Accelerating Bandgap Discovery in Hybrid Perovskites with Advanced Modeling Techniques

Hybrid perovskites are remarkable for their ability to be tuned to different bandgaps, which is crucial for creating highly efficient tandem solar cells. However, the relationship between the composition and the optical bandgap in these materials can be complex and unpredictable [1]. This makes it challenging to find the right composition that offers the desired bandgap properties. Our research introduces a novel experimental workflow using advanced Gaussian Process (GP) models to tackle this problem [2]. By applying structured and custom GP models, we can simultaneously discover new material behaviors and understand the underlying physical principles more efficiently [3]. This method has been validated with simulated datasets, showing that it significantly speeds up the discovery process. Specifically, our approach uses a few initial data points to guide further experiments, reducing the number of necessary preparations. We demonstrated that this iterative method quickly identifies the bandgap characteristics of MA_1-xGA_xPb(I_1-xBr_x)₃., revealing important relationships in the bandgap diagram. This innovative approach not only simplifies the discovery process for binary systems but can also be extended to more complex material systems, offering a powerful tool for advancing perovskite research.<br/>1.Ziatdinov, M. A., Liu, Y., Morozovska, A. N., Eliseev, E. A., Zhang, X., Takeuchi, I., & Kalinin, S. V. (2022). Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries. <i>Advanced Materials</i>, 34(20), 2201345. DOI: 10.1002/adma.202201345<br/>2. Sanchez, S. L., Foadian, E., Ziatdinov, M., Yang, J., Kalinin, S. V., Liu, Y., & Ahmadi, M. (2023). Physics-driven discovery and bandgap engineering of hybrid perovskites. <i>arXiv preprint arXiv:2310.06583</i>. (accepted in Digital Discovery)<br/>3. Ziatdinov, M., Liu, Y., Kelley, K., Vasudevan, R., & Kalinin, S. V. (2022). Bayesian Active Learning for Scanning Probe Microscopy: From Gaussian Processes to Hypothesis Learning. <i>ACS Nano</i>, 16(9), 13492-13512. DOI: 10.1021/acsnano.2c05303