> To all respected siesta users, > Thank you very much for helping me. > I have used the usual way (not supercell) for hexagonal system of mine, > using the method mostly discussed here in mailing list. In the .out file > of variable cell, I have "Naive supercell factors: 8 8 5".
OK, so you don't need any supercell calculation, after all. "Naive supercell factors" are about internal kitchen of Siesta, how to make address indices of arrays in a case of periodic system. If I am not mistaken, it depends more on lattice parameters and the extension of basis functions than on anything else. > The result I have got is that the "a" parameter is about 2% less than the > experimental value and "c" parameter is about 3% more than the > experimental value. Not so bad. From here on, to tune different parameters in order to arrive at experimental values is not a recommended way to go, because you can arrive at good values for a wrong reason. Rather, you should check the stability of your results against enhancing the cutoffs, increasing basis etc. > Reading the article, it came to my mind maybe I can change the "Naive > supercell factors" at the same time "superc: Internal auxiliary supercell" > in order to change the interactions in a way to have the parameters change > in the same manner (I mean they all decrease or increase). > Therefore, I changed some parameters and the final result was "Naive > supercell factors: 6 6 4", I could not get to "6 6 2". The result > was that I got "c" 0.5% more than experiment, but "a" changed only > slightly from the previous result. Probably the changes you have result from the fact that you changed the basis functions somehow, namely making them more extended. This can of course change your result, notably their accuracy. But just looking at the numbers 664 vs. 662 does not really give you any clue... Best regards Andrei
