Screenshot instructions:
Windows
Mac
Red Hat Linux
Ubuntu
Click URL instructions:
Rightclick on ad, choose "Copy Link", then paste here →
(This may not be possible with some types of ads)
From: Iyad ALQASIR <iyad.ne@gm...>  20130612 20:29:36

From: Iyad ALQASIR <iyad.ne@gm...>  20130619 04:11:03
Attachments:
text/plain
Group_velocity_Togo.pdf

From: Atsushi Togo <atz.togo@gm...>  20130619 05:29:45

Hi, Phonon calculation and phonopy became much more popular than I'd thought. I have implemented useful functions other than just frequencies. But some of them include implicit assumptions like finite difference or convergence parameters. In the case of group velocity, you can calculate it just with gv option, but it doesn't mean it is easy calculation. If you have some deep question, you have another choice which is reading the source code. If you don't understand the meaning, you can ask me on the mailing list. About your test calculation. Thank you for your report. I don't carefully check the convergence, but concerning to your data, delta=0.0001 would be too small and the eigenvectors are almost degenerate to distinguish. You can observe the nature of eigenvectors by watching them. The reason that I employ the formula for group velocity is that I wanted to avoid band crossing problem. There are many band crossing in the band structure usually, we have to check phonon mode symmetry and carefully connect neighboring qpoints if we calculate group velocity from the band structure frequencies. So your calculation from frequencies may be wrong though the curve is smooth. But I don't want to discuss about this because for me this is not important issue and I don't have so much time. In most of the crystal symmetry, the group velocity at the Brillouin zone boundary is zero. But in nonsymmorphic crystal symmetry (like your crystal), it can have positive of negative velocity, and oscillation like your case happens. The additional reason of your oscillation, that is more technical reason, is that the little group of the qpoint on the Brilloin zone boundary is different from neighboring qpoints used to calculate group velocity. At general points inside Brillouin zone, I think it doesn't happen in most cases because the little groups at those qpoints are same. If you want to know about this in detail, I recommend to read a practical book for group theory of crystals. Togo On Wed, Jun 19, 2013 at 1:10 PM, Iyad ALQASIR <iyad.ne@...> wrote: > Dear Togo, > > Back again to our discussion about the group velocity and the numerical > issues related to the finite difference. I would like to pay your attention > that shifting the qmesh slightly ( delta=0.0001) could produce > significantly the group velocity. I am attaching a MS word document showing > the group velocity using nonshifted, shifted meshes, and calculated from > the dispersion relations directly along high symmetry directions. > > The concern is, if someone want to calcualte the group velocity at an > arbitrary point in the BZ, how can he make sure it is calculated correctly  > no oscillating value ( in case of no qshift), or not biased ( in case of > qshift). > > I hope the attached document helps in explaining of what is going on. > > Thank you, > Iyad  Atsushi Togo http://atztogo.github.com/ atz.togo@... 
