GW+BSE

[lamia]

Dear YAMBO developpers,

I ran GW+BSE using this command: ../yambo -o b -b -k sex -y d -V qp. However, I did not get the results and the calculations are done very quickly.

Best,

Lamia

[Daniele Varsano]

Dear Taouil,
as you can see from the snapshot you posted check the WARNINGS:

1. The screening have problem: variable non compatible in PP/Em1s database
2. please note that the QP particle are not read correctly (SAVE/db.QP) probably you misspelled the name of the database (ndb.QP). Also note that reading QP from database and setting a scissor are exclusive otherwise you would assign QP correction twice in two different way.


So, because of (1) finally you are doing an independent-particle calculation which is very fast. From the report hopefully you should be able to understand incompatibility from the input and the previously calculated screening database. Please note that in the input you have both em1s (static screening) and ppa (dynamical screening in PPA), both are allowed for a BSE calculation as in the case of PPA only the static limit is read, bu you should consider only one of the two depending on how you calculated the screening previously.

Best,
Daniele

[lamia]

Dear Daniele,

Thank you for your answer. Could you explain more, please? I do not get how I could resolve the issue .

Best regards,

Lamia

[Daniele Varsano]

Dear Lamia,
please post your report files. The input file you posted anyway presents some problems as I told you in my previous answer.

Best,
Daniele

[lamia]

Dear Daniele,

You'll find the report files in the attachments.

best,

Lamia

[Daniele Varsano]

Dear Lamia,
the problem can be spotted by reading carefully the report file.
In the input you asked for:

Code: Select all

BSENGBlk= 603         RL  

But the screening was calculated with 309 RL only, so it is incompatible.
You may also note the incompatibility in the use of the coulomb cutoff potential.

Code: Select all

 [RD./SAVE//ndb.pp]------------------------------------------
    Brillouin Zone Q/K grids (IBZ/BZ):  43  144   43  144
   *WRN* RL vectors                   (WF): 14855
   *WRN* Coulomb cutoff potential         :none
    Fragmentation                    :yes
    Electronic Temperature        [K]: 300.1000
    Bosonic    Temperature        [K]: 300.1000
    PPA    diel. fun.  energies      :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
                       wavefunctions :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
    Global Gauge                     :length
   *ERR* X matrix size          : 309
    X band range           :   1   50
    X e/h energy range [ev]:-1.000000 -1.000000
    X Time ordering        :R
    X xc-Kernel            :none
    X Drude frequency      : 0.00      0.00
    X poles           [o/o]: 100.0000
    RL vectors in the sum  : 14855
    [r,Vnl] included       :yes
    Field direction        :0.1000E-4  0.000     0.000
    BZ energy Double Grid  :no
    BZ energy DbGd points  :0
    BZ Q point size factor : 1.000000

   [WARNING]Variable not compatible in PP/Em1s DB

Best,
Daniele

[lamia]

Dear Daniele,

It works but I get the warnings underlined in the snapshot. Could this warnings influence on the calculation and/or the results?

Best,
Lamia

[Daniele Varsano]

Of course as you are not considering the GW energies but the Kohn-Sham eigenvalues,
the problem here is that the name of the database is ndb.QP and not db.QP, just correct it in the input file.

Best,
Daniele

[lamia]

After the correction, the calculation is killed.

Best,
Lamia

[Daniele Varsano]

This has nothing to do with the correction.
Most probably you have memory problem as you are dealing with huge matrix dimension (26496),
you can try to avoid direct diagonalization and use iterative method.

Code: Select all

BSSmod= "h"  

Anyway, as usual, try to look carefully at the report as often you can spot where the problem is, it seems strange to me the kernel is build in few seconds.
I do not which input you are using, but the one you posted was full of inconsistencies (e.g. presence of em1s and relative variables, KfnQPdb and KfnQP_E with a finite value of scissors which are exclusive...)
Note also:

Code: Select all

% BSEBands
   1 |  50 |                   # [BSK] Bands range
%

this make the excitonic matrix very large and rarely you need to include deep states. Usually including few states across the gap is enough, and convergences are tested enlarging a window across the gap.

Daniele