FLEUR code plugin

Description

The FleurCalculation runs Fleur executable e.g. fleur or fleur_MPI.

../../_images/fleur_calc.png

Inputs

To set up an input dictionary, consider using get_inputs_fleur() which assembles input nodes in a ready-to-use single dictionary.

The table below shows all possible inputs for the FleurCalculation:

name

type

description

required

code

Code

Fleur code

yes

fleurinp

FleurinpData

Object representing inp.xml

no

parent_folder

RemoteData

Remote folder of another calculation

no

settings

Dict

special settings

no

metadata.options

Dict

computational resources

yes

  • fleurinp: FleurinpData, optional - Data structure which represents the inp.xml file and everything a Fleur calculation needs. For more information see FleurinpData.

  • parent_folder: RemoteData, optional - If specified, certain files in the previous Fleur calculation folder are copied in the new calculation folder.

Note

fleurinp and parent_folder are both optional. Depending on the setup of the inputs, one of five scenarios will happen:

  1. fleurinp: files belonging to fleurinp will be used as input for FLEUR calculation.

  2. fleurinp + parent_folder (FLEUR): files, given in fleurinp will be used as input for FLEUR calculation. Moreover, initial charge density will be copied from the folder of the parent calculation.

  3. parent_folder (FLEUR): Copies inp.xml file and initial charge density from the folder of the parent FLEUR calculation.

  4. parent_folder (input generator): Copies inp.xml file from the folder of the parent inpgen calculation.

  5. parent_folder (input generator) + fleurinp: files belonging to fleurinp will be used as input for FLEUR calculation. Remote folder is ignored.

Outputs

The table below shows all the output nodes generated by FleurCalculation:

name

type

comment

output_parameters

Dict

contains parsed out.xml

remote_folder

FolderData

represents calculation folder

retrieved

FolderData

represents retrieved folder

All the outputs can be found in calculation.outputs.

  • remote_folder: RemoteData - RemoteData which represents the calculation folder on the remote machine.

  • retrieved: FolderData - FolderData which represents the retrieved folder on the remote machine.

  • output_parameters: Dict - Contains all kinds of information of the calculation and some physical quantities of the last iteration.

    An example output node:

    # -*- coding: utf-8 -*-
    (aiidapy)% verdi data dict show 425
    {
        'CalcJob_uuid': 'a6511a00-7759-484a-839d-c100dafd6118',
        'bandgap': 0.0029975592,
        'bandgap_units': 'eV',
        'charge_den_xc_den_integral': -3105.2785777045,
        'charge_density1': 3.55653e-05,
        'charge_density2': 6.70788e-05,
        'creator_name': 'fleur 27',
        'creator_target_architecture': 'GEN',
        'creator_target_structure': ' ',
        'density_convergence_units': 'me/bohr^3',
        'end_date': {
            'date': '2019/07/17',
            'time': '12:50:27'
        },
        'energy': -4405621.1469633,
        'energy_core_electrons': -99592.985569309,
        'energy_hartree': -161903.59225823,
        'energy_hartree_units': 'Htr',
        'energy_units': 'eV',
        'energy_valence_electrons': -158.7015525598,
        'fermi_energy': -0.2017877885,
        'fermi_energy_units': 'Htr',
        'force_largest': 0.0,
        'magnetic_moment_units': 'muBohr',
        'magnetic_moments': [
            2.7677822875,
            2.47601e-05,
            2.22588e-05,
            6.05518e-05,
            0.0001608849,
            0.0001504687,
            0.0001321699,
            -3.35528e-05,
            1.87169e-05,
            -0.0002957294
        ],
        'magnetic_spin_down_charges': [
            5.8532354421,
            6.7738647125,
            6.8081938915,
            6.8073232631,
            6.8162583243,
            6.8156475799,
            6.8188399492,
            6.813423175,
            6.7733972589,
            6.6797683064
        ],
        'magnetic_spin_up_charges': [
            8.6210177296,
            6.7738894726,
            6.8082161503,
            6.8073838149,
            6.8164192092,
            6.8157980486,
            6.8189721191,
            6.8133896222,
            6.7734159758,
            6.679472577
        ],
        'number_of_atom_types': 10,
        'number_of_atoms': 10,
        'number_of_iterations': 49,
        'number_of_iterations_total': 49,
        'number_of_kpoints': 240,
        'number_of_species': 1,
        'number_of_spin_components': 2,
        'number_of_symmetries': 2,
        'orbital_magnetic_moment_units': 'muBohr',
        'orbital_magnetic_moments': [],
        'orbital_magnetic_spin_down_charges': [],
        'orbital_magnetic_spin_up_charges': [],
        'output_file_version': '0.27',
        'overall_charge_density': 7.25099e-05,
        'parser_info': 'AiiDA Fleur Parser v0.2beta',
        'parser_warnings': [],
        'spin_density': 7.91911e-05,
        'start_date': {
            'date': '2019/07/17',
            'time': '10:38:24'
        },
        'sum_of_eigenvalues': -99751.687121869,
        'title': 'A Fleur input generator calulation with aiida',
        'unparsed': [],
        'walltime': 7923,
        'walltime_units': 'seconds',
        'warnings': {
            'debug': {},
            'error': {},
            'info': {},
            'warning': {}
        }
    }
    

Errors

Errors of the parsing are reported in the log of the calculation (accessible with the verdi process report command). Everything that Fleur writes into stderr is also shown here, i.e all JuDFT error messages. Example:

(aiidapy)% verdi process report 513
*** 513 [scf: fleur run 1]: None
*** (empty scheduler output file)
*** (empty scheduler errors file)
*** 3 LOG MESSAGES:
+-> ERROR at 2019-07-17 14:57:01.108964+00:00
| parser returned exit code<302>: FLEUR calculation failed.
+-> ERROR at 2019-07-17 14:57:01.097337+00:00
| FLEUR calculation did not finishsuccessfully.
+-> WARNING at 2019-07-17 14:57:01.056220+00:00
| The following was written into std error and piped to out.error :
|  I/O warning : failed to load external entity "relax.xml"
| rm: cannot remove ‘cdn_last.hdf’: No such file or directory
| **************juDFT-Error*****************
| Error message:e>vz0
| Error occurred in subroutine:vacuz
| Hint:Vacuum energy parameter too high
| Error from PE:0/24

Moreover, all warnings and errors written by Fleur in the out.xml file are stored in the ParameterData under the key warnings, and are accessible with Calculation.res.warnings.

More serious FLEUR calculation failures generate a non-zero exit code. Each exit code has it’s own reason:

Exit code

Reason

300

One of output files can not be opened

301

No retrieved folder found

302

FLEUR calculation failed for unknown reason

303

XML output file was not found

304

Parsing of XML output file failed

305

Parsing of relax XML output file failed

310

FLEUR calculation failed due to memory issue

311

FLEUR calculation failed because atoms spilled to the vacuum

312

FLEUR calculation failed due to MT overlap

313

FLEUR calculation failed due to MT overlap during relaxation

314

Problem with cdn is suspected

315

Invalid Elements found in the LDA+U density matrix.

316

Calculation failed due to time limits.

Parallelization options

For parallel FLEUR calculations the input under metadata.options can be used. In higher level workchains this input might be present as a plain options input, but it is completely equivalent to the metadata.options input.

inputs.metadata.options = {
  'resources': {
    'num_machines': 2, #Number of computing nodes
    'num_mpiprocs_per_machine': 4, #Number of MPI processes per node
    'num_cpus_per_mpiproc': 12, #Number of OMP threads per MPI process
  },
  'withmpi': True, #This flag makes sure that the process is submitted using MPI
  'max_wallclock_seconds': 3600, #Maximum wallclock time in seconds
}

This will result in setting the following slurm Parallelization variables in the submit script.

#SBATCH --nodes=1
#SBATCH --ntasks-per-node=6
#SBATCH --cpus-per-task=8
#SBATCH --time=01:00:00
#... Further configuration options unrelated to parallelization...

'srun' '/path/to/fleur/' '<further FLEUR cmdline flags, e.g. -last_extra>'

Note, that the srun command is computer specific and is configured in verdi computer setup with the Mpirun command option.

Additional advanced features

In general see the FLEUR documentation.

While the input link with name fleurinp is used for the content of the inp.xml, additional parameters for changing the plugin behavior, can be specified in the settings input, also of type Dict.

Below we summarise some of the options that you can specify, and their effect. In each case, after having defined the content of settings_dict, you can use it as input of a calculation calc by doing:

calc.use_settings(Dict(dict=settings_dict))

Adding command-line options

If you want to add command-line options to the executable (particularly relevant e.g. ‘-hdf’ use hdf, or ‘-magma’ use different libraries, magma in this case), you can pass each option as a string in a list, as follows:

settings_dict = {
    'cmdline': ['-hdf', '-magma'],
}

The default command-line of a fleur execution of the plugin looks like this for the torque scheduler:

'mpirun' '-np' 'XX' 'path_to_fleur_executable' '-wtime' 'XXX' < 'inp.xml' > 'shell.out' 2> 'out.error'

If the code node description contains ‘hdf5’ in some form, the plugin will use per default hdf5, it will only copy the last hdf5 density back, not the full cdn.hdf file. The Fleur execution line becomes in this case:

'mpirun' '-np' 'XX' 'path_to_fleur_executable' '-last_extra' '-wtime' 'XXX' < 'inp.xml' > 'shell.out' 2> 'out.error'

Retrieving more files

AiiDA-FLEUR does not copy all output files generated by a FLEUR calculation. By default, the plugin copies only out.xml, cdn1 and inp.xml and other technical files. Depending on certain switches in used inp.xml, the plugin is capable of automatically adding additional files to the copy list:

  • if band=T : bands.1, bands.2

  • if dos=T : DOS.1, DOS.2

  • if pot8=T : pot*

  • if l_f=T : relax.xml

If you know that your calculation is producing additional files that you want to retrieve (and preserve in the AiiDA repository in the long term), you can add those files as a list as follows (here in the case of a file named testfile.txt):

settings_dict = {
  'additional_retrieve_list': ['testfile.txt'],
}

Retrieving less files

If you know that you do not want to retrieve certain files(and preserve in the AiiDA repository in the long term). i.e. the cdn1 file is to large and it is stored somewhere else anyway, you can add those files as a list as follows (here in the case of a file named testfile.txt):

settings_dict = {
  'remove_from_retrieve_list': ['testfile.txt'],
}

Copy more files remotely

The plugin copies by default the mixing_history* files if a parent_folder is given in the input.

If you know that for your calculation you need some other files on the remote machine, you can add those files as a list as follows (here in the case of a file named testfile.txt):

settings_dict = {
  'additional_remotecopy_list': ['testfile.txt'],
}

Copy less files remotely

If you know that for your calculation do not need some files which are copied per default by the plugin you can add those files as a list as follows (here in the case of a file named testfile.txt):

settings_dict = {
  'remove_from_remotecopy_list': ['testfile.txt'],
}