Hallo, some questions: in those cases when as youy say you got it converged, does everything else look reasonable? (vanadium is magnetic, graphene not, bands in due place, no crazy charge transfers etc.)
My wild guess is that, as you set spin polarization on, by default (unless you have InitSpin block whick I don't see in your fragment) SIESTA sets MAXIMAL spins on ALL atoms, including all your graphene, that is a very bad starting point. And if you switch spin out, your deposited vanadium is forced to be non-magnetic, which is probably not very stable either. Not that you have DM.UseSaveDM true in your example, which you would need to set to F in order to reset the spins. Once this is settled (if it is indeed the case, and you still have problems) take small mixing (much smaller than you have now), temperature of 300 - 600 K, to begin with. Then you must be able to force the system into convergence. Good luck Andrei Postnikov > Dear all, > > I am doing a relaxation with vanadium on graphene, and the scf loop goes > crazy: > > > siesta: iscf Eharris(eV) E_KS(eV) FreeEng(eV) dDmax > Ef(eV) > siesta: 1 -97094.38150 -99070.91635 -99070.91635 39.07032 > -4.23154 > timer: Routine,Calls,Time,% = IterSCF 1 882.442 95.03 > elaps: Routine,Calls,Wall,% = IterSCF 1 223.031 94.74 > siesta: 2 -97140.17446 -99051.35598 -99051.50902 34.72834 > -4.20753 > siesta: 3 -97226.76276 -98965.36619 -98965.50463 35.42815 > -4.12217 > siesta: 4 -94936.08243 -99371.74467 -99371.89250 35.10492 > -4.22078 > siesta: 5 -90678.74567 -99715.25553 -99715.47141 24.75070 > -4.07217 > siesta: 6 -86315.83616 -100050.89144 -100051.12562 36.83745 > -3.67364 > siesta: 7 -97702.64129 -98510.48955 -98510.724571051.40098 > -3.89029 > > > > -And it only gets worse after more iterations. > > I have tried almost all thinkable combinations of Pulay mixing, mixing > rate etc, but without success. I have also increased temperature, disabled > spin, changed energyshift, tried different basis sets for all > elements....No luck at all! Sometimes I can get it to converge, but then > after a CG step or two it goes nuts again. > If I leave out vanadium, the problem disappears. I have used this vanadium > PS many times in different calculations, but have never come across this > issue until now. > > Have you got any ideas? I have listed my .fdf file below (the long list of > coordinates is left out). > Hope you have some good advice. Thanks! > > > > NumberOfSpecies: 3 > NumberOfAtoms: 631 > > LatticeConstant 1.0 Ang > %block LatticeVectors > 55. 0. 0. > 0. 8. 0 > 0. 0. 36.84 > %endblock LatticeVectors > > %block ChemicalSpeciesLabel > 1 6 C_pbe > 2 1 H > 3 23 V > %endblock ChemicalSpeciesLabel > > > AtomicCoordinatesFormat Ang > %block AtomicCoordinatesAndAtomicSpecies > 40.470458 0.000000 0.000196 1 > 39.049818 0.000000 0.000213 1 > 36.210059 -0.000001 0.000237 1 > 0.709653 0.000000 0.000252 1 > ..... > > ..... > %endblock AtomicCoordinatesAndAtomicSpecies > > > %block kgrid_Monkhorst_Pack > 1 0 0 0.0 > 0 1 0 0.0 > 0 0 1 0.0 > %endblock kgrid_Monkhorst_Pack > > DM.UseSaveDM T > #SaveTotalPotential T > #SaveRho T > #SaveElectrostaticPotential T > #MD.UseSaveXV T > > SolutionMethod Diagon > > > %block PAO.BasisSizes > C_pbe DZ > H DZ > %endblock PAO.BasisSizes > > %block PAO.Basis > V 4 # Species label, number of l-shells > n=4 0 2 P 1 # n, l, Nzeta, Polarization, NzetaPol > > 7.497 6.783 > > 1.000 1.000 > > n=3 1 2 # n, l, Nzeta > > 3.048 1.943 > > 1.000 1.000 > > n=4 1 2 # n, l, Nzeta > > 9.870 8.285 > > 1.000 1.000 > > n=3 2 2 # n, l, Nzeta > > 4.901 2.758 > > 1.000 1.000 > > %endblock PAO.Basis > > > XC.functional GGA > LongOutput .true. > XC.authors PBE > SystemLabel bulk > > > > DM.MixSCF1 T > MaxSCFIterations 600 # Maximum number of SCF iter > DM.MixingWeight 0.01 # New DM amount for next SCF cycle > DM.Tolerance 1.d-4 # Tolerance in maximum difference > DM.UseSaveDM true # to use continuation files > DM.NumberPulay 4 > > SpinPolarized true > > WriteMullikenPop 1 > > MeshCutoff 175 Ry > > > > MD.TypeOfRun CG > MD.NumCGSteps 190 > MD.UseSaveCG true > MD.UseSaveXV true > MD.MaxCGDispl 0.02 Ang > MD.MaxForceTol 0.05 eV/Ang > WriteCoorXmol true > WriteMDXmol true > WriteMDhistory true >
