Convergence test between Yambo and BerkeleyGW

[lesheng]

Hi everyone,

Last month I attended BerkeleyGW workshop hold by UC Berkeley, and during that workshop, I was quit confused on the convergence test between these two code. Based on the tutorial of Si calculation with Yambo, when the NGsBlk goes to 7 Ry the band gap will be converged. But with BerkeleyGW, the band gap will not converge until the same parameter goes to 20 Ry. The same situation also happens on empty band number, the one for BerkeleyGW is much higher than that from Yambo. However, the final result (band gap) from these two codes are close. I tried to ask the developer of their code, but it seems like they don't want to answer this kind of question... So I wish someone could help me on this here... Any reply will be helpful.

Best,
lesheng

[Daniele Varsano]

Dear Lehseng,
sorry, but I do not have knowledge on the GW implementation in BerkleyGW. May some user with experience in both codes can help you.
Best,
Daniele

[lesheng]

Hi Daniele,

Thanks for your attention. I realize that this kind of question indeed need people who has experience in both codes. But beside of this, do you have any guess about this issue?

Thanks,
lesheng

Dear Lehseng,
sorry, but I do not have knowledge on the GW implementation in BerkleyGW. May some user with experience in both codes can help you.
Best,
Daniele

[Daniele Varsano]

Dear Lesheng,
looking at this paper:
Stankovski et al. PHYSICAL REVIEW B 84, 241201(R) (2011), in particular at Fig.1, the abinit people show different convergence parameter needed to converge GW gaps using different plasmo pole approximations. From here it looks that more bands and Gvectors are needed to converge the Hybersten-Louie model of the Plasmon Pole (which is probably the one used in BerkeleyGW) which relies in sum-rules, with respect the Godby-Needs model (the one implemented in Yambo). This could be an explication of the different convergences. But it is a just a guess: as I wrote you before, I do not have knowledge about the Berkeley GW implementation.
Best,
Daniele

[lesheng]

Dear Lesheng,
looking at this paper:
Stankovski et al. PHYSICAL REVIEW B 84, 241201(R) (2011), in particular at Fig.1, the abinit people show different convergence parameter needed to converge GW gaps using different plasmo pole approximations. From here it looks that more bands and Gvectors are needed to converge the Hybersten-Louie model of the Plasmon Pole (which is probably the one used in BerkeleyGW) which relies in sum-rules, with respect the Godby-Needs model (the one implemented in Yambo). This could be an explication of the different convergences. But it is a just a guess: as I wrote you before, I do not have knowledge about the Berkeley GW implementation.
Best,
Daniele

Hi Daniele,

Sorry I just see your reply today. And thanks for giving me this reference, I'll have a look at it.

Cheers,
Lesheng Li