Force constants

[jk262626]

Hi Nakib,
I used a q-grid of 7x7x5 for the phonon calculation in QE (which uses density functional perturbation theory), which is then fed to EPW to get the el-ph coupling. So I used 7 7 5 for coarse_qmesh in the input file.

I used 3rd-order force constants obtained from hiPhive, where atoms are displaced in a 4x4x3 supercell, and 2nd-order force constants from QE with 7x7x5 supercell. Would this be an any issue?

[nakib]

Hi Jiyoung,

coarse_qmesh is related to Wannierized quantities. This is related to 2nd order force constants only if you have use_Wannier_ifc2s = .t.. (Recall from the theory that the real-space representation of the dynamical matrix is nothing but the 2nd order force constants). However, in elphbolt, the default is use_Wannier_ifc2s = .f., in which case the 2nd order force constants from espresso.ifc2 will be used. In that case coarse_qmesh is not really relevant since the parser will read the necessary info directly from espresso.ifc2. Internally this quantity is called scell. See the following code snippet below from phonon.f90.

type, extends(particle) :: phonon
     !! Data and procedures related to phonons.
     
     character(len = 2) :: prefix = 'ph'
     !! Prefix idenitfying particle type.
     integer(i64) :: scell(3)
     !! q-mesh used in DFPT or, equivalently, supercell used in finite displencement
     !! method for calculating the 2nd order force constants.
     ...

And the quantity is set by the private procedure read_ifc2 bound to the phonon data type.

Regarding the 2nd question: if hiPhive's 3rd order force constants file's format is identical to thirdorder's, then I don't foresee any problems. Again, the relevant information related to the real space is read directly by elphbolt. See subroutine read_ifc3 of data type phonon.

I hope this answers all of your questions.

Best,
Nakib

[jk262626]

Thank you so much! That's good to know.
So it doesn't matter even if I use different supercell size for 2nd and 3rd order fc, right?

[nakib]

Yes, that's correct.

…

On Wed, Jul 17, 2024, 17:59 Jiyoung Kim ***@***.***> wrote: Thank you so much! That's good to know. So it doesn't matter even if I use different supercell size for 2nd and 3rd order fc, right? — Reply to this email directly, view it on GitHub <#146 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAN7Y2LJRS2Z5VZMNWRZDJ3ZM2IHVAVCNFSM6AAAAABK7IE6SGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMBXGU2DKMQ> . You are receiving this because you commented.Message ID: ***@***.***>

[jk262626]

Nakib, I want to make sure of more things.

1. The scell you mentioned would be 7x7x5 in my case. Is that the same as N_ph in the equation below when calculating dynamical matrix?
   

2. For the 3rd fc, it is used in calculating ph-ph interaction vertices. The i, j, k are the atoms' label in the supercell that I used for 3rd fcs from hiPhive or thirdorder?
   

3. I've already done long-range correction for polar material when I calculated 2nd fcs in QE. If I use polar = .true. in elphbolt as well, does that mean I am doing correction twice and it would be wrong?

4. If I want to recalculate after changing the polar = .true. to false, can I still reuse interaction properties by read_gq2, read_gk2, read_V?

[nakib]

1. The `scell` you mentioned would be 7x7x5 in my case. Is that the same as N_ph in the equation below when calculating dynamical matrix?

I think so. Simply put, there are N_ph number of primitive unit cells in the supercell used for the phonon calculation. You can check that this is the number of R_ph vectors in your 2nd order FC file. Refer to Eq. 32 and the discussion below it in https://journals.aps.org/prb/pdf/10.1103/PhysRevB.76.165108

2. For the 3rd fc, it is used in calculating ph-ph interaction vertices. The i, j, k are the atoms' label in the supercell that I used for 3rd fcs from `hiPhive` or `thirdorder`?

I am not familiar with the hiPhive code, but this is the case for the 3rd order FC file generated by thirdorder. Note that the sum over i is restricted to the central primitive unit cell in the supercell.

3. I've already done long-range correction for polar material when I calculated 2nd fcs in QE. If I use `polar = .true.` in elphbolt as well, does that mean I am doing correction twice and it would be wrong?

The typical procedure for capturing the polar correction is that the the long range part of the interaction is first removed from the 2nd order FCs that are passed to elphbolt. The long range corrections (the so called non-analytic term) will be added after generating the dynamical matrix at arbitrary q vectors. See discussion in Section 5 of https://iopscience.iop.org/article/10.1088/1361-648X/acd831/pdf. I believe there are some subtle differences related to whether you use a displaced supercell method or DFPT to generate the 2nd order FCs. This is discussed along with the reference in the paper shared here. At the moment elphbolt only supports 2nd order FCs generated with DFPT.

I hope this answers all your questions.

Best,
Nakib

[nakib]

Let me get back to you on this sometime next week.

…

On Wed, Jul 17, 2024, 19:15 Jiyoung Kim ***@***.***> wrote: Nakib, I want to make sure of more things. 1. The scell you mentioned would be 7x7x5 in my case. Is that the same as N_ph in the equation below when calculating dynamical matrix? image.png (view on web) <https://github.com/user-attachments/assets/cf189b1c-6d9b-4dbb-829c-0a1be3e31423> 2. For the 3rd fc, it is used in calculating ph-ph interaction vertices. The i, j, k are the atoms' label in the supercell that I used for 3rd fcs from hiPhive or thirdorder? 3. I've already done long-range correction for polar material when I calculated 2nd fcs in QE. If I use polar = .true. in elphbolt as well, does that mean I am doing correction twice and it would be wrong? 4. If I want to recalculate after changing the polar = .true. to false, can I still reuse interaction properties by read_gq2, read_gk2, read_V? — Reply to this email directly, view it on GitHub <#146 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAN7Y2KMLOG7JGGLXZJ2PUDZM2REXAVCNFSM6AAAAABK7IE6SGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMBXGYYDSMI> . You are receiving this because you commented.Message ID: ***@***.***>