From avallabh at purdue.edu Tue May 1 23:39:41 2012 From: avallabh at purdue.edu (Ajit Vallabhaneni) Date: Tue, 1 May 2012 17:39:41 -0400 (EDT) Subject: [Wannier] Query on kmesh In-Reply-To: <1955894.163993.1335905728490.JavaMail.root@mailhub016.itcs.purdue.edu> Message-ID: <209467259.164120.1335908381682.JavaMail.root@mailhub016.itcs.purdue.edu> Dear Users and developers, I am new to using W90 modules. I am trying to do some calculations on graphene using EPW (which uses wannier as a library). I did some calculations on a coarse mesh (6x6x1) successfully, when i repeated the same on a denser mesh (24x24x1), it was terminating with the following message " Exiting....... kmesh_get_bvector: Not enough bvectors found". I read a previous post which suggests to change the 'kmesh_tol' parameter. However, i am not quite sure it works from that post. Please let me know if there are any other suggestions. Also, i thought of working on the wannier functions independently without using EPW. I tried to run example05 on diamond based on the tutorial. The first three steps are fine, but in the final step, the wannier prog terminated with the following message: -------------- Cycle: 1 wann_main: ZHEEV in internal_new_u_and_m failed, info= 3 trying Schur decomposition instead wann_main: SCHUR failed, info= 4 Exiting....... wann_main: problem computing schur form 1 I havenot changed any input parameters ( I am using QE V4.0.3). Please let me know how to fix this issue. Thanks Ajit From jonathan.yates at materials.ox.ac.uk Wed May 2 00:09:47 2012 From: jonathan.yates at materials.ox.ac.uk (Jonathan Yates) Date: Tue, 1 May 2012 22:09:47 +0000 Subject: [Wannier] Query on kmesh In-Reply-To: <209467259.164120.1335908381682.JavaMail.root@mailhub016.itcs.purdue.edu> References: <209467259.164120.1335908381682.JavaMail.root@mailhub016.itcs.purdue.edu> Message-ID: <7FB4ED1D-799C-4D02-8030-72AF3F49B7D7@materials.ox.ac.uk> On 1 May 2012, at 22:39, Ajit Vallabhaneni wrote: > Dear Users and developers, > > I am new to using W90 modules. I am trying to do some calculations on graphene using EPW (which uses wannier as a library). > I did some calculations on a coarse mesh (6x6x1) successfully, when i repeated the same on a denser mesh (24x24x1), it was terminating with the following message > > " Exiting....... > kmesh_get_bvector: Not enough bvectors found". > > > I read a previous post which suggests to change the 'kmesh_tol' parameter. However, i am not quite sure it works from that post. Please let me know if there are any other suggestions. Only use kmesh_tol as a last resort. And only use it if you understand what it does. As I wrote to this list last weekend 'Most problems with finding the k-point shells are caused by inaccurate lattice parameters or kpoints in the win file.' So make sure you are giving those quantities accurately in the win file. Without more details (eg the win file) it is hard for me to give more specific guidance. However, you should know that if you set iprint=4 in the win file the kmesh routine will give you a lot more debugging information. > -------------- > Cycle: 1 > wann_main: ZHEEV in internal_new_u_and_m failed, info= 3 > trying Schur decomposition instead > wann_main: SCHUR failed, info= 4 > Exiting....... > wann_main: problem computing schur form 1 > > I havenot changed any input parameters ( I am using QE V4.0.3). Please let me know how to fix this issue. That is almost certainly a problem with the lapack / blas libraries you have linked to. Jonathan -- Department of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK tel: +44 (0)1865 612797 http://users.ox.ac.uk/~oums0549/ From avallabh at purdue.edu Wed May 2 04:58:56 2012 From: avallabh at purdue.edu (Ajit Vallabhaneni) Date: Tue, 1 May 