<div dir="ltr"><br>Dear Users<br>I have an issue related to the convergence of the static dielectric  matrix using the PH code. I have performed calculations<br>on bulk silicon using different k point meshes and I have obtained the following results:<br>
<br>k grid              diagonal component of      number of k points in<br>                         the dielectric tensor           the irreducible Brillouin zone<br><br>4*4*4                   23.668350065                        8<br>
 6 *6*6                16.297485614                       16<br>8 *8*8                  14.044830694                       29<br>10 *10*10             13.288531964                     47<br>12*12*12              13.029602882                     72<br>
16 *16*16             12.908820645                    145<br>20 *20*20             12.894538380                    256<br><br>k grid+ 1 1 1 shift              diagonal component of      number of k points in<br>
                                            the dielectric tensor           the irreducible Brillouin zone<br><br>4*4*4                                      13.840844632                           10<br>6*6*6                                     12.997009732                           28<br>
8*8*8                                     12.903849607                           60<br>10*10*10                               12.893711596                         110<br>12*12*12                               12.892685597                         182<br>
16*16*16                               12.892482798                         408<br>20*20*20                               12.892537346                         770<br> <br>I was surprised of the improvement in the convergence due to the shift of the grid. I don't think this is related to the number of k points <br>
in the IBZ (at least not<span onclick="dr4sdgryt(event)"> exclusively</span>).<br>I have observed a similar behavior in diamond. <br>The ground state energy convergence also benefits from  the shift, but the improvement is not so striking.<br>
Does someone has any hint on why the shift of the grid improves the calculation of the dielectric properties of silicon?<br>Thanks a lot<br>Dario Rocca, dept. of chemistry, UC Davis <br><br></div>