From: Aldo Humberto Romero <aromero@qr...>  20130619 12:28:04

Iyad as Togosan said, this is a very unusual solid but by looking your info, it can also be related to the fact that you have long range forces, this is a pity in the case of frozen phonon mode methods because it requires large supercells, for checking you should try a DFPT, as implemented in Abinit or Quantum Expresso, in that case, you do not need a supercell and you can go directly to evaluate the interatomic force constants.. of course, in that case you do not have PHONOPY to help you but this calculation could be very straight forward. Good luck > Dear Togo, > > Back again to our discussion about the group velocity and the numerical > issues related to the finite difference. I would like to pay your > attention that shifting the qmesh slightly ( delta=0.0001) could produce > significantly the group velocity. I am attaching a MS word document > showing the group velocity using nonshifted, shifted meshes, and > calculated from the dispersion relations directly along high symmetry > directions. > > The concern is, if someone want to calcualte the group velocity at an > arbitrary point in the BZ, how can he make sure it is calculated > correctly  no oscillating value ( in case of no qshift), or not biased ( > in case of qshift). > > I hope the attached document helps in explaining of what is going on. > > Thank you, > Iyad >  > This SF.net email is sponsored by Windows: > > Build for Windows Store. > > http://p.sf.net/sfu/windowsdev2dev_______________________________________________ > Phonopyusers mailing list > Phonopyusers@... > https://lists.sourceforge.net/lists/listinfo/phonopyusers >  Prof. Aldo Humberto Romero CINVESTAVUnidad Queretaro Libramiento Norponiente 2000 CP 76230, Queretaro, QRO, Mexico tel: 442 211 9909 fax: 442 211 9938 email: aromero@... aldorome@... www: qro.cinvestav.mx/~aromero 
From: Atsushi Togo <atz.togo@gm...>  20130612 22:33:00

Hi, The problem may come from the implemented method to calculate group velocity itself and the numerical quality of your dynamical matrices. But I need to see your calculation to be sure and fix (if possible) it. Togo On Thu, Jun 13, 2013 at 5:29 AM, Iyad ALQASIR <iyad.ne@...> wrote: > Dear Phonopy Users, > > I am trying to calculate the group velocity along high symmetry direction. > > I have a problem of getting a smooth behavior of the velocity when the > dispersions relations start to become flat in some directions in the BZ. > > I am showing in the attached document the difference between the speed of > sound for NaCl and FeSi for the LA mode. > > As we see, in case of NaCl ( I used the force constants provided in the > example directory of phonopy) the speed looks smooth. Where as in case of > FeSi, it fluctuates in the MX and XR directions. > > I am attaching the input file for NaCl (NaCl.py) I used to calculate the > group velocity ( and then to get the speed) ( lines 6279), it would be kind > of you if you could advise to improve this behavior. > > Thank You > Iyad > >  > This SF.net email is sponsored by Windows: > > Build for Windows Store. > > http://p.sf.net/sfu/windowsdev2dev > _______________________________________________ > Phonopyusers mailing list > Phonopyusers@... > https://lists.sourceforge.net/lists/listinfo/phonopyusers >  Atsushi Togo http://atztogo.github.com/ atz.togo@... 
From: Atsushi Togo <atz.togo@gm...>  20130613 06:19:03

