## phonopy-users

 [Phonopy-users] Accuracy of extremely low frequency modes From: Tian Lan - 2013-05-14 05:46:37 Attachments: Message as HTML ```Dear phonopy users, I am using phonopy to calculate the gruneisen parameters of metal oxides. I find in some cases, some acoustic modes have very low frequency around or below 10 to 20 cm-1. In this situation, the modes' response to the volume change is kinda strange, i.e, from large volume to small volume, the frequency change sometimes does not follow a linear trend, it could go up and down and up, for example, 20cm-1->15cm-1-> 21cm-1-> 17cm-1. Therefore, there is no way to define a gruneisen parameter. Obviously, in such a low frequency, an even small change would lead to a very large gruneisen parameter, but does it make any sense? For higher frequency modes, the trend is usually consistent and reasonable. For those frequencies, I am talking about the commensurate points, for example (0.5,0,0) in a 222 supercell. So there is no interpolation issue, to my understanding. And since they are perfect crystal structures, in each given volume, the atomic structure (with no displace) is perfectly relaxed--internal forces=0 for any pair of atoms. I am wondering whether this is a problem from the structure itself or DFT calculation. Or, the small displacement method has the intrinsic numerical accuracy limit in such low frequency. Any suggestions would help me evaluate my calculation. Thank you so much. Best, Tian -- Lan, Tian Ph.D. Candidate, Department of Applied Physics and Materials Science California Institute of Technology, Caltech M/C 138-78, Pasadena, CA, 91125 ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Tian Lan - 2013-05-14 05:48:40 Attachments: Message as HTML ```The volume change is not big in my calculation. Usually within the magnitude of thermal expansion. On Mon, May 13, 2013 at 10:46 PM, Tian Lan wrote: > Dear phonopy users, > > I am using phonopy to calculate the gruneisen parameters of metal oxides. > I find in some cases, some acoustic modes have very low frequency around > or below 10 to 20 cm-1. In this situation, the modes' response to the > volume change is kinda strange, i.e, from large volume to small volume, the > frequency change sometimes does not follow a linear trend, it could go up > and down and up, for example, 20cm-1->15cm-1-> 21cm-1-> 17cm-1. Therefore, > there is no way to define a gruneisen parameter. Obviously, in such a low > frequency, an even small change would lead to a very large gruneisen > parameter, but does it make any sense? For higher frequency modes, the > trend is usually consistent and reasonable. > > For those frequencies, I am talking about the commensurate points, for > example (0.5,0,0) in a 222 supercell. So there is no interpolation issue, > to my understanding. And since they are perfect crystal structures, in each > given volume, the atomic structure (with no displace) is perfectly > relaxed--internal forces=0 for any pair of atoms. > > I am wondering whether this is a problem from the structure itself or DFT > calculation. Or, the small displacement method has the intrinsic numerical > accuracy limit in such low frequency. Any suggestions would help me > evaluate my calculation. > > Thank you so much. > > Best, > Tian > > -- > Lan, Tian > Ph.D. Candidate, Department of Applied Physics and Materials Science > California Institute of Technology, > Caltech M/C 138-78, Pasadena, CA, 91125 > -- Lan, Tian Ph.D. Candidate, Department of Applied Physics and Materials Science California Institute of Technology, Caltech M/C 138-78, Pasadena, CA, 91125 ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Atsushi Togo - 2013-05-15 01:33:26 ```Hi, Low frequency modes are sensitive to numerical error. It may be a reason. In your case, anharmonicity may play a role of your strange behavior. If the acoustic mode frequency is very low already, it may be a sign that the crystal structure is close to the critical point of structural phase transition. In this case, our harmonic approximation is not valid. The finite displacement method also may not be good enough. However the default displacement distance of 0.01 A is very small enough, so you may obtain similar result using DFPT. For strongly anharmonic solid, gruneisen parameter is quite dependent on the volume change, because higher than 3rd order force constants have large value. However your frequency behavior up-down-up-down seems very strange for me. What I doubt most so far is convergence of electronic structure. Togo On Tue, May 14, 2013 at 2:48 PM, Tian Lan wrote: > The volume change is not big in my calculation. Usually within the magnitude > of thermal expansion. > > > On Mon, May 13, 2013 at 10:46 PM, Tian Lan wrote: >> >> Dear phonopy users, >> >> I am using phonopy to calculate the gruneisen parameters of metal oxides. >> I find in some cases, some