Closing the loop from generative crystal design to experimental synthesis

Demonstrate reliable, end-to-end translation of generative inverse-designed crystalline inorganic materials into experimentally realized compounds, including validated synthesis workflows that connect generated structures to practical lab procedures.

Background

Generative models are increasingly able to propose inorganic crystals conditioned on target properties, yet practical impact depends on whether such proposals can be synthesized. The paper stresses that predictions typically derive from zero-temperature DFT, often ignoring defects, disorder, finite-temperature effects, and complex chemistries that influence synthesizability.

The authors argue that automated, AI-driven laboratories and feedback loops between computation and experiment are needed, but they note that robust demonstrations of translating inverse-designed crystals into successful syntheses remain lacking.