Synthesis of Sulfonated Phenylsilsesquioxanes Guided by Machine Learning.

Abstract

Sulfonated octaphenylsilsesquioxane (SPOSS) has garnered significant interest due to its unique structural properties of containing the -SO3H group and its wide range of applications. This study introduces a novel approach to the synthesis of SPOSS, leveraging machine learning algorithms to explore new recipes and achieve higher -SO3H functionality. The focus was on synthesizing SPOSS with 2, 4, 6, and 8-SO3H functional groups on the phenyl group, marked as SPOSS-2, SPOSS-4, SPOSS-6, and SPOSS-8, respectively. The successful synthesis of SPOSS-8 was achieved by 5 training outputs based on the recipes of 21 sets of low-functionality (<4) SPOSS. The structure of SPOSS was confirmed using Fourier transform infrared (FTIR) spectroscopy, nuclear magnetic resonance (NMR) spectroscopy, and time-of-flight mass spectrometry (MALDI-TOF MS). Machine learning analysis revealed that K2SO4 is an important additive to improve the functionality of SPOSS. A synthetic mechanism was proposed and validated that K2SO4 participated in the reaction to generate sulfur trioxide (SO3), a sulfonating agent with high reactivity. SPOSS shows thermal stability superior to octaphenylsilsesquioxane (OPS) according to thermogravimetric analysis (TGA) and TG-FTIR.