UWSpace

UWSpace is the University of Waterloo’s institutional repository for the free, secure, and long-term home of research produced by faculty, students, and staff.

Depositing Theses/Dissertations or Research to UWSpace

Are you a Graduate Student depositing your thesis to UWSpace? See our Thesis Deposit Help and UWSpace Thesis FAQ pages to learn more.

Are you a Faculty or Staff member depositing research to UWSpace? See our Waterloo Research Deposit Help and Self-Archiving pages to learn more.

Photo by Waterloo staff

Communities in UWSpace

Select a community to browse its collections.

Recent Submissions

  • Thumbnail Image
    Item type: Item ,
    LLM-Based Frameworks for Information Retrieval Evaluation
    (University of Waterloo, 2026-06-29) Upadhyay, Shivani Jayantkumar
    Evaluating information retrieval (IR) systems requires a reference that captures what correct or relevant output looks like, as well as a mechanism for determining whether a system’s output matches that reference. For lexical retrieval systems, both requirements are relatively straightforward. Systems rank documents by term overlap, pooling produces a judgment file that covers most documents any system is likely to return, and determining relevance reduces to a simple membership test against that file. This evaluation paradigm relies on the assumption that relevance can be detected through surface-form overlap. When retrieval moves beyond that assumption, the framework begins to break down. Retrieval-augmented generation (RAG) systems strain this setup by synthesising free-form natural language responses from retrieved evidence. A gold answer set constructed before system execution cannot anticipate every correct phrasing, so even semantically correct outputs can fail under lexical matching. Dense retrieval systems encode queries and documents as vectors, retrieving relevant documents that might not share vocabulary with the query. Under pooling-based evaluation, these documents never receive human judgments and are instead assigned a default relevance grade of zero. Together, these failures highlight the limits of surface-form evaluation and point to the need for judgment mechanisms that reason directly about meaning. This thesis investigates whether large language models (LLMs) can fill this gap by contributing three frameworks across successive layers of the evaluation pipeline. The first contribution is an open-source QA evaluation framework that combines chain-of-thought (CoT) prompting with self-consistency decoding using instruction-tuned LLMs. When evaluated across 12 systems on NQ-open, it matches zero-shot GPT‑4 in rank correlation with human judgments while using a model more than an order of magnitude smaller, demonstrating that prompting strategy can matter as much as scale. The second contribution is a framework for patching incomplete relevance judgment sets by assigning four-level TREC-style labels to unjudged query-passage pairs via few-shot prompting. When evaluated across five TREC Deep Learning Track collections at removal rates varying from 10 to 90%, it substantially improves system ranking fidelity over the standard practice of treating unjudged documents as non-relevant. The third contribution is UMBRELA, which is a fully automated open-source relevance assessment framework deployed in the TREC 2024 RAG Track across 301 topics, achieving run-level Kendall's tau >= 0.86 against fully manual assessment. All frameworks are released as open-source tools to support reproducible and scalable IR evaluation.
  • Thumbnail Image
    Item type: Item ,
    Investigation of Thermal Behavior in Combustion of Al/Fe3O4 Nanothermite Film
    (University of Waterloo, 2026-06-29) Liu, Jingtian
    Aluminum-based nanothermites are promising energetic materials for rapid heat release and microscale combustion applications, but the mechanisms governing their combustion morphology and flame propagation remain incompletely understood. This thesis investigates Al/Fe3O4 nanothermite thin films, focusing on two controlling factors: equivalence ratio (ER) and particle morphology. Al/Fe3O4 thin films provide a suitable nanothermite platform for studying the transition between destructive multiphase burning and near-gasless condensed-phase propagation due to their high energy release and relatively limited gas production. Specifically, the work examines how ER drives the transition between destructive particle-ejection combustion and structurally preserved near-gasless propagation, and how core–shell (CS) and physically mixed (PM) architectures modify the combustion event. Polymer-assisted Al/Fe3O4 thin films were fabricated with controlled ER values and particle morphologies. Their combustion behavior was characterized using synchronized high-speed optical imaging and