Bandstructure wf #45
Bandstructure wf #45
Conversation
There was a problem hiding this comment.
Thank you @RubelMozumder. I left a few comments for improvements.
What is still needed in addition:
- you started from the stale branch
develop_workflow_conventions. Please merge the develop intoBandstructure_wfto get the latest changes (usegit merge --no-ff developon your Bandstructure_wf branch to do so) - add an entry point in the setup.json
- add a test for the bandstructure functionality see here for an example
- add documentation
- add/replace plotting functionality in plot_kkr tool to work with you workflow
- make you workflow importable:
from aiida_kkr.workflows import kkr_BS_wc
| __copyright__ = u"This is the Copyright" | ||
| __license__ = "Here is the license" | ||
| __version__ = "Here is the version" | ||
| __contributors__ = "using technique from Philpp RÜssmann's DOS wf" |
There was a problem hiding this comment.
This should contain the actual information and not just a placeholder
- copyright: FZJ
- license: MIT license
- version: start with 0.1.0
- contributors: your name
| Workchain for BandStructure calculation, starting from RemoteFolderData the previous converged KKR calculation remote folder data | ||
|
|
||
| inputs: | ||
| :param wf_parameters: (Dict), (optional); Workchain Specifications, contains npt2, tempr, emin(ev), emax(ev), kmesh | ||
| keys. The Energy emin and emax are the energy difference from the fermi energy. | ||
|
|
||
| :param options: (Dict), (optional); Computer Specifications, Schedualer command, parallel or serial | ||
| :param kpoints: (KpointsData),(optional); Kpoints data type from thestructure, | ||
| but not mendatory as it can be extrruct from structure internaly from the remote data | ||
| :param remote_data: (RemoteData)(mendaory); From the previous kkr-converged calcualtion. | ||
| :param kkr: (Code)(mendaory); KKR code specifiaction | ||
| :param label: (Str): | ||
| :param description: (Str): |
There was a problem hiding this comment.
fix types and complete missing information
|
|
||
| returns: | ||
| :out BS_Data : (ArrayData) ; Consist of (Dos_intensity matrix), k_points (list), Energy_levels (list), special_kpoints(dict) | ||
| :out result_wf: (Dict); work_chain_specifications node, BS_data node, remote_folder node |
There was a problem hiding this comment.
why do you want to return the remote folder node? It is not needed to continue a calculation from the output of the bandstructure calculation. So we should not expose this.
|
|
||
|
|
||
| returns: | ||
| :out BS_Data : (ArrayData) ; Consist of (Dos_intensity matrix), k_points (list), Energy_levels (list), special_kpoints(dict) |
There was a problem hiding this comment.
rename:
- Dos_intensity matrix -> BlochSpectralFunction
- Energy_levels -> energy_points
| 'EMAX': 5, # Energy over the fermi surface for energy contour in ev unit | ||
| "NPT2": 200, # Energy points in the energy contour | ||
| "kmesh": [30, 30, 30], | ||
| 'TEMPR': 50. # temperature |
There was a problem hiding this comment.
we should add the possibility to overwrite the cluster radius (RCLUSTZ) as well
There was a problem hiding this comment.
Why do the RCLUSTZ will have any impact on the on the DOS calc? Because, it is already converged.
There was a problem hiding this comment.
for a qdos calculation you run at much lower values of the imaginary part of the energies (given by NPOL=0 and a small value of the temperature). This puts higher demands on the numerics which in the end require larger screening clusters. Otherwise you will see negative densities.
| loaded_file = np.loadtxt(f) | ||
| total_dos[:,5:] += loaded_file[:,5:] | ||
|
|
||
| ef = fermi_level.value # in eV unit |
There was a problem hiding this comment.
this is in Ry, the conversion to eV units is only done in the next line!
There was a problem hiding this comment.
In [5]: ef = out['fermi_energy']
In [6]: ef
Out[6]: 0.6360325445
In [12]: ef = out['fermi_energy']
In [13]: eng_unit = out['energy_unit']
In [14]: eng_unit
Out[14]: 'eV'
This is from output_parameter of the parent cal. This make me a bit confused as I extract the value in the same way.
There was a problem hiding this comment.
the energy_unit refers to the energy value and the fermi_energy_unit gives you the unit of the fermi_energy which is in Ry and not in eV
There was a problem hiding this comment.
Oh, sorry. Thanks.
| total_dos[:,5:] += loaded_file[:,5:] | ||
|
|
||
| ef = fermi_level.value # in eV unit | ||
| total_dos[:,0] = total_dos[:,0]*eVscale - ef*np.ones(np.shape(total_dos[:,0])) |
There was a problem hiding this comment.
you could also write total_dos[:,0] = (total_dos[:,0]-ef)*eVscale
|
|
||
| dos_intensity = dos_intensity.T # setting eng-kpts corresponds to x-y asix | ||
| q_vec = np.asarray(q_vec) # converting q_vec into array | ||
| eng_points = np.asarray(list(eng_points)) # converting eng_popints into array |
There was a problem hiding this comment.
since eng_points is a set before, you should make sure to sort the array afterwards
| array.set_array('BS_Matrix', dos_intensity) | ||
| array.set_array('Kpts', q_vec) | ||
| array.set_array('Eng_levels',eng_points) |
|
|
||
|
|
||
|
|
||
| class kkr_BS_wf(WorkChain): |
There was a problem hiding this comment.
rename to kkr_bs_wc to be consistent with the other workflow names
…ing explanation and datatype, the removal of the 'remote_folder' from the output of the WC, using the conventional name 'Dos_intensity matrix -> BlochSpectralFunction', 'Energy_levels -> energy_points', To 'RCLUSTZ' to the bs_param, removal the k-mesh from the bs_param, changing the 'wf_option_default' to the normal specifications from the CLAIX18, to set the 'exit_codes' instead of the 'if_' condition, return exit_code instead of raising and Error in case the parent_workflow is not the WorkChainNode, fixning the error for the units of energy (eV, Ry), removing the unnecessary keyword 'BZDIVIDE', sending a report informing the structure of the unitcell, return the exit_codes fro the 'dos_run=Flase', fixining the energy units, removing the unneccessary function 'ev_To_Ry', renaming 'total_dos# to 'total_qdos'and 'num_dos_files' to the 'num_qdos_files', adding 'total_dos[:,0] = (total_dos[:,0]-ef)*eVscale' instead of 'total_dos[:,0] = total_dos[:,0]*eVscale - ef*np.ones(np.shape(total_dos[:,0]))', to make the 'eng_points' array immediately after the set of 'eng_points', to transform 'Dos_intensity matrix -> BlochSpectralFunction', 'Energy_levels -> energy_points' as convention, to rewrite the 'kkr_BS_wf' to 'kkr_bs_wc'.
Add bandstructure workflow