EPW Forum

Dear all, I have two questions regarding Wannier bands and phonon dispersion

1) I am trying to calculate the electron phonon coupling for a structure. I have calculated the frequencies for 28 q points. All are positive. However when I was trying to get the phonon dispersion by interpolation (using matdyn.x) (I used 500 points), some of the points, those close to the Gamma point, had negative frequencies (maximum of -2.8 cm^(-1)). regarding the el-ph coupling , would that affect the calculation? What could be causing this? a not so good scf convergence?

2) When comparing the electronic band structure using PWSCF and WANNIER functions, I basically obatin the same form, which is good. However, the positions of the Gamma, K, M points are not the same.. In the PWSCF bands input file, I use the following positions:


K_POINTS crystal_b
4
0.000000 0.000000 0.000000 70
0.500000 0.000000 0.000000 50
0.666666 -0.333333 0.000000 85
0.000000 0.000000 0.000000 1

Plotband.x finds that positions Gamma, M, K Gamma are 0, 0.5774, 0.9107, 1.5773 . Setting the same path in Wannier win file I get other positions for the same coordinates almost the double:
0, 1.12655, 1.77696, and 3.07778. This is problmatic when i try to plot both bands on the same graph. Could anyone tell me what is going wrong?...Is it related to different units used?

thanks very much in advance.

Hi,

Regarding the Wannier bands vs the ones obtained with plotband.x, it's only a matter of units: the k path from plotband.x is in units of 2pi/alat, whereas from Wannier it is in A-1, so you should rescale one.

As for your other question, in EPW the electron-phonon matrix element is set to zero if the phonon frequency is negative (imaginary). It is odd indeed that you get some negative frequencies even though they are positive for all the q points calculated. Have you tried changing some parameters, such as the ASR used or, in the phonon calculation, the q mesh?

Best
Carla

Dear Carla,

thanks for your reply. I have indeed tried both asr=' simple' and 'crystal'. both give negative frequencies for points close to the Gamma point. Here is a sample of their magnitude:

0.000000 -0.0000 -0.0000 0.0000 292.1863 292.1863 394.9670 394.9670 415.0476 482.2043
0.003125 -0.3436 1.6611 2.4527 292.1887 292.1951 394.9647 394.9676 415.0446 482.1986
0.006250 -0.6849 3.3219 4.9049 292.1958 292.2216 394.9578 394.9694 415.0356 482.1816
0.009375 -1.0217 4.9818 7.3560 292.2076 292.2658 394.9462 394.9725 415.0207 482.1531
0.012500 -1.3517 6.6404 9.8055 292.2241 292.3275 394.9299 394.9768 415.0001 482.1133
0.015625 -1.6725 8.2975 12.2528 292.2454 292.4067 394.9087 394.9823 414.9737 482.0622
0.018750 -1.9816 9.9525 14.6974 292.2713 292.5033 394.8824 394.9890 414.9420 481.9997
0.021875 -2.2765 11.6050 17.1387 292.3020 292.6173 394.8510 394.9968 414.9050 481.9258
.
.
.
.
My q point sampling is a 6x6x6 grid for an 18x18x18 (28 q points). I have chosen 28 points to have enough convergence to calculate the electron phonon coupling

1 0.000000000 0.000000000 0.000000000
2 0.000000000 0.000000000 0.025641026
3 0.000000000 0.000000000 0.051282051
4 0.000000000 0.000000000 -0.076923077
5 0.000000000 0.192450090 0.000000000
6 0.000000000 0.192450090 0.025641026
7 0.000000000 0.192450090 0.051282051
8 0.000000000 0.192450090 -0.076923077
9 0.000000000 0.384900179 0.000000000
10 0.000000000 0.384900179 0.025641026
11 0.000000000 0.384900179 0.051282051
12 0.000000000 0.384900179 -0.076923077
13 0.000000000 -0.577350269 0.000000000
14 0.000000000 -0.577350269 0.025641026
15 0.000000000 -0.577350269 0.051282051
16 0.000000000 -0.577350269 -0.076923077
17 0.166666667 0.288675135 0.000000000
18 0.166666667 0.288675135 0.025641026
19 0.166666667 0.288675135 0.051282051
20 0.166666667 0.288675135 -0.076923077
21 0.166666667 0.481125224 0.000000000
22 0.166666667 0.481125224 0.025641026
23 0.166666667 0.481125224 0.051282051
24 0.166666667 0.481125224 -0.076923077
25 0.333333333 0.577350269 0.000000000
26 0.333333333 0.577350269 0.025641026
27 0.333333333 0.577350269 0.051282051
28 0.333333333 0.577350269 -0.076923077

Would such small negative frequencies underlie a non convergence in the scf calculation or simply due to a bad q grid choice?

Regards

Hi,

Maybe try calculating the phonons on a smaller 4x4x4 grid and see if this still happens? Otherwise, I suggest you post this question on the QE forum.

Best
Carla