infrared thermography, combined with post-combustion SEM/EDS analysis and inverse thermal modeling. For PM films, ER < 3 produced destructive combustion with particle ejection, localized hot spots, and partial removal of the energetic layer. In contrast, ER ≥ 3 led to preserved combustion, where the reacted layer remained attached to the substrate and formed a measurable cooling zone. This transition indicates that increasing ER improves condensed-phase continuity and shifts the combustion mechanism from reactive sintering to diffusion-driven reaction. Particle morphology also strongly affected flame propagation. At ER = 3, CS films propagated faster than PM films, with velocities of 11.93 cm/s and 6.11 cm/s, respectively. The inverse-modeled reaction rate of CS films was also more than twice that of PM films, indicating stronger reaction–transport coupling due to improved fuel–oxidizer contact and shorter transport distances. Cooling-zone analysis showed that preserved reacted layers act as thermal reservoirs, redistributing heat toward the reaction and preheating zones and contributing to flame-front stability. Overall, this thesis demonstrates that both ER and particle morphology can be used to tune combustion mechanism, propagation behavior, and post-combustion structure in Al/Fe3O4 nanothermite thin films.
  • Thumbnail Image
    Item type: Item ,
    A STUDY OF THE CLASSIFICATION AND QUANTIFICATION OF MICROPLASTICS THROUGH RAMAN SPECTROSCOPY AND MACHINE LEARNING
    (University of Waterloo, 2026-06-29) Hogan, Úna Elizabeth
    Synthetic polymers, or ‘plastics’, have become an unavoidable, necessary and ubiquitous part of modern human life. Their low cost, tunable properties, durability and ease of manufacture have led to plastics use in almost every part of day-to-day life including food packaging, car manufacturing and single-use sterile medical equipment. The durability of these plastics, while an advantage in their operational life, results in substantial longevity upon their disposal. After they have been discarded, many plastic particles can exist for up to 1000 years in the environment before their eventual breakdown. Their continued use and disposal as the global population increases have led to a large accumulation of discarded plastics throughout the world. A substantial amount of these exist in sizes of 5 mm or less, and are classified as ‘microplastics’, small pervasive pollutants that have been detected in food, drink, and inside human bodies. It is necessary for researchers to determine suitable ways of characterising and quantifying microplastic particles to increase understanding of their behavior and makeup. Expanding knowledge of the sources, abundance and variety of these particles within the environment can lead to a more comprehensive understanding of the issue of microplastics pollution. The use of Raman spectroscopy as an analytical technique for classification of microplastics particles has emerged as an efficient and accurate tool for characterisation. Traditional validation of Raman spectra using library searches and comparison to reference spectra, however, is often inadequate when the plastics have faced significant environmental degradation, which can alter their Raman spectrum. Alternative validation methods such as machine learning are becoming more widely used as superior techniques to traditional library searches, and have proven fast, effective and cheap ways to identify microplastic particles from their Raman spectra. This thesis describes iterative development of machine learning based methods to semi-automate identification of microplastic particles using Raman spectroscopy.
  • Thumbnail Image
    Item type: Item ,
    Benthic macroinvertebrate assemblages and freshwater food webs of beaver-impounded streams in the eastern Canadian Arctic
    (University of Waterloo, 2026-06-29) Gao, Katelyn
    As circumpolar warming facilitates the shrubification of Arctic landscapes, the distribution of North American beavers (Castor canadensis) in Canada has been expanding northward, raising concern in Inuit communities. Though ecosystem engineering by beavers in temperate regions is well-documented, there is limited research that examines the effects of beaver impoundments in the tundra. Freshwater streams in the Arctic support subsistence fish populations and it is currently unclear how flow attenuation by dams will affect the habitat quality or prey resources of resident species. This research assesses differences in the benthic macroinvertebrate diversity and trophic structure of beaver-impounded streams above and below the treeline in Nunavik. Invertebrates and consumer stable isotopes were compared downstream and upstream of dams to characterize changes in assemblage composition, basal resource reliance, and Layman’s food