[Pw_forum] Problem with newly implemented LDA -1/2 in atomic code (ld1)
Hi Leonardo, Thanks a lot for the detailed description of LDA-1/2 in the atomic code. I appreciate it. Indeed, with CUT value of 3.67, I got the correct band-gap of Si. I went through one iteration of the paper, and I understand that the parameter CUT has to chosen such that it optimizes the excitation energy. However, It seems that there is no general rule for specifying the fraction of electronic charge to be removed from the valence electronic configuration. For group IV element it is 0.25 (C is), on the other hand for II-VI or III-V compound it is 0.5 for anion p or cation d. What about the transition metal with incomplete d orbital (for example Ti, Zr, V, Cr, Mn)? The only transition metal which was discussed in the paper is Zn, which has completely filled d orbital. Also, what about the fractional charge to be removed from the valence of f-block elements? Thank you, and best regards, Rajan On Thu, Mar 31, 2011 at 1:38 AM, Leonardo Matheus wrote: > Hi, > > I'm Leonardo M M Jorge, the one who put the LDA-1/2 method on QE. And > yes, it is working! > > You indeed have to create a new pseudopotential with ld1.x and the > flag "iswitch=4" (among other options, shown in the example file > atomic_doc/pseudo-LDA-0.5/si.lda.0.5.in ), and then use it on a > standard LDA calculation. > > You obtained results very similar to standard LDA because one of these > other options is not well set. When generating the pseudo, besides the > electronic configuration used (of a half-ionized atom, hence the name > -0.5 ), there is also another parameter needed: rcutv, which is the > radius where you cut the perturbation. The value set on the example > file (1.0) is too small; for Silicon these values should be somewhere > around 3.6. > > To obtain the right value for this rcutv parameter you need to do > several calculations, changing it by small steps, and then get the > rcutv where the gap of the system is an extremum (maximum or minimum). > The explanation why this works is written on the reference you just > cited ( L. G. Ferreira, M. Marques and L. K. Teles, Phys. Rev. B 78 > 125116 (2008) ), but what's important is that it IS NOT a free > parameter (one you keep changing until you get the result you want), > as there is a procedure to obtain it. > > Also, as small variations of the rcutv parameters should give very > small variations of the Electronic Gap (as it is an extremum for > variations), we assume that you can obtain the right rcutv value for a > small and simple system, and then use it on a bigger and more complex > system (for a given atom; different atom kinds have different rcutv!). > > I hope I could explain the method for you, if you have any more > questions feel free to write me. > > Leonardo Jorge > PhD Student > Universidade de S?o Paulo - Brazil > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110401/141f1488/attachment-0001.htm
[Pw_forum] Problem with newly implemented LDA -1/2 in atomic code (ld1)
Hi, I'm Leonardo M M Jorge, the one who put the LDA-1/2 method on QE. And yes, it is working! You indeed have to create a new pseudopotential with ld1.x and the flag "iswitch=4" (among other options, shown in the example file atomic_doc/pseudo-LDA-0.5/si.lda.0.5.in ), and then use it on a standard LDA calculation. You obtained results very similar to standard LDA because one of these other options is not well set. When generating the pseudo, besides the electronic configuration used (of a half-ionized atom, hence the name -0.5 ), there is also another parameter needed: rcutv, which is the radius where you cut the perturbation. The value set on the example file (1.0) is too small; for Silicon these values should be somewhere around 3.6. To obtain the right value for this rcutv parameter you need to do several calculations, changing it by small steps, and then get the rcutv where the gap of the system is an extremum (maximum or minimum). The explanation why this works is written on the reference you just cited ( L. G. Ferreira, M. Marques and L. K. Teles, Phys. Rev. B 78 125116 (2008) ), but what's important is that it IS NOT a free parameter (one you keep changing until you get the result you want), as there is a procedure to obtain it. Also, as small variations of the rcutv parameters should give very small variations of the Electronic Gap (as it is an extremum for variations), we assume that you can obtain the right rcutv value for a small and simple system, and then use it on a bigger and more complex system (for a given atom; different atom kinds have different rcutv!). I hope I could explain the method for you, if you have any more questions feel free to write me. Leonardo Jorge PhD Student Universidade de S?o Paulo - Brazil
[Pw_forum] Problem with newly implemented LDA -1/2 in atomic code (ld1)
Hi Nicola, Your comment prompted me to get the reference [ L. G. Ferreira, M. Marques and L. K. Teles, Phys. Rev. B 78 125116 (2008) ] and try to figure out the way authors worked out LDA - 1/2 method, in general.