Hi, In phonopy, group velocity is calculated by (see phonopy/group_velocity/__init__.py) d omega   = \ / omega d q \/q Gradient of omega in reciprocal space. d D(q) <e(q,nu)e(q,nu)> d q which is similar to modeGruneisen parameter shown here http://phonopy.sourceforge.net/gruneisen.html#methodtocalculatemodegruneisenparameters . I can't answer to your question, because I don't understand it well. If the dynamical matrix is not well built numerically in the symmetry point view, (for example doubly) degenerate TA mode splits. But the two modes have very similar eigenvectors. I guessed this may be a reason. Please send your all files to me directly if you need my help. Togo On Thu, Jun 13, 2013 at 1:11 PM, Iyad ALQASIR <iyad.ne@...> wrote: > Hi, > > What do you mean by numerical quality of the dynamical matrices. > > Can I send you the FORCE_SETS to check the dynamical matix? > > Thanks > IYAD > > > On Wed, Jun 12, 2013 at 6:32 PM, Atsushi Togo <atz.togo@...> wrote: >> >> Hi, >> >> The problem may come from the implemented method to calculate group >> velocity itself and the numerical quality of your dynamical matrices. >> But I need to see your calculation to be sure and fix (if possible) >> it. >> >> Togo >> >> On Thu, Jun 13, 2013 at 5:29 AM, Iyad ALQASIR <iyad.ne@...> wrote: >> > Dear Phonopy Users, >> > >> > I am trying to calculate the group velocity along high symmetry >> > direction. >> > >> > I have a problem of getting a smooth behavior of the velocity when the >> > dispersions relations start to become flat in some directions in the BZ. >> > >> > I am showing in the attached document the difference between the speed >> > of >> > sound for NaCl and FeSi for the LA mode. >> > >> > As we see, in case of NaCl ( I used the force constants provided in the >> > example directory of phonopy) the speed looks smooth. Where as in case >> > of >> > FeSi, it fluctuates in the MX and XR directions. >> > >> > I am attaching the input file for NaCl (NaCl.py) I used to calculate the >> > group velocity ( and then to get the speed) ( lines 6279), it would be >> > kind >> > of you if you could advise to improve this behavior. >> > >> > Thank You >> > Iyad >> > >> > >> >  >> > This SF.net email is sponsored by Windows: >> > >> > Build for Windows Store. >> > >> > http://p.sf.net/sfu/windowsdev2dev >> > _______________________________________________ >> > Phonopyusers mailing list >> > Phonopyusers@... >> > https://lists.sourceforge.net/lists/listinfo/phonopyusers >> > >> >> >> >>  >> Atsushi Togo >> http://atztogo.github.com/ >> atz.togo@... > > > > >  > ___________________________ > IYAD I. ALQASIR, PhD > Postdoc Research Associate > > XRay and Neutron Scattering and Spectroscopy Group > Materials Science and Technology Division > Oak Ridge National Lab > Oak Ridge, TN  Atsushi Togo http://atztogo.github.com/ atz.togo@... 
From: Atsushi Togo <atz.togo@gm...>  20130614 11:21:41

Hi, I see the problem occurs by the finite difference technique that I implemented. This problem only occurs in some cases when the phonon mode is degenerate at zone boundary. I don't find a best way to solve this problem right now. So I propose a way which is not rigorous but may be acceptable. It is just to shift the qpoint slightly. For example, phonopy dim="2 2 2" band="0 0 0 0.501 0.501 0 0 0.501 0 0.501 0.501 0 0.501 0.501 0.501 0 0 0 0 0.501 0 0.501 0.501 0.501" gv The slight shift of qpoint splits degenerate modes and I think you can obtain group velocity that is close to the expected value. Togo On Thu, Jun 13, 2013 at 3:18 PM, Atsushi Togo <atz.togo@...> wrote: > Hi, > > In phonopy, group velocity is calculated by (see > phonopy/group_velocity/__init__.py) > > d omega  >  = \ / omega > d q \/q > > Gradient of omega in reciprocal space. > > d D(q) > <e(q,nu)e(q,nu)> > d q > > which is similar to modeGruneisen parameter shown here > http://phonopy.sourceforge.net/gruneisen.html#methodtocalculatemodegruneisenparameters > . > > I can't answer to your question, because I don't understand it well. > If the dynamical matrix is not well built numerically in the symmetry > point view, (for example doubly) degenerate TA mode splits. But the > two modes have very similar eigenvectors. I guessed this may be a > reason. > > Please send your all files to me directly if you need my help. > > Togo > > On Thu, Jun 13, 2013 at 1:11 PM, Iyad ALQASIR <iyad.ne@...