Support calculating the XPS spectra of the atoms specific by indices #958

superstar54 · 2023-06-20T10:11:22Z

This PR implements the XPS workchain to support calculating the XPS spectra of the atoms specific by indices.

Before, the XPS workchain only supports calculating the XPS spectra for selected elements by analyzing the symmetry and finding the non-equivalent site. This is not suitable for large systems with low symmetry, e.g. supported nanoparticles, in which all atoms are non-equivalent. And user usually only interested in the spectra of some special atoms.

This PR adds a new input atoms_list, e,g,

atoms_list = [0, 1, 2, 3, 7, 8]
builder = XpsWorkChain.get_builder_from_protocol(
    atoms_list=atoms_list,
    ...
)

superstar54 · 2023-06-26T07:08:24Z

Hi @mhdzbert could you test this PR? Hi @PNOGillespie , you can also try if you are interested in. Feedback is welcomed. Thanks!

PNOGillespie · 2023-06-27T11:21:05Z

Hi @superstar54. I think this is a good, simple addition which brings some more advanced control to XPS (and other analysis types which rely on a having a marked atom in the structure). Having done a simple test, I can see one necessary change already - plus some other suggestions for the WorkChain.

PNOGillespie · 2023-06-27T11:40:00Z

src/aiida_quantumespresso/workflows/xps.py

+            atoms_list = orm.List(self.ctx.atoms_list)
+            inputs = {
+                'atoms_list' : atoms_list,
+                'marker' : self.inputs.abs_atom_marker,
+                'metadata' : {
+                    'call_link_label' : 'get_marked_structures'
+                }
            }
-        } # populate this further once the schema for WorkChain options is figured out
-        if 'structure_preparation_settings' in self.inputs:
-            optional_cell_prep = self.inputs.structure_preparation_settings
-            for key, node in optional_cell_prep.items():
-                inputs[key] = node
-        if 'spglib_settings' in self.inputs:
-            spglib_settings = self.inputs.spglib_settings
-            inputs['spglib_settings'] = spglib_settings
-        else:
-            spglib_settings = None
-
-        if 'relax' in self.inputs:
-            relaxed_structure = self.ctx.relaxed_structure
-            result = get_xspectra_structures(relaxed_structure, **inputs)
-        else:
-            result = get_xspectra_structures(self.inputs.structure, **inputs)
-
-        supercell = result.pop('supercell')
-        out_params = result.pop('output_parameters')
-        if out_params.get_dict().get('structure_is_standardized', None):
-            standardized = result.pop('standardized_structure')
-            self.out('standardized_structure', standardized)
-
-        # structures_to_process = {Key : Value for Key, Value in result.items()}
-        for site in ['output_parameters', 'supercell', 'standardized_structure']:
-            result.pop(site, None)
+            result = get_marked_structures(input_structure, **inputs)
+            self.ctx.supercell = input_structure
+            self.ctx.equivalent_sites_data = result.pop('output_parameters').get_dict()


Since get_marked_structures() doesn't return a supercell_structure or symm_analysis_dict, these can't be sent to output. This causes a problem, because these two are required outputs - thus the calculation using an atoms_list input returns error code 11 (missing required output(s)):

Property Value ----------- ------------------------------------------------------------- type XpsWorkChain state Finished [11] The process did not register a required output. pk 200 uuid edb735da-d8f8-436f-b3bf-94519255f5f6 label XpsWorkChain (default) description ctime 2023-06-27 10:06:51.852633+00:00 mtime 2023-06-27 10:18:55.460074+00:00 Inputs PK Type ------------------------ ---- ------------- ch_scf clean_workdir 190 Bool max_iterations 189 Int kpoints_force_parity 188 Bool kpoints_distance 187 Float pw parallelization 186 Dict parameters 185 Dict structure 184 StructureData code 1 InstalledCode pseudos O 12 UpfData C 32 UpfData Li 30 UpfData core_hole_pseudos O 110 UpfData C 106 UpfData Li 107 UpfData gipaw_pseudos O 109 UpfData C 105 UpfData Li 108 UpfData abs_atom_marker 191 Str atoms_list 195 List calc_binding_energy 192 Bool clean_workdir 194 Bool core_hole_treatments 196 Dict correction_energies 193 Dict spglib_settings 197 Dict structure 184 StructureData voight_gamma 198 Float voight_sigma 199 Float Outputs PK Type ------------------------ ---- ------ chemical_shifts Li_cls 250 Dict O_cls 251 Dict C_cls 252 Dict final_spectra_cls Li_cls_spectra 253 XyData O_cls_spectra 254 XyData C_cls_spectra 255 XyData output_parameters_ch_scf site_0 238 Dict site_4 242 Dict site_6 246 Dict Called PK Type --------------------- ---- ---------------------- get_marked_structures 202 get_marked_structures site_0_xps 208 PwBaseWorkChain site_4_xps 210 PwBaseWorkChain site_6_xps 212 PwBaseWorkChain compile_final_spectra 249 get_spectra_by_element

Either changing the outputs to not be required, or integrating get_marked_structures() into get_xspectra_structures() (which does return this information) should fix the issue.

Thanks for pointing out this. I changed them to be not required.

