Can not get convergence with wannier90 in EPW

[sponce]

Dear Loc,

What is the final spread for EPW?

The post you are mentioning is indeed the issue. The library mode of Wannier was not giving the same initial projection as the normal Wannier code.

EPW rely on the library mode of Wannier.

Therefore I did a patch that should have corrected that. The problem was specifically strong with spin orbit coupling. Do you have that?

So I'm not sure what the problem is at this moment.
What I suggest you to try:
1) Go inside EPW/src and recompile twice ("make" and then "make" again). This will copy and apply the patch to Wannier90. It should be correctly done during the make epw procedure though. So I do not think it is the problem but its worth checking
2) Try removing all the wdata(1) etc from your epw input file and also the corresponding Wannier input. This means removing guiding functions etc.
It could be that there is a problem in EPW at that level. Then compare both, you can post the inital spread of the new EPW and new Wannier.

If it still does not help, I can take a look. In that case, can you send me by email a very small (i.e. quick to compute on a desktop) version of your problem? You can decrease all convergence parameters like ecut, kpt etc as long as the problem persists. Please also include the psp.

Best,

Samuel

[dinhloc]

Dear Samuel,

The final spread for epw and wannier90 are:

epww:
Final State
WF centre and spread 1 ( -4.282036, -4.354539, -9.487531 ) 26.41334134
WF centre and spread 2 ( -6.532272, 2.052104, 0.883142 ) 24.63628306
WF centre and spread 3 ( 1.769598, 5.655888, 1.311409 ) 24.18132158
WF centre and spread 4 ( -0.235053, -1.296045, -3.684107 ) 26.89830748
WF centre and spread 5 ( 1.200673, 3.975507, -0.620230 ) 23.87479581
WF centre and spread 6 ( 3.253716, -2.149864, 0.306397 ) 23.78137890
WF centre and spread 7 ( 2.584093, -1.653631, -0.669229 ) 24.01218378
WF centre and spread 8 ( 2.158818, -0.023554, 1.718910 ) 24.76036179
WF centre and spread 9 ( -9.467060, 5.015509, -0.319015 ) 27.47088139
WF centre and spread 10 ( -2.147006, 0.186958, -0.252844 ) 24.29356220
WF centre and spread 11 ( 0.617143, 3.478773, -0.010687 ) 23.92708562
WF centre and spread 12 ( -1.430659, 0.743274, 1.530476 ) 23.86460126
Sum of centres and spreads (-12.510046, 11.630382, -9.293308 ) 298.11410421

Wannier90:
Final State
WF centre and spread 1 ( -0.000000, -0.000000, 0.000000 ) 1.23423780
WF centre and spread 2 ( -0.000000, -0.000000, 0.000000 ) 1.15075292
WF centre and spread 3 ( 0.000000, -0.000000, 0.000000 ) 1.16385587
WF centre and spread 4 ( 3.528003, 0.000003, 0.000000 ) 6.10919556
WF centre and spread 5 ( 0.000813, -2.038570, -2.052836 ) 1.57433755
WF centre and spread 6 ( -0.502962, -2.334567, -1.236574 ) 1.80689818
WF centre and spread 7 ( 0.516089, -2.337716, -1.227988 ) 1.78265158
WF centre and spread 8 ( 0.002501, -1.475862, -1.249503 ) 1.86748663
WF centre and spread 9 ( -1.764840, -1.016768, 2.052835 ) 1.57434083
WF centre and spread 10 ( -1.261023, -0.720786, 1.236594 ) 1.80689089
WF centre and spread 11 ( -2.280076, -0.717612, 1.227965 ) 1.78264137
WF centre and spread 12 ( -1.766505, -1.579478, 1.249505 ) 1.86750355
Sum of centres and spreads ( -3.528000,-12.221356, 0.000000 ) 23.72079274

I do not including the spin-orbit coupling at this moment. I will follow your suggestion and let you know soon.

With many thanks,
Loc

[dinhloc]

Dear Samuel,

The first step of compile twice does not help. Do I need any separated file (download) for patch?

I am testing the second step.

With best wishes,
Loc

[dinhloc]

Dear Samuel,

It is not solve the problem.

Here is the initial state of epw and wannier90 after recompile twice and remove all addition parameters.

