Problem: CSC calculation with VASP+Wannier90

[LRH123LRH123]

Hello developers,

I am testing VASP+Wannier90 for CSC calculations.
My setup is:

VASP 6.5.1

Wannier90 3.1

TRIQS 3.3.x and solid_dmft 3.3.x (also tested with the unstable branch)

I tested the example in
solid_dmft-3.3.x/doc/debugging_examples/svo-csc-w90.

In the first step I obtained converged CHGCAR and WAVECAR.
However, when I run the second step, the program stalls: it stops producing any new output files, and no error message is reported.

The last part of the solid_dmft output is:

dft parameters
dft_code vasp
dft_exec /.../vasp_std
mpi_env default
mpi_exe mpirun
n_cores 32
n_iter 6
n_iter_first 6
plo_cfg plo.cfg
projector_type w90
store_eigenvals False
w90_exec /.../wannier90-3.1.0/wannier90.x
w90_tolerance 1e-06

solid_dmft: Running VASP...

MPI executable for Vasp: /.../mpirun

Starting VASP now

running 32 mpi-ranks, on 1 nodes
distrk: each k-point on 32 cores, 1 groups
distr: one band on 1 cores, 32 groups
vasp.6.5.1 10Mar25 (build Sep 25 2025 21:19:47) complex
POSCAR found type information on POSCAR SrV O
POSCAR found : 3 types and 5 ions
Reading from existing POTCAR
scaLAPACK will be used
...
found WAVECAR, reading the header
POSCAR, INCAR and KPOINTS ok, starting setup
FFT: planning ... GRIDC
FFT: planning ... GRID_SOFT
FFT: planning ... GRID
reading WAVECAR
the WAVECAR file was read successfully
charge-density read from file: SrVO3
entering main loop
N E dE d eps ncg rms rms(c)
DAV: 1 -0.390802040041E+02 -0.39080E+02 -0.38664E-10 1504 0.907E-05

At this point the calculation hangs and does not proceed further.

Question:
Do you have any suggestions on how to resolve this issue, or possible reasons why the calculation hangs after the first step?

Thank you very much for your help!

[the-hampel]

Hi @LRH123LRH123 ,

it is a bit weird, that there is no error message printed. Can you try to remove the w90 interface here and try the VASP converter directly? This one here: https://github.com/TRIQS/solid_dmft/tree/3.3.x/doc/debugging_examples/svo-csc-plo . Maybe with the output from that one we can narrow down the problem a little bit.

Try also to spawn vasp only with a single rank first. For this small problem this should be okay. Then we can rule out any MPI comm problems.

Best,
Alex

[LRH123LRH123]

@the-hampel
Hi Alex,

Thank you for your quick reply. I have tested the svo-csc-plo example using the same versions of the programs, and it runs correctly. The final output is as follows:

============================================================
summary of energetics:
total energy:  -38.900177740769074
DFT energy:  -39.08034855236792
correlation energy:  0.09353115704355551
DFT band correction:  0.08663965455529332
============================================================

============================================================
convergence:
δμ:      3.9309e-03
δE_tot:  1.5901e-02
---
Impurity 0:
δn imp : 7.3176e-03
δn orb : 4.0253e-03  3.8565e-03  7.4864e-03
δ Gimp : 6.7881e-03
δ G0   : 9.7206e-03
δ Σ    : 5.1788e+01
============================================================

** All requested iterations finished ***
################################################################################

================================================================================
DMFT cycle took    105.442 seconds
================================================================================

solid_dmft: Stopping VASP

DAV:  62    -0.390803485524E+02   -0.46219E-05   -0.34887E-08 26368   0.103E-03

---

However, I am still unclear why svo-csc-w90 example fails to run.

Additionally, I would like to ask whether it is possible to obtain the momentum-resolved spectral function directly from a CSC calculation using VASP+PLO, similar to the figures shown in the tutorial:
https://triqs.github.io/solid_dmft/tutorials/correlated_bandstructure/plot_correlated_bands.html#

Or, alternatively, can we first perform a CSC calculation based on VASP, then use the newly obtained CHGCAR to construct the Wannier90 Hamiltonian and perform a one-shot DMFT calculation to obtain the momentum-resolved spectral function?

Thanks for your help!

[the-hampel]

@LRH123LRH123 ,

no right now it is unfortunately not possible to get a momentum resolved spectral function directly from. Vasp I am working on such feature (has to be implemented in VASP).

Let's go back to your wannier90 example and try to figure out what goes wrong. I suspect now that something goes wrong in the wannier step. Can you run VASP in the ICHARG=5 mode, as solid_dmft would start it, but start it yourself, i.e. normally with mpirun path/to/vasp_std &> log & . But you have to create a file called vasp.lock before launching VASP. Just create it beforehand and leave it empty: touch vasp.lock . Do this in a way that you can take a look at the output files. I suspect it hangs at at the point where it creates wannier functions. If everything goes correctly VASP should remove the file vasp.lock and then wait, i.e. no more output. Let me know what you see.

