Page 1 of 1

NaN in GW numerical calculation

Posted: Sat Nov 19, 2022 6:33 am
by zhouym
Dear developers,

I'm using Yambo 5.1.1 to perform the example of hBN. When I use the GW of PPA approximation : the values of (E-Eo) and Sc(Eo) are NaN for each K-point and each band.

o-gw_PPA.qp:
---------------------------------------------------------------------------------------------------------------------------------------------------------------
# K-point Band Eo [eV] E-Eo [eV] Sc|Eo [eV]
#
7 8 -0.411870 NaN NaN
7 9 3.877799 NaN NaN
---------------------------------------------------------------------------------------------------------------------------------------------------------------
r-gw_PPA_HF_and_locXC_gw0_dyson_em1d_ppa:
---------------------------------------------------------------------------------------------------------------------------------------------------------------
QP [eV] @ K [7] (iku): 0.000000 -0.500000 0.000000
B=8 Eo= -0.41 E= NaN E-Eo= NaN Re(Z)= NaN Im(Z)= NaN nlXC=-19.12909 lXC=-16.10656 So= NaN
B=9 Eo= 3.88 E= NaN E-Eo= NaN Re(Z)= NaN Im(Z)= NaN nlXC= -5.53657 lXC=-10.66983 So= NaN
---------------------------------------------------------------------------------------------------------------------------------------------------------------

There was no error message in the process of software compilation and calculation, and there was no problem when running on the virtual machine given by the official website. The problem of input file can be excluded. Can you give me some advice?

Best,

Zhou

Re: NaN in GW numerical calculation

Posted: Wed Nov 23, 2022 8:12 am
by Daniele Varsano
Dear Zhou,

it seems a compilation problem, you can check if the NaN is already present in the ndb.dipoles and/or ndb.pp filesby using the grep command.
You can try to recompile your code using the flag --enable-int-linalg in the configure process, even if it seems that the linear algebra is not the problem as the exchnage part of the self-energy is correctly computed.

If this does not solve the problem, please report here.

Please add your full name and affiliation in your signature, this is a rule of the forum, you can do it once for all by filling the signature in the user profile.

Best,
Daniele