SPARQL queries against the atomRDF graph

Every sample, simulation cell and crystallographic property created by atomRDF is a real RDF triple, so anything you can express in SPARQL is available to you. This notebook shows three styles of querying:

1. Raw SPARQL (kg.query("SELECT ...")).

2. Term-builder for ontology-aware queries without writing SPARQL (kg.query_sample, kg.query).

3. Returning a sample object and operating on it.

from atomrdf import KnowledgeGraph
import atomrdf.build as build

Build a small heterogeneous database

kg = KnowledgeGraph()
_ = build.bulk("Fe", cubic=True, graph=kg)
_ = build.bulk("Cu", cubic=True, graph=kg)
_ = build.bulk("Si", cubic=True, graph=kg)
_ = build.bulk("Mg", crystalstructure="hcp", graph=kg)
kg.n_samples

4

1. Raw SPARQL

What are the chemical species in the graph?

q = """
PREFIX cmso: <http://purls.helmholtz-metadaten.de/cmso/>
SELECT DISTINCT ?symbol
WHERE {
    ?species cmso:hasElementSymbol ?symbol .
}
"""
kg.query(q)

	symbol

Every sample with a cubic Bravais lattice and exactly two atoms in the unit cell:

q = """
PREFIX cmso: <http://purls.helmholtz-metadaten.de/cmso/>
PREFIX xsd:  <http://www.w3.org/2001/XMLSchema#>
SELECT ?sample ?symbol
WHERE {
    ?sample  cmso:hasNumberOfAtoms ?n ;
             cmso:hasMaterial      ?m .
    ?m       cmso:hasStructure     ?s .
    ?s       cmso:hasSpaceGroupSymbol ?symbol .
    FILTER (?n = "2"^^xsd:integer)
}
"""
kg.query(q)

	sample	symbol
0	sample:04478872-1802-45f3-9093-ed60d06de64d	Im-3m
1	sample:a9737877-3034-4e45-893c-4853d410d757	P6_3/mmc

2. Term builder (when the ontology network is available)

kg.terms.cmso.AtomicScaleSample lets you express the same query without typing SPARQL. It requires the ontology network to be reachable at construction time — if it is not (e.g. behind a strict firewall), kg.terms will be None and you should fall back to the raw SPARQL form above.

kg.query(
    kg.terms.cmso.AtomicScaleSample,
    [
        kg.terms.cmso.hasSpaceGroupSymbol,
        kg.terms.cmso.hasNumberOfAtoms == 2,
    ],
)

3. Return a single sample and write it out

q = """
PREFIX cmso: <http://purls.helmholtz-metadaten.de/cmso/>
PREFIX xsd:  <http://www.w3.org/2001/XMLSchema#>
SELECT ?sample
WHERE {
    ?sample cmso:hasNumberOfAtoms ?n .
    FILTER (?n = "2"^^xsd:integer)
}
"""
df = kg.query(q)
df

	sample
0	sample:04478872-1802-45f3-9093-ed60d06de64d
1	sample:a9737877-3034-4e45-893c-4853d410d757

sample = df['sample'].values[0]
kg.to_file(sample, "selected.poscar", format="vasp")
! head -10 selected.poscar

Fe
 1.0000000000000000
     2.8700000000000001    0.0000000000000000    0.0000000000000000
     0.0000000000000000    2.8700000000000001    0.0000000000000000
     0.0000000000000000    0.0000000000000000    2.8700000000000001
 Fe 
   2
Cartesian
  0.0000000000000000  0.0000000000000000  0.0000000000000000
  1.4350000000000001  1.4350000000000001  1.4350000000000001