Hi,

I don't know where U R struck. There are two ways to calculate the phonon frequencies. Here, I discussed about Density Functional Perturbation Theory (DFPT) using Phonopy code. It's already available in the website

https://atztogo.github.io/phonopy/vasp-dfpt.html#vasp-dfpt-interface

Please make sure about that U installed the Phonopy code in clean. Create one folder and make the folder path in Ur terminal. Keep Ur POSCAR in the folder by name of POSCAR-unitcell

Step 1

First we are in need to create a super cell of our system. Lattice constant of the super cell will be around 1-1.5 nm minimum. So create super cell like 3x3x3 or 6x6x6 depends upon Ur system using the comment

phonopy -d --dim="2 2 2" -c POSCAR-unitcell

(222 is a supercell size, POSCAR-unitcell - U want to rename the optimized POSCAR as POSCAR-unitcell (because it's a POSCAR of unitcell)

Now one new SPOSCAR file was generated in the folder. Rename the SPOSCAR to POSCAR.

Step 2

Take the POSCAR create POTCAR corresponds to it. If need reduce the KPOINTS. Run VASP calculation using the following INCAR

PREC = Accurate ENCUT = 500 IBRION = 8 EDIFF = 1.0e-08 IALGO = 38 ISMEAR = 0; SIGMA = 0.1 LREAL = .FALSE. ADDGRID = .TRUE. LWAVE = .FALSE. LCHARG = .FALSE.

Step 3

After the calculation successfully ended copy the vasprun.xml file in the same folder, where U keep all files.

Create the force constant by (Before that make the folder path in Ur terminal)

phonopy --fc vasprun.xml

Step 4

Copy the band.conf file which is available in the Phonopy-Examples-Phonon folder (or search band.conf in folder of Phonopy where U installed in Ur computer)

Check the line FORCE_CONSTANTS = READ in band.conf is enabled or otherwise U just included the line in band.conf. Now run the comment

phonopy --dim="2 2 2" -c POSCAR-unitcell band.conf.

U will get the Phonon spectrum.

Thank you so much Sir.

Now it's very much clear. I will try with it.

Selvarajan's answer was perfect. I just want to add few points.

If you want to use DFPT:

1) Before starting using Phonopy, please make sure the you have optimized your system very accurately. For instance, add these two flags to your optimization INCAR: EDIFF = 1.0e-08, EDIFFG = -1.0e-8

2) make sure that the ENCUT in the INCAR (both OPT and Phonopy) is at least 30% higher than ENMAX for each element in the POTCAR.

3) In Phonopy you have just ONE iteration, it means there is no optimization, Phonopy is just calculating the force on each atom. So, add NSW=1 to your Phonopy INCAR.

4) Use IBRION = 7 or 8 in your Phonopy's INCAR (see the VASP manual).

5) Adjust Kpoints. It means if you are making a supercell and increasing the size of the unit cell by 2*2*2, you need less kpoints than the kpoints that you have used in the optimization step (reciprocal lattice concept)

6) if you need thermodynamic properties, you can use:

phonopy -s -t -c POSCAR-unitcell thermo.conf > THERMO.data

7) At the beginning, when you use the below command to creat the supercell:

phonopy -d --dim="2 2 2" -c POSCAR-unitcell (you can change the size of the supercell, I just put "2 2 2" as an example)

you will get bunch of files like ...-00* . If you want to use DFPT, you caln delete these files. They can be used for "Frozen phonon" method.

8) To review the DFPT theory you can follow Stefano Baroni's papers in this matter.

Good luck!

Thank you so much Ma'am for valuable information.

Your very welcome!

Did U find the solution for that?

If not copy band.conf file in Ur home directory. Set all the parameters such as DIM and band axis (Reciprocal path like Band structure KPOINTS).

Remove all the parameters in Primitive cell and run it. I hope it will help you.

Sorry. I am not sure about that. However, it's not at all primitive axis settings. In the phonon calculation switch off the symmetry using ISYM tag (Beware that it's take long time to run). After that take the supercell as primitive cell and calculate the force constant using DIM=1 1 1. Now, U can get the supercell frequencies, which indirectly means super cell as a primitive cell.

