Prediction and experimental validation of the char yield of crosslinked polybenzoxazines.

Abstract

This study uses Molecular Operating Environment (MOE) to generate models to calculate the char yield of polybenzoxazines. A series of benzoxazine monomers were constructed to which a variety of parameters relating to the structure (e.g. water accessible surface, negative van der Waals surface area and hydrophobic volume, etc.) were obtained and a quantitative structure property relationships (QSPR) model was generated. The model was used to generate data for a new benzoxazine monomer and a comparison was made of predictions based on the QSPR models with the experimental data. This study shows the quality of predictive models and confirms how useful computational screening is prior to synthesis. In order to do that, the QSPR models were tested over a series of internal and external validation tests to explore their internal and external predictivity, prior to experimental validations which were performed later and reported in Chapter 7. The internal and external validations found out that the discrepancy in the general model (GM) which was initially thought to be a drawback to the model’s performance was actually not, as it does not compromise the model’s prediction accuracy, both internally and externally. The validation process also found that one of the structure-specific models, Ph-M (aniline-based benzoxazines) is externally predictive whilst another structure-specific model, the Ace-M (acetylenic-based polybenzoxazines) is not internally and externally predictive due to the too small training set that affects its predictivity performance. An acetylenic-based polybenzoxazine, poly(BA-apa) and a benzylamine-based polybenzoxazine, poly(BO-ba) have been successfully synthesised in this work. Both materials have been characterised using Fourier Transform – Infra Red Spectroscopy (FT-IR), Nuclear Magnetic Resonance (NMR) spectroscopy (both 1H and 13C) and Liquid Chromatography-Mass Spectrometry (LC-MS) to confirm their structures. These materials were analysed using Differential Scanning Calorimetry (DSC) to study their polymerisation behaviour and were later cured and taken further to Thermogravimetric Analysis (TGA) in order to investigate their thermal properties and the amount of char yield formed upon heating at 800 oC under an inert (nitrogen) atmosphere – which then will be used for experimental validation of the QSPR models. The study of DSC thermograms showed that both polymers exhibit a distinct polymerisation behaviour e.g. BA-apa went through two polymerisation reactions simultaneously (the oxazine ring opening polymerisation and the acetylene addition reaction) whilst BO-ba only polymerised via the ring opening reaction from the oxazine rings. It was also found that BA-apa has a lower polymerisation activation energy, consistent to its lower polymerisation temperature in comparison to the BO-ba. TGA analysis revealed that poly(BO-ba) formed an average of 44.35 % char yield and poly(BA-apa) on the other hand formed approximately 10 % higher char which is 56.28 %. The analysis also discovered that poly(BA-apa) synthesised in this work formed 15 % less char yield than previously reported in the literature (56.28 % vs. 71 %1) due to the shorter curing schedule. The final QSPR validation which is the experimental validation found that the char yield of poly(BO-ba) was predicted very well within 5-7 % error by both GM model and Ph-M. Ace-M which was reported earlier as not internally and externally predictive, has made a nearly accurate prediction towards the char yield of poly(BA-apa), close to the literature value of 71 %. The GM model has also made a close prediction to the Ace-M model, but these predictions deviated 15-17 % from the experimental poly(BA-apa) char yield measured in this work.