Yambo Community Forum

Dear Yambo developers, I'm new to Yambo, I've been trying to install Yambo 5.0.4 on my desktop, but after doing the tutorial on obtaining the band structure of a bulk material: h-BN,
I realize Yambo gives wrong band energies and when doing convergence test it usually gives a too large gap or NaN energy. I tried to install it on my laptop and it works well.
I managed to make configure file to use Yambo with internal library and put on --enable dp option (it produces bad results as before without this option) as below, this time I can
produce the band structure on the tutorial, but it still gives slightly different band energies compared to the tutorial which is correctly replicated on my laptop. This is quite annoying to me,
I hope you can help me out, for your information, my desktop is an HP z440 workstation with CPU Xeon E5 v3 2678 and 16Gb ECC ram, the config.log file is attached below.

The gap I get from my desktop and from the tutorial My configure input file

Dear Nhat Minh Pham ,

please sign your post with your name and affiliation, this is a rule of the forum and you can do once for all by filling the signature in your user profile.

The tutorial example are run in single precision, so the small difference you are finding are probably due to the DP usage.
What do you get when compiling in single precision?

Best,
Daniele

Dear Daniele,

Thank you for your reply, within single-precision I got the same result as double-precision but when changing variables like EXXRLvcs, BndsRnXp, NGsBlkXp, GbndRnge (as convergence test on the tutorial) Yambo gives E-E0 =NaN or huge numbers about 20eV for conduction band and -20eV for valance band.
Using double-precision the code gives reasonable energies on convergence test and I can plot the band structure like the tutorial.
I tried doing GW correction for 2D MoS2 following the low dimension tutorial but my desktop with DP mode gives E-E0= NaN, meanwhile, my laptop gives a correction of about 1eV for the gap.

Dear Pham,
as you said there should be a problem of compilation but it is not easy to spot, your config.log files does not show evident errors.
I would try to do a make distclean and then recompile the code and all the libraries, i.e. by running the configure without indicating libs option, something like
./configure

and see what happen.

Other possibilities could be related to the gfortran release which is very new. Do you have the possibility to try another compiler?

Best,
Daniele

Dear Daniele,

Thanks for your suggestion, I changed gfortran-11.2.0 to gfortran-8.5.0 and I got a closer band correction energy (not exact).
This time I'm able to do convergence tests for my 2D-MoS2 system without NaN error.

Best,
Minh

Dear Minh,

thanks for reporting, we will inspect what is causing problem with the gfortran-11.2.0

Best,
Daniele

Dear Minh,

gfortran 11 is not a stable release of gfortran

https://gcc.gnu.org/wiki/GFortran/News

you should not use it for production.

best
Claudio