Hope it will work.

Can u tell me what is your POSCAR and supercell POSCAR .have u reproduce anyone results?

I like to help you. But the information you given is not enough to find out what's wrong with your inputs. Please give some information.

Is there any reason why one needs a supercell for a phonon dispersion calculation? I see no reason why the unit cell wouldn't be appropriate. Thanks

Can someone help me understand the difference between "frozen phonon" and dfpt techniques, also why use POSCAR-*** for the former and SPOSCAR for the latter?

(Sorry, couldn't direct the question at anyone in specific since the "mention" feature isn't working properly for me)

Hello Hitanshu Sachania ,

I saw your personal message also. We can calculate phonon of a system using relaxed primitive structure of any system. However, in general we need large supercells to accurately calculate the force constant. As per our need, we can generate 2 x 2 x 1 or 3 x 3 x 3 supercell.

In 2 x 2 x 1 supercell,

SPOSCAR file contains exact 2 x 2 x 1 structure, whereas

POSCAR_001 might contain 2 x 1 x 1 structure.

POSCAR_002 might contain 1 x 2 x 1 structure likewise it should have four displaced POSCAR files for 2 x 2 x 1.

Now the Frozen calculation,

In frozen calculation, you have to calculate force constant of each POSCAR (POSCAR-001,etc) for your supercell without relaxing the structure. Then using phonopy interface you can calculate phonon spectra of your system. Its called Frozen phonon calculation.

In Density functional perturbation theory (DFPT) you don't need to worry about generated, POSCAR displacement files. Because its deals with exact supercell i.e SPOSCAR. You have to do DFPT calculation in VASP using defined INCAR with the generated SPOSCAR as POSCAR to get the Phonon frequencies.

Hope it might give some idea. Please let me know for further clarifications.

Hello Premkumar Selvarajan ,

Thank you for such a speedy reply. How did you calculate there being 4 displaced files for a 2x2x1 cell? Do these displacements depend on the symmetry of the cell and the type of atoms?

No its depends upon how much supercell you are creating.

Hello Premkumar Selvarajan, both of these techniques (frozen phonon and dfpt) give us thermal properties. How do we choose which to use when?

Hi Hitanshu, both should be numerically equivalent. I generally like using DFPT because of the smaller number of input files that one needs to deal with and other advantages on the level of theory.

Hello,

Can anyone please tell me how to do phonon calculations for various temperatures ???

can anyone give me a link to install VASP to calculate phonon frequency

VASP by itself does not do it, you need another software like phonopy.

Chinedu Ekuma That makes sense. I got 600+ displacements which would make the calculation too expensive. Thank you.

Hello all, why do we use a bigger cell for phonopy calculations (DIM tag)?

Hi,

Large supercells (relatively) be

used for phonon calculations to ensure that the force constants between distant atoms go to zero.

Hello Everybody,

Thank you very much for your advices. I have a few questions other than what you have answered so far:

1- If I want to calculate the band structure (BS) of phonon frequencies along a specific K-path, not an automatic K-mesh but a high symmetry path, should I use that high symmetry K-path as my KPOINTS file in VASP or should I do the VASP calculation using a general K-path like 5*5*5 and then use the high symmetry K-path later in phonopy? If in phonopy, how can I do it?

2- I have done DFPT calculation using the abovementioned INCAR file and now I am going to use phonopy, but I get the following error:

" "band.conf" was not found "

I know that I should have band.conf file in my home directory, but I cannot find it, so is there any ways to download it and put it there?

3- I used spin-orbit-coupling for my DFT calculations, should I still use it for phonon frequency calculations?

4- I haven't got any data from phonopy yet, and this question might seem a little stupid: How can I get raw data (numbers) from phonopy such that I can draw BS later using other softwares like Origin?

5- When drawing the BS in phonopy, how can I assign labels to specific k-points, e.g. Gamma or X to a point?

Thanks for your replies in advance.

Fazel

Hi Fazel,

1. You have to do vasp calculations with kmesh, i.e., 5*5*5 (say) then have to use high symmetry k-path in phonopy, which you have to put in band.conf file. I am attaching band.conf file for your reference.