2012 22:58:56 -0400 (EDT) Subject: [Wannier] Query on kmesh In-Reply-To: <7FB4ED1D-799C-4D02-8030-72AF3F49B7D7@materials.ox.ac.uk> Message-ID: <672146037.164755.1335927536237.JavaMail.root@mailhub016.itcs.purdue.edu> Prof Jonathan, Thanks for your quick reply. I used the lattice parameters obtained after equilibration( as i mentioned earlier it worked for a coarse grid). I am attaching the output file with this for (iprint=4). I am confused about your concern on k-points. Are you referring to the units of k-points or just the decimal places? I would appreciate if you can help me fix this. Thanks Ajit -------------- next part -------------- A non-text attachment was scrubbed... Name: Ghex.wout Type: application/octet-stream Size: 60075 bytes Desc: not available URL: From avallabh at purdue.edu Wed May 2 05:03:27 2012 From: avallabh at purdue.edu (Ajit Vallabhaneni) Date: Tue, 1 May 2012 23:03:27 -0400 (EDT) Subject: [Wannier] Query on kmesh In-Reply-To: <7FB4ED1D-799C-4D02-8030-72AF3F49B7D7@materials.ox.ac.uk> Message-ID: <397992779.164769.1335927807226.JavaMail.root@mailhub016.itcs.purdue.edu> Prof jonathan, Sorry forgot to attach the input file. (It is from EPW, so not all the details were written to *.win file). Thanks Ajit -------------- next part -------------- A non-text attachment was scrubbed... Name: Ghex.win Type: application/octet-stream Size: 364 bytes Desc: not available URL: From jonathan.yates at materials.ox.ac.uk Wed May 2 11:31:41 2012 From: jonathan.yates at materials.ox.ac.uk (Jonathan Yates) Date: Wed, 2 May 2012 09:31:41 +0000 Subject: [Wannier] Query on kmesh In-Reply-To: <672146037.164755.1335927536237.JavaMail.root@mailhub016.itcs.purdue.edu> References: <672146037.164755.1335927536237.JavaMail.root@mailhub016.itcs.purdue.edu> Message-ID: <18F5AB50-2F56-41F6-8B81-448E52FD92E7@materials.ox.ac.uk> On 2 May 2012, at 03:58, Ajit Vallabhaneni wrote: > Prof Jonathan, > > Thanks for your quick reply. I used the lattice parameters obtained after equilibration( as i mentioned earlier it worked for a coarse grid). I am attaching the output file with this for (iprint=4). > > I am confused about your concern on k-points. Are you referring to the units of k-points or just the decimal places? In both cases I'm referring to the numerical precision - ie the number of decimal places. It's a bit like applying symmetry operations to a crystal. Unless the lattice (and atomic positions) are given precisely the symmetry operations will not hold to within the specified tolerances. This is usually trickiest in an hexagonal system. I created a stand-alone win file from your output, and using a 24 24 1 kmesh grid generated with the kmesh.pl script. That worked fine for me (win and wout posted directly to you). Although I did need to set search_shell=20. This was because the kmesh is much denser in the x,y plane and one has to look to the 18th shell to find bvectors in the z-direction. Jonathan -- Department of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK tel: +44 (0)1865 612797 http://users.ox.ac.uk/~oums0549/ From autieri at sa.infn.it Wed May 2 15:46:30 2012 From: autieri at sa.infn.it (Carmine Autieri) Date: Wed, 02 May 2012 15:46:30 +0200 (CEST) Subject: [Wannier] hopping in supercell, wich k-points grid? In-Reply-To: Message-ID: <19ff11f3-082a-4576-bd73-fdca509d6302@mail.sa.infn.it> Dear Wannier90 users, I kindly ask a simple suggestion. I am trying to calculate the hopping parameters in a supercell with a large z-axis using wannier90. The z-axis is 10 times longer than the other axes. Morover, the system is quasi two-dimensional. I used 10 X 10 X 1 k-points grid. Is it enough for a good estimation of hopping parameters? The calculation is very expensive for my machine. Do you think that a 8 