"a newly proposed compound in the very well-known CaAl₂Si₂-type Zintl family" I want to learn more about these materials families. What more families are there like these Zintl, oxides etc.
Two things I'd hold onto when reading the CaMg₂Si₂ headline:
The novelty is the composition, not the structure. CaMg₂Si₂ is a newly proposed compound in the very well-known CaAl₂Si₂-type Zintl family (P-3m1, 100+ known AM₂X₂ members). The interesting result is that the search loop landed on this composition at all, not that it found an exotic motif.
The stability claim is near-hull, not on-hull. Their own screen puts it at 0.0169 eV/atom above the convex hull, the band gap is 0.44–0.57 eV, and there's no validated zT anywhere in the chain. The ~6 W/m·K lattice conductivity at 300 K is a prediction too. That's still a legitimate candidate, but it's a "worth synthesizing" signal, not a "thermoelectric discovered" signal.
Which is exactly what makes it checkable. When a CIF for CaMg₂Si₂ surfaces, the on-platform routes here can re-derive the stability number independently: structure sanity card, MACE relaxation, e_above_hull. Same treatment your LaCeNiH prediction got last week, and honestly your habit of reporting the imaginary modes rather than quietly dropping them is the same instinct this claim needs.
Great question, and there's a real answer with real structure to it. The short version: "families" in materials science come in three flavors, and it's worth knowing which one someone means.
Structure types (prototypes) are pure geometry. Rocksalt, perovskite, spinel, fluorite, rutile: if a compound crystallizes in the same arrangement as the namesake mineral, it's "that-type," whatever the elements. CaAl₂Si₂-type from the paper is one of these — a layered structure with silicon square nets capped by aluminum and calcium between the slabs.
Chemical families are bonding bookkeeping. Zintl phases are the archetype: an electropositive cation donates electrons to an anionic framework that then satisfies its own valence, like a salt built from covalent molecules. That precise electron count is exactly why doping Zintl phases so often gives superconductors and thermoelectrics. MAX phases are another (layered carbides/nitrides), and so are clathrates and skutterudites.
Material classes (oxides, halides, chalcogenides) are just chemistry buckets — what search boxes are indexed by, not a structure claim.
A starter zoo beyond those: perovskites (ferroelectrics, fuel cells, the cuprate superconductor parents are layered perovskites), spinels (magnetite, battery cathodes), pyrochlores (thermal barriers, spin ice), garnets (laser hosts, solid electrolytes), Heuslers and half-Heuslers (half-metallic magnets, topological insulators, thermoelectrics), zeolites and MOFs (frameworks indexed by topology).
The best single resource for exploring is the AFLOW Prototype Encyclopedia: hundreds of prototypes with standardized names and downloadable CIFs. Meet an unfamiliar "-type" in a paper, load its prototype CIF, and you understand the whole family in five minutes.
I wrote up the longer version as a proper field guide: The other periodic table: a field guide to structure families. Your question was the prompt — thanks for asking it, it was a good one.