Hi Maxim, So as I understand it now, you are running a big supercell with gamma-point. This cell is generated from a converged unit cell (or from a smaller supercell), for which the sampling is converged, and the atomic positions are also optimized down to 0.04 eV/Ang. Is that right?
Could you provide more detail on the procedure? As an old saying goes, "the devil lives in the details" :) Anyway, for very large systems (especially if they are metallic), convergence can be very difficult, indeed. So, a Mixing weight of 0.01 could still be too high - you might need to go down to 1e-3 or even lower, for the beginning of the convergence loop. Also, one thing that helped me when I started running really big systems was to increase the number of pulay steps for the mixing. Cheers, Marcos On Sat, Oct 22, 2011 at 4:55 PM, Maxim Peskov <[email protected]> wrote: > Hej Henrik, > > the main problem here, as I perceive it, is the huge forces we are getting > at the beginning for some of the atoms, when we should expect a fine scf > converged density matrix from a previous geometry step that got be > approximated to a different kpoint set, of couse, however it should > definitely be better that a random initialization of bands we are getting > when starting from scratch. Here it is worse, in fact. > Have look at the output file where I allowed SCF to proceed and watch how > energy of the systems bounces at the first fews steps. For a bigger systems > this means going off the limits and the lack of convergence - in the > consequence. > (complete output in the attachment) > > --Max > > ------ > k-point displ. along 1 input, could be: 0.00 0.50 > k-point displ. along 2 input, could be: 0.00 0.50 > k-point displ. along 3 input, could be: 0.00 0.50 > Kpoints in: 24 . Kpoints trimmed: 23 > > siesta: k-grid: Number of k-points = 23 > siesta: k-grid: Cutoff (effective) = 9.620 Ang > siesta: k-grid: Supercell and displacements > siesta: k-grid: 4 0 0 0.000 > siesta: k-grid: 0 2 0 0.000 > siesta: k-grid: 0 0 4 0.000 > Naive supercell factors: 4 2 4 > > superc: Internal auxiliary supercell: 4 x 2 x 4 = 32 > superc: Number of atoms, orbitals, and projectors: 896 11648 14336 > > * Maximum dynamic memory allocated = 3 MB > > siesta: > ============================== > Begin CG move = 0 > ============================== > > superc: Internal auxiliary supercell: 4 x 2 x 4 = 32 > superc: Number of atoms, orbitals, and projectors: 896 11648 14336 > > outcell: Unit cell vectors (Ang): > 5.440000 0.000000 0.000000 > 0.000000 13.580000 0.000000 > 0.000000 0.000000 4.810000 > > outcell: Cell vector modules (Ang) : 5.440000 13.580000 4.810000 > outcell: Cell angles (23,13,12) (deg): 90.0000 90.0000 90.0000 > outcell: Cell volume (Ang**3) : 355.3397 > New_DM. Step: 1 > Initializing Density Matrix... > > iodm: Reading Density Matrix from files > Read DM has different structure. Fixing... > > InitMesh: MESH = 54 x 144 x 48 = 373248 > InitMesh: Mesh cutoff (required, used) = 250.000 272.327 Ry > > * Maximum dynamic memory allocated = 63 MB > > stepf: Fermi-Dirac step function > > siesta: Program's energy decomposition (eV): > siesta: Ebs = -1911.686053 > siesta: Eions = 9474.841514 > siesta: Ena = 1832.919192 > siesta: Ekin = 4294.216968 > siesta: Enl = -625.333921 > siesta: DEna = -109.947618 > siesta: DUscf = 109.266231 > siesta: DUext = 0.000000 > siesta: Exc = -1627.882645 > siesta: eta*DQ = 0.000000 > siesta: Emadel = 0.000000 > siesta: Emeta = 0.000000 > siesta: Emolmec = 0.000000 > siesta: Ekinion = 0.000000 > siesta: Eharris = -6201.785840 > siesta: Etot = -5601.603306 > siesta: FreeEng = -5601.603306 > > siesta: iscf Eharris(eV) E_KS(eV) FreeEng(eV) dDmax Ef(eV) > siesta: 1 -6201.7858 -5601.6033 -5601.6033 5.0150 -2.9154 > timer: Routine,Calls,Time,% = IterSCF 1 124.835 91.50 > elaps: Routine,Calls,Wall,% = IterSCF 1 15.650 91.46 > siesta: 2 -7030.2491 -5539.8587 -5539.8587 11.3076 -7.1814 > siesta: 3 -6044.0557 -5572.8033 -5572.8033 2.8268 -3.3773 > siesta: 4 -5956.0974 -5595.5166 -5595.5166 2.7117 -2.2985 > siesta: 5 -5825.4095 -5604.4640 -5604.4640 1.1959 -3.6451 > siesta: 6 -5819.4801 -5637.6924 -5637.6924 0.9876 -4.4630 > siesta: 7 -5818.9743 -5723.0864 -5723.0864 0.5204 -3.2452 > siesta: 8 -5818.2690 -5828.9569 -5828.9569 0.2823 -4.0173 > siesta: 9 -5814.9075 -5799.5175 -5799.5175 0.1043 -5.6074 > siesta: 10 -5814.1989 -5807.7746 -5807.7746 0.0567 -5.3206 > siesta: 11 -5814.1604 -5813.2696 -5813.2696 0.0150 -5.2951 > siesta: 12 -5814.1466 -5813.9381 -5813.9381 0.0064 -5.2495 > siesta: 13 -5814.1455 -5814.4979 -5814.4979 0.0042 -5.2577 > siesta: 14 -5814.1452 -5814.2148 -5814.2148 0.0010 -5.2628 > siesta: 15 -5814.1451 -5814.1600 -5814.1600 0.0005 -5.2654 > siesta: 16 -5814.1451 -5814.1357 -5814.1357 0.0003 -5.2649 > siesta: 17 -5814.1451 -5814.1402 -5814.1402 0.0001 -5.2643 > siesta: 18 -5814.1451 -5814.1432 -5814.1432 0.0001 -5.2643 > > siesta: E_KS(eV) = -5814.1462 > > siesta: E_KS - E_eggbox = -5814.1462 > > siesta: Atomic forces (eV/Ang): > ---------------------------------------- > Tot 0.000001 0.000001 0.020157 > ---------------------------------------- > Max 1.053845 > Res 0.528280 sqrt( Sum f_i^2 / 3N ) > ---------------------------------------- > Max 1.053845 constrained > > Stress-tensor-Voigt (kbar): -62.52 -213.03 -83.88 0.00 -0.00 -0.00 > (Free)E + p*V (eV/cell) -5787.5740 > Target enthalpy (eV/cell) -5814.1462 > > -------------- > > 22 октября 2011, 17:52 от Henrik Löfås <[email protected]>: > > Dear Maxim, > > Another point, you set > MaxSCFIterations 0 > I am not sure but I think this means that you just get your DM interpolated > at the new k-points. If you would let the scf-loop reach convergence for the > new k-point set I would guess you get more reasonable forces. > > Regards > Henrik > > 2011/10/22 Marcos Veríssimo Alves > <[email protected]<http://sentmsg?compose&[email protected]> > > > >> Maxim, >> >> I believe Chun has a point here, and I beg to differ from what you say. >> >> For some materials, it is important to include special k-points, so as to >> get a correct band structure and (eventually) forces - the most "famous" >> case would perhaps be graphene, with its K point. Even if your material >> hasn't such special characteristics, making the k-point mesh denser means >> that you are sampling the BZ more accurately and thus changing the part of >> the total energy that corresponds to the electronic band occupations. Just >> think of a simple numerical integration in 1-D. >> >> To say that you have a problem with an increasing k-point mesh density is >> a bit precipitated, unless you are sure that you have a very converged >> k-mesh. Chun's e-mail is actually very sensible: you should first relax your >> structure, and then increase the k-mesh to see its effects on the forces.Not >> only that, you should also check its effects on the band structure, although >> in principle this effect should be negligible, if you have performed a >> convergence study. >> >> Marcos >> >> >> On Sat, Oct 22, 2011 at 6:34 AM, Maxim Peskov >> <[email protected]<http://sentmsg?compose&[email protected]> >> > wrote: >> >>> Thanks for the comment, Chun. However, you have missed the point. >>> I guess it will be better if I rephrase my question: how can one restart >>> from scf-preconverged density matrix and increased kpoint set? With the >>> options I used such a restart would mean the forces are going crazy which >>> clearly indicate a problem. Even more, if I change kpoints it is better to >>> start from scratch than to use DM. Any ideas? >>> >>> --Max >>> >>> >>> 20 октября 2011, 20:04 от >>> [email protected]<http://sentmsg?compose&[email protected]> >>> : >>> > Hi, >>> > >>> > From the first ourput file, I found the following lines: >>> > >>> > outcoor: Final (unrelaxed) atomic coordinates (fractional): >>> > -0.00005348 0.49985936 0.49676159 1 1 C >>> > ... >>> > ... >>> > >>> > which means your first run is not converged. >>> > >>> > Best, >>> > >>> > Chun >>> > >>> >> >> > >