In first line, DIM = you have to mention the size of supercell which you made using Phonopy. Here i am using 2 2 2 as i made supercell of 2*2*2 initially.

In second line, you have to mention the high symmetry points, which you were asking.

2. I have attached band.conf, just edit it according to your system.

3. I didn't do it, so i can't say anything about this point.

4. phonopy-bandplot --gnuplot band.yaml >1.txt, then you will have data in numbers in 1.txt file.

5. If you use band.conf as the one i attached, you will get the label of high symmetry points in band structure itself. You can plot it by phonopy-bandplot band.yaml

I hope it may help.

With Regards,

Payal

Hi,

I am trying to do the phonon calculation for a 2D material. I have done the VASP calculations with 2X2X1 (12 atoms) and 4X4X1 (48 atoms) supercell but in both cases, a large no of frequencies (12 and 48 respectively) are coming imaginary (f/i written before the frequency values). After that when I was plotting the phonon band structure there are imaginary frequencies. Is there any link between the two and how to overcome it?

Thank you very much Payal.

It worked very well.

Now, I have one more question. I am trying to get the thermal properties of my material. Would you please give me some clue how to do it? Since whatever I have found so far did not work. Thanks again.

Fazel

Dear Fazel,

Sorry, but i haven't tried the Thermal properties, so i can't help you in this regard.

Maybe someone else who are followers of this question, may help you out.

Regards,

Payal

Dear Narges Masoumi

Thanks for the information you provided above. I am going to extract the thermal properties of my 2D material and I am using the following code:

phonopy-qha -p --tmax=980 --efe fe-v.dat e-v.dat thermal_properties.yaml

But once I use it, I get the following error:

" No such file or directory: 'e-v.dat' "

What should I do with this error?

Thanks

Fazel Baniasadi

Dear Dwaipayan Chakraborty

I think the attached pdf will help you with this issue.

Bests,

Fazel Baniasadi

Fazel Baniasadi Have you created the thermo.conf file? I have messaged a sample one to you.

Dear Narges Masoumi

I couldn't use it because I'm stuck in a step of my calculations and I need to get this part done first. I was wondering maybe you can help me in this regard. Thanks in advance.

I am a little confused with what I should do for Raman and IR vibrational modes calculation.

Actually, I want to calculate Raman and IR active modes and assign them to their corresponding irreducible representations , so I need the born effective charges, right?

I calculated the born effective charges; however, I don't know how to extract them from my vasprun.xml file. Should I use phonopy or phono3py? Do you know the script for doing so?

Also, if I need to install phono3py, how should I do it? since it seems it's a different package from phonopy.

I would appreciate it if you can help me for this matter.

Regards,

Fazel Baniasadi

Hello everyone,

Phonon calculations (Phonon Band Structures, Phonon DOS, and thermal properties) in materials science using VASP and phonopy are explained on the Youtube Channel. Please find the link below:

https://bit.ly/2UtvbHE

https://www.youtube.com/watch?v=FX7WjL074g4

Best regards,

Rasoul Khaledialidusti

Dear Fazel Baniasadi ,

You can calculate phonon dispersion curve with using phonopy and VASP with following instructions:

1.You can take your studying material unitcell from material project : https://materialsproject.org/

for example in my case of study ( monolayer ZrS2) i use this:

https://materialsproject.org/materials/mp-1186/

2. you need to relax this unitcell (POSCAR)

-First you should optimize cutoff (ENCUT) my case was 450 meV(attached ECut file)

-second you should optimize kpoints (my case was 7x7x1) (attached Kpoint file)

-third you can relax the structure using atteched INCAR file.

3.to obtain force constant we need the displacments of supercell:

use phonopy -d --dim="N N N" for NxNxN supercell size ( i used 6 6 1)

this step phonopy generates few SPOSCAR files with name of POSCAR-001... POSCAR-00N based on your crystal symmetry and size.(For ZrS2 generates 6 POSCARS)

4. you need to move each POSCAR-00X files into a disp-X file with new name of

POSCAR . then use this INCAR:

IBRION = -1 ENCUT = 450 #Optimized value EDIFF = 1.0e-08 ISMEAR = 0; SIGMA = 0.01 IALGO = 38 LREAL = .FALSE. LWAVE = .FALSE. LCHARG = .FALSE.

and then use VASP to calculate Forces... .

