Abstract
Polyhedral oligomeric silsesquioxane (POSS) compounds are defined by the chemical formula (RSiO3/2)8 with R being an organic fragment. They display versatile features due to the combination of both their stable Si-O-Si inorganic cores and the large number of possible organic groups that can be attached to them. The present work aims at characterizing a highly-thermoresistant POSS, the octa(aminophenyl)silsesquioxane (OAPS). This siloxane-based cage has three different isomers depending on the meta, ortho and para positions of the amines with respect to the phenyl groups and can be obtained using two synthesis routes. However, the presence of the isomers depends on the synthesis route and remains up to now an open question.
Experimental characterizations including pycnometry, infrared spectroscopy (IR), 1-dimensional and 2-dimensional nuclear magnetic resonance (NMR) have been performed for a commercial OAPS containing all three isomers and a controlled OAPS containing only the para and meta isomers. The density is found to be insensitive to the nature of the isomers, unlike the IR, 13C-NMR and 1H-NMR spectra that are isomer-dependent. To better identify the isomers, the experimental IR and NMR spectra were compared to predictions from Density Functional Theory (DFT) quantum mechanical methods and by machine-learning analyses. Within this context, quantum mechanical methods were found to be clearly superior to machine-learning methods, despite being computationally much more expensive. As a result, several peaks in the IR spectra and each peak in both the 13C-NMR and 1H-NMR spectra could be assigned to a specific OAPS isomer.
Experimental characterizations including pycnometry, infrared spectroscopy (IR), 1-dimensional and 2-dimensional nuclear magnetic resonance (NMR) have been performed for a commercial OAPS containing all three isomers and a controlled OAPS containing only the para and meta isomers. The density is found to be insensitive to the nature of the isomers, unlike the IR, 13C-NMR and 1H-NMR spectra that are isomer-dependent. To better identify the isomers, the experimental IR and NMR spectra were compared to predictions from Density Functional Theory (DFT) quantum mechanical methods and by machine-learning analyses. Within this context, quantum mechanical methods were found to be clearly superior to machine-learning methods, despite being computationally much more expensive. As a result, several peaks in the IR spectra and each peak in both the 13C-NMR and 1H-NMR spectra could be assigned to a specific OAPS isomer.
