Identifying catalyst and adsorbate

[chaitjo]

Hi! I was interested in identifying the catalyst and adsorbate atoms within the PyG Batch/Data object:

Batch(atomic_numbers=[172], batch=[172], cell=[2, 3, 3], cell_offsets=[7584, 3], distances=[7584], edge_index=[2, 7584], fixed=[172], force=[172, 3], natoms=[2], neighbors=[2], pos=[172, 3], pos_relaxed=[172, 3], tags=[172], y_init=[2], y_relaxed=[2])

Intuitively, I understand that all the catalyst atoms are packed at the end of the Data object attribute lists, e.g. the atomic numbers 6, 1, 1, 1, and 8:

> data.atomic_numbers[data.batch==0]
tensor([20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20.,
        20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20., 20.,
        20., 20., 20., 20., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28.,
        28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28.,
        28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28.,
        28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28.,
        28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28., 28.,  6.,  1.,
         1.,  1.,  8.], device='cuda:0')

> data.fixed[data.batch==0]
tensor([1., 1., 1., 1., 1., 0., 0., 1., 1., 1., 1., 1., 1., 0., 0., 1., 1., 1.,
        1., 1., 1., 0., 0., 1., 1., 1., 1., 1., 1., 0., 0., 1., 1., 1., 1., 0.,
        1., 1., 0., 1., 1., 0., 1., 1., 0., 1., 1., 1., 1., 1., 1., 0., 1., 1.,
        0., 1., 1., 0., 1., 1., 0., 1., 1., 1., 1., 1., 1., 0., 1., 1., 0., 1.,
        1., 0., 1., 1., 0., 1., 1., 1., 1., 1., 1., 0., 1., 1., 0., 1., 1., 0.,
        1., 1., 0., 1., 1., 1., 0., 0., 0., 0., 0.], device='cuda:0')

> data.tags[data.batch==0]
tensor([0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0,
        0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0,
        0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0,
        0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0,
        2, 2, 2, 2, 2], device='cuda:0')

Am I correct in my understanding that the 'tags' attribute is the reverse of the 'fixed' attribute, but also identifies the catalyst atoms with the tag = 2?

[mshuaibii]

Hi @chaitjo

Great question. Each system (or Data object in this case) consists of a catalyst + adsorbate. Catalysts consist of a surface and a subsurface. During the DFT relaxation process subsurface atoms were fixed, preventing them from moving during the relaxation.

data.fixed is used to identify those fixed atoms, 1 being fixed and 0 being free. data.tags is used to identify the type of atom with 0, 1, 2 corresponding to subsurface, surface, and adsorbate atoms respectively. Connecting it back to data.fixed we can see that all atoms with data.tags = 0 have data.fixed = 1.

Am I correct in my understanding that the 'tags' attribute is the reverse of the 'fixed' attribute, but also identifies the catalyst atoms with the tag = 2?

Catalyst atoms would then correspond to tags = 0,1

Check out this notebook we put together to help better interact/understand the data: https://github.com/Open-Catalyst-Project/ocp/blob/master/docs/source/tutorials/data_playground.ipynb.

I hope this helps clarify

[chaitjo]

Ah, sorry, when I wrote 'catalyst' atoms, I meant the 'adsorbate' atoms!

[chaitjo]

Thanks for the clarification.