Re: [SIESTA-L] Crazy SCF with vanadium
-Original Message- From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED] Sent: 6. november 2008 17:59 To: SIESTA-L@listserv.uam.es Subject: Re: [SIESTA-L] Crazy SCF with vanadium Deal All, I know tried to replace the V atom with a C instead. After many CG steps, the same problem in the SCF run suddenly occured. Can it be an issue of compilation? -Joachim Dear Joachim, let's be systematic: 1) can you please confirm that the problem is not due to what I suggested to check several days ago - the maximal spin set up by default on ALL atoms? -Yes, that is not the case 2) does the electronic structure (either with V, or without) - position of carbon 2s and 2p bands, charges - remains reasonable until the same problem in the SCF run suddenly occured? Or are you having weird results all the way through and only notice it when they start to diverge? Sorry... -No, I have normal charges and forces before it goes wrong. 3) what happens afer many CG steps it terms of structure; in which respect does the crashed structure differ from the good one? You may have the .ANI file stored, or just .XV files from the beginning and from immediately before the crash. If it happens suddenly then some crazy forces calculated on atoms might be responsible, whatever the reason for this. -I don't see anything unusual with the CG move of the structure. The crazy forces are only, of course, a result of the non-converged SCF. For the prior, well-converging CG runs, the forces are fine. Thank you for your time. Best, Joachim
Re: [SIESTA-L] Crazy SCF with vanadium
Deal All, I know tried to replace the V atom with a C instead. After many CG steps, the same problem in the SCF run suddenly occured. Can it be an issue of compilation? -Joachim -Original Message- From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta [mailto:[EMAIL PROTECTED] On Behalf Of Joachim Fürst Sent: 5. november 2008 11:06 To: SIESTA-L@listserv.uam.es Subject: Re: [SIESTA-L] Crazy SCF with vanadium Dear All, The V3p are included in the PS I used (it should be anyway!!). So, yes, I agree, it is best to have it in there. I also agree that the SCF shouldn't behave like that with it being in there or not. -Original Message- From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED] Sent: 5. november 2008 10:48 To: SIESTA-L@listserv.uam.es Subject: Re: [SIESTA-L] Crazy SCF with vanadium Quoting Joachim Fürst [EMAIL PROTECTED]: Dear all, I followed the discussion about the vanadium deposited on a graphene sheet. My experience with vanadium is that it is necessary to insert semi-core states the 3p with the atom program. Yes I fully agree - V3p are important in order to obtain accurate results, but I don't think their absence was responsible for the crazy SCF behaviour, if (even not so good) V potential was working in other systems. By the way is it scheduled in SIESTA to include soon or later ultrasoft pseudo? Note that in Siesta the softness of the pseudopotential does not matter to such extent as it does in planewave codes (where this becomes expensive in terms of basis size). In Siesta, even deep pseudos can be (with a bit of luck) treated with appropriate basis sets. Hard cases which demand high mesh cutoffs are not a priori those where the pseudo is hard. Andrei Postnikov
Re: [SIESTA-L] Crazy SCF with vanadium
Deal All, I know tried to replace the V atom with a C instead. After many CG steps, the same problem in the SCF run suddenly occured. Can it be an issue of compilation? -Joachim Dear Joachim, let's be systematic: 1) can you please confirm that the problem is not due to what I suggested to check several days ago - the maximal spin set up by default on ALL atoms? 2) does the electronic structure (either with V, or without) - position of carbon 2s and 2p bands, charges - remains reasonable until the same problem in the SCF run suddenly occured? Or are you having weird results all the way through and only notice it when they start to diverge? Sorry... 3) what happens afer many CG steps it terms of structure; in which respect does the crashed structure differ from the good one? You may have the .ANI file stored, or just .XV files from the beginning and from immediately before the crash. If it happens suddenly then some crazy forces calculated on atoms might be responsible, whatever the reason for this. Good luck Andrei
Re: [SIESTA-L] Crazy SCF with vanadium
Quoting Joachim Fürst [EMAIL PROTECTED]: Dear all, I followed the discussion about the vanadium deposited on a graphene sheet. My experience with vanadium is that it is necessary to insert semi-core states the 3p with the atom program. Yes I fully agree - V3p are important in order to obtain accurate results, but I don't think their absence was responsible for the crazy SCF behaviour, if (even not so good) V potential was working in other systems. By the way is it scheduled in SIESTA to include soon or later ultrasoft pseudo? Note that in Siesta the softness of the pseudopotential does not matter to such extent as it does in planewave codes (where this becomes expensive in terms of basis size). In Siesta, even deep pseudos can be (with a bit of luck) treated with appropriate basis sets. Hard cases which demand high mesh cutoffs are not a priori those where the pseudo is hard. Andrei Postnikov
