Low-hysteresis shape-memory ceramics designed by multimode modelling.

Zirconia ceramics exhibit a martensitic phase transformation that enables large strains of order 10%, making them prospects for shape-memory and superelastic applications at high temperature 1-5 . Similarly to other martensitic materials, this transformation strain can be engineered by carefully alloying to produce a more commensurate transformation with reduced hysteresis (difference in transformation temperature on heating and cooling) 6-11 . However, such 'lattice engineering' in zirconia is complicated by additional physical constraints: there is a secondary need to manage a large transformation volume change 12 , and to achieve transformation temperatures high enough to avoid kinetic barriers 6 . Here we present a method of augmenting the lattice engineering approach to martensite design to address these additional constraints, incorporating modern computational thermodynamics and data science tools to span complex multicomponent spaces for which no data yet exist. The result is a new zirconia composition with record low hysteresis of 15 K, which is about ten times less transformation hysteresis compared to typical values (and approximately five times less than the best values reported so far). This finding demonstrates that zirconia ceramics can exhibit hysteresis values of the order of those of widely deployed shape-memory alloys, paving the way for their use as viable high-temperature shape-memory materials.