[Pw_forum] HYBRID XC not allowed in non-scf calculations
Dear Eduardo, ecut_ws is a real space cutoff on the wigner Seitz cell for coulomb. We (Giovanni Bussi Amdrea Ferretti and myself) have developped this for SaX (GW calculations and BSE). The issue was that for strongly anysotropic cells the coulomb potential is not behaving any more as the FFT of 1/q**2 cloes to q=0. Usual techniques like Baldereschi-massidda and others "correct " only q=0, and this is not enough. So the ecut_ws is a cutoff that says to the code since how big q+G it has to correct coulomb: the more anysotropic the more you have to have big this ecut_ws, but it will take more time. The method is similar to Martyna-Tuckermans one withe the difference that we just cutoff the fock operator (this is alsways valid as the overlaps goe to 0 when r-r' -> infinity). If you are interested I have a presentation, and we can discuss these questions by skype. Unfortunatelly, we did nerver write the corresponding paper (always to much work to do). But it is planned to do it one day. hope this can help cheers Layla 2012/10/29 Eduardo Ariel Menendez Proupin > Hi Layla and everybody, > > Let us forget the nscf calculation for the moment, as my scf calculation > was probably not fine. > I have a strongly anisotropic unit cell, and large enough not to dare to > attempt to build a quasi-cubi one, that would have about 360 atoms. For > anysotropic cell, INPUT_PW instructs to use > > exxdiv_treatment = 'vcut_ws', > for this to work one must use > x_gamma_extrapolation = .false., (b.t.w., it > is not mentioned at INPUT_PW.txt ) > then I have to set the value > ecutvcut = X , ! which is 0.7 Ry in the Si > example. > so, what is the method followed by using vcut_ws , and how to set > ecutvcut ? Are low or high values of ecutwfc more or less accurate or more > or less expensive? Is it the method given by Sorouri, Folkes and Hine in > JCP124, 064105 (2006)? > > Thanks > -- > > > Eduardo Menendez Proupin > Departamento de Qu?mica Fisica Aplicada > Facultad de Ciencias > Universidad Aut?noma de Madrid > 28049 Madrid, Spain > Phone: +34 91 497 6706 > > On leave from: Departamento de Fisica, Facultad de Ciencias, Universidad > de Chile URL: http://fisica.ciencias.uchile.cl/~emenendez > > > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum > -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20121029/dc45d3b8/attachment.html
[Pw_forum] HYBRID XC not allowed in non-scf calculations
Hi Layla and everybody, Let us forget the nscf calculation for the moment, as my scf calculation was probably not fine. I have a strongly anisotropic unit cell, and large enough not to dare to attempt to build a quasi-cubi one, that would have about 360 atoms. For anysotropic cell, INPUT_PW instructs to use exxdiv_treatment = 'vcut_ws', for this to work one must use x_gamma_extrapolation = .false., (b.t.w., it is not mentioned at INPUT_PW.txt ) then I have to set the value ecutvcut = X , ! which is 0.7 Ry in the Si example. so, what is the method followed by using vcut_ws , and how to set ecutvcut ? Are low or high values of ecutwfc more or less accurate or more or less expensive? Is it the method given by Sorouri, Folkes and Hine in JCP124, 064105 (2006)? Thanks -- Eduardo Menendez Proupin Departamento de Qu?mica Fisica Aplicada Facultad de Ciencias Universidad Aut?noma de Madrid 28049 Madrid, Spain Phone: +34 91 497 6706 On leave from: Departamento de Fisica, Facultad de Ciencias, Universidad de Chile URL: http://fisica.ciencias.uchile.cl/~emenendez -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20121029/26b781c0/attachment.html
[Pw_forum] HYBRID XC not allowed in non-scf calculations
Dear Simon, one can think to do an interpolation calculating psi_k+q' and V(q'), but pf, Let'S first stabilize exx.f90! cheers Layla 2012/10/25 Simon Binnie > I've always wondered about this. Surely if you've converged your initial > scf calculation with respect to q and k-point sampling this change of grid > doesn't matter, for your new q' points you just end up setting your V(q') > to be the same as the nearest original V(q) ? > > Obviously I'm not volunteering to code this up :) > > Simon > > On Thu, 25 Oct 2012 15:01:22 +0200, Layla Martin-Samos > wrote: > > > Dear Eduardo, it is also related to kpoint samplig and q vectors. If one > > changes k sampling between the scf and the nonscf you can not have the > > same > > v(q) (the same fock). I order to avoid confusion nscf has been disable in > > all the cases. > > > > cheers > > > > Layla > > > > -- > Simon Binnie | Post Doc, Condensed Matter Sector > Scuola Internazionale di Studi Avanzati (SISSA) > Via Bonomea 256 | 34100 Trieste | sbinnie at sissa.it > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum > -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20121025/24fa74c6/attachment.html
[Pw_forum] HYBRID XC not allowed in non-scf calculations
I've always wondered about this. Surely if you've converged your initial scf calculation with respect to q and k-point sampling this change of grid doesn't matter, for your new q' points you just end up setting your V(q') to be the same as the nearest original V(q) ? Obviously I'm not volunteering to code this up :) Simon On Thu, 25 Oct 2012 15:01:22 +0200, Layla Martin-Samos wrote: > Dear Eduardo, it is also related to kpoint samplig and q vectors. If one > changes k sampling between the scf and the nonscf you can not have the > same > v(q) (the same fock). I order to avoid confusion nscf has been disable in > all the cases. > > cheers > > Layla > -- Simon Binnie | Post Doc, Condensed Matter Sector Scuola Internazionale di Studi Avanzati (SISSA) Via Bonomea 256 | 34100 Trieste | sbinnie at sissa.it
[Pw_forum] HYBRID XC not allowed in non-scf calculations
Dear Eduardo, it is also related to kpoint samplig and q vectors. If one changes k sampling between the scf and the nonscf you can not have the same v(q) (the same fock). I order to avoid confusion nscf has been disable in all the cases. cheers Layla 2012/10/25 Eduardo Ariel Menendez Proupin > Hi, > Why ? > Error in routine setup (1): > HYBRID XC not allowed in non-scf calculations > > I guess that to compute the exchange potential one needs to use the SCF > wavefunctions, and not only the charge density, and for this the > input/output of the wfc must be restructured. Is it implemented in trunk > version. Is there a more fundamental reason for not doing it? > > -- > > > Eduardo Menendez Proupin > Departamento de Qu?mica Fisica Aplicada > Facultad de Ciencias > Universidad Aut?noma de Madrid > 28049 Madrid, Spain > Phone: +34 91 497 6706 > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum > -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20121025/f525dcc4/attachment.html
[Pw_forum] HYBRID XC not allowed in non-scf calculations
Hi, Why ? Error in routine setup (1): HYBRID XC not allowed in non-scf calculations I guess that to compute the exchange potential one needs to use the SCF wavefunctions, and not only the charge density, and for this the input/output of the wfc must be restructured. Is it implemented in trunk version. Is there a more fundamental reason for not doing it? -- Eduardo Menendez Proupin Departamento de Qu?mica Fisica Aplicada Facultad de Ciencias Universidad Aut?noma de Madrid 28049 Madrid, Spain Phone: +34 91 497 6706 -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20121025/d6191938/attachment.html