Re: [SIESTA-L] Crazy SCF with vanadium
Dear All, The V3p are included in the PS I used (it should be anyway!!). So, yes, I agree, it is best to have it in there. I also agree that the SCF shouldn't behave like that with it being in there or not. -Original Message- From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED] Sent: 5. november 2008 10:48 To: SIESTA-L@listserv.uam.es Subject: Re: [SIESTA-L] Crazy SCF with vanadium Quoting Joachim Fürst [EMAIL PROTECTED]: Dear all, I followed the discussion about the vanadium deposited on a graphene sheet. My experience with vanadium is that it is necessary to insert semi-core states the 3p with the atom program. Yes I fully agree - V3p are important in order to obtain accurate results, but I don't think their absence was responsible for the crazy SCF behaviour, if (even not so good) V potential was working in other systems. By the way is it scheduled in SIESTA to include soon or later ultrasoft pseudo? Note that in Siesta the softness of the pseudopotential does not matter to such extent as it does in planewave codes (where this becomes expensive in terms of basis size). In Siesta, even deep pseudos can be (with a bit of luck) treated with appropriate basis sets. Hard cases which demand high mesh cutoffs are not a priori those where the pseudo is hard. Andrei Postnikov
Re: [SIESTA-L] Crazy SCF with vanadium
631 atoms? whew... Thee could be a few things that might help. One thing is to set an electronic temperature higher than the default (about 25 meV). Even though you say you have raised the temperature (which I assume to be the electronic temperature), how high have you raised? Maybe you'll have to set it to a really high value for your calculation to converge initially, after which you'll be able to lower it gradually, in steps of about 10 meV and re-relaxing everything again. Another thing that helps convergence is that V atoms are in a good initial position. My personal experience is that when the initial positions of the atoms are way off the equilibrium ones, scf can be problematic, especially if you have d electrons (in my case I have Ni slabs). Supposing that V atoms would be adsorbed over the graphene surface, you could try doing an initial relaxation with the Harris functional - which would overcome scf convergence problems in a first step. It will most probably not give very reliable results, but it could help in determining an initial set of positions which could be not too off from equilibrium. Make a test with a V atom over a 32-atom graphene sheet to compare the outcomes of full scf relaxation and a Harris functional calculation; if it gives decent results, you could try doing this to your larger sheet. Finally, I would use a larger MeshCutoff. It might help to improve SCF convergence. Try at least 220-250 Ry. Best of luck, Marcos Vous avez écrit / You have written / Lei ha scritto / Você escreveu... Joachim Fürst Dear all, I am doing a relaxation with vanadium on graphene, and the scf loop goes crazy: siesta: iscfEharris(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,% = IterSCF1 882.442 95.03 elaps: Routine,Calls,Wall,% = IterSCF1 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 elementsNo 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 0.00 0.000196 1 39.049818 0.00 0.000213 1 36.210059 -0.01 0.000237 1 0.709653 0.00 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
Re: [SIESTA-L] Crazy SCF with vanadium
Hi, Thank you for your advices. I have also cut the system down to only a 100 atoms or so, without any improvement. I have a pretty good idea about V position, what is less known to me is the distortion of the sheet. I have done runs with 220 Ry mesh, but no change. I will try very high temperature as a final try. Thansk again. Joachim -Original Message- From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta [mailto:[EMAIL PROTECTED] On Behalf Of Marcos Verissimo Alves Sent: 4. november 2008 17:42 To: SIESTA-L@listserv.uam.es Subject: Re: [SIESTA-L] Crazy SCF with vanadium 631 atoms? whew... Thee could be a few things that might help. One thing is to set an electronic temperature higher than the default (about 25 meV). Even though you say you have raised the temperature (which I assume to be the electronic temperature), how high have you raised? Maybe you'll have to set it to a really high value for your calculation to converge initially, after which you'll be able to lower it gradually, in steps of about 10 meV and re-relaxing everything again. Another thing that helps convergence is that V atoms are in a good initial position. My personal experience is that when the initial positions of the atoms are way off the equilibrium