X 8 X 2 grid should be really better? thanks Carmine Autieri Universita' di Salerno (Italy) From nicolas.lorente at cin2.es Wed May 2 16:15:35 2012 From: nicolas.lorente at cin2.es (Nicolas Lorente) Date: Wed, 02 May 2012 16:15:35 +0200 Subject: [Wannier] hopping in supercell, wich k-points grid? In-Reply-To: <19ff11f3-082a-4576-bd73-fdca509d6302@mail.sa.infn.it> References: <19ff11f3-082a-4576-bd73-fdca509d6302@mail.sa.infn.it> Message-ID: <4FA14187.9000307@cin2.es> On 02/05/12 15:46, Carmine Autieri wrote: > Dear Wannier90 users, I kindly ask a simple suggestion. > > I am trying to calculate the hopping parameters in a supercell with a large z-axis using wannier90. > > The z-axis is 10 times longer than the other axes. Morover, the system is quasi two-dimensional. > I used 10 X 10 X 1 k-points grid. Is it enough for a good estimation of hopping parameters? > The calculation is very expensive for my machine. Do you think that a 8 X 8 X 2 grid should be really better? > > thanks > Carmine Autieri > Universita' di Salerno (Italy) > _______________________________________________ > Wannier mailing list > Wannier at quantum-espresso.org > http://www.democritos.it/mailman/listinfo/wannier > I'd use the same k-point sampling as for the DFT calculation to start with. Plot the bands and check if the wannier bands are in good agreement. I would not use 8x8x2, but 8x8x1 eventually. Start with the smallest k-point sampling, check the bands and increase if the agreement should be better. It will depend on what you want to do, etc ... of course, you should also pay attention to the energy windows, disentanglement, spread, etc that are probably more important than the k-point sampling to get good bands. Try a smaller system first. Hope this helps, Nicol?s -- Nicol?s Lorente Centro de Investigaci?n en Nanociencia y Nanotecnolog?a Campus de la UAB E-08193 Bellaterra Tel. (+34) 935868410 -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: nicolas_lorente.vcf Type: text/x-vcard Size: 179 bytes Desc: not available URL: From autieri at sa.infn.it Wed May 2 18:12:15 2012 From: autieri at sa.infn.it (Carmine Autieri) Date: Wed, 02 May 2012 18:12:15 +0200 (CEST) Subject: [Wannier] hopping in supercell, wich k-points grid? In-Reply-To: <4FA14187.9000307@cin2.es> Message-ID: <3e060d93-1485-448c-b72f-bfb44eb390df@mail.sa.infn.it> Great, thank you very much. Carmine Autieri ----- Messaggio originale ----- Da: "Nicolas Lorente" A: wannier at quantum-espresso.org Inviato: Mercoled?, 2 maggio 2012 16:15:35 Oggetto: Re: [Wannier] hopping in supercell, wich k-points grid? On 02/05/12 15:46, Carmine Autieri wrote: Dear Wannier90 users, I kindly ask a simple suggestion. I am trying to calculate the hopping parameters in a supercell with a large z-axis using wannier90. The z-axis is 10 times longer than the other axes. Morover, the system is quasi two-dimensional. I used 10 X 10 X 1 k-points grid. Is it enough for a good estimation of hopping parameters? The calculation is very expensive for my machine. Do you think that a 8 X 8 X 2 grid should be really better? thanks Carmine Autieri Universita' di Salerno (Italy) _______________________________________________ Wannier mailing list Wannier at quantum-espresso.org http://www.democritos.it/mailman/listinfo/wannier I'd use the same k-point sampling as for the DFT calculation to start with. Plot the bands and check if the wannier bands are in good agreement. I would not use 8x8x2, but 8x8x1 eventually. Start with the smallest k-point sampling, check the bands and increase if the agreement should be better. It will depend on what you want to do, etc ... of course, you should also pay attention to the energy windows, disentanglement, spread, etc that are probably more important than