> wrote: >> Hi, >> >> What do you mean by numerical quality of the dynamical matrices. >> >> Can I send you the FORCE_SETS to check the dynamical matix? >> >> Thanks >> IYAD >> >> >> On Wed, Jun 12, 2013 at 6:32 PM, Atsushi Togo <atz.togo@...> wrote: >>> >>> Hi, >>> >>> The problem may come from the implemented method to calculate group >>> velocity itself and the numerical quality of your dynamical matrices. >>> But I need to see your calculation to be sure and fix (if possible) >>> it. >>> >>> Togo >>> >>> On Thu, Jun 13, 2013 at 5:29 AM, Iyad ALQASIR <iyad.ne@...> wrote: >>> > Dear Phonopy Users, >>> > >>> > I am trying to calculate the group velocity along high symmetry >>> > direction. >>> > >>> > I have a problem of getting a smooth behavior of the velocity when the >>> > dispersions relations start to become flat in some directions in the BZ. >>> > >>> > I am showing in the attached document the difference between the speed >>> > of >>> > sound for NaCl and FeSi for the LA mode. >>> > >>> > As we see, in case of NaCl ( I used the force constants provided in the >>> > example directory of phonopy) the speed looks smooth. Where as in case >>> > of >>> > FeSi, it fluctuates in the MX and XR directions. >>> > >>> > I am attaching the input file for NaCl (NaCl.py) I used to calculate the >>> > group velocity ( and then to get the speed) ( lines 6279), it would be >>> > kind >>> > of you if you could advise to improve this behavior. >>> > >>> > Thank You >>> > Iyad >>> > >>> > >>> >  >>> > This SF.net email is sponsored by Windows: >>> > >>> > Build for Windows Store. >>> > >>> > http://p.sf.net/sfu/windowsdev2dev >>> > _______________________________________________ >>> > Phonopyusers mailing list >>> > Phonopyusers@... >>> > https://lists.sourceforge.net/lists/listinfo/phonopyusers >>> > >>> >>> >>> >>>  >>> Atsushi Togo >>> http://atztogo.github.com/ >>> atz.togo@... >> >> >> >> >>  >> ___________________________ >> IYAD I. ALQASIR, PhD >> Postdoc Research Associate >> >> XRay and Neutron Scattering and Spectroscopy Group >> Materials Science and Technology Division >> Oak Ridge National Lab >> Oak Ridge, TN > > > >  > Atsushi Togo > http://atztogo.github.com/ > atz.togo@...  Atsushi Togo http://atztogo.github.com/ atz.togo@... 
From: Atsushi Togo <atz.togo@gm...>  20130614 22:02:40

Hi, I ran your script, but I didn't understand how I read the output values. If you want to me to understand it, please plot it, or send it as a gnuplot data. The smaller q_length is the better, but too small value induces numerical error. So you have to choose it by yourself for your system. Togo On Fri, Jun 14, 2013 at 10:39 PM, Iyad ALQASIR <iyad.ne@...> wrote: > Hi, > > Many thanks for the explanation. I will try and let you know. > > One more question please, How can I use larger q_length ( default is 1e4) > to calculate group velocity. I am trying to do that through a python script > (please see the attached file FeSi_group_velocity.py). The reason for that, > I am thinking maybe using coarser q_length will smooth the calculated group > velocity too. > > Kindest Regards, > IYAD > > > > On Fri, Jun 14, 2013 at 7:21 AM, Atsushi Togo <atz.togo@...> wrote: >> >> Hi, >> >> I see the problem occurs by the finite difference technique that I >> implemented. This problem only occurs in some cases when the phonon >> mode is degenerate at zone boundary. I don't find a best way to solve >> this problem right now. So I propose a way which is not rigorous but >> may be acceptable. It is just to shift the qpoint slightly. For >> example, >> >> phonopy dim="2 2 2" band="0 0 0 0.501 0.501 0 0 0.501 0 0.501 >> 0.501 0 0.501 0.501 0.501 0 0 0 0 0.501 0 0.501 0.501 0.501" >> gv >> >> The slight shift of qpoint splits degenerate modes and I think you >> can obtain group velocity that is close to the expected value. >> >> Togo >> >> On Thu, Jun 13, 2013 at 3:18 PM, Atsushi Togo <atz.togo@...