web metrics. Shannon-Weiner diversity and the percentage of lotic invertebrates were lower upstream of beaver dams. Filter-feeders and EPT taxa (Ephemeroptera, Plectoptera, Trichoptera) decreased with variables associated with lentic conditions, such as reduced stream velocity, increased depth, and finer substrates. Geomorphic-driven differences in assemblage composition, without exhibiting changes in richness or abundance, suggest restructuring in response to upstream habitat transformation. In subarctic forest sites, reliance on terrestrially derived carbon in consumer diets was greater upstream of beaver dams but no effect was observed in shrub tundra sites. Additionally, upstream averages of consumer carbon were more enriched and similar to riparian vegetation than epilithic algae. Although a resource shift was observed, overall food web metrics were not affected by beaver dams. Collectively, the findings presented in this study demonstrate that beaver dam effects below the treeline generally resemble the lotic taxa replacement and dietary shifts reported within their historical range, while recently colonised streams above the tree line appear to be marginally less affected.
  • Thumbnail Image
    Item type: Item ,
    Powder Bed Fusion of Difficult-to-Print Ni-Based Superalloys: Microstructural Evolution and Cracking Behavior
    (University of Waterloo, 2026-06-26) Aghajani, Hamidreza
    IN738 is a precipitation-strengthened nickel-based superalloy that is widely valued in industry due to its excellent creep resistance and good corrosion performance. However, it exhibits poor manufacturability, primarily due to its complex alloy chemistry and the challenges associated with solidification during processing. In this study, a mechanistic process–microstructure–cracking relationship was first established for LPBF-processed IN738LC. It was observed that melt pool geometry plays a critical role in crack formation, with an optimal width-to-depth ratio governing crack susceptibility. Reducing hatch spacing or increasing laser power resulted in grain coarsening, with the effect being more pronounced for hatch spacing reduction. Nevertheless, crack density was significantly reduced with decreasing hatch spacing, which is attributed to improved part densification and a more favorable melt pool geometry. The as-built microstructure was found to be highly non-equilibrium due to the rapid cooling rates inherent to LPBF. It consisted of a γ matrix with cellular/dendritic solidification substructures, submicron carbides located at interdendritic regions, and dispersed oxide particles with Al-rich cores. No γ′ precipitates were detected in the as-built condition. Elemental segregation and oxide formation, combined with thermal stresses, contributed to reduced ductility and promoted crack initiation during processing. In the second stage of the study, heat treatment was employed to develop a high-temperature-capable microstructure, particularly aiming for a controlled distribution and size of γ′ precipitates with spherical, cuboidal, and irregular morphologies. A series of heat treatments with varying solutionizing (S) and ageing (A) temperatures were performed to promote crack healing and to develop an optimized microstructure, particularly γ′ precipitates with desirable size, distribution, and morphology for high-temperature applications. The heat treatment conditions mainly included solutionizing at different temperatures (S), solutionizing followed by low-temperature ageing at 845 °C (SLA), solutionizing followed by high-temperature ageing at 1120 °C (SHA), solutionizing followed by double-ageing at high and low temperatures (SDA), and the industry-recommended standard heat treatment for this material (ST: S1120-A845). It was observed that high-temperature solutionizing promotes a more homogeneous microstructure, whereas at lower temperatures (around or below 1120 °C), homogenization is only partial. It was observed that varying the solutionizing and ageing conditions led to the development of diverse γ′ precipitate size distributions, ranging from unimodal to bimodal and multimodal. Unimodal distributions were dominated by fine secondary γ′ precipitates, while multimodal structures consisted of fine secondary γ′ in conjunction with coarse primary γ′. It was further demonstrated that high-temperature ageing (≈1120 °C) facilitates γ′ coarsening. In contrast, low-temperature ageing (≈850 °C) stabilizes the secondary γ′, resulting in a fine, well-defined, and coherent γ′ distribution within the γ matrix. In addition to γ′ precipitation, other secondary phases were identified. Carbide precipitates, including those enriched in alloying elements such as Ti, were predominantly located along grain boundaries. Moreover, Cr-rich phases were observed to preferentially form