ones, scf can be problematic, especially if you have d electrons (in my case I have Ni slabs). Supposing that V atoms would be adsorbed over the graphene surface, you could try doing an initial relaxation with the Harris functional - which would overcome scf convergence problems in a first step. It will most probably not give very reliable results, but it could help in determining an initial set of positions which could be not too off from equilibrium. Make a test with a V atom over a 32-atom graphene sheet to compare the outcomes of full scf relaxation and a Harris functional calculation; if it gives decent results, you could try doing this to your larger sheet. Finally, I would use a larger MeshCutoff. It might help to improve SCF convergence. Try at least 220-250 Ry. Best of luck, Marcos Vous avez écrit / You have written / Lei ha scritto / Você escreveu... Joachim Fürst Dear all, I am doing a relaxation with vanadium on graphene, and the scf loop goes crazy: siesta: iscfEharris(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,% = IterSCF1 882.442 95.03 elaps: Routine,Calls,Wall,% = IterSCF1 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 elementsNo 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 0.00 0.000196 1 39.049818 0.00 0.000213 1 36.210059 -0.01 0.000237 1 0.709653 0.00 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
Re: [SIESTA-L] Crazy SCF with vanadium
Could you please post your V pseudo? The mesh cutoff doesn't vary the scf, it only controls the rippling of energy/forces. This rippling is due to the FFT aliasing of those magnitudes (orbs^2,neutral atom potential, core charge for non linear core corrections) which are projected into the grid. As the aliasing is constant (ie during the scf the atoms don't move with respect to the grid) the mesh cutoff shouldn't play any role. Best regards, Eduardo On 04/11/2008, at 18:05, Joachim Fürst wrote: Hi, Thank you for your advices. I have also cut the system down to only a 100 atoms or so, without any improvement. I have a pretty good idea about V position, what is less known to me is the distortion of the sheet. I have done runs with 220 Ry mesh, but no change. I will try very high temperature as a final try. Thansk again. Joachim -Original Message- From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta [mailto:[EMAIL PROTECTED] On Behalf Of Marcos Verissimo Alves Sent: 4. november 2008 17:42 To: SIESTA-L@listserv.uam.es Subject: Re: [SIESTA-L] Crazy SCF with vanadium 631 atoms? whew... Thee could be a few things that might help. One thing is to set an electronic temperature higher than the default (about 25 meV). Even though you say you have raised the temperature (which I assume to be the electronic temperature), how high have you raised? Maybe you'll have to set it to a really high value for your calculation to converge initially, after which you'll be able to lower it gradually, in steps of about 10 meV and re-relaxing everything again. Another thing that helps convergence is that V atoms are in a good initial position. My personal experience is that when the initial positions of the atoms are way off the equilibrium ones, scf can be problematic, especially if you have d electrons (in my case I have Ni slabs). Supposing that V atoms would be adsorbed over the graphene surface, you could try doing an initial relaxation with the Harris functional - which would overcome scf convergence problems in a first step. It will most probably not give very reliable results, but it could help in determining an initial set of positions which could be not too off from equilibrium. Make a test with a V atom over a 32-atom graphene sheet to compare the outcomes of full scf relaxation and a Harris functional calculation; if it gives decent results, you could try doing this to your larger sheet. Finally, I would use a larger MeshCutoff. It might help to improve SCF convergence. Try at least 220-250 Ry. Best of luck, Marcos Vous avez écrit / You have written / Lei ha scritto / Você escreveu... Joachim Fürst Dear all, I am doing a relaxation with vanadium on graphene, and the scf loop goes crazy: siesta: iscfEharris(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,% = IterSCF1 882.442 95.03 elaps: Routine,Calls,Wall,% = IterSCF1 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 elementsNo 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 0.00 0.000196 1 39.049818 0.00 0.000213 1 36.210059 -0.01 0.000237 1 0.709653 0.00 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
Re: [SIESTA-L] Crazy SCF with vanadium
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: iscfEharris(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,% = IterSCF1 882.442 95.03 elaps: Routine,Calls,Wall,% = IterSCF1 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 elementsNo 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 0.00 0.000196 1 39.049818 0.00 0.000213 1 36.210059 -0.01 0.000237 1 0.709653 0.00 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