Pan Bin, Madani Majed S, Forsberg Allison P, Brutchey Richard L, Malmstadt Noah
Mork Family Department of Chemical Engineering and Materials Science, University of Southern California, 925 Bloom Walk, Los Angeles, California 90089-1211, United States.
Department of Chemical and Materials Engineering, King Abdulaziz University, Jeddah 21589, Saudi Arabia.
ACS Nano. 2024 Sep 17;18(37):25542-25551. doi: 10.1021/acsnano.4c05807. Epub 2024 Sep 5.
Colloidal platinum nanoparticles (Pt NPs) possess a myriad of technologically relevant applications. A potentially sustainable route to synthesize Pt NPs is polyol reduction in ionic liquid (IL) solvents; however, the development of this synthetic method is limited by the fact that reaction kinetics have not been investigated. In-line analysis in a flow reactor is an appealing approach to obtain such kinetic data; unfortunately, the optical featurelessness of Pt NPs in the visible spectrum complicates the direct analysis of flow chemistry products ultraviolet-visible (UV-vis) spectrophotometry. Here, we report a machine learning (ML)-based approach to analyze in-line UV-vis spectrophotometric data to determine Pt NP product concentrations. Using a benchtop flow reactor with ML-interpreted in-line analysis, we were able to investigate NP yield as a function of residence time for two IL solvents: 1-butyl-1-methylpyrrolidinium triflate (BMPYRR-OTf) and 1-butyl-2-methylpyridinium triflate (BMPY-OTf). While these solvents are structurally similar, the polyol reduction shows radically different yields of Pt NPs depending on which solvent is used. The approach presented here will help develop an understanding of how the subtle differences in the molecular structures of these solvents lead to distinct reaction behavior. The accuracy of the ML prediction was validated by particle size analysis and the error was found to be as low as 4%. This approach is generalizable and has the potential to provide information on various reaction outcomes stemming from solvent effects, for example, differential yields, orders of reaction, rate coefficients, NP sizes,
胶体铂纳米颗粒（Pt NPs）具有众多与技术相关的应用。一种潜在的可持续合成Pt NPs的方法是在离子液体（IL）溶剂中进行多元醇还原；然而，这种合成方法的发展受到尚未研究反应动力学这一事实的限制。在流动反应器中进行在线分析是获取此类动力学数据的一种有吸引力的方法；不幸的是，Pt NPs在可见光谱中的光学无特征性使得通过紫外 - 可见（UV - vis）分光光度法对流动化学产物进行直接分析变得复杂。在此，我们报告一种基于机器学习（ML）的方法来分析在线UV - vis分光光度数据，以确定Pt NP产物浓度。使用带有ML解释的在线分析的台式流动反应器，我们能够研究两种IL溶剂（1 - 丁基 - 1 - 甲基吡咯烷鎓三氟甲磺酸盐（BMPYRR - OTf）和1 - 丁基 - 2 - 甲基吡啶鎓三氟甲磺酸盐（BMPY - OTf））中NP产率与停留时间的函数关系。虽然这些溶剂在结构上相似，但多元醇还原显示出取决于使用哪种溶剂的截然不同的Pt NPs产率。本文提出的方法将有助于理解这些溶剂分子结构中的细微差异如何导致不同的反应行为。通过粒度分析验证了ML预测的准确性，发现误差低至4%。这种方法具有通用性，有可能提供关于溶剂效应导致的各种反应结果的信息，例如，不同的产率、反应级数、速率系数、NP尺寸。
