A spatiotemporal experimental route is reported for the antisolvent vapor diffusion crystal growth of metal halide perovskites. A computational analysis combining automated image capture and diffusion modeling enables the determination of the critical concentrations required for nucleation and crystal growth from a single experiment. Five different solvent systems and ten distinct organic ammonium iodide salts were investigated with lead iodide, from which nine previously unreported compounds were discovered. Automated image capture of the mother liquor and antisolvent vials were used to determine changes in solution meniscus positions and detect nucleation event location. Matching the observations to a numerical solution of Fick’s second law diffusion model enables the calculation of reactant, solvent and antisolvent concentrations at both the time and position of the first stable nucleation and crystal growth. A machine learning model was trained on the resulting data reveals solvent- and amine-specific crystallization tendencies. Solvent systems that interact more weakly with dissolved lead species promote crystallization, while those with stronger interaction can prevent crystallization through increased solubilities. Organic amines that interact more strongly with inorganic components and exhibit greater rigidity are more likely to be incorporated into crystalline products.
CCDC 2121660-2121668 contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/structures. Solubility measurement procedure and data; detailed reaction, image analysis description, images of all reaction outcomes, time and position data for first observed nucleation and crystal growth, LLPS images; powder diffraction data; single crystal synthesis details, crystal packing, layer images, and bond valence table for compounds 1 - 9; detailed description of refractive index measurement and diffusion model calculation, calculated critical concentration values for all reactions along with the chemical descriptors and their descriptions are given in the Supporting Information file. A Github repository (https://github.com/darkreactions/rapid2) contains the following files and code: reaction block CAD file, image analysis process and Python code for: diffusion heights and crystal growth, laser diffraction for diffusion coefficient measurement and analysis and diffusion model scripts.