This thesis identifies a facile and versatile technique for creating multi-scale porous titania with tunable meso-scale morphology. Three templating approaches were simultaneously utilized in achieving this; namely, soft-templating by template-free self-assembly of an aligned macroporous structure, freeze-casting for the preservation of particle dispersion found in suspension, and hard-templating by the use of cellulose nanocrystals (CNCs) as sacrificial material. A systematic study was conducted wherein three synthesis parameters (water content, alcohol solvent content, and drying method) were varied in the hydrolysis of titanium tetra-isopropoxide (TTIP) by the sol-gel method to determine their contribution to the formation of multi-scale porous titania exhibiting aligned macrochannels and mesoporosity. The optimal synthesis settings for producing multi-scale porous titania were identified as H2O/TTIP molar ratio of 30, without any isopropanol (acting as solvent), and freeze-drying after freezing at -40°C. Subsequently, CNCs were added in various quantities (0-50vol%) to the hydrolysis of TTIP using these optimized settings to achieve more direct and precise control of the final titania meso-structure. Morphological studies revealed that the final titania bodies maintained the formation of macrochannels 1-3 μm in diameter as a result of hydrolysis in excess water in the absence of an organic solvent and exhibited successful templating mutually affected by CNC addition and freeze-casting. Freeze-drying preserved particle dispersion in the colloid suspension, hindering agglomeration otherwise found after oven-drying and enhanced the CNCs' role of disrupting titania aggregation and increasing interconnectivity. Thus, meso-structured multi-scale porous titania was prepared by a combined templating strategy using template-free self-assembly, freeze-casting, and CNC hard-templating. / MS / Titanium dioxide (TiO₂) has been shown to exhibit desirable properties including physical and chemical stability and biocompatibility making it a material of great interest in a variety of fields including pigments and biomedicine. Furthermore, the material’s photocatalytic activity (i.e. ability to absorb light energy to generate usable charge) has led to its implementation in solar cells, in the production of hydrogen as an eco-friendly fuel, and in decontaminating water from organic pollutants. While TiO₂ has shown great promise in these applications, there remains a need to identify a simple strategy to synthesize TiO₂ with a tunable multi-scale porous structure with pores of different sizes and shapes to improve its performance. To this end, a facile and versatile procedure was used to prepare multi-scale porous TiO₂ with tunable morphology. In investigating the effect of water content, alcohol content and drying method on the final morphology, a multi-scale structure was achieved by synthesizing TiO₂ in the absence of an alcohol solvent and within a new moderate range of water content that had not been previously explored. Lacking an effective and easy strategy to further manipulate the multi-scale morphology, this self-assembly technique was modified by incorporating cellulose nanocrystals (CNCs) into the synthesis procedures to further tune the structure on the nanometric scale by altering the final porosity and surface area. The final TiO₂ samples exhibited multi-scale porous structures that could be manipulated by combining the self-assembly and CNC-templating techniques in an adaptable strategy to tailor the TiO₂ morphology for its various uses in photocatalysis and biomedicine.

Paper forms are a commonly used format for collecting information, including information that ultimately will be added to a digital database. This work focuses on the automatic extraction of information from form images. It examines what can be achieved at parsing forms without any textual information. The resulting model, FUDGE, shows that computer vision alone is reasonably successful at the problem. Drawing from the strengths and weaknesses of FUDGE, this work also introduces a novel model, Dessurt, for end-to-end document understanding. Dessurt performs text recognition implicitly and is capable of outputting arbitrary text, making it a more flexible document processing model than prior methods. Dessurt is capable of parsing the entire contents of a form image into a structured format directly, achieving better performance than FUDGE at this task. Also included is a technique to generate synthetic handwriting, which provides synthetic training data for Dessurt.