acoustic modes have very low frequency around or >> below 10 to 20 cm-1. In this situation, the modes' response to the volume >> change is kinda strange, i.e, from large volume to small volume, the >> frequency change sometimes does not follow a linear trend, it could go up >> and down and up, for example, 20cm-1->15cm-1-> 21cm-1-> 17cm-1. Therefore, >> there is no way to define a gruneisen parameter. Obviously, in such a low >> frequency, an even small change would lead to a very large gruneisen >> parameter, but does it make any sense? For higher frequency modes, the >> trend is usually consistent and reasonable. >> >> For those frequencies, I am talking about the commensurate points, for >> example (0.5,0,0) in a 222 supercell. So there is no interpolation issue, to >> my understanding. And since they are perfect crystal structures, in each >> given volume, the atomic structure (with no displace) is perfectly >> relaxed--internal forces=0 for any pair of atoms. >> >> I am wondering whether this is a problem from the structure itself or DFT >> calculation. Or, the small displacement method has the intrinsic numerical >> accuracy limit in such low frequency. Any suggestions would help me evaluate >> my calculation. >> >> Thank you so much. >> >> Best, >> Tian >> >> -- >> Lan, Tian >> Ph.D. Candidate, Department of Applied Physics and Materials Science >> California Institute of Technology, >> Caltech M/C 138-78, Pasadena, CA, 91125 > > > > > -- > Lan, Tian > Ph.D. Candidate, Department of Applied Physics and Materials Science > California Institute of Technology, > Caltech M/C 138-78, Pasadena, CA, 91125 > > ------------------------------------------------------------------------------ > AlienVault Unified Security Management (USM) platform delivers complete > security visibility with the essential security capabilities. Easily and > efficiently configure, manage, and operate all of your security controls > from a single console and one unified framework. Download a free trial. > http://p.sf.net/sfu/alienvault_d2d > _______________________________________________ > Phonopy-users mailing list > Phonopy-users@... > https://lists.sourceforge.net/lists/listinfo/phonopy-users > -- Atsushi Togo http://atztogo.github.com/ atz.togo@... ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Aldo Humberto Romero - 2013-05-15 10:12:27 ```I completely agree with Togo. Could you check your k-mesh convergence of your gamma point? cutoff energy is also converged?... just to be sure... this is unusual and they are some suspects but let rules out the simplest ones... some PAW data is not also good enough.. hopefully it is only Kpoint.. I would do the DFPT instead of the frozen ion approach.. Good luck. -- Prof. Aldo Humberto Romero CINVESTAV-Unidad 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 ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Tian Lan - 2013-05-15 19:07:59 Attachments: Message as HTML ```Thank you ! I used the small displacement method. The problem mostly comes from cuprite Ag2O. For those low frequency acoustic modes, I found an annoying thing, that is, the frequency is dependent on the magnitude of the displacement quite a bit. The electronic convergence has been done carefully. 0.01 A , 0.02 A or 0.04 A displacement would result in quite different outcomes, though they are all small. I can get positive frequency for one displacement and negative frequency for another displacement. I think basically the modes eigenvalues are not converged to any value even in such a small displacement range. Does it mean that the phonon potential is not in a good harmonic regime, or unstable, so the quasiharmonic model fails? Would DFPT approach solve this problem? Thanks, Tian On Wed, May 15, 2013 at 3:08 AM, Aldo Humberto Romero < aromero@...> wrote: > I completely agree with Togo. Could you check your k-mesh convergence of > your gamma point? > cutoff energy is also converged?... just to be sure... this is unusual and > they are some suspects > but let rules out the simplest ones... some PAW data is not also good > enough.. hopefully it is > only Kpoint.. I would do the DFPT instead of the frozen ion approach.. > > Good luck. > > > -- > > Prof. Aldo Humberto Romero > CINVESTAV-Unidad 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 > -- Lan, Tian Ph.D. Candidate, Department of Applied Physics and Materials Science California Institute of Technology, Caltech M/C 138-78, Pasadena, CA, 91125 ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Tian Lan - 2013-05-15 19:20:46 Attachments: Message as HTML ```Those modes are not at the Gamma point, but some commensurate points for the 222 supercell, such as (0.5,0.5,0). For example, at this point, a 0.01A displament gives a negative 12 cm-1, but 0.04 displacement may give a positive 15 cm-1 or so. Tian On Wed, May 15, 2013 at 12:07 PM, Tian Lan wrote: > Thank you ! > > I used the small displacement method. The problem mostly comes from > cuprite Ag2O. For those low frequency acoustic modes, I found an annoying > thing, that is, the frequency is dependent on the magnitude of the > displacement quite a bit. The electronic convergence has been done > carefully. > > 0.01 A , 0.02 A or 0.04 A displacement would result in quite different > outcomes, though they are all small. I can get positive frequency for one > displacement and negative frequency for another displacement. I think > basically the modes eigenvalues are not converged to any value even in such > a small displacement range. Does it mean that the phonon potential is not > in a good harmonic regime, or unstable, so the quasiharmonic model fails? > Would DFPT approach solve this problem? > > Thanks, > Tian > > > On Wed, May 15, 2013 at 3:08 AM, Aldo Humberto Romero < > aromero@...> wrote: > >> I completely agree with Togo. Could you check your k-mesh convergence of >> your gamma point? >> cutoff energy is also converged?... just to be sure... this is unusual and >> they are some suspects >> but let rules out the simplest ones... some PAW data is not also good >> enough.. hopefully it is >> only Kpoint.. I would do the DFPT instead of the frozen ion approach.. >> >> Good luck. >> >> >> -- >> >> Prof. Aldo Humberto Romero >> CINVESTAV-Unidad 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 >> > > > > -- > Lan, Tian > Ph.D. Candidate, Department of Applied Physics and Materials Science > California Institute of Technology, > Caltech M/C 138-78, Pasadena, CA, 91125 > -- Lan, Tian Ph.D. Candidate, Department of Applied Physics and Materials Science California Institute of Technology, Caltech M/C 138-78, Pasadena, CA, 91125 ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Atsushi Togo - 2013-05-15 21:21:40 ```Then it seems quite anharmonic. The harmonic approximation is appropriate for your system. I can't say quasiharmonic model fails, but clearly you can't rely on quasiharmonic approximation to solve your problem. DFPT approach may give an answer but that's, I guess, not what you want, because the phonon frequency would be temperature dependent due to temperature dependent atomic displacements. You need special treatment, which phonopy can't offer. Togo On Thu, May 16, 2013 at 4:07 AM, Tian Lan wrote: > Thank you ! > > I used the small displacement method. The problem mostly comes from cuprite > Ag2O. For those low frequency acoustic modes, I found an annoying thing, > that is, the frequency is dependent on the magnitude of the displacement > quite a bit. The electronic convergence has been done carefully. > > 0.01 A , 0.02 A or 0.04 A displacement would result in quite different > outcomes, though they are all small. I can get positive frequency for one > displacement and negative frequency for another displacement. I think > basically the modes eigenvalues are not converged to any value even in such > a small displacement range. Does it mean that the phonon potential is not in > a good harmonic regime, or unstable, so the quasiharmonic model fails? Would > DFPT approach solve this problem? > > Thanks, > Tian > > > On Wed, May 15, 2013 at 3:08 AM, Aldo Humberto Romero > wrote: >> >> I completely agree with Togo. Could you check your k-mesh convergence of >> your gamma point? >> cutoff energy is also converged?... just to be sure... this is unusual and >> they are some suspects >> but let rules out the simplest ones... some PAW data is not also good >> enough.. hopefully it is >> only Kpoint.. I would do the DFPT instead of the frozen ion approach.. >> >> Good luck. >> >> >> -- >> >> Prof. Aldo Humberto Romero >> CINVESTAV-Unidad 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 > > > > > -- > Lan, Tian > Ph.D. Candidate, Department of Applied Physics and Materials Science > California Institute of Technology, > Caltech M/C 138-78, Pasadena, CA, 91125 -- Atsushi Togo http://atztogo.github.com/ atz.togo@... ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Aldo Humberto Romero - 2013-05-15 21:28:46 ```Well, I am not sure... one thing is clear (if the convergence has been done correctly), that the harmonic regime is much smaller and you should go to DFPT. the phonons you are interest it are very soft and those are sometime kind of tricky. what is know experimentally about them? is there any specific dependence with temperature or pressure?... you can try SCAILD and see if there are important changes with temperature.. of course, that is not also the truth but you can check your anharmonicities. Phonopy is an excellent and beautiful implementation that relies on a good calculation, if it fails is not much what phonopy can do for you. You have to dig and find in your DFT.... did you try different pseudos?.. are you including semi core electrons? Best of luck! -> Then it seems quite