Self‐Driving Platform for Metal Nanoparticle Synthesis: Combining Microfluidics and Machine Learning

Abstract

Many applications of inorganic nanoparticles (NPs), including photocatalysis, photovoltaics, chemical and biochemical sensing, and theranostics, are governed by NP optical properties. Exploration and identification of reaction conditions for the synthesis of NPs with targeted spectroscopic characteristics is a time-, labor-, and resource-intensive task, as it involves the optimization of multiple interdependent reaction conditions. Integration of machine learning (ML) and microfluidics (MF) offers accelerated identification and optimization of reaction conditions for NP synthesis. Here, an autonomous ML-driven, oscillatory MF platform for the synthesis of NPs is reported. The platform utilized multiple recipes and reaction times for the synthesis of NPs with different dimensions, conducted spectroscopic NP characterization, and employed ML approaches to analyze multiple yet prioritized spectroscopic NP characteristics, and identified reaction conditions for the synthesis of NPs with targeted optical properties. The platform is also used to develop an understanding of the relationship between reaction conditions and NP properties. This study shows the strong potential of ML-driven oscillatory MF platforms in materials science and paves the way for automated NP development.

Reaching critical MASS: Crowdsourcing designs for the next generation of materials acceleration platforms

“…8 The Matter Lab is also developing MAPs for a number of different materials classes, including for metal nanoparticle synthesis. 9 From the initial proof-of-concept tools like ARES, Ada, and BEAR to more recent systems like CAMEO and ll Matter 5, 1972Matter 5, -1976, July 6, 2022 1973 SARA, proof of principle for single and multi-modal optimizations and physical constraints have been demonstrated. As a community, we should begin to ask ourselves how these concepts can be applied to more significant problems with societal impact.…”

Autonomous retrosynthesis of gold nanoparticles via spectral shape matching

“…Application of BO to structure optimization of nanoparticles has recently been studied in problems that involve optimizing a characteristic response collected through experiments such as UV-Vis spectroscopy incorporated into high-throughput frameworks. [3][4][5] Oentimes, the characteristic responses collected as a spectrum (i.e., a signal over a discrete sample of a stimulus e.g. : wavelength) are not suited for direct usage in BO.…”

Autonomous Nanocrystal Doping by Self‐Driving Fluidic Micro‐Processors

“…[31][32][33] Over the past 5 years, the concept of ML-guided experimentation has been applied to a wide range of organic and inorganic syntheses, including pharmaceuticals, [34] metal electrocatalysts, [35] carbon dots, [36] and magnetic resonance imaging agents, [37] as well as nanoparticles. [38][39][40][41] The closed-loop ML-guided optimization techniques vary from black-box [42] to informed [43,44] modeling techniques.…”