Dear Marcos, Thank you once again for the detailed description of the possible sources of errors in my calculations. Unfortunately, I cannot still overcome this problem even after implementation of your advice. I checked the coordinates and I found a mistake but it did not help, I increased number of k-point but the dipole moment is the same, its dependence on the meshcutoff level is very weird (increasing of meshcutoff leads to oscillations with rising amplitude like secular terms, there is no any convergence). I paid my attention to one more thing. I expected that the x,y,z-projections of the atomic forces should be equal to zero (the total sum of the atomic forces along the particular axis should be also equal to zero) but I was surprised that this was not so. Moreover, in case of infinite (along x-axis) armchair ribbon passivated by H-atoms the non-zero dipole moment (along y-axis) depends on the x-coordinates of the H-atoms!!!
Next thing is about the finite flake of the graphene layer (for example, zigzag stripe) that I was trying to study. Even in case of absolutely symmetrical coordinates of C-,H-atoms there is non-zero dipole moment but now along x- and y-directions. (I prepared system and put the length of lattice vectors long enough to prevent the overlapping of the adjacent flakes). And one more related thing, when Siesta iterates the atomic and orbital populations on every atoms I found that the distribution of charges over the symmetrical H-atoms on the edges of the infinite armchair ribbon H-atoms is not symmetrical. Is it related to dipole moment problem? Eventually, is it possible that all problems come from incorrect definition of unit cell (lattice vectors, lattice constant? I don't know what to come up with more than I have already done!!! My account of reasonable hypotheses is almost depleted. Thank you,Marcos, for your guidance. Best regards Artem Baskin, PhD student, University of Illinois at Chicago On Tue, March 16, 2010 10:03 am, Marcos Veríssimo Alves wrote: > About the coordinates, I haven't really checked them. If they are > incorrect then it would be the most obvious source of error. > > For the dipole moment, I don't see how the origin of coordinates and > k-point sampling could possibly be related. Surely you are right, the > dipole moment is independent of the origin of coordinates, but the > origin of coordinates and the k-point sampling are ultimately > completely un-related to each other. The issue with k-point sampling > is that it is used for performing a sum over the energies of occupied > bands in k-space, which accounts for the part of the total energy. If > this is wrong, then your total energy is wrong as well. If your total > energy is wrong, then by the Kohn-Sham theorems you have a wrong > charge density distribution. And, if you have a wrong charge density > distribution, then definitely your dipole moment would be wrong. > Fundamentally, this is it. > > Nevertheless, as it has been discussed rather recently in the list, > bear in mind that, in a solid, the dipole would be ill-defined, and to > get a meaningful value you'd need to resort to the Berry's phase > approximation. So, it would be advisable (1) to check your initial > coordinates just in case and (2) see if you get the dipole moment to > converge with an increasing number of k-points along the ribbon's > length. > > Cheers, > > Marcos > > > On Tue, Mar 16, 2010 at 12:22 PM, Roland Gillen > <[email protected]> wrote: >> Artem, >> >> I might be terribly wrong mistaken, but your geometry doesn't look like >> it's >> symmetric. Look at the y-coordinates of your H-atoms. They are about >> +5.8 >> and -5.8. >> Your left-most carbon atoms are at y=-4.9, but the carbon atoms at the >> other >> edge are at y=7.3. >> Btw. the x-positions of the hydrogen atoms don't seem reasonable as >> well. >> >> Maybe that is the reason for your wrong band structures? >> >> Cheers >> Roland >> >> [email protected] schrieb: >>> >>> Thank you, Marcos, for such a quick response. >>> >>> You are absolutely right about the low level of mesh cutoff that I >>> used, I >>> used it on purpose to cut down on the calculation time. But there is a >>> thing that I still don't understand in your explanations: how may any >>> k-point