. 

The explicit water molecules are placed on one side of the slab while the other side of the slab is vacuum. Therefore this structure is asymmetric and the dipole correction should be applied. 

However, once the dipole correction was applied to the system the SCF cycle cannot get converged (the mixing alpha value is smaller than 0.1) although the calculation without the correction can converge very well (The input file, output file without dipole correction and output file with dipole correction are attached in this email).

From the above electrostatic potential figure we can find that the potential in the vacuum region is not that flat due to the net dipole on the surface. According to the reference provided by the CP2K manual. This surface net dipole is cancelled via adding one internal-origin electric field/potential, which is proportional to the surface-dipole density of the surface.

 

From the above equation, the dipole-correction related potential only depends on the surface dipole density (I guess it is a constant for a certain system) and the cell size along the direction which is perpendicular to the surface. The magnitude of this potential tends to contribute to the total energy and might cause some instability of the SCF. I tried to vary the cell size of the system but it is useless for the convergence problem. I also tried some general method for a better convergence, increasing E_cutoff, decreasing mixing alpha while they cannot work in this case.

Another idea is to set the position of the dipole correction manually (the default is setting the position at the electron density minimum). This is not available for me because the CP2K version I use is 6.1, which does not support setting the dipole correction position manually. 

I was wondering that could anyone has some ideas about this SCF convergence problem stems from the dipole correction can help me on this? I am looking forward to any suggestion from you!

Best Regards,

Bingxin