> wrote: >> > Hi, >> > >> > In phonopy, group velocity is calculated by (see >> > phonopy/group_velocity/__init__.py) >> > >> > d omega  >> >  = \ / omega >> > d q \/q >> > >> > Gradient of omega in reciprocal space. >> > >> > d D(q) >> > <e(q,nu)e(q,nu)> >> > d q >> > >> > which is similar to modeGruneisen parameter shown here >> > >> > http://phonopy.sourceforge.net/gruneisen.html#methodtocalculatemodegruneisenparameters >> > . >> > >> > I can't answer to your question, because I don't understand it well. >> > If the dynamical matrix is not well built numerically in the symmetry >> > point view, (for example doubly) degenerate TA mode splits. But the >> > two modes have very similar eigenvectors. I guessed this may be a >> > reason. >> > >> > Please send your all files to me directly if you need my help. >> > >> > Togo >> > >> > On Thu, Jun 13, 2013 at 1:11 PM, Iyad ALQASIR <iyad.ne@...> >> > wrote: >> >> Hi, >> >> >> >> What do you mean by numerical quality of the dynamical matrices. >> >> >> >> Can I send you the FORCE_SETS to check the dynamical matix? >> >> >> >> Thanks >> >> IYAD >> >> >> >> >> >> On Wed, Jun 12, 2013 at 6:32 PM, Atsushi Togo <atz.togo@...> >> >> wrote: >> >>> >> >>> Hi, >> >>> >> >>> The problem may come from the implemented method to calculate group >> >>> velocity itself and the numerical quality of your dynamical matrices. >> >>> But I need to see your calculation to be sure and fix (if possible) >> >>> it. >> >>> >> >>> Togo >> >>> >> >>> On Thu, Jun 13, 2013 at 5:29 AM, Iyad ALQASIR <iyad.ne@...> >> >>> wrote: >> >>> > Dear Phonopy Users, >> >>> > >> >>> > I am trying to calculate the group velocity along high symmetry >> >>> > direction. >> >>> > >> >>> > I have a problem of getting a smooth behavior of the velocity when >> >>> > the >> >>> > dispersions relations start to become flat in some directions in the >> >>> > BZ. >> >>> > >> >>> > I am showing in the attached document the difference between the >> >>> > speed >> >>> > of >> >>> > sound for NaCl and FeSi for the LA mode. >> >>> > >> >>> > As we see, in case of NaCl ( I used the force constants provided in >> >>> > the >> >>> > example directory of phonopy) the speed looks smooth. Where as in >> >>> > case >> >>> > of >> >>> > FeSi, it fluctuates in the MX and XR directions. >> >>> > >> >>> > I am attaching the input file for NaCl (NaCl.py) I used to calculate >> >>> > the >> >>> > group velocity ( and then to get the speed) ( lines 6279), it would >> >>> > be >> >>> > kind >> >>> > of you if you could advise to improve this behavior. >> >>> > >> >>> > Thank You >> >>> > Iyad >> >>> > >> >>> > >> >>> > >> >>> >  >> >>> > This SF.net email is sponsored by Windows: >> >>> > >> >>> > Build for Windows Store. >> >>> > >> >>> > http://p.sf.net/sfu/windowsdev2dev >> >>> > _______________________________________________ >> >>> > Phonopyusers mailing list >> >>> > Phonopyusers@... >> >>> > https://lists.sourceforge.net/lists/listinfo/phonopyusers >> >>> > >> >>> >> >>> >> >>> >> >>>  >> >>> Atsushi Togo >> >>> http://atztogo.github.com/ >> >>> atz.togo@... >> >> >> >> >> >> >> >> >> >>  >> >> ___________________________ >> >> IYAD I. ALQASIR, PhD >> >> Postdoc Research Associate >> >> >> >> XRay and Neutron Scattering and Spectroscopy Group >> >> Materials Science and Technology Division >> >> Oak Ridge National Lab >> >> Oak Ridge, TN >> > >> > >> > >> >  >> > Atsushi Togo >> > http://atztogo.github.com/ >> > atz.togo@... >> >> >> >>  >> Atsushi Togo >> http://atztogo.github.com/ >> atz.togo@... > > > > >  > ___________________________ > IYAD I. ALQASIR, PhD > Postdoc Research Associate > > XRay and Neutron Scattering and Spectroscopy Group > Materials Science and Technology Division > Oak Ridge National Lab > Oak Ridge, TN  Atsushi Togo http://atztogo.github.com/ atz.togo@... 
Sign up for the SourceForge newsletter:
No, thanks