Many properties of nanoparticles are governed by their shape, size, polydispersity and surface chemistry. To apply nanoparticles in chemical sensing, medical diagnostics, catalysis, thermoelectrics, photovoltaics or pharmaceutics, they have to be synthesized with precisely controlled characteristics. This is a time-consuming, laborious and resource-intensive task, because nanoparticle syntheses often include multiple reagents and are conducted under interdependent experimental conditions. Machine learning (ML) offers a promising tool for the accelerated development of efficient protocols for nanoparticle synthesis and, potentially, for the synthesis of new types of nanoparticles. In this Review, we discuss ML algorithms that can be used for nanoparticle synthesis and highlight key approaches for the collection of large datasets. We examine ML-guided synthesis of semiconductor, metal, carbon-based and polymeric nanoparticles, and conclude with a discussion of current limitations, advantages and perspectives in the development of ML-assisted nanoparticle synthesis.

纳米粒子的许多特性受其形状、大小、多分散性和表面化学性质的控制。要将纳米颗粒应用于化学传感、医学诊断、催化、热电、光伏或制药，它们必须以精确控制的特性合成。这是一项耗时、费力和资源密集型的任务，因为纳米颗粒合成通常包括多种试剂，并且是在相互依赖的实验条件下进行的。机器学习 (ML) 为加速开发用于纳米粒子合成的有效协议以及可能用于合成新型纳米粒子提供了一种很有前途的工具。在这篇评论中，我们讨论了可用于纳米粒子合成的 ML 算法，并重点介绍了收集大型数据集的关键方法。