anharmonic. The harmonic approximation is -> appropriate for your system. I can't say quasiharmonic model fails, -> but clearly you can't rely on quasiharmonic approximation to solve -> your problem. DFPT approach may give an answer but that's, I guess, -> not what you want, because the phonon frequency would be temperature -> dependent due to temperature dependent atomic displacements. You need -> special treatment, which phonopy can't offer. -> -> Togo -> -> On Thu, May 16, 2013 at 4:07 AM, Tian Lan wrote: ->> Thank you ! ->> ->> I used the small displacement method. The problem mostly comes from ->> cuprite ->> Ag2O. For those low frequency acoustic modes, I found an annoying thing, ->> that is, the frequency is dependent on the magnitude of the displacement ->> quite a bit. The electronic convergence has been done carefully. ->> ->> 0.01 A , 0.02 A or 0.04 A displacement would result in quite different ->> outcomes, though they are all small. I can get positive frequency for ->> one ->> displacement and negative frequency for another displacement. I think ->> basically the modes eigenvalues are not converged to any value even in ->> such ->> a small displacement range. Does it mean that the phonon potential is ->> not in ->> a good harmonic regime, or unstable, so the quasiharmonic model fails? ->> Would ->> DFPT approach solve this problem? ->> ->> Thanks, ->> Tian ->> ->> ->> On Wed, May 15, 2013 at 3:08 AM, Aldo Humberto Romero ->> wrote: ->>> ->>> I completely agree with Togo. Could you check your k-mesh convergence ->>> of ->>> your gamma point? ->>> cutoff energy is also converged?... just to be sure... this is unusual ->>> and ->>> they are some suspects ->>> but let rules out the simplest ones... some PAW data is not also good ->>> enough.. hopefully it is ->>> only Kpoint.. I would do the DFPT instead of the frozen ion approach.. ->>> ->>> Good luck. ->>> ->>> ->>> -- ->>> ->>> Prof. Aldo Humberto Romero ->>> CINVESTAV-Unidad 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 ->> ->> ->> ->> ->> -- ->> Lan, Tian ->> Ph.D. Candidate, Department of Applied Physics and Materials Science ->> California Institute of Technology, ->> Caltech M/C 138-78, Pasadena, CA, 91125 -> -> -> -> -- -> Atsushi Togo -> http://atztogo.github.com/ -> atz.togo@... -> -> ------------------------------------------------------------------------------ -> AlienVault Unified Security Management (USM) platform delivers complete -> security visibility with the essential security capabilities. Easily and -> efficiently configure, manage, and operate all of your security controls -> from a single console and one unified framework. Download a free trial. -> http://p.sf.net/sfu/alienvault_d2d -> _______________________________________________ -> Phonopy-users mailing list -> Phonopy-users@... -> https://lists.sourceforge.net/lists/listinfo/phonopy-users -> -- Prof. Aldo Humberto Romero CINVESTAV-Unidad 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 ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Aldo Humberto Romero - 2013-05-15 21:37:18 ```Well, I am not sure... one thing is clear (if the convergence has been done correctly), that the harmonic regime is much smaller and you should go to DFPT. the phonons you are interest it are very soft and those are sometime kind of tricky. what is know experimentally about them? is there any specific dependence with temperature or pressure?... you can try SCAILD and see if there are important changes with temperature.. of course, that is not also the truth but you can check your anharmonicities. Phonopy is an excellent and beautiful implementation that relies on a good calculation, if it fails is not much what phonopy can do for you. You have to dig and find in your DFT.... did you try different pseudos?.. are you including semi core electrons? Best of luck! -> Then it seems quite anharmonic. The harmonic approximation is -> appropriate for your system. I can't say quasiharmonic model fails, -> but clearly you can't rely on quasiharmonic approximation to solve -> your problem. DFPT approach may give an answer but that's, I guess, -> not what you want, because the phonon frequency would be temperature -> dependent due to temperature dependent atomic displacements. You need -> special treatment, which phonopy can't offer. -> -> Togo -> -> On Thu, May 16, 2013 at 4:07 AM, Tian Lan wrote: ->> Thank you ! ->> ->> I used the small displacement method. The problem mostly comes from ->> cuprite ->> Ag2O. For those low frequency acoustic modes, I found an annoying thing, ->> that is, the frequency is dependent on the magnitude of the displacement ->> quite a bit. The electronic convergence has been done carefully. ->> ->> 0.01 A , 0.02 A or 0.04 A displacement would result in quite different ->> outcomes, though they are all small. I can get positive frequency for ->> one ->> displacement and negative frequency for another displacement. I think ->> basically the modes eigenvalues are not converged to any value even in ->> such ->> a small displacement range. Does it mean that the phonon potential is ->> not in ->> a good harmonic regime, or unstable, so the quasiharmonic model fails? ->> Would ->> DFPT approach solve this problem? ->> ->> Thanks, ->> Tian ->> ->> ->> On Wed, May 15, 2013 at 3:08 AM, Aldo Humberto Romero ->> wrote: ->>> ->>> I completely agree with Togo. Could you check your k-mesh convergence ->>> of ->>> your gamma point? ->>> cutoff energy is also converged?... just to be sure... this is unusual ->>> and ->>> they are some suspects ->>> but let rules out the simplest ones... some PAW data is not also good ->>> enough.. hopefully it is ->>> only Kpoint.. I would do the DFPT instead of the frozen ion approach.. ->>> ->>> Good luck. ->>> ->>> ->>> -- ->>> ->>> Prof. Aldo Humberto Romero ->>> CINVESTAV-Unidad 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 ->> ->> ->> ->> ->> -- ->> Lan, Tian ->> Ph.D. Candidate, Department of Applied Physics and Materials Science ->> California Institute of Technology, ->> Caltech M/C 138-78, Pasadena, CA, 91125 -> -> -> -> -- -> Atsushi Togo -> http://atztogo.github.com/ -> atz.togo@... -> -> ------------------------------------------------------------------------------ -> AlienVault Unified Security Management (USM) platform delivers complete -> security visibility with the essential security capabilities. Easily and -> efficiently configure, manage, and operate all of your security controls -> from a single console and one unified framework. Download a free trial. -> http://p.sf.net/sfu/alienvault_d2d -> _______________________________________________ -> Phonopy-users mailing list -> Phonopy-users@... -> https://lists.sourceforge.net/lists/listinfo/phonopy-users -> -- Prof. Aldo Humberto Romero CINVESTAV-Unidad 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 ```
 Re: [Phonopy-users] Accuracy of extremely low frequency modes From: Atsushi Togo - 2013-05-15 23:55:29 ```I re-read what I wrote and I found that I made mistake in writing English. The harmonic approximation is appropriate for your system. --> The harmonic approximation is "not" appropriate for your system. I can't say quasiharmonic model fails, but clearly you can't rely on quasiharmonic approximation to solve your problem. --> I can't say quasiharmonic model fails, but clearly you can't rely on "only" quasiharmonic approximation to solve your problem. Togo On Thu, May 16, 2013 at 6:21 AM, Atsushi Togo wrote: > Then it seems quite anharmonic. The harmonic approximation is > appropriate for your system. I can't say quasiharmonic model fails, > but clearly you can't rely on quasiharmonic approximation to solve > your problem. DFPT approach may give an answer but that's, I guess, > not what you want, because the phonon frequency would be temperature > dependent due to temperature dependent atomic displacements. You need > special treatment, which phonopy can't offer. > > Togo > > On Thu, May 16, 2013 at 4:07 AM, Tian Lan wrote: >> Thank you ! >> >> I used the small displacement method. The problem mostly comes from cuprite >> Ag2O. For those low frequency acoustic modes, I found an annoying thing, >> that is, the frequency is dependent on the magnitude of the displacement >> quite a bit. The electronic convergence has been done carefully. >> >> 0.01 A , 0.02 A or 0.04 A displacement would result in quite different >> outcomes, though they are all small. I can get positive frequency for one >> displacement and negative frequency for another displacement. I think >> basically the modes eigenvalues are not converged to any value even in such >> a small displacement range. Does it mean that the phonon potential is not in >> a good harmonic regime, or unstable, so the quasiharmonic model fails? Would >> DFPT approach solve this problem? >> >> Thanks, >> Tian >> >> >> On Wed, May 15, 2013 at 3:08 AM, Aldo Humberto Romero >> wrote: >>> >>> I completely agree with Togo. Could you check your k-mesh convergence of >>> your gamma point? >>> cutoff energy is also converged?... just to be sure... this is unusual and >>> they are some suspects >>> but let rules out the simplest ones... some PAW data is not also good >>> enough.. hopefully it is >>> only Kpoint.. I would do the DFPT instead of the frozen ion approach.. >>> >>> Good luck. >>> >>> >>> -- >>> >>> Prof. Aldo Humberto Romero >>> CINVESTAV-Unidad 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 >> >> >> >> >> -- >> Lan, Tian >> Ph.D. Candidate, Department of Applied Physics and Materials Science >> California Institute of Technology, >> Caltech M/C 138-78, Pasadena, CA, 91125 > > > > -- > Atsushi Togo > http://atztogo.github.com/ > atz.togo@... -- Atsushi Togo http://atztogo.github.com/ atz.togo@... ```