AI-accelerated scalable synthesis of nanocrystals for cost-effective photovoltaics

Abstract

Metal halide perovskite nanocrystals (PNCs) are promising candidates for next-generation optoelectronics. However, their commercial application necessitates low-cost synthesis and processing methods that surpass conventional lab-scale techniques. To address this, we report a ligand triggered room-temperature injection (LTRI) method that initiates nanocrystal nucleation by tuning precursor reactivity with addition of short-chain octanoic acid (OcA). Subsequently, integrating automation synthesis and machine learning, we further develop an efficient AI-accelerated framework towards fast and target synthesis of the FAPbI3 (FA: formamidinium) PNCs. Leveraging machine learning modeling, we identify that the critical size and distribution of FAPbI3 PNCs can be effectively optimized by tuning the feed ratio of OcA to the precursor formamidinium acetate. Furthermore, our LTRI method is fully compatible with flow-synthesis system for scale-up production, enabling a substantial reduction in production cost to US$2.16 g−1, significantly lower than the widely adopted hot-injection method (US$118.81 g−1). For practical application, we achieve a power conversion efficiency of 19.37% in photovoltaic applications with significant enhancement in device stability, matching the highest reported results for nanocrystal-based solar cells. The commercialization of perovskite nanocrystals requires low-cost synthesis and processing. Zhao et al. developed an AI-accelerated framework to optimize the FAPbI3 nanocrystal synthesis under room-temperature, achieving a production cost of US$ 2.16 g−1 with a power conversion efficiency of 19.37%.