The non-Hamiltonian of seedname_r.dat

[LiuyichenYanwushang]

When I used wannier90_r.dat and wannier90_hr.dat to calculate transport, I noticed that _r.dat is not Hermitian, so it will bring some wrong results. When I force it to be Hermitian, its The results are also inconsistent with the results of wannier90.

I noticed that the reason is that when calculating AA_R in postw90, the S matrix is ​​hermitized and then Fourier transformed. However, when writing seedname_r.dat, there is a lack of a hermitization step, so the calculation results differ.

Due to my ability, I have no way to correct this problem yet. I would like to know if it can first convert into hermit when writing seedname_r.dat? This way we can use the results of wannier90 for the next step of calculation.

[stepan-tsirkin]

As far as I remember, there is some more inconsistency between the calculations using postw90, and those using _r.dat file.

In particular, the minimal distance replica selection ("use_ws_distance=True") is not reflected in the _r.dat file.

Alternatively, you may look int using WannierBerri to read the chk ile and the write _tb file (which combines _hr and )r)

[matthew7879]

The wannier90_r.dat gives the position operator matrix elements of r_x, r_y and r_z, but it must satisfy the relation:

( [ r_{\alpha} (a,b,c) ]{m,n} ) * = [ r{\alpha}(-a,-b,-c) ]_{n,m} where * is conjugate.

I must use these position operator matrix elements to construct the non-abelian Berry connection matrix. These Berry connection matrices must be Hermitian. But as the position operators matrix elements do not satisfy the above relation, that is if r_{\alpha} (a,b,c) ]{m,n} * != r{\alpha} (-a,-b,-c)]_{n,m} the Berry connection matrix will not be a Hermitian matrix. Has anyone found a way to resolve this issue or a workaround?