different QP correction in stable and devel version

[arlonne]

Hi all,

Recently, I just noticed that devel(v3.4.0) have different behaviors in different level of response function calculations. And the QP correction is different even in two version of YAMBO: stable and devel.

for devel:

in IP & RPA:
=== QP/DB derived ===
Gap correction (database) [ev]: 1.292792
(FIT) [ev]: 0.901168
Energies 0th order c/v [ev]: 1.869547 -0.324412
Energies 1st order : 1.244195 1.321030
Error:0.4685E-3 0.4675E-3
=== Additional ===
------------------------------------------------------------
[QP] Fermi level variation [ev]: 0.772568
[QP] Last Filled/Metallic band : 272 272
[QP] Ind. Gap Correction [ev]: 2.193960

-----------------------------------------------------------------------------------------
in BSE:
=== QP/DB derived ===
Gap correction (database) [ev]: 1.292792
(FIT) [ev]:-0.209299
Energies 0th order c/v [ev]: 2.035653 0.952160
Energies 1st order : 1.194278 1.756682
Error: 0.0013 0.0014
=== Additional ===
------------------------------------------------------------
[QP] Fermi level variation [ev]: 1.493906
[QP] Last Filled/Metallic band : 272 272
[QP] Ind. Gap Correction [ev]: 1.083493


for stable:
IP & RPA:
[QP] Kpts covered exactly [o/o]: 100.0000
[QP] Fermi level variation [ev]: 1.134224
[QP] Last Filled/Metallic band : 272 272
[QP] Ind. Gap Correction [ev]: 1.292792
-------------------------------------------------------------------

The resulted Gaps of devel are 2.19, 1.08 eV for IP and BSE, respectively. The resulted Gap in stable is 1.29+0.50=1.79 eV, which is exactly the gap from G0W0 calculations. It is hard to understand that the devel gives a large band gap than G0W0 in IP and RPA calculations. of course, I think the correction in BSE is much underestimated in devel version although I don't have the result for stable version yet. in the gap I trust the stable version. But why use different strategy for applying QP correction in the devel version? what's the meaning of '(FIT)' in the devel version?

Thanks advanced.

[Daniele Varsano]

Dear Arlonne,
we are aware of this problem, and we will clarify this issue soon.
Best,

Daniele

[arlonne]

Dear Daniele,

Thanks for reply. in this stage, how can I make the devel-yambo use the same QP_apply as in the stable-yambo.
I have to use devel-yambo because I found the devel uses less memory than the stable one, i.e., BSE calculations stop in BSE solver (insufficient virtual memory) in the stable version but successful in the devel version.

Thanks,
arlonne

[Daniele Varsano]

Dear Arlonne,
sorry for the delay, but it is a very busy period. We still did not have time to fix the bug, we will do it soon.
In the meanwhile you can try to use the Haydock solver with the stable version, it should use less memory.
I will write you once the bug is fixed, hope very soon.
Best,
Daniele