Is this peak real?

[chinaye]

Dear all,
We calculated the complex dielectric function of GaAs at the BSE theory level. K-points are 11*11*11.We found that there is a peak located at 1 eV (scissor energies were used), as shown in the figure. However, the experimental results and theoretical resuts from (PRB 62 4927) do not show this peak. Will this peak disappear by incrasing K-point? Thank you.

[Conor Hogan]

Dear Jianfei,
I think it's very possible that the peaks are arising from the k-point sampling. How many points have you in the IBZ and in the full BZ? Do you also see the peak in the RPA spectrum? By increasing/decreasing the mesh size you should see a shift in these peak positions if they are just k-mesh related. Note that to converge well the spectrum of a bulk semiconductor you need many thousands of points if you are just doing a simple sum over k-points. There is no easy answer: you just have to do the convergence tests, I'm afraid.

[chinaye]

Dear Jianfei,
I think it's very possible that the peaks are arising from the k-point sampling. How many points have you in the IBZ and in the full BZ? Do you also see the peak in the RPA spectrum? By increasing/decreasing the mesh size you should see a shift in these peak positions if they are just k-mesh related. Note that to converge well the spectrum of a bulk semiconductor you need many thousands of points if you are just doing a simple sum over k-points. There is no easy answer: you just have to do the convergence tests, I'm afraid.

Dear Conor Hogan,
Thank you for your reply. Your suggestion that the spectrum is not converged is possible. We used 146 and 11*11*11 points in the the IBZ and in the full BZ, respectively. The position of the peak will change when different k-point shift is used. We now increase the K points to test the convergence of our results. However,on the other hand the peak is somewhat reasonable. When we use the k-point mesh directly involving the gama point, after using the scissor operator the position of the peak indictes that it is derived from the electronic transition from the top of the valence band to the bottom of the conduction band at the gama point. GaAs is a direct-gap semiconductor, so at the gama point the top of the valence band and the bottom of the conduction band both have the feature dE/dk=0. The excitaion is very likely to happen at this kind of point.