Hi Fei The mesh cutoff controls the accuracy in the calculation of some integrals in real space. It is no a variational parameter in the sense that more accurate does not mean lower energy. 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.
Emilio On 21 Nov 2012, at 13:19, 毛飞 wrote: > Hi, all > > I have question about the variable MeshCutoff used in Siesta. I know if I > increase the MeshCutoff higher, the results will be more accurate. > > I am doing a geometry optimization of a given system, sometimes the total > energy increase a little (about 0.2 ~ 0.3 eV) when I increase the MeshCutoff, > and the optimization is well converged (0.01 eV/Ang). So, in that situation, > How could I choose a reasonable value for the total energy of the system? And > How could I choose the optimized geometry (the one with lower energy or the > one with higher MeshCutoff, which are all well converged)? > > Any comment or suggestion is appreciated. > Best > > Fei Mao > > > -- Emilio Artacho CIC nanoGUNE Consolider, and Cavendish Laboratory, University of Cambridge Tolosa Hiribidea 76, E-20018 Donostia - San Sebastián, Spain, e.arta...@nanogune.eu, +34 943 574039, http://theory.nanogune.eu
