ABINIT Discussion Forums

I realized that for the finite field Berry phase calculation that number of k-points is increased by the Berry phase routine when a finite field is applied. For example, in the response tutorial tffield_6.in, the WFK of AlAs is calculated, and a zero field Berry phase is calculated using 28 k-points. For subsequent cycles with a finite field, the old WFK is read in an a new k-point mesh is instituted leading to a total of 91 k-points. Presumably this is to make the Berry phase path integral more tractable. My question is then -- is it possible to parallelize the iterative process of increasing the external electric field in a Berry phase calculation? This would seem to be even more critical for optimization of the ion positions under the influence of an electric field. Does anyone have an input file they would be willing to share to give me a working example to study from? Is k-point (or other) parallelization even possible?


Thanks,
Paul Fons

The increase in k points is coming from the fact that the presence of the external field lowers the symmetry of the system from its original, field free symmetry. The code is parallelized over k points, automatically. Just run on more processors, but not with paral_kgb (I am not sure the code is fully debugged for parallelization of bands and fft planes, I will have to try some tests).