Autonomous nanomanufacturing of lead-free metal halide perovskite nanocrystals using a self-driving fluidic lab

Abstract

We present a self-driving fluidic lab for accelerated synthesis science studies of lead-free metal halide perovskite nanocrystals.

Active Learning of Ligands That Enhance Perovskite Nanocrystal Luminescence

“…The sequential process enables the iterative refinement of models and increases predictive performance by using fewer labeled samples. With these advantages, AL has found widespread application from accelerating molecular simulations to materials discovery , and synthesis optimization, − e.g., for perovskite NCs. , …”

“…The existing understanding of how ligands interact with particle surfaces may not translate readily to particles with different compositions, requiring repeated tedious trial-and-error experimentation in vast and complex chemical space. Highthroughput (HT) automated workflows 15 and Bayesian optimization 16 have been used to accelerate optimization of perovskite NC synthesis conditions, such as ligand ratios and reaction temperatures for a defined set of precursors. Beyond optimizing reaction conditions, however, it remains an open challenge how to efficiently select and prioritize ligands to investigate, e.g., for postsynthetic treatments that optimize optical properties.…”

“…With these advantages, AL has found widespread application from accelerating molecular simula- tions 21 to materials discovery 22,23 and synthesis optimization, 24−26 e.g., for perovskite NCs. 16,27 Here, we combine AL with HT experimentation to develop an efficient screening workflow for identifying ligands that enhance the photoluminescence (PL) of CsPbBr 3 perovskite NCs. To represent the large chemical space of possible ligands, we formulated a pool of unlabeled samples consisting of 29,904 small organic candidate molecules.…”

Engineering a Sustainable Future: Harnessing Automation, Robotics, and Artificial Intelligence with Self-Driving Laboratories

“…Furthermore, choosing the appropriate optimization technique can be challenging when dealing with multiple considerations such as the number of experiments, computational cost, sizes of the input and output spaces, and performance metrics. In recent years, algorithm benchmarking has gained popularity to help future researchers decide which algorithms work best for a particular project.85,96 …”