p2y error

[mferri]

Dear developers,

I'm using QE v5.2.0 and Yambo v3.4.1 in order to calculate optical properties of silicene nanoribbons.
I've already performed some calculations on this system and I'm trying to increase the number of k-points to reach convergence in optical spectrum.
in the attachments you can find PWSCF input and output files. The differences between the I/O files in working/ folder and the ones in not_working/ folder are the number of k points (10 in scf, 25 in nscf vs 15 in scf and 35 in nscf) and the number of bands (250 vs 350).
Using the files in working/ folder, everything works and i can calculate optical properties, etc.
Using the files in not_working/ I see the following error message in p2y runlevel:

_| _| _|_| _| _| _|_|_| _|_|
_| _| _| _| _|_| _|_| _| _| _| _|
_| _|_|_|_| _| _| _| _|_|_| _| _|
_| _| _| _| _| _| _| _| _|
_| _| _| _| _| _|_|_| _|_|

<---> P(W) 2 Y(ambo) Ver. 5.0
<---> DBs path set to .


<---> Index file set to data-file.xml
[ERROR] STOP signal received while in :
[ERROR]Error in Read pwscf version variables pwscf_current_version not present ierr: 1

It seems to be a problem with data-file.xml. In the not_working/ folder you can find it.

Could you please help me?

Best,

Matteo

[Conor Hogan]

Hi Matteo,
Thanks for pointing this out. It might be something specific to 5.2.0.
Can you post also the data-file.xml for the run that worked?
And can you tell me which iotk library you link to at compilation - the one automatically downloaded by the code or one that you linked to by hand (via configure)?
Conor

[Conor Hogan]

I think I understand the problem: unless you somehow cut the end of the file when posting it, the data-file.xml file was not created correctly by QE.
The nscf reports 35 k-points and seems to terminate correctly after 23 hours, but the data-file.xml ends with:

Code: Select all

      <WFC.1 iotk_link="./K00014/evc1.dat">
        <!--This is a link to the file indicated in the iotk_link attribute-->
      </WFC.1>
      <WFC.2 iotk_link="./K00014/evc2.dat">
        <!--This is a link to the file indicated in the iotk_link attribute-->
      </WFC.2>
    </K-POINT.14>
    <K-POINT.15>
      <NUMBER_OF_GK-VECTORS type="integer" size="1">
    244542
      </NUMBER_OF_GK-VECTORS>
      <GK-VECTORS iotk_link="./K00015/gkvectors.dat">
        <!--This is a link to the file indicated in the iotk_link attribute-->
      </GK-VECTORS>
      <WFC.1 iotk_link="./K0001

I don't know what to suggest....if you still have all the .wfc files you can try to restart and make the .save directory again. Perhaps the code was still writing files when it hit some 24h time limit, or ran out of disk space. Check that all the .save/K*/evc1.dat etc files exist and are the same size.

Also: do you really need 75Ry for silicon? The most I've ever used is 30Ry, but usually less.

Hope that helps,
Conor
PS Let me shamelessly promote my recent article on silicene nanoribbons, in case you are interested: http://journals.aps.org/prb/abstract/10 ... .92.115439

[mferri]

Dear Conor,
thank you for your reply. Regarding the iotk library, this is the one automatically downloaded by the code.
Unfortunately I don't have the data-file.xml you've requested, but in the attachments you can find the data-file of a small test performed this morning that worked.
It seems that data-file.xml was not created correctly as you said. I checked the .save/K000* files and they seem correct up to K00019.

mferri@alan:/globalscratch/mferri/silicene/Si-4zig-dist.save$ ls -lrt K00019
total 2677096
-rw-rw-r-- 1 mferri mferri 17115 Apr 19 17:10 eigenval.xml
-rw-rw-r-- 1 mferri mferri 3911583 Apr 19 17:18 gkvectors.dat
-rw-rw-r-- 1 mferri mferri 1368700755 Apr 19 17:18 evc1.dat
-rw-rw-r-- 1 mferri mferri 1368700755 Apr 19 17:18 evc2.dat

From K00020 forward, evc1.dat and evc2.dat become dimensionless:

mferri@alan:/globalscratch/mferri/silicene/Si-4zig-dist.save$ ls -lrt K00020
total 773364
-rw-rw-r-- 1 mferri mferri 17115 Apr 19 17:10 eigenval.xml
-rw-rw-r-- 1 mferri mferri 3912703 Apr 19 17:18 gkvectors.dat
-rw-rw-r-- 1 mferri mferri 785682432 Apr 19 17:18 evc1.dat
-rw-rw-r-- 1 mferri mferri 3350528 Apr 19 17:18 evc2.dat
mferri@alan:/globalscratch/mferri/silicene/Si-4zig-dist.save$ ls -lrt K00021
total 28
-rw-rw-r-- 1 mferri mferri 17115 Apr 19 17:10 eigenval.xml
-rw-rw-r-- 1 mferri mferri 8192 Apr 19 17:18 gkvectors.dat
-rw-rw-r-- 1 mferri mferri 0 Apr 19 17:19 evc1.dat
-rw-rw-r-- 1 mferri mferri 0 Apr 19 17:19 evc2.dat
mferri@alan:/globalscratch/mferri/silicene/Si-4zig-dist.save$ ls -lrt K00022
total 20
-rw-rw-r-- 1 mferri mferri 17115 Apr 19 17:10 eigenval.xml
-rw-rw-r-- 1 mferri mferri 0 Apr 19 17:19 gkvectors.dat
-rw-rw-r-- 1 mferri mferri 0 Apr 19 17:19 evc1.dat
-rw-rw-r-- 1 mferri mferri 0 Apr 19 17:19 evc2.dat

and so on. I can try to restart the 25 k-points calculation mantaining 350 bands in order to see what happens.

However, i'm using 75 Ry cutoff because the PP is a fully relativistic PP. In the attachments you can find a picture representing the convergence test.

Best,

Matteo

PS: Thank you for the article you suggested. I will surely check it as soon as possible

[mferri]

Hi Conor,
I perfomed on CINECA/FERMI the same calculation that had not worked on our cluster.
Now data-file.xml is complete and each .save/K000* folder contains complete evc.dat files.
QE version is 5.0.3b
I'm trying anyway to reproduce again the calculation on our cluster but it runs much slower. I'll let you know the results.

Cheers,

Matteo

[Conor Hogan]

Ok good luck! Of course you can afterwards transfer the netCDF formatted SAVE files on Fermi back to your cluster, if you prefer to work there.
Regards,
Conor