Merging High-Throughput Autonomous Experiments and Machine Learning Supercharge Discovery of Two-Dimensional Halide Perovskites

Two-dimensional halide perovskites (HPs) combine the richness of physical functionalities of inorganic materials and complexity of organic molecules (spacer cations) in a single dynamic material. However, discovery and optimization of these materials require joint optimization of the composition of the inorganic components and selection of the molecular moieties, to harness the phase formation and self-assembly processes on the material level, and extend it to micro- and macroscale functional devices¹.<br/>In this talk, I will discuss the potential of high throughput automated experiments to expedite the discovery of 2D and quasi 2D halide perovskites (HPs), optimize processing pathways, and enhance understanding of formation kinetics^2,3. This approach requires maximal acceleration of the synthesis-characterization-prediction cycle, enabled by the incorporation of rapid characterization and machine learning (ML) methods in the discovery loop. In many cases, the intrinsic latencies of theoretical modeling for sufficiently complex systems favor the experiment-first discovery approach. I will showcase how high throughput automated synthesis and characterization provides a comprehensive guide for designing optimal precursor stoichiometry to achieve functional quasi-2D perovskite phases in films capable of realizing high-performance optoelectronics^2,3. With excellent agreement between theoretical and experimental observations, I show that with judicious selection of spacer cations, 2D HP can manifest self-assembly of twisted Moire structure, which has not been observed from conventional 2D HP systems with linear spacers⁴. These studies exemplify how merging high-throughput automated experimental workflow and ML effectively expedites discoveries and processing optimizations in complex materials systems with multiple functionalities, facilitating their realization in scalable optoelectronic manufacturing processes⁵.<br/><br/><b>References: </b><br/><br/>1. Yang J., Kalinin S.V., Cubuk E.D. Ziatdinov M., Ahmadi M. Toward self-organizing low-dimensional organic–inorganic hybrid perovskites: Machine learning-driven co-navigation of chemical and compositional spaces. <i>MRS Bulletin </i>(2023) <b>48</b>, 164–172. DOI: 10.1557/s43577-023-00490-y<br/>2. Yang J., Lawrie B.J., Kalinin S.V., Ahmadi M. High-throughput automated exploration of phase growth kinetics in quasi-2D formamidinium metal halide perovskites. <i>Advanced Energy Materials</i> (2023) 13, 2302337. DOI: 10.1002/aenm.202302337<br/>3. Yang J., Hidalgo J., Li R., Kalinin S.V., Correa-Baena J.-P., Ahmadi M. Accelerating materials discovery by high-throughput GIWAXS characterization of quasi-2D formamidinium metal halide perovskites. <i>ChemRxiv</i> (2023). DOI: 10.26434/chemrxiv-2023-zcvl0<br/>4. Yang J., Fuhr A.S., Roccapriore K.M., Dryzhakov B., Hu B., Sumpter B.G., Kalinin S.V., Ahmadi M. Ligand-induced self-assembly of twisted two-dimensional halide perovskites. <i>ChemRxiv</i> (2024). DOI: 10.26434/chemrxiv-2024-wwwb9<br/>5. Yang J., Ahmadi M. Empowering scientists with data-driven automated experimentation. <i>Nature Synthesis </i>(2023) 2, 462–463. DOI: 10.1038/s44160-023-00337-z