5.When your VASP calculation finished, use phonopy to extract forces from vasprun.xml

phonopy -f disp-{001..006}/vasprun.xml

So you would have FORCE_SET file.

6.now in order to plot Band structures we need to create band.conf file with key words:

ATOM_NAME = Zr S #Your system

DIM = 6 6 1 #SUPERCELL size in step 3

PRIMITIVE_AXIS = 1 0 0 0 1 0 0 0 1

BAND = 0.0 0.0 0.0 0.0 0.5 0.0 0.333333333333 0.333333333333 0.0 0.0 0.0 0.0 #full symmetry point

BAND_LABELS = Gamma M K Gamma

MP= 50 50 1

7. now there is a question, how we should take BAND and BAND_LABELS?

you can use this paper to use full symmetry point. (Article Setyawan, W. & Curtarolo, S. High-throughput electronic band...

for example in my case of study : crystal path was : G –M–K– G –A–L–H–A|L–M|K–H

8. in the last step use phonopy -p band.conf and Band structure will be plotted.(attached Figure)

Hi Premkumar Selvarajan , Nima Ghafari Cherati , Rasoul Kh , Narges Masoumi , and all,

similar to Dwaipayan Chakraborty 's situation. I tried many times for a 2D system. and always ended-up with negative frequencies. But the same material when done Phonon with Quantum Espresso (not linked to Phonopy), there were no such negative frequencies. Now the problem is that QE work done by a teammate long back and not known how was it done.

attaching both figures. Can anyone could suggest what could went wrong with Phonopy/VASP combination? i could generate Phonon -DOS, Thermal properties with Phonopy commands, but not this spectrum.

Dear SK M ,

How did you calculate the phonon band via QE?

For Phonopy&VASP what was your method? DFPT or finite difference method?

before start Phonopy, did you optimize the structure?

what was the accuracy of optimizing calculation?

last but not least, what was your supercell size? supercell size is quite important.

I'll be waiting for your answers.

Best,

Nima

Dear Nima Ghafari Cherati

i am answering to each of your question below, copied your query as well:

How did you calculate the phonon band via QE? The calculation for the picture i attached was not done by me, but by a teammate many years back, so no clue on this.

For Phonopy&VASP what was your method? DFPT or finite difference method? i followed https://phonopy.github.io/phonopy/vasp-dfpt.html#vasp-dfpt-interface.

before start Phonopy, did you optimize the structure? yes

what was the accuracy of optimizing calculation? EDIFF (1e08) & EDIFFG (-0.005)

last but not least, what was your supercell size? supercell size is quite important.

15.0399999619999996 0.0000000000000000 0.0000000000000000

0.0000000000000000 10.3970003127999995 0.0000000000000000

0.0000000000000000 0.0000000000000000 21.0000000000000000

I'll be waiting for your answers.

Best,

Nima

Nima Ghafari Cherati further to inform. I did another run with EDIFF (1e08) & EDIFFG (-0.0005) and ENCUT 550.

there are 72 POSCAR files generated and i generated Hessian data . and each run took about 7.5 hrs x 72 POSCARs. so much time it took. even then i got finally negative frequency for Phonon spectrum. but this is drawn through vasprun.xml files. it also generated DYNMAT files. Can we use these files?

Hi all and Nima Ghafari Cherati

finally i could get the phonon spectrum.

Nima Ghafari Cherati i got the spectrum with the very slight negative frequencies at Gamma point (almost similar to your graph posted above). Has anyone applied Acoustic Sum Rule? can anyone help me how to apply the same?

Regards

Hi SK M , In VASP+DFPT (finite displacement method) based calculating phonon spectrum generating normally more the 50 POSCAR file (there are 72 POSCAR files generated and i generated Hessian data . and each run took about 7.5 hrs x 72 POSCARs) if it's optimized means it takes more time. you have tried any alternate method or using the same method. Which method do you get the phonon spectrum?

