Vasilii, Thanks for your reply. Yes, I did turn the MD.UseSaveCG flag on. Something is not working with the CG algorithm. Will try (second order?) Broyden's method instead. Andrey
________________________________ From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta on behalf of Vasilii Artyukhov Sent: Tue 10/16/2007 4:07 AM To: [email protected] Subject: Re: [SIESTA-L] Lattice relaxation and band structure calculations for In2O3 1) Did you turn the MD.UseSaveCG flag on? Also, you might want to try MD.TypeOfRun Broyden, in my experience this is really superior to CG. 1) ??? 2007/10/16, Semichaevsky, Andrey < [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]> >: Dear Siesta users, I have two questions related to the calculation of the electronic structure of In2O3 in Siesta: 1) The optimization of the atomic positions using the MD/CG method results in very long runs for 80 atoms using the DZP basis, and the wallclock time limit is exhausted. The CG after a restart seems to take the last updated DM results, but, for some reason, it i s not able to reuse all the optimization-related parameters. So, after each restart, the CG iterations start essentially from the the new initial point every time. This makes it very difficult to achieve an unambiguous convergence to just one (global) minimum. Is there any way for the CG to reuse its most recent sets of parameters. 2) The number of bands in the valence is 192, for spin-unpolarized calculations, irrespective of the numerical orbital basis used (SZ or DZP). Does this number have any physical explanation? - I know that other codes, i.e., FLAPW produce 92 valence bands instead. Thanks. Regrads, Andrey Semichaevsky
