BSE calculation stopped

[ezekiel]

Dear Moderators,
I am running a BSE calculation for a 2D heterostructure system. But after, sometimes the run stopped uncompleted and without an error. Could you please assist on the way out?
See part of the content of my report file:

[WARNING]Non periodic geometry and unit cell are incompatible

[05] Coloumb potential CutOff :box z
====================================

[RD./2D_WR_WC//ndb.cutoff]----------------------------------
Brillouin Zone Q/K grids (IBZ/BZ): 802 1600 802 1600
CutOff Geometry :box z
Coulomb cutoff potential :box z 56.000
Box sides length [au]: 0.00000 0.00000 56.00000
Sphere/Cylinder radius [au]: 0.000000
Cylinder length [au]: 0.000000
RL components : 31395
RL components used in the sum : 182079
RIM corrections included :yes
RIM RL components : 105
RIM random points : 1000000
- S/N 003636 -------------------------- v.04.05.00 r.09122 -

[05.01] External/Internal QP corrections
========================================

[K]Fermi Level [ev]: 0.000000
[K]VBM / CBm [ev]: 0.000000 3.222326
[K]Electronic Temp. [ev K]: 0.00 0.00
[K]Bosonic Temp. [ev K]: 0.00 0.00
[K]Finite Temperature mode: no
[K]El. density [cm-3]: 0.129E+24
[K]States summary : Full Metallic Empty
0001-0018 0019-0400
[K]Indirect Gaps [ev]: 3.222326 5.421104
[K]Direct Gaps [ev]: 4.379939 6.607738
[QP apply] Ind. Gap Correction [ev]: 1.603000

Timing [Min/Max/Average]: 02s/02s/02s

[06] Dipoles
============


[WARNING] DIPOLES database not correct or not present
[RD./SAVE//ns.kb_pp_pwscf]----------------------------------
Fragmentation :yes
- S/N 003636 -------------------------- v.04.05.00 r.09122 -

[WARNING] [x,Vnl] slows the Dipoles computation. To neglect it rename the ns.kb_pp file
[WF-Oscillators/G space loader] Normalization (few states) min/max :0.111E-10 1.00

[WR./2D_WR_WC//ndb.dipoles]---------------------------------
Brillouin Zone Q/K grids (IBZ/BZ): 802 1600 802 1600
RL vectors (WF): 31395
Fragmentation :yes
Electronic Temperature [K]: 0.000000
Bosonic Temperature [K]: 0.000000
X band range : 1 400
X band range limits : 18 19
X e/h energy range [ev]:-1.000000 -1.000000
RL vectors in the sum : 31395
[r,Vnl] included :yes
Bands ordered :yes
Direct v evaluation :no
Field momentum norm :0.1000E-4
Approach used :G-space v
Dipoles computed :R V P
Wavefunctions :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
- S/N 003636 -------------------------- v.04.05.00 r.09122 -

Timing [Min/Max/Average]: 24m-19s/30m-58s/26m-09s

[07] Dynamic Dielectric Matrix (PPA)
====================================

[WR./2D_WR_WC//ndb.pp]--------------------------------------
Brillouin Zone Q/K grids (IBZ/BZ): 802 1600 802 1600
RL vectors (WF): 31395
Coulomb cutoff potential :box z 56.000
Fragmentation :yes
Electronic Temperature [K]: 0.000000
Bosonic Temperature [K]: 0.000000
PPA diel. fun. energies :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
wavefunctions :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
Global Gauge :length
X matrix size : 1371
X band range : 1 300
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 : 31395
[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
- S/N 003636 -------------------------- v.04.05.00 r.09122 -
[RD./2D_WR_WC//ndb.dipoles]---------------------------------
Brillouin Zone Q/K grids (IBZ/BZ): 802 1600 802 1600
RL vectors (WF): 31395
Fragmentation :yes
Electronic Temperature [K]: 0.000000
Bosonic Temperature [K]: 0.000000
X band range : 1 400
X band range limits : 18 19
X e/h energy range [ev]:-1.000000 -1.000000
RL vectors in the sum : 31395
[r,Vnl] included :yes
Bands ordered :yes
Direct v evaluation :no
Field momentum norm :0.1000E-4
Approach used :G-space v
Dipoles computed :R V P
Wavefunctions :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
- S/N 003636 -------------------------- v.04.05.00 r.09122 -

Regards,
Oyeniyi Ezekiel

[Daniele Varsano]

Dear Ezekiel,

from the report only it is not easy to spot the problem.
Can you upload your input, complete report file and eventually one of the log files? You can use the attachment button below, but you need to rename the files with an allowed format (e.g. .txt or .zip). Nest please fill your signature with your affiliation, you can do once for all filling the signature in the user profile.