PNOGillespie · 2023-06-27T12:02:41Z

One other point: I see that for get_builder_from_protocol() some of the inputs have to be included via **kwargs rather than having specific inputs:

        calc_binding_energy = kwargs.pop('calc_binding_energy', False)
        correction_energies = kwargs.pop('correction_energies', orm.Dict())

For one thing, calc_binding_energy should use orm.Bool(False) as a default value - as the default Python bool type isn't accepted as input. The other point is that both of these should be provided via specific inputs and then mapped to the proper AiiDA types as needed. That way, these optional inputs don't need to be included at all if they're not needed by the user.

These points aren't specific for this PR, so I wouldn't ask to address this now, but should be addressed at some point.

superstar54 · 2023-07-27T15:13:36Z

For one thing, calc_binding_energy should use orm.Bool(False) as a default value

I changed it to orm.Bool

The other point is that both of these should be provided via specific inputs and then mapped to the proper AiiDA types as needed. That way, these optional inputs don't need to be included at all if they're not needed by the user.

Could explain more about what this means? There is an input parameter called calc_binding_energy in the spec.inputs.

PNOGillespie · 2023-07-27T16:19:42Z

src/aiida_quantumespresso/workflows/xps.py

    def get_builder_from_protocol(
        cls, code, structure, pseudos, core_hole_treatments=None, protocol=None,
-        overrides=None, elements_list=None, options=None,
+        overrides=None, elements_list=None, atoms_list=None, options=None,
        structure_preparation_settings=None, **kwargs
    ):


The other point is that both of these should be provided via specific inputs and then mapped to the proper AiiDA types as needed. That way, these optional inputs don't need to be included at all if they're not needed by the user.

Could explain more about what this means? There is an input parameter called calc_binding_energy in the spec.inputs

You have defined these in the process spec, yes, however in get_builder_from_protocol(), there are no inputs fields in the function itself and instead we must provide both the correction energies and the switch to calculate the binding energies via kwargs instead - which isn't as easy to use or provide documentation for.

PNOGillespie · 2023-07-27T16:21:55Z

src/aiida_quantumespresso/workflows/xps.py

+        overrides=None, elements_list=None, atoms_list=None, options=None,
        structure_preparation_settings=None, **kwargs
    ):


Adding at least the correction_energies in the form of a standard Python dict object would probably be enough - since they are required for a calculation where calc_binding_energy = True, the function could just set calc_binding_energy = orm.Bool(True) if correction_energies is defined at all. If they aren't defined, then the function should logically set calc_binding_energy = orm.Bool(False)

def get_builder_from_protocol( ... structure_preparation_settings=None, correction_energies=None, **kwargs ):

if correction_energies: builder.correction_energies = orm.Dict(correction_energies) builder.calc_binding_energies = orm.Bool(True) else: builder.calc_binding_energies = orm.Bool(False)

Thanks! I changed the code as you suggested.

PNOGillespie

All good. I ran a quick live-fire test on my end with no issues. The latest change makes this a lot more intuitive to use IMO.

superstar54 added 2 commits June 20, 2023 07:39

support mark atoms by index

4fd922d

add test

3fdc492

superstar54 changed the title ~~Support~~ Support calculating the XPS spectra of the atoms specific by indices Jun 20, 2023

superstar54 added the type/enhancement label Jun 20, 2023

PNOGillespie reviewed Jun 27, 2023

View reviewed changes

set some outputs as non required

6f219a2

superstar54 requested a review from PNOGillespie July 27, 2023 15:13

PNOGillespie reviewed Jul 27, 2023

View reviewed changes

superstar54 added 2 commits July 28, 2023 14:04

add correction_energies explicitly

496ee15

Merge branch 'main' into fix_xps

9299947

superstar54 requested a review from PNOGillespie July 28, 2023 14:09

Merge branch 'main' into fix_xps

32d62b8

PNOGillespie approved these changes Sep 14, 2023

View reviewed changes

mbercx self-requested a review September 19, 2023 14:55

mbercx approved these changes Sep 19, 2023

View reviewed changes

superstar54 merged commit fc1a940 into aiidateam:main Sep 19, 2023
6 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support calculating the XPS spectra of the atoms specific by indices #958

Support calculating the XPS spectra of the atoms specific by indices #958

superstar54 commented Jun 20, 2023

superstar54 commented Jun 26, 2023

PNOGillespie commented Jun 27, 2023

PNOGillespie Jun 27, 2023

PNOGillespie Jun 27, 2023

superstar54 Jul 27, 2023

PNOGillespie commented Jun 27, 2023

superstar54 commented Jul 27, 2023

PNOGillespie Jul 27, 2023

PNOGillespie Jul 27, 2023

superstar54 Jul 28, 2023

PNOGillespie left a comment

Support calculating the XPS spectra of the atoms specific by indices #958

Support calculating the XPS spectra of the atoms specific by indices #958

Conversation

superstar54 commented Jun 20, 2023

superstar54 commented Jun 26, 2023

PNOGillespie commented Jun 27, 2023

PNOGillespie Jun 27, 2023

Choose a reason for hiding this comment

PNOGillespie Jun 27, 2023

Choose a reason for hiding this comment

superstar54 Jul 27, 2023

Choose a reason for hiding this comment

PNOGillespie commented Jun 27, 2023

superstar54 commented Jul 27, 2023

PNOGillespie Jul 27, 2023

Choose a reason for hiding this comment

PNOGillespie Jul 27, 2023

Choose a reason for hiding this comment

superstar54 Jul 28, 2023

Choose a reason for hiding this comment

PNOGillespie left a comment

Choose a reason for hiding this comment