The synthesis and performance of MHP NCs are highly sensitive to surface chemistry, which plays a pivotal role in dictating their colloidal stability and optoelectronic properties. Central to this is the acid–base equilibrium that governs surface ligation, where organic acids and bases act as capping ligands to stabilize NCs in solution. The choice of organic acid introduces distinct ligand environments that influence nucleation and growth pathways, leading to varied nanocrystal morphologies and emission characteristics. Furthermore, post-synthetic halide exchange reactions provide an effective means to finely tune the optical properties of MHP NCs without altering their core structure. Consequently, the resulting properties are governed by a complex interplay between discrete parameters, such as ligand identity, and continuous parameters, including reaction temperature, concentration, and time. This high-dimensional, mixed-variable, and interdependent chemical space poses significant challenges for elucidating the ligand structure–synthesis–property relationships. Conventional batch-based synthesis approaches, relying on manual operation and trial-and-error exploration, are limited by low throughput, poor reproducibility, and a lack of real-time feedback, thereby impeding the systematic discovery and optimization of high-performing MHP NCs.
To address the challenges posed by the complex and interdependent synthesis space of MHP NCs, we present Rainbow—a self-driving laboratory (SDL) designed for autonomous bandgap optimization of NCs. Rainbow integrates a multi-robot synthesis and characterization platform with a fine-tuned artificial intelligence (AI) agent to enable accelerated and systematic exploration of MHP NCs' synthesis–structure–property relationships. Leveraging parallelized, miniaturized batch reactors, robotic sample handling, and real-time spectroscopic feedback, Rainbow performs high-throughput synthesis and characterization with high reliability and reproducibility. This closed-loop system enables autonomous decision-making and execution of experiments, positioning Rainbow as a next-generation platform for data-driven nanomaterials discovery.
Through continuous closed-loop experimentation, Rainbow autonomously investigated the effects of varying ligand structures and precursor concentrations on key optical performance metrics, including PLQY, FWHM, and emission wavelength. A data-driven modeling framework and a bespoke AI-guided experiment selection algorithm empowered Rainbow to efficiently navigate the intricate parameter landscape, unlocking previously unexplored regions of the MHP NC chemical space. This intelligent exploration revealed new insights into the influence of surface ligands on MHP NC growth and emission characteristics, and also enabled the identification of Pareto-optimal formulations for targeted bandgaps within 24 h per target bandgap—corresponding to over a 350-fold acceleration compared to conventional approaches.
By establishing a closed-loop, autonomous workflow that combines multi-robotic experimentation with AI-driven decision-making, Rainbow effectively overcomes the limitations of traditional nanomaterials research. It represents a transformative advancement in both applied and fundamental studies of MHP NCs, allowing for precise, on-demand synthesis of NCs with optimal optical properties. This work demonstrates, for the first time, the autonomous discovery of ligand–condition combinations that yield MHP NCs with target emission energies, laying the foundation for next-generation photonic and optoelectronic technologies.