sampling be related with calculation of dipole moment of my >>> system? Dipole moment is invariant (with respect of the used system of >>> coordinates) for electrically neutral system, therefore it should not >>> depend on the lattice vector determination and, consequently, on any >>> k-point sampling. >>> Is it?. Moreover, band structure calculations and dipole moment are >>> separated things (if only they are not connected somehow through >>> plane-wave method). So, I can understand why the magnitude of non-zero >>> dipole moment may vary depending on the mesh cutoff value, but zero >>> should >>> be equal zero. >>> The question about dipole moment is important for me because I look for >>> the reasons which result in errors of my band structure calculation for >>> graphene ribbon. (Initially I thought that there was something wrong >>> with >>> my input file information or something like this). >>> >>> Thank you once again, hope you will help me with that weird thing. >>> >>> Artem Baskin, >>> PhD student, >>> University of Illinois at Chicago >>> >>> >>> >>> >>> >>> >>> >>>> >>>> Artem, >>>> >>>> Two things could be hindering your calculation. First, your mesh >>>> cutoff is quite low. I would use at least 200 Ry to get only decent >>>> results. In my experience with 300 Ry you get pretty good results for >>>> a graphitic system. Second, I don't see any k-point sampling along the >>>> ribbon length. This definitely could be the cause of your fluctuating >>>> dipole value. >>>> >>>> Cheers, >>>> >>>> Marcos >>>> >>>> On Sat, Mar 13, 2010 at 3:08 AM, [email protected] <[email protected]> >>>> wrote: >>>> >>>>> >>>>> Hi all! >>>>> I am a new user of Siesta and I'm involved in graphene band structure >>>>> calculations. I can not obtain the correct band structure for a >>>>> graphene >>>>> nanoribon (N=3m+2, armchair passivated by H-atoms). Trying to do that >>>>> I >>>>> was overwhelmed with the fact that the system absolutely symmetrical >>>>> in >>>>> Y-axis (finite size, 5 honeycomb rings) has non-zero dipole moment >>>>> along >>>>> Y-axix which depends sharply on the Meshcutoff value. >>>>> Could anyone explain me where it comes from? What are the possible >>>>> sources >>>>> of this "effect" and how it may affect the band structure? >>>>> >>>>> The replica of my .fdf and .out files are attached below. >>>>> >>>>> Thanks for any reasonable help. >>>>> >>>>> From input-file >>>>> >>>>> NumberOfAtoms 26 >>>>> NumberOfSpecies 2 >>>>> >>>>> %block ChemicalSpeciesLabel >>>>> 1 6 C # Species index, atomic number, species label >>>>> 2 1 H >>>>> %endblock ChemicalSpeciesLabel >>>>> Meshcutoff 200.00 Ry >>>>> LongOutput T >>>>> WriteDenchar T >>>>> AtomicCoordinatesFormat Ang >>>>> %block AtomicCoordinatesAndAtomicSpecies >>>>> 0.00 7.37853644 0.000 1 >>>>> 1.42 7.37853644 0.000 1 >>>>> -0.71 6.148780367 0.000 1 >>>>> 2.13 6.148780367 0.000 1 >>>>> 0.00 4.919024294 0.000 1 >>>>> 1.42 4.919024294 0.000 1 >>>>> -0.71 3.68926822 0.000 1 >>>>> 2.13 3.68926822 0.000 1 >>>>> 0.00 2.459512147 0.000 1 >>>>> 1.42 2.459512147 0.000 1 >>>>> -0.71 1.229756073 0.000 1 >>>>> 2.13 1.229756073 0.000 1 >>>>> 0.00 0.00 0.000 1 >>>>> 1.42 0.00 0.000 1 >>>>> -0.71 -1.229756073 0.000 1 >>>>> 2.13 -1.229756073 0.000 1 >>>>> 0.00 -2.459512147 0.000 1 >>>>> 1.42 -2.459512147 0.000 1 >>>>> -0.71 -3.68926822 0.000 1 >>>>> 2.13 -3.68926822 0.000 1 >>>>> 0.00 -4.919024294 0.000 1 >>>>> 1.42 -4.919024294 0.000 1 >>>>> -0.485 8.218581082 0.000 2 >>>>> -0.485 -5.759068935 0.000 2 >>>>> 1.905 8.218581082 0.000 2 >>>>> 1.905 -5.759068935 0.000 2 >>>>> %endblock AtomicCoordinatesAndAtomicSpecies >>>>> >>>>> %block PAO.BasisSizes >>>>> C DZP >>>>> H DZP >>>>> %endblock PAO.BasisSizes >>>>> >>>>> LatticeConstant 2.459512147 Ang >>>>> %block LatticeVectors >>>>> 1.732050808 0.00 0.0 >>>>> 0.0 80.0 0.0 >>>>> 0.0 0.00 80.0 >>>>> >>>>> %endblock LatticeVectors >>>>> >>>>> %block BandLines >>>>> 1 -1.0 0.0 0.0 >>>>> 40 0.0 0.0 0.0 >>>>> 40 1.0 0.0 0.0 >>>>> %endblock BandLines >>>>> >>>>> >>>>> From out-file >>>>> >>>>> siesta: Electric dipole (a.u.) = 0.000000 -12.769435 0.000000 >>>>> siesta: Electric dipole (Debye) = 0.000000 -32.456688 0.000000 >>>>> >>>>> >>>>> Artem Baskin, >>>>> PhD student, >>>>> University of Illinois at Chicago
