sisl error running CrI3

[Tong Liu]

Hello, Dr.He Xu

    I am running tb2j on monolayer CrI3 then i get error below:

siesta2J.py --fdf_fname siesta-x.fdf --elements Cr --kmesh 4 4 1

TB2J version 0.7.4
Copyright (C) 2018-2020  TB2J group.
This software is distributed with the 2-Clause BSD License, without any warranty. For more details, see the LICENSE file delivered with this software.

Traceback (most recent call last):
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/bin/siesta2J.py", line 139, in <module>
    run_siesta2J()
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/bin/siesta2J.py", line 134, in run_siesta2J
    orb_decomposition=args.orb_decomposition
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/TB2J/manager.py", line 268, in gen_exchange_siesta
    H = fdf.read_hamiltonian()
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/io/siesta/fdf.py", line 2056, in read_hamiltonian
    H = getattr(self, f"_r_hamiltonian_{f.lower()}")(*args, **kwargs)
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/io/siesta/fdf.py", line 2092, in _r_hamiltonian_hsx
    H = hsxSileSiesta(f).read_hamiltonian(*args, **kwargs)
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/io/siesta/binaries.py", line 1141, in read_hamiltonian
    return getattr(self, f"_r_hamiltonian_v{version}")(**kwargs)
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/io/siesta/binaries.py", line 1103, in _r_hamiltonian_v1
    geom = self.read_geometry(**kwargs)
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/io/siesta/binaries.py", line 1048, in read_geometry
    return getattr(self, f"_r_geometry_v{version}")(**kwargs)
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/io/siesta/binaries.py", line 1031, in _r_geometry_v1
    atoms = self._read_atoms(**kwargs)
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/io/siesta/binaries.py", line 1005, in _read_atoms
    orbs.append(AtomicOrbital(n=n, l=l, m=m, zeta=zeta, R=-1.))
  File "/data/home/nm1021/nm1021/liuchang/software/anaconda3/envs/tb2j-new/lib/python3.7/site-packages/sisl/orbital.py", line 882, in __init__
    raise ValueError(f'{self.__class__.__name__} does not implement shell h and above!')
ValueError: AtomicOrbital does not implement shell h and above!

and i install sisl  0.12.2    pypi_0    pypi 

  tb2j        0.7.4                    pypi_0    pypi

siesta: siesta-master  https://gitlab.com/siesta-project/siesta

I guess sisl post-procedure is not matched with siesta-master version. and there are too many releases. could you tell me  right version of sisl and siesta??

 Thank you!!

[Tong Liu]

all input files from https://docs.siesta-project.org/projects/siesta/tutorials/tb2j/index.html  Exercise 2: isotropic/anisotropic exchange and DMI of CrI3 monolayer

and i also tried psml pesduo files and got same error...

[Xu He]

Hi, 

Could you send the siesta output file? I have encountered the same bug before. It was not a incompatibility issue but a bug in siesta, which  seems to be fixed in the most recent master version.  The other released versions should work, for example the Max versions unless you want to try some specific features from the git master version.  You can find some information here:

https://gitlab.com/siesta-project/siesta/-/wikis/Guide-to-Siesta-versions

From the error message it is using the binary HSX format file. Is there a file named siesta.nc? Both the HSX and the nc file contain the Hamiltonian, but the nc file is recommended . It requires siesta to be compiled with the netcdf library.  

Best regards,

HeXu

[Tong Liu]

Hello, I re-installed siesta with netcdf, yes, and i got same results with your excellent  tutorials. Thank you, Dr.Xu He!!!

  Next i tested monolayer CrI3 with different Hubbard U used TB2J combined with SIESTA-master. Jij seem changed significantly. So my question is when there is no experimental data, how to choose the right value of U.  I know we can get it with linear response or CRPA, but it is also possible to get questionable results. and I find three refereces used MFT to calcuate Jij and get good results by using downfolding methods (ref1&ref2 without Hubbard U for CrI3 ) and unspin-polarized DFT+U (ref3).

  So, I think it would be easy to use the PEB0 or HSE functional, which included exact exchange, althoug it can't slove ligand spin probrems. however my problem is that the current SIESTA does not calculate exact Hartree-Fock Exchange (although the author says there is a special version that can use HSE, but I can't find it in gitlab).

1.Exchange interactions and magnetic force theorem. PHYSICAL REVIEW B 103, 104428 (2021)

2.Microscopic origin of ferromagnetism in the trihalides CrCl3 and CrI3. PHYSICAL REVIEW B 99, 104432 (2019)

3.Relativistic exchange interactions in CrX3 (X = Cl, Br, I) monolayers.PHYSICAL REVIEW B 102, 115162 (2020)

  Thank you very much

  Tong Liu

[Xu He]

Hi, 

Great to know that you succeeded to make it work. 

- Indeed selecting the U value is  difficult, and sometimes the J values are very sensitive to the U values and +U methodologies. 

 - The hybrid functional is indeed a possible solution, although there is no systematic benchmark to show that it is superior when computing the exchange parameters.  

For the moment the hybrid is not openly accessible in SIESTA as far as I know. One possible way is to use another code which supports it, then use the Wannier90 interface. 

- Note that part of the downfolding algorithm in Ref1 is implemented in TB2J.  The usage of this technique is still quite rare so I think more tests are required. Please send the  feedback (good or bad) if you try it.

Best wishes,

HeXu