Hi, Emilio and Alex Thank you for your kind reply.
I have understood your ideas. It only improves the accurate of the results by increasing the variable MeshCutoff. There is a specific ground state energy within a certain MeshCutoff (May be the difference of these energies is not very small, like 0.2~0.3 eV), and the absolute values of the ground state energy is not very important. But if I compare the energy difference of two configurations of the system (like calculating the binding energy of an adatom or the formation of the point defect) without the same calculation conditions (such as different MeshCutoffs), it will bring about 0.2~0.3 eV errors artificially. What about your opinions? Any comment or suggestion is appreciated. Regards Fei Mao > -----原始邮件----- > 发件人: "Oleksandr Voznyy" <[email protected]> > 发送时间: 2012年11月22日 星期四 > 收件人: [email protected] > 抄送: > 主题: Re: [SIESTA-L] Any comments on the MeshCutoff > > > The way to converge it is just by making it larger and larger until > the property you are interested in is converged within the tolerance you > need. > > I want to comment that Etot vs Ecutoff is converging very slowly, and > looking at that curve one might want to go as high as 800Ry. > On practice people usually say that mesh should be good enough to avoid > the eggbox effect. > Anyhow, using %block GridCellSampling is a good alternative to > increasing the mesh cutoff. > Alex.
