problems in dipole calculation using covariant method

[baiyunfei]

Dear developers:

I am trying to calculate BSE on top of PBE+U calculation . So I followed the suggestion using covariant method to calculate dipole matrix. However, yambo complains when reading dipole bases:

[size=85]<01h-04m> P2-hpcc028: [DIP] Checking overlaps header
P2-hpcc028: [ERROR] STOP signal received while in[07] Static Dielectric Matrix
P2-hpcc028: [ERROR] Reading overlaps with (i_sp_pol,ik)= (2,1)[/size]

I checked the output and found :

[WARNING] Covariant Dipoles not in 3d. Assuming non periodic system. Using forth order approximation for covariant dipoles.

I guess yambo may not support 2D covariant method. Further, report in LOG directoy I found:

<58m-02s> P2-hpcc028: [FFT-Oscillators/R space/Real Space] Mesh size: 42 42 150
<58m-11s> P2-hpcc028: Dipoles in real space |##################### | [052%] 02m-35s(E) 04m-55s(X)

It seems yambo regard my 2D system as isolated system even though I set coulmb cutoff and ''slab z'' in my inputfile. It then (I guess) used ''R-space x'' method to calculate dipole matrix ,which caused inconsistency. When I simplely changed back to default method(G-space v) , yambo can successfully run. Could you please offer me some suggestions how to fix this problem?

Best wishes!

[Daniele Varsano]

Dear Yunfei,

can you post your log and report file? You can attach them after adding them as suffix e.g. .txt
It seems yambo correctly see the system as 2D, in this case the covariant method is aplpied for the dipoles along the periodic system while the non periodic direction is then treated in real space (as in that direction the covariant algorithm cannot be applied, and the real space is well-defined).
In order to spot the problem further investigation is needed and the report and log can help.
Best,

Daniele

[baiyunfei]

Dear developers:
Here are my report file and LOG file
Best wishes !

[Daniele Varsano]

Dear Yunfei,

the dipoles are calculated correctly. When applying covariant method for a non 3D system, the dipole along the non-periodic direction is calculated in real time.

The code seems to stop when reading the database, can you also post your input file?

Daniele

[baiyunfei]

Dear Daniele:
Here is my set_up input file and BSE inputfile. There are two things worth mentioning , One is I turned off all symmetries in nscf and subsequent yambo calculation for certain reason, The other is I delete all lines related to Hubbard U in 'data-file-schema.xml' in order to have a successful p2y conversion.
Best Wishes !

[Daniele Varsano]

Dear Yunfei,

it is not really clear to me what is causing the error, anyway from your input file I can see you set:

Code: Select all

% BSEBands
    1 |  500 |                       # [BSK] Bands range
% 

That's not needed, BSE converges usually considering few bands around the gap. Your setting implies an excitonic matrix with a dimension (nc x nv x nk) over 1 millions which is tremendously large.

Can you try to repeat your calculation using few bands? e.g. something like

Code: Select all

% BSEBands
    30 |  50 |                       # [BSK] Bands range
% 

Best,
Daniele