Dear Chiu Fang Yu: The criterion of convergence varies from case to case. It depends on which property you are after. In the test I suggest, this is the minimum in the Etot(z-step) curve. You want to get your chain step with a certain accuracy. Each set of parameters will give you a different curve. Enhance the parameters till the minimum of the curve (also its curvature) gets stable. The absolute value of energies in each curve is not important, it may drift with parameters.
Best regards Andrei [email protected] > sorry, it is > 'change the kpoint until the energy/atom (eV) change is less than 0.01eV > ?' > > -- > Best Regards > > 邱芳瑜 Chiu Fang Yu > 國立成功大學 材料科學與工程學系碩二 > MOBILE：0930287221(中華) > GMAIL：[email protected] <[email protected]> > > > > 2014-12-17 14:55 GMT+08:00 joyce79928cc . <[email protected]>: >> >> I was wondering how to decide whether it is converge? >> change the kpoint until the energy(eV) change is less than 0.01eV or...? >> Thanks! >> >> -- >> Best Regards >> >> 邱芳瑜 Chiu Fang Yu >> 國立成功大學 材料科學與工程學系碩二 >> MOBILE：0930287221(中華) >> GMAIL：[email protected] <[email protected]> >> >> >> >> 2014-12-17 14:51 GMT+08:00 joyce79928cc . <[email protected]>: >>> >>> Thanks for reply. I will try to fix it! >>> >>> -- >>> Best Regards >>> >>> 邱芳瑜 Chiu Fang Yu >>> 國立成功大學 材料科學與工程學系碩二 >>> MOBILE：0930287221(中華) >>> GMAIL：[email protected] <[email protected]> >>> >>> >>> >>> 2014-12-17 14:38 GMT+08:00 <[email protected]>: >>>> >>>> Dear joyce79928: >>>> >>>> just some ideas. >>>> Your results may be noise due to insufficient basis and/or >>>> insufficient k-mesh ans/or insufficient MeshCutoff >>>> (probably all three). >>>> Does your DOS have any resemblance with what is expected for a 1-dim. >>>> chain? >>>> Does your stress definition in the GeometryConstrains block >>>> serve any purpose? >>>> There is a number of works with Siesta on Au chains around, >>>> maybe scroll through them first. >>>> Anyway, you can systematically proceed in the following way: >>>> >>>> Take just one atom along the chain step, not three >>>> (make the third lattice vector shorter :-). >>>> Fix it. >>>> For the k-mesh, you need just 1 point in the plane but more than 4 >>>> along Z. >>>> Run a sequence of calculations just varying the 3d lattice vector >>>> (chain step) - single point, no relaxation. >>>> Trace the total energy curve as function of chain step. >>>> This will give you an approximation to "true Siesta" result. >>>> Study how it depends on basis, k-mesh and MeshCutoff. >>>> Find "good" values for all three. >>>> Make the usual "egg box" test. >>>> Then hopefully your forces will be OK so that >>>> you can add more atoms to your unit, release them, let them relax etc. >>>> >>>> Good luck >>>> >>>> Andrei >>>> >>>> [email protected] >>>> >>>> >>>> > Dear: >>>> > >>>> > I want to relax the chain with Siesta. >>>> > but different atom position with same parameter will get different >>>> > structure. >>>> > for example, I set the distance between chain to 2.93 A and get a >>>> final >>>> > structure of 2.569 A between chain. >>>> > Then I change initial chain distance to 2.88 A and get a final >>>> structure >>>> > of >>>> > 2.758 A between chain. >>>> > >>>> > why this happened? >>>> > >>>> > Thanks for reading my mail and it is welcome to discuss with me. >>>> > >>>> > below's are my fdf input file. >>>> > >>>> > SystemName bulk_au >>>> > SystemLabel bulk_au >>>> > >>>> > ================================================== >>>> > ================================================== >>>> > # SPECIES AND BASIS >>>> > >>>> > # Number of species >>>> > NumberOfSpecies 1 >>>> > %block ChemicalSpeciesLabel >>>> > 1 79 Au >>>> > %endblock ChemicalSpeciesLabel >>>> > >>>> > PAO.BasisSize SZP >>>> > PAO.EnergyShift 0.04 Ry >>>> > >>>> > >>>> > # K-points >>>> > >>>> > %block kgrid_Monkhorst_Pack >>>> > 4 0 0 0.0 >>>> > 0 4 0 0.0 >>>> > 0 0 4 0.5 >>>> > %endblock kgrid_Monkhorst_Pack >>>> > >>>> > >>>> > # UNIT CELL AND ATOMIC POSITIONS >>>> > >>>> > # UNIT CELL >>>> > LatticeConstant 1.0 Ang >>>> > %block LatticeVectors >>>> > 8.81553253 0.00000000 0.00000000 >>>> > 4.40776625 7.63447409 0.00000000 >>>> > 0.00000000 0.00000000 8.79434817 >>>> > %endblock LatticeVectors >>>> > >>>> > # Atomic coordinates >>>> > NumberOfAtoms 3 >>>> > AtomicCoordinatesFormat ScaledCartesian >>>> > %block AtomicCoordinatesAndAtomicSpecies >>>> > 4.407766232 4.241374236 0.00000000 1 >>>> > 4.407766232 4.241374236 2.93144939 1 >>>> > 4.407766232 4.241374236 5.86289878 1 >>>> > %endblock AtomicCoordinatesAndAtomicSpecies >>>> > >>>> > ================================================== >>>> > ================================================== >>>> > # General variables >>>> > >>>> > ElectronicTemperature 300 K >>>> > MeshCutoff 200. Ry >>>> > xc.functional LDA # Exchange-correlation >>>> functional >>>> > xc.authors CA >>>> > SpinPolarized .false. >>>> > SolutionMethod Diagon >>>> > >>>> > >>>> > # SCF variables >>>> > >>>> > DM.MixSCF1 T >>>> > MaxSCFIterations 300 # Maximum number of SCF iter >>>> > DM.MixingWeight 0.03 # 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 5 >>>> > #Diag.ParallelOverK yes >>>> > >>>> > # MD variables >>>> > >>>> > MD.FinalTimeStep 1 >>>> > MD.TypeOfRun CG >>>> > MD.NumCGsteps 50 >>>> > MD.VariableCell T >>>> > UseSaveData T >>>> > >>>> > # Output variables >>>> > >>>> > WriteMullikenPop 1 >>>> > WriteBands .false. >>>> > SaveRho .false. >>>> > SaveDeltaRho .false. >>>> > SaveHS .false. >>>> > SaveElectrostaticPotential True >>>> > SaveTotalPotential no >>>> > WriteCoorXmol .true. >>>> > WriteMDXmol .true. >>>> > WriteMDhistory .false. >>>> > WriteEigenvalues yes >>>> > >>>> > %block GeometryConstraints >>>> > stress 4 5 6 >>>> > position 1 >>>> > %endblock GeometryConstraints >>>> > >>>> > >>>> > >>>> > -- >>>> > Best Regards >>>> > >>>> > 邱芳瑜 Chiu Fang Yu >>>> > 國立成功大學 材料科學與工程學系碩二 >>>> > MOBILE：0930287221(中華) >>>> > GMAIL：[email protected] <[email protected]> >>>> > >>>> >>>> >>>> >>>> >>>> >
