Cannot generate mesh.yaml

[liuyuhang599]

Cannot generate mesh.yaml

[JMSkelton]

With the --hdf5 flag, newer versions of Phonopy write mesh.hdf5 in place of mesh.yaml.

The code has been updated to read mesh.hdf5, if present, and the a-SiO2 example has been updated to reflect this.

Could you check this fix works for you, and if so I'll close the issue?

[JMSkelton]

Making sure the comments on commit 8fbef70 get added to the issue.

---

Oops -- forgot to include the updated Constants module in the commit. Please check whether the latest push works for you.

---

On 10 April 2018 at 15:41, Giuncan [email protected] wrote:

I tried to run phonopy-ir but ZeroTolerance is not defined in Constants file

[liuyuhang599]

Thank you. I have slove this by using --readfc-format to generate mesh.yaml. Jonathan Skelton <[email protected]>于2018年4月10日 周二02:50写道：

…

With the --hdf5 flag, newer versions of Phonopy write mesh.hdf5 in place of mesh.yaml. The code has been updated to read mesh.hdf5, if present, and the a-SiO2 example has been updated to reflect this. Could you check this fix works for you, and if so I'll close the issue? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Acmlt7mcXtp_blSnFEg23zILKiSTBRrXks5tnIBwgaJpZM4TMENK> .

[JMSkelton]

Tagging additional comments on 8fbef70 and closing issue.

---

OK - glad to hear it.

No problem. I'll add this conversation to the issue and close it.

---

On 11 April 2018 at 09:48, Giuncan [email protected] wrote:

thanks, now it works
(Do I have to delete the comment?)

[liuyuhang599]

Dear Jonathan, I have noticed that the relative Raman intensities have discrepancies with experimental measurements in bands with low-frequency. Could you tell what is the possible reason for that? I’m calculating Raman spectra using Phonopy-Spectroscopy, it will be better if I could explain the discrepancies in my paper. Best, Yuhang

[JMSkelton]

Hi Yuhang, I presume you're referring to your own calculations rather than the SiO2 example? The first question to ask is probably "how large is the discrepancy"? If the correspondence between the experiments and theory is sufficient to assign the bands, I'd personally say the agreement is good enough. I can think of a number of potential reasons why the two may disagree quantitatively: 1. Problems during your calculation -- if you're using DFPT to calculate the dielectric tensors, make sure the initial SCF converges properly, as I've found in the past that if it doesn't it can lead to unphysical intensities. 2. Errors in the calculation of the dielectric tensor with your chosen DFT functional -- the dielectric constant depends on the bandgap, which is underestimated with DFT; although I think taking the derivative should, to some extent, cancel this, artefacts due to this problem may still be an issue. 3. Numerical errors in the calculation of the intensities -- you could test selected bands with a smaller/larger step to check this isn't the problem. 4. Errors/issues in the calculation of the phonon frequencies - bands that should/shouldn't overlap don't/do, giving a different intensity distribution; this could happen if your calculation was done with the relaxed (0 K) cell volume, but your measurements were done at room temperature. 5. Not including calculated linewidths (if you didn't), or errors in those calculations (if you did). 6. Instrumental artefacts in the experimental spectra -- these are entirely possible, and depend on how good the Raman instrument used to take the measurements was; in particular, you sometimes see artefacts in bands close to the laser line. I hope this is helpful, and I'd be happy to elaborate on any of these points if they're not clear (I don't know how much of a Raman expert you are - you may well know a lot more than I do!). Also, if you want to send me your results, I might be able to make some more specific comments. Best wishes, Jonathan

…

On 17 May 2018 at 08:14, liuyuhang599 ***@***.***> wrote: Dear Jonathan, I have noticed that the relative Raman intensities have discrepancies with experimental measurements in bands with low-frequency. Could you tell what is the possible reason for that? I’m calculating Raman spectra using Phonopy-Spectroscopy, it will be better if I could explain the discrepancies in my paper. Best, Yuhang — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AISuFD-VOnNXBoCuLAmsQICko43_TkMKks5tzSNQgaJpZM4TMENK> .

[JMSkelton]

Hi again, Three more possibilities: 7. If your material isn't cubic, the Raman tensor won't be isotropic, and unless the experimental spectra were measured on a powder sample with no preferred orientations you may see differences in intensities due to polarisation. 8. If the Raman laser wavelength is close to an electronic state, you may see some resonance, which isn't accounted for in the method Phonopy-Spectroscopy uses to calculate the intensities. An IR laser (typically 785 nm or 1064 nm for an FT-Raman instrument) are usually the best to avoid such things. 9. You may be seeing overtones and/or combination bands in the experimental spectra, which would affect your intensity distribution; however, this is a lot less likely for Raman than for IR, as Raman scattering is a much weaker phenomenon. As I said - if the agreement is good enough to assign the bands, I think you can make a convincing argument in your publication. Best wishes, Jonathan

