'Exceeded job memory limit' when using NonCollinearSpin and DM.InitSpin (Siesta)

, ,
1

I’m doing a DOS/PDOS/Band structure calculation on a 2x2x2 bulk supercell where I replace an atom in the system with a Mn. However w/ the electrons in the d-orbital of Mn, I got my fermi energy shifting below my VBM due to the fact of how the energy states were being filled. Nevertheless, the job successfully converged and performed all requested calculations. As a solution, I sought out to control the spin of those electrons using the below addition to my .fdf file.

######################
NonCollinearSpin .true.

%block DM.InitSpin
16 + 0. 0.
%endblock DM.InitSpin
######################

When I ran the calculation I received an ‘Exceeded job memory limit’ error and the job quit after the .bands file was written (i.e. at the start of the DOS/PDOS calc.). After reading ([memory] pdos calculations and eigenvalue array (#45) · Issues · siesta-project / siesta · GitLab) I figured the number of k-points I was using was too large, so in order to test the extent that the error would go I chose extremely low k-point values (i.e. not reasonable results) in all possible facets with the following inputs:

######################
%block kgrid_Monkhorst_Pack
4 0 0 0.0
0 4 0 0.0
0 0 4 0.0
%endblock kgrid_Monkhorst_Pack

NonCollinearSpin .true.

%block DM.InitSpin
16 + 0. 0.
%endblock DM.InitSpin

BandLinesScale ReciprocalLatticeVectors
%block Bandlines
1 0.000 0.000 0.000 \Gamma
2 0.000 0.500 0.000 X
2 0.500 0.500 0.000 M
2 0.000 0.000 0.000 \Gamma
2 0.500 0.500 0.500 R
%endblock Bandlines

%block ProjectedDensityOfStates
-8.00 4.00 0.200 10 eV
%endblock ProjectedDensityOfStates
######################

Even w/ such few points, I received the same ‘Exceeded job memory limit’ error.

Is this a known ‘bug’ when using NonCollinearSpin and DM.InitSpin, or is there an obvious piece that I am missing? Any guidance is much appreciated.

Brian Leininger

2

Hi,
NonCollinearSpin and Initspin can be used together.
Could you share the whole fdf input file and the pseudopotentials you were using?

3

Thank you for the response. Is there a way to attach files on this platform? The fdf is short enough that I can copy/paste here, but is there another way to provide the pseudos?

Again the inputs for k-path, Monkhorst, and DOS are low for the purpose of testing the extent of the error. Also, this file runs fine w/o the NonCollinearSpin and DM.InitSpin.
.fdf file:
####################################
SystemName Mn-CsPbCl3
SystemLabel Mn-CsPbCl3

NumberOfAtoms 40

LatticeConstant 1 Ang
%block LatticeVectors
11.1600 0.00000 0.00000
0.00000 11.1600 0.00000
0.00000 0.00000 11.1600
%endblock LatticeVectors

XC.functional GGA
XC.authors PBEsol
MeshCutoff 200 Ry
MaxSCFIterations 200

NumberOfSpecies 4
%block ChemicalSpeciesLabel
1 55 Cs
2 82 Pb
3 17 Cl
4 25 Mn
%endblock ChemicalSpeciesLabel
%block AtomicMass
1 132.91
2 207.2
3 35.54
4 54.94
%endblock AtomicMass
%block Ps.lmax
Cs 2
Pb 2
Cl 1
Mn 2
%endblock Ps.lmax

PAO.SplitNorm 0.25

%block PAO.Basis
Pb 3
n=6 0 2
0.000 0.000
1.000 1.000
n=6 1 2 P 1
0.000 0.000
1.000 1.000
n=5 2 1
0.000
1.000
%endblock PAO.Basis

