Hafnia — hafnium oxide, HfO₂ — has too many polymorphs and no agreed genealogy. The monoclinic ground state is stable at ambient conditions. Under pressure it transforms to orthorhombic phases. At high temperature, to tetragonal and cubic. Epitaxial strain stabilizes a ferroelectric orthorhombic phase that has made hafnia one of the most important materials in microelectronics. But the relationships between these phases — which ones are parents, which are daughters, which share a common ancestor — have been proposed rather than demonstrated.
Cernov, Íñiguez-González, and Aramberri (arXiv:2603.22402) use pressure as the organizing principle. First-principles calculations trace each polymorph's response to hydrostatic compression, watching how structures evolve continuously along the pressure coordinate. Some phases that appear unrelated at ambient conditions converge at high pressure to the same parent structure. Some that were thought to share a parent turn out to descend from different ancestors. The family tree is not what was assumed.
The method also discovers new phases — higher-energy structures that don't exist at ambient conditions but serve as common ancestors of widely discussed reference phases. These are transition states in the crystallographic sense: they're never the thermodynamic ground state at any pressure, but the family relationships between observed phases pass through them. Remove the ancestors from the tree and the connections between daughters become inexplicable.
The structural insight is that pressure is a better organizing axis than temperature for solid-state genealogy. Temperature produces thermal motion that blurs structural distinctions — high-temperature phases are often disordered averages of low-temperature ones. Pressure preserves order while changing the potential energy surface, making parent-daughter relationships visible as continuous deformations rather than order-disorder transitions. The clarity comes from what pressure doesn't destroy.
What's interesting beyond hafnia is the method itself: use a single external parameter as a probe of structural kinship, and let the family tree emerge from the responses rather than from assumed symmetry relationships. The phases that seemed unrelated weren't wrong about their own structures. They were just being compared at the wrong pressure.