…

On 17 May 2018 at 08:39, Jonathan Skelton ***@***.***> wrote: Hi Yuhang, I presume you're referring to your own calculations rather than the SiO2 example? The first question to ask is probably "how large is the discrepancy"? If the correspondence between the experiments and theory is sufficient to assign the bands, I'd personally say the agreement is good enough. I can think of a number of potential reasons why the two may disagree quantitatively: 1. Problems during your calculation -- if you're using DFPT to calculate the dielectric tensors, make sure the initial SCF converges properly, as I've found in the past that if it doesn't it can lead to unphysical intensities. 2. Errors in the calculation of the dielectric tensor with your chosen DFT functional -- the dielectric constant depends on the bandgap, which is underestimated with DFT; although I think taking the derivative should, to some extent, cancel this, artefacts due to this problem may still be an issue. 3. Numerical errors in the calculation of the intensities -- you could test selected bands with a smaller/larger step to check this isn't the problem. 4. Errors/issues in the calculation of the phonon frequencies - bands that should/shouldn't overlap don't/do, giving a different intensity distribution; this could happen if your calculation was done with the relaxed (0 K) cell volume, but your measurements were done at room temperature. 5. Not including calculated linewidths (if you didn't), or errors in those calculations (if you did). 6. Instrumental artefacts in the experimental spectra -- these are entirely possible, and depend on how good the Raman instrument used to take the measurements was; in particular, you sometimes see artefacts in bands close to the laser line. I hope this is helpful, and I'd be happy to elaborate on any of these points if they're not clear (I don't know how much of a Raman expert you are - you may well know a lot more than I do!). Also, if you want to send me your results, I might be able to make some more specific comments. Best wishes, Jonathan On 17 May 2018 at 08:14, liuyuhang599 ***@***.***> wrote: > Dear Jonathan, > > I have noticed that the relative Raman intensities have discrepancies > with experimental measurements in bands with low-frequency. Could you tell > what is the possible reason for that? I’m calculating Raman spectra using > Phonopy-Spectroscopy, it will be better if I could explain the > discrepancies in my paper. > > Best, > Yuhang > > — > You are receiving this because you modified the open/close state. > Reply to this email directly, view it on GitHub > <#1 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AISuFD-VOnNXBoCuLAmsQICko43_TkMKks5tzSNQgaJpZM4TMENK> > . >

[liuyuhang599]

Dear Jonathan, I am trying to use HSE06 functional to calculate the Raman spectrum of VO2Cl molecule by using Phonopy-Spectriscopy, but the calculated Raman tensors of several peaks are null matrix. Do you know any possible reasons for that error? Some useful files are attached. Best, Yuhang

[JMSkelton]

Hi Yuhang,

Did you mean to attach something? I don't see any files with your last message.

When you say "VO2Cl molecule", I presume you mean an isolated (gas-phase) molecule in a vacuum box? I presume also that you expect the bands with zero intensity to be Raman active?

I've never actually tested Raman spectra on a molecule-in-a-box model (IR seems to work reasonably well), so I'd have to see your output files.

Best wishes,

Jonathan

[liuyuhang599]

I calculated a VO2Cl molecule in vacuum, I could get non-zero intensities by using PBE but only get zero intensities by HSE06. The output files of PBE and HSE06 are attached.

…

On Thu, Jul 12, 2018 at 12:08 AM, Jonathan Skelton ***@***.*** > wrote: Hi Yuhang, Did you mean to attach something? I don't see any files with your last message. When you say "VO2Cl molecule", I presume you mean an isolated (gas-phase) molecule in a vacuum box? I presume also that you expect the bands with zero intensity to be Raman active? I've never actually tested Raman spectra on a molecule-in-a-box model (IR seems to work reasonably well), so I'd have to see your output files. Best wishes, Jonathan — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Acmlt6OCsovQaRGdIXgbUqkYRzi9xHaJks5uFvXdgaJpZM4TMENK> .

[JMSkelton]

OK -- this is weird, but I still can't see the attachments. Can you message me directly: j.m.skelton[at]bath[dot]ac[dot]uk?

Jonathan

[JMSkelton]

Hi Yuhang,

Thanks for sending the files -- and apologies for my apparent inability to use the GitHub mailing system properly (!).

I believe I've found the problem. All the calculated dielectric tensors in your HSE 06 Raman.yaml are the same (i.e. all displacements, for all the bands you calculated).

I can think of three possible causes:

(1) Your band selection (--bands) for the HSE 06 calculation somehow triggered a bug that causes phonopy-raman to write out the same structure repeatedly -- the easiest way to check would be to open the Raman-POSCAR-*.vasp files in a text editor and compare them.

(2) You (accidentally) copied the wrong POSCAR file when you did your HSE 06 calculations -- if #1 checks out, perhaps check your scripts, etc. to make sure this isn't the issue.

(3) You (or phonopy-raman) accidentally read in the same OUTCAR file for all the modes -- I think this is unlikely, but if neither #1 or #2 apply then it's worth investigating.

Can you take a look and let me know what you find?

Best wishes,

Jonathan