PROBLEM WITH CONVERTED SECOND ORDER HARMONIC FORCE CONSTANTS

[vivekgusain09]

Dear Dr. Tadano,
Warm greetings. I am writing this query regarding the strange results i am getting when i am using the second order force constants (harmonic as well as temperature normalized) obtained using alamode. When computing the phonon dispersion in alamode and in phonopy seperately it gives the correct results and there are no imaginary frequencies. However, the moment I am using the second order force constants obtained using alamode to computed phonon dispersions and lattice thermal conductivty in phonopy and ShengBTE respectively, i am getting wrong phonon dispersion with imaginary frequencies and unphysically low values of lattice thermal conductivity.

The procedure i am following is as follows:

1. obtained the second order force constants using alamode in .xml format.
2. Converted the .xml file into shengbte/phonopy format using the convert_fc2.py script that has been provided with the alamode in the tools directory.
3. Command used to extract the force constants: python3 convert_fc2.py -i s222.xml -o FORCE_CONSTANTS --full

After comparing the converted force constants with the force constants obtained using phonopy i can see the values are different for the same pair of atoms. I computed the lattice thermal conductivity and phonons with phonopy obtained second order force constants and they are giving the correct and physically meaningfull results, similar to the results for lattice thermal conductivity obtained using the alamode code. I think there is some problem with the conversion which i am not able to identify.

I kindly request for your help and guidance in this regard as it is extremely important for my research. Any response is highly appreciated.
Thank you for your time and consideration.

I have uploaded both the harmonic force constant files here (one generating using phonopy DFPT and other converted from alamode .xml format)

converted_FORCE_CONSTANTS.txt
phonopy_FORCE_CONSTANTS.txt

[ttadano]

Did you check whether the supercell structures (as in SPOSCAR) are consistent between the two codes?

[vivekgusain09]

Dear sir,
Thank you for your prompt response. No, both the supercell structures are different in phonopy and alamode. Can it be the reason for the imaginary frequencies obtained in the phonon dispersions when using the alamode converted 2nd order IFCs ?
My main confusion is then when i am using the third and fourth order IFCs (generated by alamode using FC_3/4_SHENGBTE option) with these converted second order harmonic force constants in ShengBTE the lattice thermal conductivity results are coming unphysically low (0.07 W/mK at 300 K) whereas, the values obtained using the alamode code alone are consistent.
I computed the 3 phonon lattice thermal conductivity using alamode and shengbte (this time with phonopy dfpt second order force constants) and the results are both physically reasonable and consistent. Only problem is occuring when i am using the converted second order force constants. What could be the reason for this ? Could you please guide me regarding this ?

[ttadano]

All the issues seem to result from inconsistencies in the supercell atom indices between the two codes. Atom index i in one code must correspond to atom i in the other code to get meaningful results.

For example, if you prefer to use the SPOSCAR definition from phonopy, I recommend that you give the same supercell structures in &position field of ALM and calculate everything using that structure.

If you have already performed DFT calculations for various configurations to get DFSET and you want to avoid repeating that using the phonopy-generated SPOSCAR, I recommend that you make a small script that maps the atomic indices from alamode to phonopy and then arrange the orders in DFSET to make it consistent with the phonopy's structure.

[vivekgusain09]

Dear sir,
The problem has now been resolved and the cause was indeed the supercell mismatch as i computed the lattice thermal conductivity for other material in ShengBTE using the converted second order force constants along with the third and fourth order IFCs, I got the correct results.
I am still confused about the relationship of the supercell mismatch between alamode and phonopy leading to the errors in shengBTE as i am using the 2nd, 3rd and 4th order IFCs generated using the same supercell in alamode (which is different from the phonopy) to compute the lattcie thermal conductivity. Is there any relation of the conversion step of IFCs 2nd to phonopy format ?
Nevertheless, the problem has been resolved! I sincerely thank you for your time and valuable guidance.