Doctoral Dissertations (Winter 2014 to Present)

Permanent URI for this collection

New submissions to the University of Delaware Doctoral Dissertations collection are added as they are released by the Office of Graduate & Professional Education. The Office of Graduate & Professional Education deposits all dissertations from a given semester after the official graduation date.

Doctoral dissertations from 1948 to present are also available online through Dissertations & Theses @ University of Delaware. Check DELCAT Discovery to locate print or microform copies of dissertations that are not available online.

More information is available at Dissertations & Theses.


Recent Submissions

Now showing 1 - 5 of 2163
  • Item
    Creating an inclusive community of practice with micro-credentials
    (University of Delaware, 2023) Jentzsch, Tracy H.
    Digital competencies have become an essential skill set for most professions. The University of Delaware (UD) needs to recognize and give credit for the digital skills students bring to their graduate studies, and for digital skills learned while in graduate programs. The current structure of humanities graduate programs at UD does not offer this. Further, there is a gap between the number of students that express an interest in graduate-level programs in the humanities, and those who matriculate and graduate from these humanities programs. This gap is especially pronounced for underrepresented minority students. This is particularly striking since UD’s students of color do show an initial interest in digital competencies, and since research shows that people of color contribute significantly to digital spaces. There is a need for UD to better support students of color to maximize and professionalize their digital competencies. In this Educational Leadership Portfolio, I justify the creation of a micro-credential module on digital competencies and describe the design of the scope and sequence of modules that can be utilized in graduate-level public humanities courses to verify personalized digital learning and mastery of discrete competencies. This module will provide graduate students with portable credentials that can be used to demonstrate their skill mastery to an external audience. The design planning for this module will address the engagement and retention of underrepresented minority students to create an intentional community of practice.
  • Item
    Amino-modified porous silica as adsorbents for the removal of aqueous uranium(VI) species: adsorbent design, synthesis, and characterization
    (University of Delaware, 2023) Wei, Kegang
    U(VI), a radionuclides pollution despised in the past decades, has now been frequently detected in various water bodies worldwide with a concentration higher than WHO and EPA. Thus, it sparked interest in developing new approaches to remove it from aqueous solutions. This study produces amino-functionalized porous silica (AFPS) as the adsorbent to remove U(VI) from aqueous solutions. First, mesoporous silica (MPS) materials with different surface areas and average pore sizes were produced by applying urea-formaldehyde resin as the template. After modifying MPS with AEPTES and APTES (AE@MPS and AP@MPS, respectively), the surface properties of the obtained materials are characterized for surface chemical properties via SEM, XPS, NMR, and zeta potential and specific surface area by BET. The surface acidity of AE@MPS and AP@MPS are determined based on electrophoretic mobility measurements, i.e., zeta potential as a function of pH. Second, though used AE@MPS and AP@MPS as adsorbents to remove aqueous U(VI), how the surface properties of adsorbents, such as total pore volume, average pore size, and functional group density will influence its maximum U(VI) adsorption capacity was evaluated and discussed. The result also indicates that, with a given functional group density, AE@MPS and AP@MPS have the highest U(VI) adsorption capacity at a pore size of 4.1 nm and 2.7 nm, respectively. Third, the experimental results are fitted by the Langmuir adsorption isotherm, Potential of Mean Force (PMF) model, and Surface Complex Formation Model (SCFM), as to estimate the adsorption energy ΔG0ads, which consists of specific chemical energy, ΔG0chem, coulombic energy, ΔG0coul, solvation energy, ΔG0solv, and lateral interaction energy, ΔG0lat. From the result of SCFM, for AE@MPS, the values of ΔG0coul and ΔG0chem are close, and both are the main contributor to ΔG0ads; for AP@MPS, the values of ΔG0coul are much larger than ΔG0chem and are the main contributors of ΔG0ads. Fourth, the mechanism of U(VI) adsorption on AE@MPS and AP@MPS was drawn. The U(VI) adsorption capacity of AE@MPS and AP@MPS is greatly affected by pH. At pH < 2, AE@MPS and AP@MPS exhibit no U(VI) adsorption capacity. At 2 < pH < 4, the U(VI) adsorption capacity is mainly brought by SiO-. At 4 < pH < 8, the OH- in the solution rapidly increases. During this phase, U(VI) will desorb from SiO- and form hydrated U(VI) with OH-. Because the structure of hydrated U(VI) is the atomic cluster of UO22+ cation that surrounded by OH- anions. The surface layer of the hydrated U(VI) atomic cluster will be negatively charged, thereby becoming easier adsorbed by the positively charged amino group. Fifth, keep in mind that the adsorption energy distributions of AE@MPS and AP@MPS against aqueous U(VI) are different. By comparing the molecular structure of amino groups on the surface of AE@MPS and AP@MPS, it can be estimated that if the number of amino groups on a single functional branch increase, the proportion of ΔG0chem will increase, leading to larger U(VI) adsorption capacity and more robust selectivity over other aqueous ions. ☐ With the detailed understanding of AE@MPS and AP@MPS as adsorbents against aqueous U(VI), this study is crucial for the design of similar adsorbents.