[Pw_forum] Problem with newly implemented LDA -1/2 in atomic code (ld1)
Dear Quantum Espresso Community, I built the latest release (perhaps alpha) of QE-4.3a. Tried to look into the latest LDA -1/2 (assumed to be similar to SIC) in atomic code (ld1), and from the available test case of Silicon, I decided to check if I get a corrected band gap of bulk Si (from example05). First I generated LDA - 0.5 pseudo (input deck for the same is supplied with the distribution). Subsequently, to run a band structure calculation of Si, I used the template in example05. First I did band structure calculation with Si.pz-vbc.UPF, and then another calculation with the LDA - 0.5 pseudo (Si.LDA.0.5.UPF). I checked the output of scf calculation of the latter (I am copying the portion from output): PseudoPot. # 1 for Si read from file Si.LDA.0.5.UPF MD5 check sum: ac6493edff0fd6b486246c8fa11d17dc Pseudo is Norm-conserving, Zval = 4.0 Generated using LDA-1/2 implemented by Leonardo Matheus Marion Jorge Using radial grid of 431 points, 2 beta functions with: l(1) = 0 l(2) = 1 atomic species valencemass pseudopotential Si 4.0028.08600 Si( 1.00) 48 Sym.Ops. (with inversion) However, the end results of the two calculations are the same. That is the band structure and hence the band gap in two cases are same - nearly 50% underestimated. Not sure if I am missing something. I have attached the two scripts: run_example-0 ==> uses LDA pseudo, and run_example ==> uses LDA - 1/2 pseudo. I shall appreciate any suggestion. Thanks, and regards, Rajan Rajan K. Pandey, Ph.D. Advisory Research Engineer, Semiconductor Research & Development Center India Systems & Technology Engineering Lab IBM India Pvt. Ltd. MD3 1F B354 Manyata Embassy Business Park Nagawara, Outer Ring Road Bangalore - 560045, India Phone: +91-80-28061262 Mobile: +91-9901850981 Email: rajapand at in.ibm.com rajanpandey at gmail.com -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110329/a4c79c6c/attachment-0001.htm -- next part -- A non-text attachment was scrubbed... Name: run_example-0 Type: application/octet-stream Size: 5836 bytes Desc: not available Url : http://www.democritos.it/pipermail/pw_forum/attachments/20110329/a4c79c6c/attachment-0002.obj -- next part -- A non-text attachment was scrubbed... Name: run_example Type: application/octet-stream Size: 5839 bytes Desc: not available Url : http://www.democritos.it/pipermail/pw_forum/attachments/20110329/a4c79c6c/attachment-0003.obj
[Pw_forum] Problem with newly implemented LDA -1/2 in atomic code (ld1)
H, sometime ago LDA-1/2 in pwscf was working. I saw some numbers for TiO2 anatase, But I do not remember the details. i'll ask around bests regards Layla 2011/3/29 Nicola Marzari > > > Hi, > > If I understand correctly, you generated a lda-1/2 pseudo > with the atomic code, but did a lda calculation on silicon > with pwscf - so you will have gotten very similar but not perfectly > identical results, of lda bulk silicon done with a consistent > pseudo (lda) or an inconsistent pseudo . > > 2) not sure if lda-1/2 is in pwscf (I think not) or if it > can be applied to solids. > >nicola > > > > On 3/29/11 3:28 PM, Rajan Pandey wrote: > > Dear Quantum Espresso Community, > > > > I built the latest release (perhaps alpha) of QE-4.3a. Tried to look > > into the latest LDA -1/2 (assumed to be similar to SIC) in atomic code > > (ld1), and from the available test case of Silicon, > > I decided to check if I get a corrected band gap of bulk Si (from > > example05). First I generated LDA - 0.5 pseudo (input deck for the same > > is supplied with the distribution). Subsequently, > > to run a band structure calculation of Si, I used the