EPW:
Initial State
WF centre and spread 1 ( 0.159165, -1.034464, -0.024546 ) 61.62042931
WF centre and spread 2 ( -0.220840, -0.456448, -0.229825 ) 64.18611836
WF centre and spread 3 ( 0.498244, 0.157664, -0.591081 ) 70.71307323
WF centre and spread 4 ( -0.110570, -0.068055, -0.058749 ) 45.48953389
WF centre and spread 5 ( 1.112359, 0.347153, -0.918744 ) 51.30670062
WF centre and spread 6 ( -0.722285, -0.755601, 1.138105 ) 59.09773727
WF centre and spread 7 ( -0.436002, 0.606401, 1.039276 ) 65.26531976
WF centre and spread 8 ( -1.530822, -0.561120, 0.124265 ) 43.84596623
WF centre and spread 9 ( -0.748247, 0.028336, 0.463830 ) 53.71225170
WF centre and spread 10 ( 0.680561, -1.990758, -0.084900 ) 66.36694223
WF centre and spread 11 ( -0.295961, 1.483920, 1.237902 ) 55.88726337
WF centre and spread 12 ( -0.315100, 0.325623, -0.054156 ) 65.11850018
Sum of centres and spreads ( -1.929499, -1.917348, 2.041378 ) 702.60983614

Wannier90:
Initial State
WF centre and spread 1 ( -0.338225, -0.340303, -0.010592 ) 5.54166690
WF centre and spread 2 ( 0.277507, -0.314435, -0.009587 ) 5.54144413
WF centre and spread 3 ( 0.000943, -0.165925, -0.000735 ) 1.75629687
WF centre and spread 4 ( 0.419591, -0.484985, 0.007673 ) 43.84177548
WF centre and spread 5 ( 0.001362, -2.015545, -1.562470 ) 1.57026361
WF centre and spread 6 ( 0.012635, -1.958710, -1.389535 ) 3.23674857
WF centre and spread 7 ( -0.000405, -2.082241, -1.328443 ) 3.13738905
WF centre and spread 8 ( 0.009001, -1.990323, -1.369166 ) 3.17753002
WF centre and spread 9 ( -1.763887, -1.018538, 1.564809 ) 1.56959823
WF centre and spread 10 ( -1.767245, -0.995840, 1.392216 ) 3.21325727
WF centre and spread 11 ( -1.777614, -0.938354, 1.338846 ) 3.03588044
WF centre and spread 12 ( -1.764063, -1.017367, 1.365842 ) 2.89448688
Sum of centres and spreads ( -6.690400,-13.322564, -0.001142 ) 78.51633744

I am making a test with small ecut and send you by email.
Thanks,
Loc

[dinhloc]

HI Samuel,

When I prepare the file for you and test with single core running, the error is gone, The initial state of epw and wannier90 is quite similar.The big error is reproduced with running in multicore.

The initial state of epw running with 1 core and 16 core:
1 core:
Initial State
WF centre and spread 1 ( -0.622963, -0.331013, 0.009105 ) 6.66356580
WF centre and spread 2 ( 0.214766, -0.336396, 0.008987 ) 6.74146571
WF centre and spread 3 ( -0.000001, -0.247941, 0.006389 ) 2.96736550
WF centre and spread 4 ( -0.000011, 0.027759, -0.008292 ) 26.26084124
WF centre and spread 5 ( -0.007526, -2.013505, -1.575266 ) 1.63933925
WF centre and spread 6 ( -0.027812, -1.957905, -1.496766 ) 3.48799708
WF centre and spread 7 ( 0.000563, -2.089378, -1.330214 ) 3.21557174
WF centre and spread 8 ( -0.003841, -1.939084, -1.345998 ) 3.18113310
WF centre and spread 9 ( -1.770826, -1.018923, 1.577400 ) 1.63916847
WF centre and spread 10 ( -1.790176, -1.017583, 1.467878 ) 3.42809229
WF centre and spread 11 ( -1.765652, -0.883547, 1.334386 ) 2.93155433
WF centre and spread 12 ( -1.768032, -1.089022, 1.348684 ) 3.13108856
Sum of centres and spreads ( -7.541510,-12.896540, -0.003709 ) 65.28718307


