Permanent URI for this collection

Duke migrated to an electronic-only system for dissertations between 2006 and 2010. As such, dissertations completed between 2006 and 2010 may not be part of this system, and those completed before 2006 are not hosted here except for a small number that have been digitized. For access to dissertations created prior to 2006 and those not submitted electronically, please see: https://library.duke.edu/find/theses-dissertations.

Policies & procedures governing electronic theses & dissertations can be found on the graduate school website.


Recent Submissions

Now showing 1 - 20 of 5763
  • ItemEmbargo
    Improved Localization Precision in 3D Single-Particle Localization Microscopy via Off-Center Sampling and Its Applications in Living Systems
    (2023) Zhang, Chen

    Optimization of sampling patterns in 3D Real-time Single Particle Tracking (3D-RT-SPT) systems is crucial due to the limited photon yield and irreversible photobleaching of fluorescent materials commonly used in biological imaging. This study focuses on the evaluation of different sampling patterns in a highly flexible and robust 3D Single Molecule Active Real-time Tracking (3D-SMART) microscope, a representative 3D-RT-SPT system, and further application of an information-efficient version of 3D-SMART. The investigation reveals that optimal sampling patterns in the XY-plane and Z-axis, are off-centered, leading to doubled precision. Theoretical analysis further demonstrates that information-efficient sampling patterns coincide with areas of high Fisher information. These findings not only enhanced the precision of the 3D-SMART system but also established a roadmap for achieving information-efficient sampling in similar 3D-RT-SPT methodologies. Utilizing the information-efficient 4-Corners sampling pattern implemented in the 3D-SMART system, further investigations focused on intracellular events, specifically filopodium dynamics, by employing silver nanoparticles (AgNPs) as probes. Filopodia are thin extracellular protrusions that have important functions in various biological events. Due to their small scale and highly dynamic nature, it is difficult to investigate their activities in a physiologically relevant condition. AgNPs offer advantages over conventional fluorescent materials, providing ample photon flux due to their unique photoluminescent mechanisms known as surface plasmon resonance. Upon incubation with live cells, AgNPs exhibited constrained cylindrical diffusion, revealing significant heterogeneity in residence time across the cylindrical surface. Notably, localized "hot spots" repeatedly visited by AgNPs were observed. Force analysis indicated that these hot spots corresponded to local potential wells, where diffusion speed remained relatively stable. These findings provided evidence of nanoscale structures (~50 nm scale) on filopodia, demonstrating charge-dependent interactions with AgNPs. Furthermore, this study presented a framework for extracting hidden information from 3D trajectories of particles interacting with cylindrical structures, surpassing the capabilities of conventional techniques.

  • ItemEmbargo
    Pore-Scale Flow Mechanisms and the Hydrodynamic Porosity of Porous Media in Surface Water Treatment and Groundwater Remediation
    (2023) Frechette, August

    As climate change and growing demand exacerbate water scarcity, it will become more imperative than ever to remediate our natural resources and treat our waste streams. This is especially true if we are to successfully provide clean water for all and ensure the future of endangered species and habitats. Thus, we look to surface water treatment technologies (e.g., granular media and filtration membranes) and groundwater remediation strategies (e.g., the vertical circulation well, rapidly pulsed pump and treat, and bioremediation) to add to our freshwater stores and reduce environmental pollution.

    Complicating the matter is the fact that both surface water treatment and groundwater remediation are reliant, to varying degrees, on flow through porous media. Even without the added complexities of multiphase flows, immiscible fluids, and the time-dependent processes associated with chemical reactions and biofouling, characterizing flow through porous media, properly, is a cumbersome and arduous task. Heterogeneities in the morphology of the medium range from the pore scale, to, in the case of groundwater flows, meters. Resulting is a random distribution of the shape, size, and connectivity of the pore space. To quantify flow through porous media, researchers are forced to either make a set of simplifying assumptions, some more appropriate than others, or more recently, use black-box machine learning models that have little basis in the physicality of the flow. In this work, we choose to focus on one of the standard assumptions researchers make when calculating the pore-scale velocity (i.e., the supposed “static” nature of flow porosity). In relaxing this assumption, we provide a paradigm shift in the modeling of flows through porous media. We apply our theory to flow through and along the walls of microporous membranes, granular media, and aquifer substrates.