  • Item
    Learning from high-order data: hypergraphs and smart codes
    (University of Delaware, 2023) Pena Pena, Karelia
    Currently, there is an increasing need to develop tools that allow the processing and exploitation of the massive amount of data available in many fields, such as computer vision, biology, social sciences, computational image, and others. The proposed research focuses on developing novel theories and algorithms that enable learning from data containing high-order inter-relationships. In particular, in data modeled by graphs and hypergraphs structures, the applications range from embedding images into QR codes to classification and denoising in a myriad of fields. Graph signal processing (GSP) techniques are powerful tools that model complex relationships within large datasets and are now used in a number of applications in different areas, including data science, communication networks, epidemiology, and sociology. Simple graphs, however, can only model pairwise relationships among data, preventing their application in modeling networks with higher-order relationships. In representation learning, for instance, considering high-order relationships in data has recently shown to be superior in many applications. For this reason, some efforts have been made to generalize well-known graph signal processing techniques to more complex graphs, such as hypergraphs, which allow for capturing higher-order relationships among data. In this dissertation, we provide a new hypergraph signal processing framework (t-HGSP) based on a novel tensor-tensor product algebra that has emerged as a powerful tool for preserving the intrinsic structures of tensors. The proposed framework allows the generalization of traditional GSP techniques while keeping the dimensionality characteristic of the complex systems represented by hypergraphs. To this end, the core elements of the t-HGSP framework are introduced, including the shifting operators and the hypergraph signal. The hypergraph Fourier space is also defined, followed by the concept of bandlimited signals and sampling. These tools are not only useful in the areas of hypergraph signal processing but also enable representation learning applications such as tensor-based hypergraph neural networks. ☐ Key to the success of hypergraph-based methods is having a meaningful hypergraph that may only be readily available for some applications. Nonetheless, having laid down the essential tools for t-HGSP opens the door for learning the hypergraph topology that dictates the higher-order relationships among the data. Hence, in this work, we also address the challenge of learning the underlying hypergraph topology from the data. As in graph signal processing applications, we consider the case in which the data possesses certain regularity or smoothness on the hypergraph. Given the hypergraph spectrum and frequency coefficient definitions provided by the t-HGSP framework, we propose a method to learn the hypergraph Laplacian from a set of smooth signals by minimizing their total variation on the hypergraph (TVL-HGSP). Additionally, we introduce an alternative approach (PDL-HGSP) that takes advantage of primal-dual-based algorithms and approximations to reduce time and space complexity. Finally, in representation learning, we depart from existing matrix representations of hypergraphs and combine the proposed learning algorithms with novel tensor-based hypergraph convolutional neural networks to propose hypergraph learning-convolutional neural networks (t-HyperGLNN). Compared to state-of-the-art, the learned adjacency tensor provides a more robust representation in high dimensions and the hypergraph signal model joint effects among connected nodes. ☐ Throughout this dissertation, we validate, theoretically and experimentally, that the proposed methods offer significant gains over the state-of-the-art in different applications, which include clustering, denoising, classification, and embedding of QR codes. For the latter, we proposed a fast generation of visually pleasant and robust QR codes. The proposed framework leverages the proposed hypergraph-based algorithms and state-of-the-art deep-learning algorithms to embed a color image into a baseline QR code in seconds while keeping a maximum error probability during the decoding procedure.
  • Item
    Essays on environmental valuation: applications of the travel cost demand model
    (University of Delaware, 2023) Dalvand, Kaveh