16 cores:
Initial State
WF centre and spread 1 ( 0.097921, -0.347903, 0.762817 ) 62.20491022
WF centre and spread 2 ( 0.767947, 0.623923, 1.676170 ) 54.26499128
WF centre and spread 3 ( 0.653325, 0.170062, 1.339667 ) 54.48759153
WF centre and spread 4 ( -0.540982, -0.009957, 1.267121 ) 46.01592629
WF centre and spread 5 ( -0.011920, -1.448857, -1.907058 ) 46.41011089
WF centre and spread 6 ( 0.018823, -0.460921, -0.180905 ) 68.13083516
WF centre and spread 7 ( -0.514440, -0.108092, 0.323779 ) 63.43241126
WF centre and spread 8 ( -0.193539, -0.935828, -0.448985 ) 52.33886097
WF centre and spread 9 ( -1.044090, 0.613524, 0.107676 ) 45.49526424
WF centre and spread 10 ( -0.487439, -0.192411, -0.510953 ) 62.15953477
WF centre and spread 11 ( 0.023671, -0.763981, -0.890079 ) 67.47118796
WF centre and spread 12 ( -0.651690, 0.192471, 1.952490 ) 52.95603267
Sum of centres and spreads ( -1.882413, -2.667972, 3.491739 ) 675.36765723

I am sending you the input file.


With best wishes,
Loc

[sponce]

Hello Loc,

I've tried the TiSe2 files you send me. In my case it works perfectly using EPW and 4 cores. See below.
There is a small difference of the final spread but this should be fine. This might be the difference between using 4 cores for EPW and 1 core for Wannier.

What I have modify from your input files:
1) Increased the k grid from 4x4x1 to 8x8x1 in scf.in
2) increase conv_thr = 1.D-12 in nscf.in
3) put the same number of iteration for Wannier in both cases.

You parallel compilation might be wrong? I'm using gcc 5.4 +openmpi 1.8.8 for the test.

Best,

Samuel

---------------------------------------------------------------------

EPW using 4 cores:
Initial State
WF centre and spread 1 ( -0.416503, -0.018644, 0.005284 ) 6.60242200
WF centre and spread 2 ( 0.421913, -0.018698, 0.005293 ) 6.59960174
WF centre and spread 3 ( -0.000001, -0.248914, 0.006618 ) 2.99609615
WF centre and spread 4 ( -0.000012, 0.035701, -0.008192 ) 26.37565579
WF centre and spread 5 ( 0.000137, -2.013235, -1.575417 ) 1.63694573
WF centre and spread 6 ( 0.000450, -1.956759, -1.498830 ) 3.47426794
WF centre and spread 7 ( -0.000015, -2.061709, -1.330437 ) 3.08375174
WF centre and spread 8 ( 0.000070, -1.937933, -1.348893 ) 3.17115444
WF centre and spread 9 ( -1.763909, -1.019032, 1.577571 ) 1.63615356
WF centre and spread 10 ( -1.763591, -1.017728, 1.472546 ) 3.41538598
WF centre and spread 11 ( -1.763965, -0.862998, 1.339528 ) 2.93453583
WF centre and spread 12 ( -1.763941, -1.088194, 1.351282 ) 3.12352687
Sum of centres and spreads ( -7.049367,-12.208144, -0.003645 ) 65.04949775

Final State
WF centre and spread 1 ( 0.011632, -0.000400, -0.017633 ) 2.26099166
WF centre and spread 2 ( -0.012449, -0.004731, -0.026494 ) 2.49159721
WF centre and spread 3 ( 0.000831, -0.001321, 0.003940 ) 2.19340963
WF centre and spread 4 ( 0.028154, -0.594878, -0.059581 ) 9.56459447
WF centre and spread 5 ( -0.006054, -2.030320, -2.005170 ) 1.02930075
WF centre and spread 6 ( 0.007688, -2.159781, -1.005290 ) 2.21687346
WF centre and spread 7 ( 0.015119, -2.165875, -1.339928 ) 2.05431433
WF centre and spread 8 ( -0.015262, -2.086143, -1.416930 ) 4.54812393
WF centre and spread 9 ( -1.758924, -0.895471, 2.071019 ) 1.07548364
WF centre and spread 10 ( -1.748047, -0.841172, 0.954770 ) 1.69366543
WF centre and spread 11 ( -1.791741, -0.885951, 1.338501 ) 2.01014919
WF centre and spread 12 ( -1.767567, -1.371494, 1.364770 ) 2.67139979
Sum of centres and spreads ( -7.036619,-13.037538, -0.138024 ) 33.80990348

