Dear Min Cheol Kim, in order to get a reasonable result on your (magnetic) system, you should of course run it spin-polarized, and not just as you say "even tried the spinpolarization" but this is the main (and almoet the only) thing that you really need. Forget playing around with the basis - it is important for getting numerically sensitive results (later on), but has almost no effect in getting things basically right. Forget the MeshCutoff either, unless you are already happy with setting up things in the electronic structure part and proceed with the structure relaxation. My suggestions: 1. Switch on spin polarization 2. Initialize maximal spins on both Cu atoms, but not on anything else. 3. Set the spins of both Cu in opposite 4. Swith on fixing the total spin, and fix it to zero 5. Take small enough mixing parameter- 0.01 or less I bet it will work, unless you have an error in the structure. Once it works, it will work with any basis which you can tune afterwards.
Good luck Andrei Postnikov > Hi all siesters,I'm performing a convergence test before geometry > optimization for a copper-ligand (chloro and methylpyridine ligands) > crystal-like structure with periodic boundary conditions.The unit cell > contains total of 62 atoms[2 copper atoms and 4 chloride atoms], in > GGA-PBE scheme,Cu and Cl are DZP, while N, C, H are all DZWhen executing > the simulation, the energies and the dDmax value didn't converged even on > thefirst CG run. So I played around with the paramters.First I applied the > electronic temperature of 0.1 eV, used different Mesh Cutoffs, different > Mixing Weights, Number pulays, and even tried the spinpolarization but yet > had no luck.I was kind of frustrated, but after looking at the tutorials > in the siesta page I realized I haven't changed the basis setSo I tuned > the PAO.EnergyShift Value to 0.30 Ry, and PAO.SplitNorm to 0.20, and > finally the systemconverged. But looking at the output file, the r_c > values seemed to be changed really dramatically, so I wantto find the > optimum value for the energyshift. Right now I'm trying several > simulations increasing the energyshift by 0.01 Ry,[0.21 Ry, 0.22 Ry,0.23 > Ry and so on] and leaving the SplitNorm value to 0.15The question is, is > there any systematic way to determine, or atleast predict the level of > optimum energyshift for the