at grain boundaries. These Cr-rich precipitates were shown to develop during low-temperature ageing (≈850 °C) and may contribute to the degradation of tensile properties. Solutionizing was identified as the primary factor governing recrystallization. At elevated temperatures, the microstructure underwent full recrystallization, resulting in pronounced grain coarsening. In contrast, at lower temperatures (e.g., ~1120 °C), the grain structure remained largely similar to the as-built condition with minor modifications. Additionally, crack healing was observed at higher solutionizing temperatures and was directly associated with the recrystallization of the material. A solid-state crack healing mechanism was proposed, whereby the high-energy state of the as-built microstructure—characterized by cracks, free surfaces, and high grain boundary density—provides a strong thermodynamic driving force for energy reduction. Upon heating above a critical temperature, this driving force promotes recrystallization and crack closure, leading to a more stable microstructure. In the subsequent phase, mechanical performance was systematically evaluated through room-temperature tensile testing of both as-built and heat-treated samples. Tensile tests at room temperature were performed along the vertical direction (i.e., loading direction parallel to the build direction). The as-built condition exhibited the lowest yield strength, while the standard heat-treated sample showed the highest. Overall, the tensile properties at room temperature were found to be governed by a combination of factors, including dislocation density, LAGB structures, grain size, γ′ precipitation (precipitation strengthening), anisotropy, residual cracking, crack healing, recrystallization, grain coarsening, and the presence of detrimental grain boundary phases. In the as-built state, the relatively lower strength and higher ductility were primarily attributed to strengthening mechanisms dominated by high dislocation density and low-angle grain boundary (LAGB) networks. The superior strength of the standard heat-treated sample in the vertical direction resulted from the synergistic effect of γ′ precipitation strengthening and the retained as-built microstructural characteristics (e.g., columnar grain structure and high LAGB density). In the other heat-treated conditions, strengthening was mainly controlled by γ′ precipitation together with crack healing during high-temperature solution treatment. The higher yield strength of the SLA sample relative to the other highly solutionized conditions was primarily attributed to the finer γ′ precipitates and reduced interparticle spacing. Furthermore, Cr-rich grain boundary phases formed during ageing at 845 °C contributed to intergranular embrittlement and fracture, which was confirmed by EDS analysis. This was consistent with the lower ductility and reduced UTS values observed in the SLA and SDA conditions relative to the SHA condition. The modified heat treatment strategies developed in this study produced a crack-free and nearly isotropic microstructure while providing improved room-temperature mechanical properties compared with the as-built condition. The combination of recrystallization and complete crack healing highlights their potential for high-temperature service, making these heat treatment routes promising alternatives to the conventional industrial heat treatment. In another case study, CM247LC, a non-weldable Ni-base superalloy, was fabricated by electron beam powder bed fusion (EB-PBF) at a wide range of energy levels. For this purpose, variable process parameters were adjusted to investigate their effect on microstructure and crack formation. Samples fabricated at both low and high area energies exhibited pronounced crack susceptibility. At very low energy densities, lack of fusion (LoF) and porosities were observed, while higher energy densities produced denser samples. Adjustments to energy density and process parameters resulted in a grain structure transition from fine-columnar to coarse-columnar and near-single crystal morphologies. Despite these changes, the cracking issue persisted, with micro-cracks observed in low-energy samples and macro-scale cracks, several millimeters long, forming at higher energy densities, highlighting the material’s high sensitivity to crack formation. Both solidification and liquation cracking were identified— the former showing dendritic crack surfaces, and the latter associated with eutectic phases and grain boundary precipitates. Severe recrystallization around cracks was observed at high energy densities, characterized by elevated dislocation densities. EDS analysis revealed hafnium- and silicon-rich precipitates in interdendritic regions and near cracks, contributing to severe hot cracking in the material.