    We choose to study pore-scale flow velocity because it is an essential parameter in determining transport through porous media, but it is often miscalculated. Researchers use a static porosity value to relate volumetric or superficial velocities to pore-scale flow velocities. We know this modeling assumption to be an oversimplification. The porosity conducive to flow, what we define as hydrodynamic porosity, exhibits a quantifiable dependence on Reynolds number (i.e., pore-scale flow velocity) in the laminar flow regime. This fact remains largely unacknowledged in the literature. In this work, we quantify the dependence of hydrodynamic porosity on Reynolds number via numerical flow simulation at the pore scale. We demonstrate that, for the tested flow geometries, hydrodynamic porosity decreases by as much as 42% over the laminar flow regime. Moreover, hydrodynamic porosity exhibits an exponential dependence on Reynolds number. The fit quality is effectively perfect, with a coefficient of determination of approximately 1 for each set of simulation data. We then demonstrate the applicability of this model by validating a high fit quality for a range of rectangular and non-rectangular cavity geometries. Finally, we show that this exponential dependence can be easily solved for pore-scale flow velocity using only a few Picard iterations, even with an initial guess that is over 10 orders of magnitude off. Not only is this relationship a more accurate definition of pore-scale flow velocity, but it is also a necessary modeling improvement that can be easily implemented.

    In the chapters that follow our introduction of hydrodynamic porosity, we apply the concept to subsurface flow modeling for groundwater remediation via the vertical circulation well and flows over patterned membrane surfaces for surface water treatment – supposing that a hydrodynamic porosity parameter could be defined for the surface pattern of a membrane and then correlated to the rate of particle deposition (and therefore fouling) at the membrane surface.

    In the future, we aim to explore the applicability of the hydrodynamic porosity model to microporous membrane wall flows. Although the characteristic length scale of the membrane wall is admittedly much smaller than the characteristic length scale of granular media, microporous membranes, like granular media, have dead-end pores. Thus, it remains necessary to determine the effect of these dead-end pore volumes on membrane wall flows. Preliminary experimental data previously collected from a hollow-fiber ultrafiltration membrane will be used to verify our numerical results.

    Following our study of steady flows, we pivot to the analysis of rapidly pulsed flows and the mixing mechanisms these flows induce at the pore scale (i.e., the deep sweep and vortex ejection) in cavities and other effectively immobile zones. These mechanisms have been shown to significantly reduce contaminant recovery time in media with significant immobile zone volume. This finding suggests substantial cost-savings for treatment and remediation methods that utilize rapidly pulsed flows.

    Regarding groundwater remediation, we estimate that the cost savings from utilizing rapidly pulsed flows could be on the order of magnitude of 100 billion USD. But this calculation assumes that we can remediate the entirety of a contaminated groundwater matrix with the mixing mechanisms induced by rapidly pulsed pump-and-treat. In application, induced oscillations will only reach a small volume of the flow field before dissipating to a negligible amplitude. Equally important, these oscillations will only induce a deep sweep or vortex ejection if the mean pore-scale flow velocity is above a Reynolds number of 0.1. Following, we use our model of hydrodynamic porosity to determine the magnitude of the volume we expect to benefit from rapidly pulsed pumping in a vertical circulation well.

    Finally, given the similarity in characteristic length scale, we liken flow in the dead-end pore space of groundwater matrices, to flow past the channels in patterned membrane surfaces. We find that for the studied surface pattern, the vortex ejection and deep sweep are still present in highly laminar flows (i.e., a Reynolds number of 1600 for pipe flows). We hypothesize that these mechanisms can prevent particle deposition at the membrane surface, and when used as a cleaning mechanism, can remove loose deposits that would otherwise adhere to the membrane surface. It is also likely that these mechanisms would speed up the regeneration of fouled granular media used to remove suspended solids, microorganisms, and organics (i.e., sand and granulated activated carbon) from wastewater.