the k-point sampling to get good bands. Try a smaller system first. Hope this helps, Nicol?s -- Nicol?s Lorente Centro de Investigaci?n en Nanociencia y Nanotecnolog?a Campus de la UAB E-08193 Bellaterra Tel. (+34) 935868410 _______________________________________________ Wannier mailing list Wannier at quantum-espresso.org http://www.democritos.it/mailman/listinfo/wannier From avallabh at purdue.edu Wed May 2 22:10:21 2012 From: avallabh at purdue.edu (Ajit Vallabhaneni) Date: Wed, 2 May 2012 16:10:21 -0400 (EDT) Subject: [Wannier] Query on kmesh In-Reply-To: <18F5AB50-2F56-41F6-8B81-448E52FD92E7@materials.ox.ac.uk> Message-ID: <15895605.166517.1335989421631.JavaMail.root@mailhub016.itcs.purdue.edu> Prof Yates, Thanks for your reply. I think the no. of shells is the issue. The no.of shells in my *.wout file is only 12. As you might know, the EPW generates the prefix.win file, so i can't include it directly using search_shells. I have to modify the source code file to include this. Does the coordinates in the .win file always have to be in crystal coordinates as in your input file? can i use cartesian (tpiba)? Thanks Ajit From jonathan.yates at materials.ox.ac.uk Thu May 3 01:46:16 2012 From: jonathan.yates at materials.ox.ac.uk (Jonathan Yates) Date: Wed, 2 May 2012 23:46:16 +0000 Subject: [Wannier] Query on kmesh In-Reply-To: <15895605.166517.1335989421631.JavaMail.root@mailhub016.itcs.purdue.edu> References: <15895605.166517.1335989421631.JavaMail.root@mailhub016.itcs.purdue.edu> Message-ID: <7924762C-4A1C-4CA7-B338-4579A0C49B6A@materials.ox.ac.uk> On 2 May 2012, at 21:10, Ajit Vallabhaneni wrote: > > > Does the coordinates in the .win file always have to be in crystal coordinates as in your input file? can i use cartesian (tpiba) Wannier90 accepts both. You will find this information in the user guide. http://www.wannier.org/user_guide.html Jonathan -- Department of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK tel: +44 (0)1865 612797 http://users.ox.ac.uk/~oums0549/ From brad.malone at gmail.com Thu May 3 18:13:40 2012 From: brad.malone at gmail.com (Brad Malone) Date: Thu, 3 May 2012 09:13:40 -0700 Subject: [Wannier] Query on kmesh Message-ID: > > > > Prof Yates, > > Thanks for your reply. I think the no. of shells is the > issue. The no.of shells in my *.wout file is only 12. As you might know, > the EPW generates the prefix.win file, so i can't include it directly using > search_shells. I have to modify the source code file to include this. > > Dear Ajit, If adding more search shells is what you need to do, the changes required to do so within EPW are minimal. If you look in EPW/src/wannierize.f90 you will see a section like so: WRITE (iuwinfil,'("num_wann ",i3)') nbndsub > WRITE (iuwinfil,'("iprint ",i3)') iprint > ! > WRITE (iuwinfil, '("dis_win_min ", f9.3)') dis_win_min > WRITE (iuwinfil, '("dis_win_max ", f9.3)') dis_win_max > WRITE (iuwinfil, '("dis_froz_min ", f9.3)') dis_froz_min > WRITE (iuwinfil, '("dis_froz_max ", f9.3)') dis_froz_max > WRITE (iuwinfil, '("num_iter ", i7)') num_iter This is where EPW writes in some basic information to the *.win file. You can add what you'd like to change here and it should work. If you'd like to pass your new information through the epw.in input file, you can easily do so by doing the same thing that was done with the variable num_iter. Just grep num_iter and you will see it only shows up in three places: in wannierize.f90, in epw_readin.f90, and epwcom.f90. Do the same with your search_shells option and you should be good to go. Hope this helps! Best, Brad Malone UC Berkeley -------------- next part -------------- An HTML attachment was scrubbed... URL: From joukj at hrem.nano.tudelft.nl Fri May 4 14:58:19 2012 From: joukj at hrem.nano.tudelft.nl (Jouk Jansen) Date: Fri, 4 May 2012 14:58:19 +0200 (MEST) Subject: [Wannier] Wannier90 for OpenVMS Message-ID: <12050414581896_21C00162@hrem.nano.tudelft.nl> a.mostofi at imperial.ac.uk wrote on 28-APR-2012 01:28:47.57 >Regarding the character strings in plot.F90, we will modify them using >the more compact solution that Giovanni has suggested below. Fine to me. >Regarding the other comments, I would prefer not to modify the main >release with platform-dependent compiler directives, especially for >platforms that we do not have the means to test ourselves. When we >release the new version of the code, I hope it would not be too much >trouble for you to update your patch for it. No problem. But maybe a link to my web-site can be created on your web-site, telling about the "not official supported" version for OpenVMS. Regards Jouk Pax, vel iniusta, utilior est quam iustissimum bellum. (free after Marcus Tullius Cicero (106 b.Chr.-46 b.Chr.) Epistularum ad Atticum 7.1.4.3) >------------------------------------------------------------------------------< Jouk Jansen joukj at hrem.nano.tudelft.nl Technische Universiteit Delft tttttttttt uu uu ddddddd Kavli Institute of Nanoscience tttttttttt uu uu dd dd Nationaal centrum voor HREM tt uu uu dd dd Lorentzweg 1 tt uu uu dd dd 2628 CJ Delft tt uu uu dd dd Nederland tt uu uu dd dd tel. 31-15-2782272 tt uuuuuuu ddddddd >------------------------------------------------------------------------------< From huanghq at phys.tsinghua.edu.cn Sun May 13 05:20:58 2012 From: huanghq at phys.tsinghua.edu.cn (Hua-Qing Huang) Date: Sun, 13 May 2012 11:20:58 +0800 Subject: [Wannier] bands structure of graphene In-Reply-To: References: Message-ID: <20120513030019.M98915@phys.tsinghua.edu.cn> Dear all, Recently?I want to calculate MLWF of graphene and I have got the sp2-like and pz-like MLWF. However, I found that the band strucutre obtained using MLWF interpolation doesn't agree well with the band structure obtained form first- principles calculation (pwscf). What's worse, there is a gap near the fermi level, which should be a Dirac cone. I have tried different dis_win_max, dis_froz_max, but still can't solve the problem. So could you give me some suggestions? -- Huaqing Huang Ph.D. candidate Group of Condensed Matter Theory Department of Physics Tsinghua University Beijing, 100084 P.R.CHINA Tel: 86-10-62772784 From jonathan.yates at materials.ox.ac.uk Sun May 13 13:57:24 2012 From: jonathan.yates at materials.ox.ac.uk (Jonathan Yates) Date: Sun, 13 May 2012 11:57:24 +0000 Subject: [Wannier] bands structure of graphene In-Reply-To: <20120513030019.M98915@phys.tsinghua.edu.cn> References: <20120513030019.M98915@phys.tsinghua.edu.cn> Message-ID: On 13 May 2012, at 04:20, Hua-Qing Huang wrote: > Dear all, > Recently?I want to calculate MLWF of graphene and I have got the sp2-like and > pz-like MLWF. However, I found that the band strucutre obtained using MLWF > interpolation doesn't agree well with the band structure obtained form first- > principles calculation (pwscf). What's worse, there is a gap near the fermi > level, which should be a Dirac cone. > I have tried different dis_win_max, dis_froz_max, but still can't solve the > problem. So could you give me some suggestions? Dear Huaqing, Without seeing more information (input/output) it is difficult to say what the problem is. My general advice would be to first start with example10 which is AB stacked graphite. Check this gives what you expect. Then remove one of the planes of atoms. The MLWF should be very similar - as should the interpolated band structure. You could then increase the plane separation. If this works (and it has for me in the past), and your inputs don't, you should be able to work out what went