Mahendiran Durairaj are you asking me to know to apply for your calculations? or to help me out and you want more information? lol...

in either case, let me admit, i was by mistake messing up few finer points because of which i did not get good results or wasted long calculation times, i suppose.

Anyway, i have done VASP with super-cell method with Finite Displacement Method. DFPT is different method.

i think the following changes helped me for my system of 2D material with 3 types of atoms.

used Phonopy to make supercell and then post processing for drawing Phonon spectrum.

Hope that helps you.

Dear SK M

In the setting section you can find force symmetry tag named FC_SYMMETRY.(https://phonopy.github.io/phonopy/setting-tags.html#symmetry)

FC_SYMMETRY imposes acoustic sum rule and force constant symmetry of exchanging a couple of atom indices repeatedly.

You can use it to apply acoustic sum rule.

Best regards,

Nima

dear Nima Ghafari Cherati

thanks for the reply.

but where to use the FC_SYMMETRY tag?

i guess it can be mentioned in band.conf file. So i put the tag but no effect.

i think let me tell you i used VASP for optimizing the cell. then used Phonopy for super-cell creating and thus get displacement POSCAR files. Then all these files again run for single point calculation to get vasprun.xml files. then used phonopy code to get FORCE_SETS, then run phonopy command for band graph.

in this process where can use this FC_SYMMETRY?

regards

Dear SK M ,

Your process is completely correct. The FC_SYMMETRY tag can be used in band.conf file:

FC_SYMMETRY = .TRUE.

and then use phonopy -p band.conf .

You can check its effect on the modes by comparing the old band.yaml and new band.yaml. If you check the frequency section you can see that numbers changed. However, It might be a very small change.

Bests,

Nima

Nima Ghafari Cherati thank you.

there is no change absolutely (comparing the final spectrum). there might be very finer number changes in the file, but nothing visible.

I am stuck at that stage now. no other option to cure this data, it seems.

Tell me in case any other option or method we can try or check this.

Regards

Dear SK M ,

If you need to remove the imaginary modes, I recommend you to try higher dimension for supercell and relax the structure with higher accuracy.

If the number of generated POSCAR files was too much, try hiphive code.

hiphive is an effective tool for extracting high-order force constant.

https://hiphive.materialsmodeling.org/

DOI:10.1002/adts.201800184

Bests,

Nima

Dear SK M ,

If you need to remove the imaginary modes, I recommend you to try higher dimension for supercell and relax the structure with higher accuracy.

If the number of generated POSCAR files was too much, try hiphive code.

hiphive is an effective tool for extracting high-order force constant.

https://hiphive.materialsmodeling.org/

DOI:10.1002/adts.201800184

Bests,

Nima

Dear Nima Ghafari Cherati

Could you please tell the steps to calculate phonon dispersion using hiphive package using DFT ?

Dear All,

I do not know the reason, However, by following the above procedure I am unable to get the phonon spectrum at the end. Although the calculations are completed without any error. Kindly need guidance from experts on how to get the final plot.

Dear Experts,

I am facing issue for band.conf.

I have attached error file and band.conf. Please find the attachment.

Your suggestion will be very helpful.

Dear Shivam Kansara,

Have you ever plotted a band structure with another POSCAR and FORCE CONSTANTS?

First, by using phonopy in the example folder, make sure that you installed your phonopy correctly. If it is working, then check your POSCAR file.

Could you share your POSCAR file here?

It would be better if you tell me how you did your calculation?

Did you follow the instruction that I explained above?

Best,

Nima

Dear Nima Ghafari Cherati yes I checked my installation by doing calculation of https://github.com/phonopy/phonopy/tree/master/example/NaCl-VASPdfpt. and I get the graph.

I am attaching POSCAR file of my system.

I am following http://phonopy.github.io/phonopy/vasp-dfpt.html

Dear Shivam Kansara , i assume you have followed the procedure correctly. I think its syntax error in band.conf file as follows,

ValueError: invalid literal for int() with base 10: '#SUPERCELL'

Delete the comment lines from DIM and BAND rows (Keep only numbers) and give a try.

