Periodicity: StructureData #907

AndresOrtegaGuerrero · 2023-04-11T08:49:26Z

This PR implements changes in the base, and relax workchain so it can create modifications when dealing with 1D or 2D StructureData. This implementation addresses the issues highlighted in #896

AndresOrtegaGuerrero · 2023-04-14T12:50:28Z

@mbercx let me know what you think

mbercx

Sorry @AndresOrtegaGuerrero! Had to look into the cell_dofree options a bit, since I'm not super familiar, I've left some questions. :)

As for the implementation, maybe it's a matter of preference, but I'm more of a fan of using dictionaries to deal with "cases" instead of if-elif-else statements:

            pbc_cell_dofree_map = {
                (True, True, True): 'all',
                (True, False, False): 'x',
                (False, True, False): 'y',
                (False, False, True): 'z',
                (True, True, False): '2Dxy',
            }
            try:
                base.pw.parameters['CELL']['cell_dofree'] = pbc_cell_dofree_map[structure.pbc]
            except KeyError:
                raise ValueError(
                    f'Structures with periodic boundary conditions `{structure.pbc}` are not supported.'
                )

mbercx · 2023-04-14T16:23:24Z

src/aiida_quantumespresso/workflows/pw/base.py

@@ -180,6 +180,10 @@ def get_builder_from_protocol(
        parameters['SYSTEM']['ecutwfc'] = cutoff_wfc
        parameters['SYSTEM']['ecutrho'] = cutoff_rho

+        #If 2D periodicity set assume_isolate
+        if (True, True, False) == structure.pbc:
+            parameters['SYSTEM']['assume_isolated'] = '2D'


I see, so this correction can only be used when the first two lattice vectors are entirely in the x-y plane. So running with (False, True, True) should not be allowed for this reason? Should the third lattice vector be perpendicular to this plane?

@mbercx Thank you for the question. So I included this option given the suggestion from @demiranda-gabriel , Usually If i were to do a slab simulation i would omit this part and just be sure that my structure has enough vacuum in the perpendicular direction of the slab. According to the manual this option is only for xy-plane. Though I am pro removing this line.

Hmm, since the get_builder_from_protocol method is meant to provide a set of sensible defaults, it might be reasonable to keep it since the user can always override it. One sentence in the QE documentation still worries me:

Total energy, forces and stresses are computed in a two-dimensional framework.

@demiranda-gabriel So QE would not consider any z-components of the forces? Wouldn't we want to include these when e.g. optimizing the surface layers?

Nevermind, it clearly does. The initial forces and stresses:

Forces acting on atoms (cartesian axes, Ry/au): atom 1 type 1 force = 0.00000000 0.00000000 -0.01049299 atom 2 type 1 force = 0.00000000 0.00000000 0.01049299

total stress (Ry/bohr**3) (kbar) P= -4.00 -0.00003190 -0.00000000 0.00000000 -4.69 -0.00 0.00 -0.00000000 -0.00003190 0.00000000 -0.00 -4.69 0.00 0.00000000 0.00000000 -0.00001767 0.00 0.00 -2.60

are reduced to

Forces acting on atoms (cartesian axes, Ry/au): atom 1 type 1 force = 0.00000000 0.00000000 0.00047838 atom 2 type 1 force = 0.00000000 0.00000000 -0.00047838

total stress (Ry/bohr**3) (kbar) P= 0.07 0.00000070 0.00000000 0.00000000 0.10 0.00 0.00 0.00000000 0.00000070 0.00000000 0.00 0.10 0.00 0.00000000 0.00000000 0.00000006 0.00 0.00 0.01

Input/output cell:

In [1]: relax_wc = load_node(16355) In [2]: relax_wc.inputs.structure.cell Out[2]: [[3.6086769266, 0.0, 0.0], [-1.8043384633, 3.1252058924, 0.0], [0.0, 0.0, 21.3114930844]] In [3]: relax_wc.outputs.output_structure.cell Out[3]: [[3.5615977430325, 0.0, 0.0], [-1.7807988715162, 3.0844341234181, 0.0], [0.0, 0.0, 21.311492898049]]

mbercx · 2023-04-14T16:25:25Z

src/aiida_quantumespresso/workflows/pw/relax.py

+            if x_dim and not (y_dim or z_dim):
+                base.pw.parameters['CELL']['cell_dofree'] = 'x'
+            elif y_dim and not (x_dim or z_dim):
+                base.pw.parameters['CELL']['cell_dofree'] = 'y'


Maybe you are more of an expert, but I'm wondering: assume you have a structure with pbc == [False, True, False], and the following CELL_PARAMETERS:

CELL_PARAMETERS angstrom 13.866974650 0.0000000000 0.0000000000 1.9334873250 3.3488982827 0.0000000000 1.9334873250 1.1162994276 13.157371580

I’m wondering if I set cell_dofree to 'y', will only the second component of the second lattice parameter (i.e. 3.3488982827) be allowed to change? If only the second component is allowed to change, also the direction of the lattice vector will change, which could deform the chain structure unless all atoms are on the second lattice vector, no?

In this scenario , and according to the manual 'y' : only the y component of axis 2 (v2_y) is moved given the cell parameters you are giving there is an angle in that direction so perhaps your pbe = True, False, True is not what you want

@mbercx Perhaps if there are so many options for cell_dofree the best for QEApp, is just to define the StructureData.pbc (via widget) and just use the logic from this PR externally (not in the RelaxWorkchain) to override the QEworkchain. Since for the QEApp we might just need to consider only a few cases that are 1D (along one direction x,y,z) , 2D (in xy) and 3D. In anycase I know that create_kpoints_from_distance will take into account StructureData.pbc to generate the mesh. Let me know your opinion I can close the PR

In this scenario , and according to the manual 'y' : only the y component of axis 2 (v2_y) is moved given the cell parametersyou are giving there is an angle in that direction so perhaps yourpbe = [True, False, True]` is not what you want

Yeah, it seems you typically would make sure the Xth lattice vector only has an X component. E.g.

CELL_PARAMETERS angstrom -24.711115283 0.0000 0.00000 0.00000 -24.808651431 0.00000 0.00000 0.000000 4.165100096

Would be a reasonable cell with cell_dofree = 'z'. So a user has to be careful, but we can also help by adding validation on the PwCalculation class (but I won't subject you to that ^^).

@mbercx Perhaps if there are so many options for cell_dofree the best for QEApp, is just to define the StructureData.pbc (via widget) and just use the logic from this PR externally (not in the RelaxWorkchain) to override the QEworkchain. Since for the QEApp we might just need to consider only a few cases that are 1D (along one direction x,y,z) , 2D (in xy) and 3D. In anycase I know that create_kpoints_from_distance will take into account StructureData.pbc to generate the mesh. Let me know your opinion I can close the PR

I understand this will probably be quicker for your use case, and it possible avoids creating bugs in the main aiida-quantumespresso plugin that other users could suffer from. However, long term I think it's definitely something that should be implemented in the aiida-quantumespresso plugin, and since it relies on the pbc it doesn't affect any user running 3D structures at all. So I'd be inclined to push through and get it in here, with a rapid release promised by yours truly.

@mbercx Ok, I will do the changes you suggested, I am pro removing the parameters['SYSTEM']['assume_isolated'] = '2D'

@AndresOrtegaGuerrero if it's ok for you, I was already doing the changes and adding a test, but got distracted by aiidateam/aiida-core#5967 when I actually tried to import a 2D structure. 😅

@mbercx no problem! , thank you

Hmm, seems I can't push to the nanotech-empa fork 😭. Could you give me "Write" permission? 😇 See

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/about-permissions-and-visibility-of-forks#about-permissions-of-forks

and

https://docs.github.com/en/organizations/managing-user-access-to-your-organizations-repositories/repository-roles-for-an-organization#permissions-for-each-role

…m/nanotech-empa/aiida-quantumespresso into feature/periodicity-structuredata

mbercx

Thanks a lot @AndresOrtegaGuerrero! This LGMT 🚀

Currently none of the `get_builder_from_protocol()` methods consider the periodic boundary conditions (`pbc`) of the input `StructureData`. Here we add support for several cases to the `PwBaseWorkChain` and `PwRelaxWorkChain`: * 2D structures in the x-y plane with periodic boundary conditions for the first two lattice vectors (`pbc == [True, True, False]`). For this case, the `PwBaseWorkChain` protocol will set `SYSTEM.assume_isolated` to `2D`, and the `PwRelaxWorkChain` protocol will use `CELL.do_free = '2Dxy'` in case the `RelaxType` is `CELL` or `POSITIONS_CELL`. * 1D structures defined on an arbitrary lattice vector will use the corresponding `CELL.do_free` setting (`'x'`, `'y'`, `'z'`) if `RelaxType` is `CELL` or `POSITIONS_CELL`. So far, no validation is done to see if the structure has the correct `cell` specified. This could potentially be added to the `PwCalculation` validation. Co-authored-by: Marnik Bercx <mbercx@gmail.com>

AndresOrtegaGuerrero and others added 3 commits April 11, 2023 08:40

changing variables naming style

da5c4cd

Merge branch 'main' into feature/periodicity-structuredata

d330787

Merge branch 'main' into feature/periodicity-structuredata

fa70e86

AndresOrtegaGuerrero mentioned this pull request Apr 13, 2023

Periodicity #896

Closed

Merge branch 'main' into feature/periodicity-structuredata

bf1fd9c

mbercx requested changes Apr 14, 2023

View reviewed changes

mbercx mentioned this pull request Apr 15, 2023

Importing .xyz structure with ase does not respect pbc aiidateam/aiida-core#5967

Open

AndresOrtegaGuerrero added 2 commits April 17, 2023 10:55

pre-commit fix

a3dab3b

Merge branch 'feature/periodicity-structuredata' of https://github.co…

52b69fc

…m/nanotech-empa/aiida-quantumespresso into feature/periodicity-structuredata

AndresOrtegaGuerrero requested a review from mbercx April 17, 2023 11:07

Merge branch 'main' into feature/periodicity-structuredata

13fe734

mbercx approved these changes Apr 17, 2023

View reviewed changes

mbercx merged commit 5687800 into aiidateam:main Apr 17, 2023

This was referenced Apr 17, 2023

Allow more options for cell_dofree in cell optimization #871

Open

Adapt implementation and add tests nanotech-empa/aiida-quantumespresso#1

Closed

This was referenced Jun 12, 2023

Monkhorst-Pack grid for 1D/2D structures in ph.x #954

Closed

Matdyn Q2R 2D Structure/ loto_2d #955

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Periodicity: StructureData #907

Periodicity: StructureData #907

AndresOrtegaGuerrero commented Apr 11, 2023

AndresOrtegaGuerrero commented Apr 14, 2023

mbercx left a comment

mbercx Apr 14, 2023

AndresOrtegaGuerrero Apr 14, 2023

mbercx Apr 15, 2023

mbercx Apr 16, 2023

mbercx Apr 14, 2023 •

edited

Loading

AndresOrtegaGuerrero Apr 14, 2023

AndresOrtegaGuerrero Apr 14, 2023

mbercx Apr 15, 2023

AndresOrtegaGuerrero Apr 15, 2023

mbercx Apr 15, 2023

AndresOrtegaGuerrero Apr 16, 2023

mbercx Apr 16, 2023

mbercx left a comment

Periodicity: StructureData #907

Periodicity: StructureData #907

Conversation

AndresOrtegaGuerrero commented Apr 11, 2023

AndresOrtegaGuerrero commented Apr 14, 2023

mbercx left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mbercx Apr 14, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mbercx left a comment

Choose a reason for hiding this comment

mbercx Apr 14, 2023 •

edited

Loading