Dimension mismatch between L and G

[hellozhaoming]

Description

I am leaning DMFT with this example: https://triqs.github.io/dft_tools/unstable/tutorials/nio_csc_vasp/nio_csc.html
The NiO case didn't show any errors during the one-shot dmft run.
However, there will be an error if I replace Ni to Eu.

Steps to Reproduce

1. replace Ni to Eu in VASP files, and conduct a structural optimization and static scf.
2. add the follow parameter in INCAR and rerun the scf:
   LOCPROJ ="1 : f : Pr
   2 : p : Pr"
3. add a plo.cfg file

plo.cfg:
——————————————————————————————
[General]

HK = True
COMPLEMENT = False

[Group 1]
SHELLS = 1 2
EWINDOW = -5 10
NORMION = False
NORMALIZE = True

[Shell 1]
LSHELL = 3
IONS = 1
CORR = True

[Shell 2]
LSHELL = 1
IONS = 2
CORR = False
————————————————————————

4. run "python conver.py"
   conver.py:

from triqs_dft_tools.converters.vasp import *
import triqs_dft_tools.converters.plovasp.converter as plo_converter
plo_converter.generate_and_output_as_text('plo.cfg', vasp_dir='./')
Converter = VaspConverter(filename = 'vasp', proj_or_hk = 'hk')
Converter.convert_dft_input()

4. run "python nio.py"

the original nio.py is from https://triqs.github.io/dft_tools/unstable/tutorials/nio_csc_vasp/nio.py.html#nio-py. And replace "l=2" into "l=3" in the two lines

U_sph = U_matrix_slater(l=3, U_int=U, J_hund=J)
U_cubic = transform_U_matrix(U_sph, spherical_to_cubic(l=3, convention='vasp'))

There will be an error:

File "/home/xzm/.conda/envs/triqs/lib/python3.12/site-packages/triqs/gf/gf.py", line 695, in from_L_G_R
assert L.shape[1] == G.target_shape[0], "Dimension mismatch between L and G"
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/xzm/.conda/envs/triqs/lib/python3.12/site-packages/triqs_dft_tools/sumk_dft.py", line 1892, in symm_deg_gf
helper.from_L_G_R(v.conjugate().transpose(), gf_to_symm[key], v)
File "/home/xzm/data/2025/test-dmft/nio.py", line 123, in
SK.symm_deg_gf(S.Sigma_iw, ish=0)
File "/home/xzm/.conda/envs/triqs/lib/python3.12/runpy.py", line 88, in _run_code
exec(code, run_globals)
File "/home/xzm/.conda/envs/triqs/lib/python3.12/runpy.py", line 198, in _run_module_as_main (Current frame)
return _run_code(code, main_globals, None,
AssertionError: Dimension mismatch between L and G

Versions

triqs and triqs_dft_tools 3.3.1
vasp 6.5.1

input files

temp.zip

[the-hampel]

Hi @hellozhaoming ,

I moved this to discussions. There is no problem in the code here.

Your problem is that in the vasp.h5 file you provided is already a Sigma_itX stored. This Sigma is loaded in your nio.py script around line 107:

        if 'iteration_count' in ar['DMFT_results']:
            iteration_offset = ar['DMFT_results']['iteration_count']+1
            S.Sigma_iw = ar['DMFT_results']['Iterations']['Sigma_it'+str(iteration_offset-1)]
            SK.dc_imp = ar['DMFT_results']['Iterations']['dc_imp'+str(iteration_offset-1)]
            SK.dc_energ = ar['DMFT_results']['Iterations']['dc_energ'+str(iteration_offset-1)]
            SK.chemical_potential = ar['DMFT_results']['Iterations']['chemical_potential' +
                                                                     str(iteration_offset-1)].real

then this Sigma with has 5, not 7 orbitals, is attempted to be plugged into sumk. This does not work. Simply delete the vasp.h5 file and rerun the converter via python conver.py, and finally rerun the nio.py script should resolve your problem.

Another comment: your INCAR convergence settings are not tight enough for high precision projectors. I would set:

EDIFF = 1.E-10
NELMIN = 30

and check than rms converges to at least 1E-6 when running VASP. This should improve the projectors. You can check this when running the converter. Pay attention to the output of Density matrix and Local Hamiltonian and always make sure that complex parts are small and symmetries between orbitals are as expected.

Best,
Alex