From alexys at fc.unesp.br Mon May 5 20:08:05 2008 From: alexys at fc.unesp.br (Alexys Bruno Alfonso) Date: Mon, 05 May 2008 15:08:05 -0300 Subject: [Wannier] recommended Linux installation Message-ID: <20080505150805.daxkan6hc0wkcw0w@webmail.fc.unesp.br> Dear colleagues, please, I would like to know which of the available Linux installations is usually recommended for running the Fortran files of Wannier90. Best regards, A. Bruno-Alfonso ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program. From jry20 at cam.ac.uk Tue May 6 11:19:17 2008 From: jry20 at cam.ac.uk (Jonathan Yates) Date: Tue, 6 May 2008 10:19:17 +0100 (BST) Subject: [Wannier] recommended Linux installation In-Reply-To: <20080505150805.daxkan6hc0wkcw0w@webmail.fc.unesp.br> References: <20080505150805.daxkan6hc0wkcw0w@webmail.fc.unesp.br> Message-ID: On Mon, 5 May 2008, Alexys Bruno Alfonso wrote: > please, I would like to know which of the available Linux > installations is usually recommended for running the Fortran files of > Wannier90. Alexys, Wannier90 simply requires a good f95 compiler, so there is no requirement for a particular linux distribution. The best choice is probably whatever your institution supports (or the one the local computer-guru uses!) There are notes on compiler choice in the wannier90 file README.INSTALL. The g95 and gfortran compilers will work with any distribution. On their websites the commercial compilers (intel and pathscale) list a number of distributions with which they are known to be compatible. (as does the semi-free Sun compiler). As a data point: locally we're running Suse with intel, pathscale, sun, g95 and gfortran - w90 works fine with all of those. In the past I've found both ubuntu and fedora straight-forward to install. earlier you asked: > I am a new user of Wannier90. I have not succeeded to compile the > source codes in Digital Visual Fortran, Windows XP. I believe I did > not understand the installation procedures. Can anybody tell me how to > do it? I assume by the silence, nobody has. I will say the DVF is a pretty old compiler (it was replaced by CVF, and then superceded by intel). I'm not sure if it supports f95, as opposed to just f90. If you still want to try this please post the version number of the compiler, what you tried to do, and any error messages. Although I think it a much better plan to move to linux and a more recent compiler. Jonathan -- Dr Jonathan Yates | Theory of Condensed Matter Group Corpus Christi College | Cavendish Laboratory Cambridge, CB2 1RH, UK | Cambridge, CB3 OHE, UK email jry20 at cam.ac.uk | Tel +44 (0)1223 337461 From marzari at MIT.EDU Tue May 6 14:53:34 2008 From: marzari at MIT.EDU (Nicola Marzari) Date: Tue, 06 May 2008 08:53:34 -0400 Subject: [Wannier] recommended Linux installation In-Reply-To: References: <20080505150805.daxkan6hc0wkcw0w@webmail.fc.unesp.br> Message-ID: <482054CE.3010508@mit.edu> Jonathan Yates wrote: > earlier you asked: >> I am a new user of Wannier90. I have not succeeded to compile the >> source codes in Digital Visual Fortran, Windows XP. I believe I did >> not understand the installation procedures. Can anybody tell me how to >> do it? > > I assume by the silence, nobody has. I will say the DVF is a pretty old > compiler (it was replaced by CVF, and then superceded by intel). I'm not > sure if it supports f95, as opposed to just f90. Hi Jonathan, I had briefly mentioned to Alfonso to install Linux on a machine, and use a standard compiler. This is probably good advice for everyone reading this mailing list - the earlier you bury your Windows distribution, the happier you will be. You would all be amazed how pervasive e.g. Windows is in the developing world (Ethiopia, where I'm writing this) and how it messes up everyone's head. nicola -- --------------------------------------------------------------------- Prof Nicola Marzari Department of Materials Science and Engineering 