Concern over the occurrence of micropollutants in drinking water and their health effects is increasing. Therefore, there is a growing interest in understanding micropollutant removal during drinking water treatment. Ozonation and advanced oxidation processes (AOPs) have been found to be effective in the degradation of many micropollutants. Ozonation involves reactions with both molecular ozone (direct pathway) and hydroxyl radicals (indirect pathway), while hydroxyl radicals are the main oxidants in advanced oxidation processes. Reaction rate constants of micropollutants with molecular ozone (kO3) and hydroxyl radicals (kOH) are indicators of their reactivity and are therefore useful in assessing their removal efficiency in ozonation and AOPs. However, to date, only a limited number of rate constants are available for micropollutants, especially emerging micropollutants such as endocrine disrupting chemicals (EDCs) and pharmaceuticals. Quantitative structure-property relationships (QSPR) are therefore desirable for predicting rate constants of numerous untested micropollutants without experimentation. The overall objective of this thesis was to develop predictive QSPR models which correlate the rate constants of a wide range of structural diverse micropollutants to their structural characteristics.
To ensure the wide applicability of the QSPR models, the training set compound selection is critical and a group of heterogeneous compounds which are structurally representative of many others is preferred. A systematic compound selection approach which involves principal component analysis (PCA) and D-optimal onion design was applied for the first time in water treatment research. As a result, 22 micropollutants with diverse structures were selected as representatives from a large pool of micropollutants of interest (182 compounds). In addition, 12 molecular descriptors were identified which link relevant structural features to the removal mechanisms of oxidation processes.
The kO3 and kOH values of the 22 selected micropollutants were then determined experimentally in bench-scale reactors at neutral pH using high performance liquid chromatography equipped with a photodiode array detector (HPLC-PDA). Three methods, competition kinetics, compound monitoring, and ozone monitoring were used for kO3 measurement, and competition kinetics was used for kOH measurement. As expected, kO3 values span a wide range from 10-2 to 107 M-1 s-1 because of the selective nature of molecular ozone. The general trends of micropollutant reactivity with ozone can be explained by the micropollutant structures and the electrophilic nature of ozone reactions. The kOH values range from 108 to 1010 M-1 s-1 because hydroxyl radicals are relatively non-selective in their reactions. For the majority of these micropollutants kO3 and kOH values were not reported prior to this study. Thus they provide valuable information for modeling and designing of ozonation and AOP treatment.
QSPR models for kO3 and kOH prediction were then developed with special attention to model validation, applicability domain and mechanistic interpretation. With the experimentally determined rate constants, QSPR models were developed for predicting kO3 values using the selected 22 micropollutants as the training set and the 12 identified descriptors as model variables. As a result, two QSPR models were developed using piecewise linear regression (PLR) both showing an excellent goodness-of-fit. Model 1 was governed by average molecular weight and number of phenolic functional groups, and Model 2 was dominated by two principal components extracted from the descriptor matrix. The models were then validated using an external validation set collected from the literature, showing good predictive power of both models. Prior to applying these models to unknown micropollutants they need to be classified as high-reactive (logkO3 > 2 M-1 s-1) or low-reactive (logkO3 2 M-1 s-1), so that the appropriate submodel of the PLR can be applied. A classification function using linear discriminant analysis (LDA) was therefore developed which worked very well for both training and validation sets. With the help of additional compounds collected from the literature, and DRAGON molecular descriptors, a QSPR model for kOH prediction in the aqueous phase was developed using multiple linear regression. As a result, 7 DRAGON descriptors were found to be significant in modeling kOH, which related kOH of micropollutants to their electronegativity, polarizability, presence of double bonds and H-bond acceptors. The model fitted the training set very well and showed great predictive power as assessed by the external validation set. In addition, the model is applicable to a wide range of micropollutants. The model’s applicability domain was defined using a leverage approach.
The main contributions of this thesis lie in the successful development of QSPR models for kO3 and kOH value prediction which, for the first time, can be used for a wide range of structurally diverse micropollutants. In addition, all QSPR models were externally validated to verify their predictive power, and the applicability domains were defined so that the applicability of the models to new compounds can be determined.
Finally, the applicability of the model to natural water was explored by combining the QSPR models with the established Rct concept which predicts micropollutant removals during ozone treatment of natural water but requires kinetic data as input. Results show that the kinetic data from the QSPR model predictions worked well in the Rct model providing reliable estimations for most of the selected micropollutants. This approach can therefore be used in water treatment for initial assessment and estimation of ozonation efficiency.
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:OWTU.10012/6776 |
| Date | 16 May 2012 |
| Creators | Jin, Xiaohui |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | English |
| Detected Language | English |
| Type | Thesis or Dissertation |
Page generated in 0.0026 seconds