LDA+DMFT calculation problem

[mljphy]

Hi everyone After installation successfully of TRIQS and WIEN2K_11, I want to make a test calculation following the online tutorial on LDA+DMFT of Ce-gamma .After runing wien2k, I got successfully all need five files(Ce-gamma.ctqcout, Ce-gamma.symqmc...) for the following DMFT calculation. But I always the following error messages:

mlj>pytriqs Ce-gamma.py Starting on 1 Nodes at : 2012-10-01 04:20:10.860876 Traceback (most recent call last): File "Ce-gamma.py", line 2, in from pytriqs.Wien2k.SumK_LDA_Wien2K_input import * ImportError: No module named SumK_LDA_Wien2K_input

It seems tell me miss" Sumk_LDA_Wien2K_input" module, but the compilation of TRIQS is smooth without any errors (including all tests passed) Could you help me and tell me what happened ? Thanks.

[leopo]

It seems that the LDA+DMFT tutorial was not updated since some time and was not compatible with the newest TRIQS version. This has now been corrected, please update your TRIQS and try once more

[mljphy]

Thanks Yes, I update my TRIQS by using 'git pull'. Actually I found a bug in online script file( 'Ce-gamma.py'). The second line "from pytriqs.Wien2k.SumK_LDA_Wien2K_input import * "
should be "from pytriqs.Wien2k.SumK_LDA_Wien2k_input import *" Substituting the uppercase letter 'K' for lowercase letter 'k' .

But now I got the following error message when I do full self consistent LDA+DMFT calculation by using 'run -qdmft 1' (a simgle CUP) . ...... writing the file case.outbwin...

END OF THE PRGM python_with_DMFT: Command not found

stop error

It seems that anything is OK if I replace one-shot calculation with commands (init_lapw,run_lapw,x lapw2 -almd,dmftproj,pytriqs in turn) according to the example of Ce-gamma on the online tutorial. What happened?

[aichhorn]

Hi, Please look at issue #77. There is a problem with the name of the executable. When Wien2k 11.1 was released, the name of the executable was chosen as python_with_DMFT. This changed to pytriqs. The solution is to open the run_lapw, runsp_lapw, ... files of Wien2k and change the names python_with_DMFT to pytriqs. Of course, pytriqs has to be in the PATH.

[parcollet]

Hi, Why was the name of the executable encoded hard in the Wien2k files in the first place ? A quick fix could be also to add a link in our bin : python_with_DMFT -> pytriqs for backard compatibility until a new Wien version fixes this...

[leopo]

Perhaps to solve this problem we may include the corrected scripts (like run_para_triqs, runsp_para_triqs) to the TRIQS package, and ask the user to copy them to the Wien-2k root directory (as we do now for the cubic harmonics transformation files, case.cf_f_mm2 and case.cf_p_cubic).

[parcollet]

Looks even simpler. Markus ?

[mljphy]

Thanks. aichhorn. I have fix this problem now. I want to repruduce some results on TRIQS website http://ipht.cea.fr/triqs/doc/user_manual/green/tutorial.html#a-simple-example or other similar example by using pytriqs command. Although it seems that anything is OK without any error and warning message. But I cannot got the plot of results and even no any information occurs in running directory. why?

By the way, if I can get the resulting data to plot with other plot tools ?

[aichhorn]

@Olivier,Leonid: Indeed, that is the most practical solution for the moment. Leonid, can you do this? I have hardly any time the next days to look at this.

@mljphy: Do you mean that you just cannot plot the data, or that there is no data? The Greens functions, self energies, and so on should be in the hdf5 file (case.h5). you can easily check by opening the archive and looking whats inside (you don't even need pytriqs for that, you can use any h5viewer to look if there is data). Now if you just cannot plot it, then maybe your matplotlib does not work properly. Solution 1: You fix matplotlib. Solution 2 (workaround): You write the Greens function into a text file (the name of the Greens function in the Archive is supposed to be 'G'):

from pytriqs.Base.Archive.HDF_Archive import HDF_Archive from pytriqs.Base.GF_Local import * ar=HDF_Archive('case.h5','r') ar['G'].save('G')

That produces files G*.dat, which contain now real and imaginary part of G. You can plot this with any tool you like. The output of the run is put to standard out. If you don't pipe this output into a file when starting the job in the background, then this output is lost. Otherwise you should get the output on the screen.

[leopo]

OK, I can do this. However, I first need to install the current version Wien-2k, because I am still using my own 'old version' with different flags (alm instead of almp in lapw2 etc.). Markus, if you have ready scripts that works with the official release could you send them to me?

[leopo]

I added these scripts to pytriqs/Wien2k/dmftproj/SRC_templates

However, I cannot push my updated version back, it gives

git push origin master error: The requested URL returned error: 403 while accessing https://github.com/TRIQS/TRIQS.git/info/refs

fatal: HTTP request failed

Second, we do not know how MPI wrappers (mpprun, mpirun) are called in the user's system, because this is system-depended. So I just added comments and error message printing into run_triqs, telling the user he has to insert a proper pytriqs call manually

[mferrero]

@leopo: did you manage to push your changes in the end?

[leopo]

Yes, finally I did