13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu From vranjan at ncsu.edu Fri May 30 22:38:14 2008 From: vranjan at ncsu.edu (Vivek Ranjan) Date: Fri, 30 May 2008 16:38:14 -0400 (EDT) Subject: [Wannier] spread Message-ID: <53402.152.14.74.117.1212179894.squirrel@webmail.ncsu.edu> Hello, I am trying to calculate wannier function of BN sheet. The way the unit cell is constructed, I have chosen (12,1,1) kpoint mesh for scf calculation. Then I do an nscf calculation with 4 kpoints (4,1,1). The spread of one of wannier function is 26 A. I run several steps (more than 20000) and still it does not converge. The question is how to reduce this spread ? I tried putting more kpoints by doing (12,2,2) scf calculations and 16 kpoints (4,2,2) for nscf and wannier parts. The problem is instead of 16 kpoints in the nscf that I hand put in the file, pwscf always generates 24 kpoints. The additional kpoints are shifted kpoints that I put. It never happened to me before and wondering why it is happening ? (12,1,1) scf, and (4,1,1) kpoints has no such problems. Thanks, Vivek, Physics, NCSU. From marzari at MIT.EDU Fri May 30 22:58:29 2008 From: marzari at MIT.EDU (Nicola Marzari) Date: Fri, 30 May 2008 16:58:29 -0400 Subject: [Wannier] spread In-Reply-To: <53402.152.14.74.117.1212179894.squirrel@webmail.ncsu.edu> References: <53402.152.14.74.117.1212179894.squirrel@webmail.ncsu.edu> Message-ID: <48406A75.1050303@mit.edu> Dear Vivek, first of all, is your unit cell a long strip ? I'm a bit surprised by the fact that you only sample 1 k-point in 2 out of 3 directions. Also, have you tried a simple BN-sheet unit cell with only two atoms ? My guess it that the WFs should look like bonds between the atoms in the plane, and then on each N you have a p_z like orbital. Not sure on the multiplication of points in the nscf - that usually shouldn't happen, exactly because nscf should just do the points it's told. nicola Vivek Ranjan wrote: > Hello, > > I am trying to calculate wannier function of BN sheet. The way the unit > cell is constructed, I have chosen (12,1,1) kpoint mesh for scf > calculation. Then I do an nscf calculation with 4 kpoints (4,1,1). The > spread of one of wannier function is 26 A. I run several steps (more than > 20000) and still it does not converge. The question is how to reduce this > spread ? > > I tried putting more kpoints by doing (12,2,2) scf calculations and 16 > kpoints (4,2,2) for nscf and wannier parts. The problem is instead of 16 > kpoints in the nscf that I hand put in the file, pwscf always generates 24 > kpoints. The additional kpoints are shifted kpoints that I put. It never > happened to me before and wondering why it is happening ? > > (12,1,1) scf, and (4,1,1) kpoints has no such problems. > > Thanks, > > Vivek, > Physics, NCSU. -- --------------------------------------------------------------------- Prof Nicola Marzari Department of Materials Science and Engineering 13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu From jry20 at cam.ac.uk Sat May 31 16:41:51 2008 From: jry20 at cam.ac.uk (Jonathan Yates) Date: Sat, 31 May 2008 15:41:51 +0100 (BST) Subject: [Wannier] spread In-Reply-To: <48406A75.1050303@mit.edu> References: <53402.152.14.74.117.1212179894.squirrel@webmail.ncsu.edu> <48406A75.1050303@mit.edu> Message-ID: On Fri, 30 May 2008, Nicola Marzari wrote: > Not sure on the multiplication of points in the nscf - that usually > shouldn't happen, exactly because nscf should just do the points it's > told. If you put nosym=.true. in the system namelist then you really will get the same kpoints that you specified (sometimes pwscf will generate extra symmetry related kpoints from your list during an nscf - I forget the exact conditions) Jonathan -- Dr Jonathan Yates | Theory of Condensed Matter Group Corpus Christi College | Cavendish Laboratory Cambridge, CB2 1RH, UK | Cambridge, CB3 OHE, UK email jry20 at cam.ac.uk | Tel +44 (0)1223 337461