template in > > example05. First I did band structure calculation with Si.pz-vbc.UPF, > > and then another calculation with the LDA - 0.5 pseudo (Si.LDA.0.5.UPF). > > I checked the output of scf calculation of the latter (I am copying the > > portion from output): > > > > > > PseudoPot. # 1 for Si read from file Si.LDA.0.5.UPF > > MD5 check sum: ac6493edff0fd6b486246c8fa11d17dc > > Pseudo is Norm-conserving, Zval = 4.0 > > Generated using LDA-1/2 implemented by Leonardo Matheus Marion > Jorge > > Using radial grid of 431 points, 2 beta functions with: > > l(1) = 0 > > l(2) = 1 > > > > atomic species valencemass pseudopotential > > Si 4.0028.08600 Si( 1.00) > > > > 48 Sym.Ops. (with inversion) > > > > However, the end results of the two calculations are the same. That is > > the band structure and hence the band gap in two cases are same - nearly > > 50% underestimated. > > Not sure if I am missing something. I have attached the two scripts: > > run_example-0 ==> uses LDA pseudo, and run_example ==> uses LDA - 1/2 > > pseudo. > > I shall appreciate any suggestion. > > > > Thanks, and regards, > > > > Rajan > > > -- > -- > Prof Nicola MarzariDepartment of MaterialsUniversity of Oxford > Chair of Materials Modelling Director, Materials Modelling Laboratory > nicola.marzari at materials.ox.ac.uk http://mml.materials.ox.ac.uk/NM > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110329/bf9f7c2b/attachment.htm
[Pw_forum] Problem with newly implemented LDA -1/2 in atomic code (ld1)
Hi, If I understand correctly, you generated a lda-1/2 pseudo with the atomic code, but did a lda calculation on silicon with pwscf - so you will have gotten very similar but not perfectly identical results, of lda bulk silicon done with a consistent pseudo (lda) or an inconsistent pseudo . 2) not sure if lda-1/2 is in pwscf (I think not) or if it can be applied to solids. nicola On 3/29/11 3:28 PM, Rajan Pandey wrote: > Dear Quantum Espresso Community, > > I built the latest release (perhaps alpha) of QE-4.3a. Tried to look > into the latest LDA -1/2 (assumed to be similar to SIC) in atomic code > (ld1), and from the available test case of Silicon, > I decided to check if I get a corrected band gap of bulk Si (from > example05). First I generated LDA - 0.5 pseudo (input deck for the same > is supplied with the distribution). Subsequently, > to run a band structure calculation of Si, I used the template in > example05. First I did band structure calculation with Si.pz-vbc.UPF, > and then another calculation with the LDA - 0.5 pseudo (Si.LDA.0.5.UPF). > I checked the output of scf calculation of the latter (I am copying the > portion from output): > > > PseudoPot. # 1 for Si read from file Si.LDA.0.5.UPF > MD5 check sum: ac6493edff0fd6b486246c8fa11d17dc > Pseudo is Norm-conserving, Zval = 4.0 > Generated using LDA-1/2 implemented by Leonardo Matheus Marion Jorge > Using radial grid of 431 points, 2 beta functions with: > l(1) = 0 > l(2) = 1 > > atomic species valencemass pseudopotential > Si 4.0028.08600 Si( 1.00) > > 48 Sym.Ops. (with inversion) > > However, the end results of the two calculations are the same. That is > the band structure and hence the band gap in two cases are same - nearly > 50% underestimated. > Not sure if I am missing something. I have attached the two scripts: > run_example-0 ==> uses LDA pseudo, and run_example ==> uses LDA - 1/2 > pseudo. > I shall appreciate any suggestion. > > Thanks, and regards, > > Rajan > -- -- Prof Nicola MarzariDepartment of MaterialsUniversity of Oxford Chair of Materials Modelling Director, Materials Modelling Laboratory nicola.marzari at materials.ox.ac.uk http://mml.materials.ox.ac.uk/NM