Calculating vibrational frequencies of a charged defect

[John Buckeridge]

Hello,
I am attempting to calculate the vibrational frequencies of a Si-N defect in GaAs. The defect can be in a neutral or charged state. To do the calculation with (for example) the defect in the positive state, I first relax the structure with the NELECT tag in the INCAR file set to a value of one less than that of the nuetral state. I then determine the dynamical matrix by moving the N and Si atoms by a small amount in x,y,z, running VASP for one iteration and determine the force derivatives.
However, when I do the force calculations with the NELECT tag set the same as for the relaxation calculation after the fifth electronic iteration I get the following error message in the output file: BRMIX: very serious problems
the old and the new charge density differ
old charge density: 567.00000 new 566.00000
The calculation then continues on fine and gives me the forces. Is this a serious error or can I trust the resulting forces?

[alex]

Hi,

I think the message is just because charged unit cells and solid state is unphysical. In a normal calculation (e.g. no charges in the cell) the calc. would stop if it experiences a large difference in charge density.

I'd check if the total energy behaves as expected. Then I'd check the DOS, if the additional electron is at its place. Well, does it differ very much from the neutral defect?
And what does experiment say ...?

Cheers

alex