Best,
Alex

[LRH123LRH123]

@the-hampel
Hi Alex,

Thank you for your reply. I tried running VASP in the ICHARG=5 mode as you suggested. The program indeed gets stuck at that point, and the vasp.lock file is automatically removed. However, no Wannier90-related files are generated.
For reference, my VASP normally works fine when calculating a standard Wannier Hamiltonian with ICHARG=11.

The VASP output looks like this:

 running   16 mpi-ranks, on    1 nodes
 distrk:  each k-point on   16 cores,    1 groups
 distr:  one band on    1 cores,   16 groups
 vasp.6.5.1 10Mar25 (build Sep 25 2025 21:19:47) complex                         
 POSCAR found type information on POSCAR SrV O 
 POSCAR found :  3 types and       5 ions
 Reading from existing POTCAR
 scaLAPACK will be used
 -----------------------------------------------------------------------------
|                                                                             |
|           W    W    AA    RRRRR   N    N  II  N    N   GGGG   !!!           |
|           W    W   A  A   R    R  NN   N  II  NN   N  G    G  !!!           |
|           W    W  A    A  R    R  N N  N  II  N N  N  G       !!!           |
|           W WW W  AAAAAA  RRRRR   N  N N  II  N  N N  G  GGG   !            |
|           WW  WW  A    A  R   R   N   NN  II  N   NN  G    G                |
|           W    W  A    A  R    R  N    N  II  N    N   GGGG   !!!           |
|                                                                             |
|     For optimal performance we recommend to set                             |
|       NCORE = 2 up to number-of-cores-per-socket                            |
|     NCORE specifies how many cores store one orbital (NPAR=cpu/NCORE).      |
|     This setting can greatly improve the performance of VASP for DFT.       |
|     The default, NCORE=1 might be grossly inefficient on modern             |
|     multi-core architectures or massively parallel machines. Do your        |
|     own testing! More info at https://www.vasp.at/wiki/index.php/NCORE      |
|     Unfortunately you need to use the default for GW and RPA                |
|     calculations (for HF NCORE is supported but not extensively tested      |
|     yet).                                                                   |
|                                                                             |
 -----------------------------------------------------------------------------

 Reading from existing POTCAR
 LDA part: xc-table for (Slater+PW92), standard interpolation
 found WAVECAR, reading the header
 POSCAR, INCAR and KPOINTS ok, starting setup
 FFT: planning ... GRIDC
 FFT: planning ... GRID_SOFT
 FFT: planning ... GRID
 reading WAVECAR
 the WAVECAR file was read successfully
 charge-density read from file: SrVO3                                   
 entering main loop
       N       E                     dE             d eps       ncg     rms          rms(c)
DAV:   1    -0.390802040045E+02   -0.39080E+02   -0.36235E-10  1280   0.907E-05

In addition, the INCAR file I used is as follows:

SYSTEM = SrVO3
ENCUT = 400
ISPIN = 1
ICHARG = 5 ! 5 for fcsc calculations (communication with python)
NBANDS = 32

NELM = 1000 ! to prevent VASP from stopping too early (either because EDIFF is reached or 60 iterations were done)
NELMIN = 1000
NELMDL = -2 ! first perform 2 iterations. 1 is enough when starting from converged WAVECAR

!! Smearing procedure
ISMEAR = -5 ! -4 (-5) Tetrahedron (with Bloechl), 0 Gaussian, 1,... Methfessel-Paxton
SIGMA = 0.01

LMAXMIX = 4
LOCPROJ = 1 : s : Pr

!! write WAVECAR, CHGCAR
LWAVE = .FALSE.
LCHARG = .FALSE.

LWANNIER90 = .TRUE.
NUM_WANN = 3
WANNIER90_WIN = "
guiding_centres = T

begin projections
V:dxy,dyz,dxz
end projections

dis_win_min = 4.1
dis_win_max = 6.9

write_hr = true
write_u_matrices = true

num_iter = 0
conv_tol = 1E-9
conv_window = 10

dis_num_iter = 2000
dis_conv_tol = 1E-9
dis_conv_window = 10

bands_plot      =  true
begin kpoint_path
G  .0  .0  .0   X  .0  .5  .0
X  .0  .5  .0   M  .5  .5  .0
M  .5  .5  .0   G  .0  .0  .0
end kpoint_path
bands_num_points 40
"

Thanks a lot for your help!

