segment fault error in EPW 6.0 but not in 5.9

[tyj]

Dear developer,
My job was finished with QE-7.3.1 and EPW 5.9 without error. Now I want to rerun the same job and test the new features in v6.0, ran with QE-v7.5, but it reports 'segmentation fault' error when I run it. Is there some change in the new version that causes this? The input/outputs of these two versions are attached below.

Best,
Yujia

[stiwari]

Hello Yujia,

Did you run everything including phonon calculations using QE7.5 or you reused the output from QE7.3? I did not see a phonon calculation in your v6.0 folder.

Best regards,
Sabya.

[tyj]

Dear Sabya,
I just reused the results from QE7.3 with EPW6.0 in my original attached file. The phonon calculation was done but I didn't attach it.

To avoid possible version incompatibility problem, I redid all the calculation with QE7.5 and EPW6.0, but the same error is still here. I attach the files of both phonon and epw calculation below. There're still plenty of free memory on the node, so it's strange that it reports segmentation fault error.

Best,
Yujia

[stiwari]

Hello Yujia,

I see, you are using PAW pseudopotential; Can you use ONCV/NC pseudopotential instead of PAW?

Best,
Sabya.

[tyj]

Hi Sabya,
I changed two of my elements to ONCV, but one is still PAW, because I didn't find fully-relativistic ONCV/NC potential for rare-earth element. This is same as my original attached file. I redid the calculation and still doesn't work. I also tried USPP potential and has same problem with PAW (Not sure why I can't attach extra file during edit).

To check NC potential, I ran Pb wSOC example which uses NC pseudopotential, and it works.

So looks like EPW6.0 doesn't support PAW and USPP pseudopotential which caused the problem I have, but previous version like EPW5.8.1 supports? (The attachment epw 5.8 is the same input/ouput I attached at first).

I mainly want to use the q-point parallelization feature in v6.0, which could speed up the calculation a lot. Other results should be same. So in my case, where there's no NC/ONCV pseudopotential for rare-earth element, is there a way to use EPW6.0 or I can only use the previous versions?

Best,
Yujia

[stiwari]

Hi Yujia,

Yes, I have taken a look and there is an issue in EPW v6.0 regarding PAW pseudos. If EPW 5.8.1 works fine, you can perform step 1 with EPW v5.8.1 and generate the epmatwp files (electron-phonon matrix in Wannier representation). You then use the EPW v6.0 for the fine-grid calculations (enabling q-parallelization).

Let us know if this route gives you any problem.

Best,
Sabya.

[tyj]

Dear Sabya,
Thanks very much for the help. I'm glad to know that we can use an earlier version in step 1 and v6.0 for fine-grid calculations. I would try that, and hopefully it works.

Best,
Yujia