Wannier :
Initial State
Initial State
WF centre and spread 1 ( -0.429467, -0.375204, 0.002758 ) 6.72261584
WF centre and spread 2 ( 0.362948, -0.344434, 0.002728 ) 6.87935930
WF centre and spread 3 ( -0.000001, -0.248914, 0.006618 ) 2.99609602
WF centre and spread 4 ( -0.000013, 0.035697, -0.008192 ) 26.37560370
WF centre and spread 5 ( -0.002643, -2.011710, -1.575223 ) 1.64186554
WF centre and spread 6 ( -0.009478, -1.948955, -1.492082 ) 3.48569088
WF centre and spread 7 ( 0.000205, -2.087240, -1.333893 ) 3.19831404
WF centre and spread 8 ( -0.001338, -1.928582, -1.349868 ) 3.18602316
WF centre and spread 9 ( -1.766439, -1.019870, 1.577751 ) 1.63655760
WF centre and spread 10 ( -1.773297, -1.016381, 1.479276 ) 3.42247700
WF centre and spread 11 ( -1.764591, -0.886265, 1.336054 ) 2.91881357
WF centre and spread 12 ( -1.765448, -1.089719, 1.350343 ) 3.14134562
Sum of centres and spreads ( -7.149561,-12.921576, -0.003730 ) 65.60476227

Final State
WF centre and spread 1 ( -0.005349, -0.006822, 0.010683 ) 3.04651763
WF centre and spread 2 ( 0.001228, -0.001761, 0.002823 ) 2.14622731
WF centre and spread 3 ( 0.000007, -0.001797, 0.000985 ) 2.17820075
WF centre and spread 4 ( -0.002997, -0.063445, -0.007690 ) 9.30630516
WF centre and spread 5 ( 0.006017, -2.101399, -2.073092 ) 1.05038318
WF centre and spread 6 ( 0.018098, -2.196108, -0.968585 ) 1.78319436
WF centre and spread 7 ( -0.022728, -2.177268, -1.329322 ) 2.01614114
WF centre and spread 8 ( -0.002560, -1.960227, -1.398962 ) 3.32336612
WF centre and spread 9 ( -1.771757, -0.981622, 2.103758 ) 1.09199093
WF centre and spread 10 ( -1.778809, -0.903362, 0.920549 ) 1.75674028
WF centre and spread 11 ( -1.737962, -0.879070, 1.330054 ) 2.00788062
WF centre and spread 12 ( -1.762915, -1.037494, 1.395422 ) 2.76290400
Sum of centres and spreads ( -7.059728,-12.310375, -0.013376 ) 32.4698514

[dinhloc]

Dear Samuel,

I use intel-openmpi1.8.5 with mkl.
I recompile again with the same compiler, still the same error.

I will check with gnu-openmpi and let you know.

With many thanks,
Loc

[dinhloc]

Hi Samuel,

Still I have issue even I recompile the epw with gcc5.4-openmpi1.8.8.

Quantum espresso and phonon is till working fine. I completely have no idea why it happens.

Is there any difference between the source code you are using and the code in the package?
Here is the procedure I compile epw:
1. ./confgure
2. make pwall
3. make ph
4. make epw.x

Compiling process is smoothly fine.

Could you have any suggestion?

With many thanks,
Loc

[sponce]

Dear Loc,

This is very strange indeed.

I'm usually compiling in the following way (but the way you are describing should work as well):
1. ./confgure
2. make pwall
3. make ph
4. cd EPW/src
5. make
6. make

If that does not work, it could be a hardware issue? Maybe you could try on a different machine (for example your desktop or laptop)?
I'm a bit puzzled ...

[dinhloc]

Dear Samuel and Carla,

I figured out the reason. EPW require the same number of cores and npool for nscf and epw calculation. I did not follow this rule.

Thank you so much for your time and effort to help me figure out the error.

With many thanks,
Loc