    Sound natural resource management takes into account the entirety of costs and benefits associated with policy action. Understanding these consequences in relation to nonmarket products or externalities requires a precise evaluation of consumer preferences and behavior. The chapters of this dissertation establish this information in relation to offshore wind development on the East Coast of the United States and recreational freshwater fishing in the state of Delaware. ☐ In chapter 1, we assess the impact of offshore wind power projects on recreational beach use along the East Coast of the United States using contingent behavior data from a stated preference survey. The data are from a probability-based sample of beachgoers (n = 1,725) who visited beaches from Massachusetts to South Carolina in 2015. The contingent-behavior results are based on responses to visual simulations of wind power projects at seven various offshore distances (2.5 to 20 miles) under clear, hazy, and nighttime conditions. As expected, the closer a project is to the shore, the greater its negative impact on beach enjoyment and visitation. For instance, at 2.5 miles offshore, the aggregate cost to beach visitors in a frequently visited beach (such as Rehoboth Beach, Delaware) may reach $35 million, whereas at 20 miles offshore, the loss would be just $11 million. In our estimations, we account for curiosity trips to view wind turbines. ☐ In chapter 2, we estimate a recreation demand model for warm-water fishing in Delaware and use it to measure welfare gains associated with improved fishing quality as measured by catch rate of fish, diversity of species, and clarity of water. We use a “linked” site choice - trip frequency model with data gathered by the Delaware Division of Fish and Wildlife. Our site choice model includes 118 rivers and lakes in the state with detailed characteristics on each. We develop hypothetical scenarios of fishing quality improvement involving combinations of fish catch, fish diversity, and water clarity and apply it to individual water bodies, water basins, selected water body groupings, and statewide. Values are reported in per-person, per-season, and aggregate terms. ☐ And finally, in chapter 3, we estimate the benefits and costs associated with Delaware’s put-and-take trout fishery. We use existing data gathered by the state to estimate a Random Utility Model of trout fishing and then use the model to simulate the economic benefits associated with existing and possible new trout fishing programs. The costs are estimated using per-fish market rates of stocking in the region by hatcheries employed by the state. We find that all current programs easily pass a benefit-cost test and that modest expansions at existing stocking sites and new sites also make sense from an economic perspective. Although current fees for annual trout licenses produce revenue that falls well below the cost of stocking, from the perspective of social efficiency an even larger program is recommended. The model also implies that anglers are willing to pay a higher fee for purchasing trout licenses.
  • Item
    Computations of structures of protein assemblies from experimental magic angle spinning NMR restraints
    (University of Delaware, 2023) Russell, Ryan W.
    This dissertation concerns with computational aspects of protein structure determination from experimental magic angle spinning nuclear magnetic resonance (MAS NMR) data and by integrating MAS NMR experimental restraints with information obtained by other structural biology techniques, such as cryogenic electron microscopy (cryo-EM) and X-ray crystallography. ☐ In Chapter 1, protein structure calculation approaches are introduced. ☐ Although the framework for NMR protein structure determination has existed for quite some time, the general requirements for obtaining accurate and precise structures, particularly in the solid state, have not been established until recently. Therefore, we have performed a systematic model study to quantify accuracy and precision with varying numbers of distance restraints. The results of this work are discussed in Chapter 2. ☐ Chapter 3 focuses on structure calculations of two crystalline systems: (1) the N-terminal domain (NTD) of SARS-CoV-2 nucleocapsid protein, and (2) the crystalline array of HIV-1 CACTD-SP1 protein bound with assembly co-factor IP6 and a maturation inhibitor Bevirimat (BVM). ☐ Chapters 4 and 5 concern with an integrated approach to determine atomic-resolution structures of large biological assemblies, whereas MAS NMR restraints are combined with information from other experimental and computational methods. In Chapter 4, the structure of tubular assemblies of HIV-1 CA capsid protein is presented, determined by integrating MAS NMR restraints with cryo-EM density in data-driven molecular dynamics (MD) simulations. In Chapter 5, this general approach is expanded and adapted to determine the structure of a motor domain of conventional kinesin-1, KIF5B, bound to polymerized microtubules. The studies presented in these two chapters establish the integrative structural biology framework for determination of structures of large biological systems inaccessible by any single technique in isolation.