Best regards

[the-hampel]

I think I know what the problem is here for you. Did you add the modification mentioned here: https://triqs.github.io/solid_dmft/md_notes/vasp_csc.html#enabling-csc-calculations-with-wannier90-projectors to your VASP source code? To be able to use w90 for CSC you have to add this part before compiling VASP. Actually for VASP 6.5.1 the code changes is slightly different. Please change the lines in electron.F after 774 to:

      IF (MOD(INFO%ICHARG,10)==5) THEN
         IF (IO%LORBIT==14) CALL SPHPRO_FAST( &
              GRID,LATT_CUR, P,T_INFO,W, WDES, 71,IO%IU6,&
              INFO%LOVERL,LMDIM,CQIJ, LDIMP, LDIMP,LMDIMP,INNER_STEP, IO%LORBIT,PAR, &
              EFERMI, KPOINTS%EMIN, KPOINTS%EMAX)
         CALL LPRJ_PROALL(W,WDES,P,CQIJ,LATT_CUR,LPRJ_functions,LPRJ_COVL,T_INFO,INFO,IO)
         CALL LPRJ_WRITE(LPRJ_functions,LPRJ_COVL,IO%IU6, IO%IU0, W)
         IF (WANNIER90()) THEN
            CALL KPAR_SYNC_ALL(WDES,W)
            CALL MLWF_MAIN(WDES,W,P,KPOINTS,CQIJ,T_INFO,LATT_CUR,INFO,IO,MLWF_GLOBAL)
         ENDIF

The last part IF (WANNIER90()) THEN ... ENDIF has to be added. I assume you did not do this? Otherwise you would see some other crashes. Also note that you still have to change the order of files in src/.objects as mentioned in the above link. With this I am at least able to run now on one MPI rank VASP and get the wannier90 output. Make sure to add now in the INCAR LWRITE_MMN_AMN = .TRUE. (otherwise no mmn file is written for w90 to run). If I do this I can see how vasp runs the w90 interface before removing the vasp.lock file and waiting for new input:

DAV:   1    -0.390748323474E+02   -0.39075E+02   -0.31007E-10  2600   0.954E-05 Calling wannier_setup of wannier90 in library mode
 LOCPROJ mode                                                                                                                                          
 Computing MMN (overlap matrix elements)  

This should then work when launched from solid_dmft. However, I do see crashes with more than one mpi rank. Let me know if this works. I can look in the mpi issue at some point. Note however that this interface is not officially supported currently. But I can make a note to add this.

Best,
Alex

[LRH123LRH123]

Hi Alex,

Thank you very much for your reply and the suggested modification. With your changes, I am now able to generate the Wannier90-related files. However, there is still an issue:

When I run VASP, I get the following error:

running    4 mpi-ranks, on    1 nodes
distrk:  each k-point on    4 cores,    1 groups
distr:  one band on    1 cores,    4 groups
vasp.6.5.1 10Mar25 (build Sep 30 2025 18:04:32) complex
POSCAR found type information on POSCAR SrV O
POSCAR found :  3 types and       5 ions
Reading from existing POTCAR
scaLAPACK will be used
Reading from existing POTCAR
LDA part: xc-table for (Slater+PW92), standard interpolation
POSCAR, INCAR and KPOINTS ok, starting setup
FFT: planning ... GRIDC
FFT: planning ... GRID_SOFT
FFT: planning ... GRID
WAVECAR not read
charge-density read from file: SrVO3
entering main loop
      N       E                     dE             d eps       ncg     rms          rms(c)
DAV:   1     0.160222025210E+03    0.16022E+03   -0.88187E+03  2264   0.755E+02  
Calling wannier_setup of wannier90 in library mode
LOCPROJ mode
Computing MMN (overlap matrix elements)
Computing AMN (projections onto localized orbitals)

DAV:   2    -0.215148220985E+02   -0.18174E+03   -0.15857E+03  2512   0.220E+02  
At line 509 of file mlwf.F
Fortran runtime error: Attempting to allocate already allocated variable 'p_mlwf'

It seems that there is still a problem in mlwf.F, possibly related to multiple allocations of P_MLWF. I will continue to test and try to fix it.

Additionally, I have a question regarding the workflow: can we first perform a CSC calculation in VASP, then use the newly obtained CHGCAR to construct the Wannier90 Hamiltonian and perform a one-shot DMFT calculation to obtain the momentum-resolved spectral function? Would this procedure be considered a CSC DMFT calculation?

Thank you again for your help.

Best regards

[the-hampel]

Hi @LRH123LRH123 ,

I will also ask internally if there is any easy fix for this.

Regarding your second question: Yes I would agree if you mean by that 1) do every DMFT calculation with W90 (do not mix PLOs and W90) and 2) add to the final W90 of course the self-energy of the final DMFT calculation. That should work. However, you have to be very well converged in your CSC DFT+DMFT calculation (rms(c) in OSZICAR close to 0) to able to rely only on the CHGCAR of the final run. And also node that VASP only writes the CHGCAR when it finishes. Typically we run CSC DFT+DMFT in a way that we do not gracefully exit VASP in the end. So you might never get the final CHGCAR. I think I managed this with VASP 6.5.x once but it is a bit delicate to do. You can add a STOPCAR file at the right moment to make VASP stop correctly and write out WAVECAR / CHGCAR . Typically it would be better to just run your CSC DFT+DMFT until it is converged and use the last w90 output together with your final self-energy.

Best,
Alex