AtomicCoordinatesFormat Ang
%block AtomicCoordinatesAndAtomicSpecies
Pb 0.0000000000 0.0000000000 5.5800000000 2
Pb 0.0000000000 5.5800000000 0.0000000000 2
Pb 0.0000000000 5.5800000000 5.5800000000 2
Pb 5.5800000000 0.0000000000 0.0000000000 2
Pb 5.5800000000 0.0000000000 5.5800000000 2
Pb 5.5800000000 5.5800000000 0.0000000000 2
Pb 5.5800000000 5.5800000000 5.5800000000 2
Cs 2.7900000000 2.7900000000 2.7900000000 1
Cs 2.7900000000 2.7900000000 8.3700000000 1
Cs 2.7900000000 8.3700000000 2.7900000000 1
Cs 2.7900000000 8.3700000000 8.3700000000 1
Cs 8.3700000000 2.7900000000 2.7900000000 1
Cs 8.3700000000 2.7900000000 8.3700000000 1
Cs 8.3700000000 8.3700000000 2.7900000000 1
Cs 8.3700000000 8.3700000000 8.3700000000 1
Mn 0.0000000000 0.0000000000 0.0000000000 4
Cl 0.0000000000 0.0000000000 2.7900000000 3
Cl 0.0000000000 2.7900000000 0.0000000000 3
Cl 2.7900000000 0.0000000000 0.0000000000 3
Cl 0.0000000000 0.0000000000 8.3700000000 3
Cl 0.0000000000 2.7900000000 5.5800000000 3
Cl 2.7900000000 0.0000000000 5.5800000000 3
Cl 0.0000000000 5.5800000000 2.7900000000 3
Cl 0.0000000000 8.3700000000 0.0000000000 3
Cl 2.7900000000 5.5800000000 0.0000000000 3
Cl 0.0000000000 5.5800000000 8.3700000000 3
Cl 0.0000000000 8.3700000000 5.5800000000 3
Cl 2.7900000000 5.5800000000 5.5800000000 3
Cl 5.5800000000 0.0000000000 2.7900000000 3
Cl 5.5800000000 2.7900000000 0.0000000000 3
Cl 8.3700000000 0.0000000000 0.0000000000 3
Cl 5.5800000000 0.0000000000 8.3700000000 3
Cl 5.5800000000 2.7900000000 5.5800000000 3
Cl 8.3700000000 0.0000000000 5.5800000000 3
Cl 5.5800000000 5.5800000000 2.7900000000 3
Cl 5.5800000000 8.3700000000 0.0000000000 3
Cl 8.3700000000 5.5800000000 0.0000000000 3
Cl 5.5800000000 5.5800000000 8.3700000000 3
Cl 5.5800000000 8.3700000000 5.5800000000 3
Cl 8.3700000000 5.5800000000 5.5800000000 3
%endblock AtomicCoordinatesAndAtomicSpecies

%block kgrid_Monkhorst_Pack
4 0 0 0.0
0 4 0 0.0
0 0 4 0.0
%endblock kgrid_Monkhorst_Pack

NonCollinearSpin .true.

%block DM.InitSpin
16 + 0. 0.
%endblock DM.InitSpin

BandLinesScale ReciprocalLatticeVectors
%block Bandlines
1 0.000 0.000 0.000 \Gamma
2 0.000 0.500 0.000 X
2 0.500 0.500 0.000 M
2 0.000 0.000 0.000 \Gamma
2 0.500 0.500 0.500 R
%endblock Bandlines

%block ProjectedDensityOfStates
-8.00 4.00 0.200 10 eV
%endblock ProjectedDensityOfStates

DM.MixingWeight 0.1
DM.NumberPulay 10

KB.New.Reference.Orbitals .true.

SaveElectrostaticPotential .true.
WriteCoorXmol .true.
SaveTotalCharge .true.

WriteBands .true.
WriteKbands .true.

save-rho T
save-delta-rho T
save-total-potential T
save-neutral-atom-potential T
save-hs T
####################################

Thank you again.

Brian Leininger

4

Thanks for posting the fdf.

Now you should be able to upload psf, psml and fdf file format.

5

Thank you. Now it’s saying ‘Sorry, new users can not upload attachments.’ when I try to upload files.

Brian Leininger

6

I am sorry for the inconvenience, please try again.

7

No worries, thank you. Here is the fdf and pseudos (psf). I’m only allowed 2 attachments per post, so I will post the pseudos in following replies.

Mn-CsPbCl3.fdf (4.6 KB)

8

Cl and Cs.

Cl.psf (129.6 KB) Cs.psf (234.6 KB)

9

Pb and Mn.

Mn.psf (222.1 KB) Pb.psf (192.7 KB)

10

I could run the calculation with no problem.
It seems to be related to your specific environment.
Maybe you can try again with “ulimit -s unlimited” and/or “ulimit -u unlimited” before running the calculation.

11

Thanks for help. I had increased the number of processors several times, but unfortunately since I am a user on the environment (supercomputer), I don’t have permissions to use those commands.

That said, I moved to Siesta 4.1-b4 and began using some features such as the +U parameters that the manual states ‘Currently, the LDA+U implementation does not support non-colinear, nor spin-orbit coupling.’ As a result, the calculations run w/o any memory issues.

Thanks again for all the assistance.

12

From the link below you can have access to an experimental version of SIESTA with DFT+U with SOC.

Hope this can help.