Best,
Daniele

[ezekiel]

Dear Daniele,
Thank you for your prompt response. Attached are the report, log, and input files.

Regards

[Daniele Varsano]

Dear Ezekiel,
even if not totally sure, it seems that yambo is not able to calculate the screening for memory reason.
You have a very large number of k-points and you should try to reduce the parameters of your calculation, at least to see if the problem is solved.
What I suggest you is the following, you can try all of them or a combination in sparse order

*Reduce the volume of the supercell in the orthogonal direction, you are using a coulomb cutoff and most probably you can get reasonable results with a smaller cell
* Consider to use a more balanced parallelization over bands (i.e. X_CPU= "1 1 1 40 4"), in general Nc/CPU_c \sim Nv/CPU_v is a good choice)
* reduce the energy cutoff in the screening (you are asking for 12Ry, but it seems you plan to use 7Ry)
* if you are not interested in GW correction, you can calculate static screening, here you have a plasmon pole screening.
* update to yambo 5.0 (git clone git@github.com:yambo-code/yambo.git)
* compile the code using the --enable-memory-profile flag in the configure, this will not solve the problem but it will allow to track the needed memory

Hopefully with some of these measures you can reduce the needed memory. Note that compiling enabling the memory profile, could help to see how much memory is needed and consequently tune the number of cpu on each node to fulfil the requirements.

Best,
Daniele

[ezekiel]

Dear Daniele,
I reduced values of parameters as you suggested but simultaneously. The screening part of the calculation was successful but the calculations again stopped at '[08.04] BSE solver(s) @q1' Attached are the report, log, and input files. Please, what is the way out?
Thank you.

[Daniele Varsano]

Dear Ezekiel,
I can see that the BSE matrix is built, then there are problems in the diagonalization.
You matrix is rather large 16kx16k, so it could also be a memory issue.


My suggestion is:

1)Check first in your job output if you have any error message, e.g. if it was a seg. fault or the job expired because of the wall time. I say this because the digoanalization of such a large matrix takes time.

2) Try to diagonaliza iteratively

Code: Select all

BSSmod= "h"

in this way you will have the spectrum, but not the eigenvectors.

Next, if you need the eigenvector, you can still extract the first eigenvectors linking the slepc libraries.

3) You can still try to diagonalize directly the matrix using parallel linear algebra: you need to compile yambo linking the scalapack libs and the use

Code: Select all

BS_nCPU_LinAlg_DIAGO=ncpu

4) Other option is to allow more memory per core, by using a smaller number of cpu per node. In order to do that you would have to recalculate the BSE matrix as the databases depends on the number of used cpus.

Best,
Daniele

[ezekiel]

Dear Daniele,
I diagonalized the BSE matrix iteratively and it ran successfully. I have the following output files:
o-2D_WR_WC.eel_q1_haydock_bse
o-2D_WR_WC.eps_q1_haydock_bse

I also expected to have a file labelled: o-2D_WR_WC.eps_alpha_q1_haydock_bse since I'm dealing with a reduced dimensional system (2D-system) but I can't find such file. Please, is any wrong with my calculations?

[ezekiel]

Dear Daniele,
I diagonalized the BSE matrix iteratively and it ran successfully. I have the following output files:
o-2D_WR_WC.eel_q1_haydock_bse
o-2D_WR_WC.eps_q1_haydock_bse

But, I also expected to have a file labelled: o-2D_WR_WC.eps_alpha_q1_haydock_bse since I'm dealing with a reduced dimensional system (2D-system) but I can't find such file. Please, is anything wrong with my calculations?

[Daniele Varsano]

Dear Ezekiel,
indeed it should be present, can you post the report file of your calculation?

Best,
Daniele

[ezekiel]

Dear Daniele,
I have attached the report file. Thank you.