Regards,

Thank you Swapnil Deshpande and Nima Ghafari Cherati , I solved my problem by removing # from .conf.

Dear all,

I completed successfully all steps till phonopy --dim="2 2 2" -c POSCAR-unitcell band.conf

now please help me to plot and save graphs e.g. phonon dispersion curve, POS and combine graph of Phonon and dos.

Thank you

Dear @Shivam Kansara , Auxiliary tools can be used for that purpose.

Hi Shivam Kansara

Use this command

phonopy-bandplot --gnuplot band.yaml >band.txt

then you will have your data in band.txt file, which you can plot in any plotting software.

Thank you Payal Wadhwa Swapnil Deshpande

Nowadays, I am trying phonon using phonopy-vasp of alpha-SiO2. Reported phonon graph has no imaginary frequency but I am getting a single phonon line in imaginary part. I did very precise calculation but even I am getting the same result.

A list of attempts:

KPOINTS optimization

ENCUT optimization

Change Pseudopotential

Various super cell (222, 333, 223, 332)

I would like share latest file with GGA pp here, please have a look.

Thanking you

Try with IBRION = 6 once, whether it brings some change or not. Also try to skip NPAR and KPAR flag, sometimes it also creates problem, i have observed it many times.

Dear Payal Wadhwa I tried the same but no changes. I got the similar graph.

Thanks for your input.

I used the INCAR below, using a supercell replicated by phonopy, and I got very good agreement with experimental results. See here:

https://iopscience.iop.org/article/10.1088/1361-648X/ac4b2a/pdf

This is how I prepared the files using Phonopy:

----------------------------------------------------------------------------------

% phonopy -d --dim="4 4 4" -c POSCAR-unitcell

% #Run the VASP calculation

% phonopy --fc vasprun.xml # band.conf

% #Run phonopy from the command line:

%. phonopy -p -c POSCAR_unitcell_phonopy band.conf > out.phonopy

----------------------------------------------------------------------------------

############

# INCAR #

############

System = Be hcp

NWRITE = 1

#Startparameter for this Run (values of parameters are equalt to default):

ISTART = 0

ICHARG = 2

LWAVE = .FALSE.

LCHARG = .FALSE.

ENCUT = 1000

EDIFF = 1E-8

ADDGRID = .TRUE.

IALGO = 38

LREAL = .FALSE. #Evaluate projection operators in real space

NELMIN = 6

BMIX = 2.0

MAXMIX = 50

IBRION = 6 # PHONONS: DF-PT

ISYM = 1 # symmetry: 0 for no symmetry

NSW = 1 # CHANGE : number of MD steps

NBLOCK = 1

KBLOCK = 1

ISIF = 2 # 2: V= no, Pos= no, Rest= yes

#SMASS = 0.0

#POTIM = 0.5 # CHANGE : timestp size

#TEBEG = 5000

ISMEAR = -1

NBANDS = 400 # For 14*144 = 2016 electrons, and divisible by 96 (paralellization)

SIGMA = 0.025851991011651636 # 300 K

LPLANE = .TRUE.

#NCORE = 8

KPAR = 2

@Premkumar-Selvarajan can you please tell the way to calculate the group velocity and eigenvectors of a mode.

Dear Experts,

I am facing issue for band.conf.

I have attached error file and band.conf. Please find the attachment.

Your suggestion will be very helpful.

Renfei Cheng what is the the actual problem? its not clear. Can you elaborate? how the conf, file giving you the error? i cant find anything related to .conf file in error file you sent, (as per my limited knowledge)

still i try whatever i know.

Dear Renfei Cheng

It seems that your phonopy has problem with loading the libraries. Please install it over again. Did you use pip to install it? The best way is using conda.

Please make sure you installed all the prerequsitions then just run it.

prerequsitions:

• Python (>=3.7)
• numpy (>=1.15)
• matplotlib (>=2.2.2)
• python-yaml (pyyaml)
• python-h5py (h5py)
• spglib

Bests,

Nima