wrong. Jonathan -- Department of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK tel: +44 (0)1865 612797 http://users.ox.ac.uk/~oums0549/ From julen.ibanez at ehu.es Mon May 14 11:14:14 2012 From: julen.ibanez at ehu.es (Julen Ibanez Azpiroz) Date: Mon, 14 May 2012 11:14:14 +0200 Subject: [Wannier] bands structure of graphene In-Reply-To: References: <20120513030019.M98915@phys.tsinghua.edu.cn> Message-ID: Dear Huaqing, Have you explicitly included the K point where the Dirac cone is expected in the original k-point mesh in the .win file? If not, you may be getting an interpolation error between k-points nearby K that leads to the gap. You could try to use a 3*3*3 grid or so in the .win file that includes the K point explicitly and see if you get degenerate bands at K. best regards -- ?======================================== Julen Iba?ez Azpiroz Materia Kondentsatuaren Fisika Saila Zientzia eta Teknologia Fakultatea Euskal Herriko Unibertsitatea ======================================== -------------- next part -------------- An HTML attachment was scrubbed... URL: From nicola.marzari at epfl.ch Mon May 14 11:41:17 2012 From: nicola.marzari at epfl.ch (Nicola Marzari) Date: Mon, 14 May 2012 11:41:17 +0200 Subject: [Wannier] bands structure of graphene In-Reply-To: References: <20120513030019.M98915@phys.tsinghua.edu.cn> Message-ID: <4FB0D33D.6060005@epfl.ch> Dear Hua-Qing, Julen, actually missing the K-point shouldn't give rise to any problem. I attach a few slides for a metallic carbon nanotube, but the concept is the same: first slide: use a frozen window below the first green line (the states you want to keep) and a disentanglement window just above the top of the pi/pi^star manifold (the second green line). This would give you very good p_z states on every carbon, and bonds on, well, ever bond (so 2.5 WFs per carbon atom) second slide: This shows that missing the crossing point does not have any effect. third slide: this is how band structures should be reproduced (red - full pwscf calculation; black - wannier interpolation, done using the k-points denoted by the blue lines). nicola On 14/05/2012 11:14, Julen Ibanez Azpiroz wrote: > Dear Huaqing, > > Have you explicitly included the K point where the Dirac cone is > expected in the original k-point mesh in the .win file? If not, you may > be getting an interpolation error between k-points nearby K that leads > to the gap. You could try to use a 3*3*3 grid or so in the .win file > that includes the K point explicitly and see if you get degenerate bands > at K. > > best regards > > > -- > ?======================================== > Julen Iba?ez Azpiroz > Materia Kondentsatuaren Fisika Saila > Zientzia eta Teknologia Fakultatea > Euskal Herriko Unibertsitatea > ======================================== > > > _______________________________________________ > Wannier mailing list > Wannier at quantum-espresso.org > http://www.democritos.it/mailman/listinfo/wannier -- ---------------------------------------------------------------------- Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL -------------- next part -------------- A non-text attachment was scrubbed... Name: Marzari OSU Feb09.ppt.pdf Type: application/pdf Size: 701325 bytes Desc: not available URL: From zheng_fawei at iapcm.ac.cn Sun May 20 07:38:30 2012 From: zheng_fawei at iapcm.ac.cn (=?gb2312?B?1qO3qM6wMjA2Mw==?=) Date: Sun, 20 May 2012 13:38:30 +0800 Subject: [Wannier] Fourier Coefficients of a Wannier Function Message-ID: Dear Wannier90 developers and users, Is it possible to get the Fourier coefficients or Bloch wave coefficients of a wannier function from the Wanner90 output files directly ? Thanks, Fawei -------------- next part -------------- An HTML attachment was scrubbed... URL: