Structure fields in 'materials.search'

Materials Project Materials Project Data/API
1

I was trying to find the structure of a compound (Ta2NiSe5) using the API function materials.search(), and queried both ‘structure’ and ‘initial_structures’ as fields.

I saw that the ‘structure’ field returned the structure displayed on the interactive website, while ‘initial_structures’ returned a list of structures which didn’t contain the structure returned in the first case.

Could anyone help me understand the nature of these structures that ‘initial_structures’ is returning? (are they unrelaxed, unstable etc.). I really appreciate the help.

2

The initial_structure contains the unique starting structures used in all calculations (tasks) that go into building a material on MP.

Looking at a materials document:

from mp_api.client import MPRester

with MPRester() as mpr:
  mat_doc = mpr.materials.search(material_ids=['mp-149'])[0]
print(len(mat_doc.task_ids))
>>> 42 # will vary over time

So we see that 42 individual calculations / “tasks” build up the material mp-149.

If we then query the tasks for those materials, and de-duplicate them:

from pymatgen.analysis.structure_matcher import StructureMatcher

with MPRester() as mpr:
  tasks = mpr.materials.tasks.search(task_ids=mat_doc.task_ids, fields=["input"])

sm = StructureMatcher(
  ltol=0.1, stol=0.1, angle_tol=0.1, scale=False, attempt_supercell=False
)

unique_structure_groups = sm.group_structures(
  [task.input.structure for task in tasks if task.input and task.input.structure])

initial_structures = [group[0] for group in unique_structure_groups]
print(len(initial_structures))
>>> 6 # will vary depending on numpy + spglib version
print(len(mat_doc.initial_structures))
>>> 5 # will vary by database version

Ideally those two numbers at the end would be the same, but can differ depending on the numpy, spglib, etc. versions used in building data

3

Thank you for your response Aaron!

So, as I mentioned in my original post, I was looking at Ta2NiSe5: mp-541070. When I queried the ‘initial_structures’ field, it returned 3 structures. However, the third structure surprisingly had a different lattice symmetry from the first 2 structures - it was slightly monoclinic while the others were orthorhombic.

From what I could tell, Materials Projects usually presents different-symmetry lattices for the same compound in separate materials project entries. Am I correct in understanding that, in my case, the third structure was also classified as orthorhombic due to the thresholds used in the StructureMatcher class in the code you presented?

4

Correct, the symprec (linear tolerance for identifying structures) used to generate Materials Project data is looser than the default in pymatgen:


from mp_api.client import MPRester

with MPRester() as mpr:
    init_structs = mpr.get_structure_by_material_id("mp-541070",final=False)
print({s.get_space_group_info(symprec=0.1) for s in init_structs})
>>> {('Cmcm', 63)}

So all of the structures, including the one that is borderline another space group, are classified as Cmcm in MP’s data.