  • ItemEmbargo
    Self-assembly of frustrated and disordered systems -- equilibrium microphases and out-of-equilibrium active matter
    (2023) Zheng, Mingyuan

    The potential for matter to self-assemble in complex morphologies and to follow intricate dynamics is of great fundamental interest as well as for a broad array of applications in nanotechnology, catalysis, drug delivery, and beyond. The inherent complexity of these processes, however, is a challenge both to theories and to numerical simulations. This dissertation focuses on two such complex systems: (i) equilibrium microphases and (ii) out-of-equilibrium active matter. In the first part, a model of microphase formers with short-range attractive and long-range repulsive (SALR) interactions is investigated. First, using various advanced Monte Carlo methods, a thorough characterization of the disordered regime of this model is obtained. Given that this regime had thus far eluded systematic characterization, the results serve as a benchmark for evaluating algorithmic performances. Around the order-disorder transition, existing algorithms nevertheless remain inefficient at sampling configurations, due to the severe critical slowdown. To understand the limitations of these algorithms in microphase formers, simple spin models are studied in a one-dimensional chain, on a two-dimensional square lattice and on the Bethe lattice. The results reveal that the existing cluster algorithms overestimate the correlation length, and therefore its divergence no longer coincides with the critical point. In fact, because frustration depresses the correlation length, a negative bonding probability would formally be needed for a cluster scheme to succeed. Two cluster algorithms that approximate this effect are proposed and shown markedly to improve sampling, albeit only for small to moderate system sizes. In the second part, the out-of-equilibrium nature of active matter is confronted with the goal of obtaining first-principle descriptions of its behavior. Sluggish dynamics (and arrest) at high densities in these systems is a particularly challenging concern. To obtain insight into the thus-far unsolved dynamic mean-field theory (DMFT), which is exact in the high-dimensional limit $d\rightarrow\infty$, I have conducted simulations of active Brownian particles (ABP) in the heterogeneous random Lorentz gas environment, using event-driven Brownian dynamics algorithm in $d$ spatial dimension. The results reveal that activity shifts the glass transition to higher density and saturates around the percolation threshold. The non-Gaussian parameter is also markedly different from that of the passive systems. These findings suggest that non-trivial processes might be at play in the arrest of active matter, which helps chart the way for eventually solving and extending the DMFT.

  • ItemEmbargo
    The Impacts of Disruptive Environmental Change on Vital Microbial Ecosystems
    (2023) Kilner, Christopher

    Microbes both affect the global nutrient and carbon cycles that influence climate change and our modern environment, and are in turn influenced by environmental change. Through resource acquisition, metabolism, symbiosis with plants and other organisms, and functional & taxonomic composition, microbial communities determine ecosystem services and stability. Through a mix of experiments, observational studies, and theory, we investigate the dynamic interplay between microbes and their changing environment.

    In our first research chapter, we address these questions using data from a longterm whole-ecosystem warming experiment at a boreal peatland. We explore how temperature and CO2 jointly influence communities of abundant, diverse, yet poorly understood non-fungi microbial Eukaryotes (protists). Using a combination of high-throughput fluid imaging and 18S amplicon sequencing, we demonstrate a taxonomic convergence but a functional divergence of microbes in response to warming and elevated CO2 ; we find novel evidence that warming effects on functional composition are reversed by elevated CO2 and amplified in larger microbes. These findings show how the interactive effects of warming and rising CO2 could alter the structure and function of peatland microbial food webs — a fragile ecosystem that stores 25% of terrestrial carbon.

    In our second research chapter, we examine National Ecological Observatory Network (NEON) 16S amplicon sequencing data from a forest fire to examine the impact of fire on soil microbial communities. We report drastic fire-induced shifts in bacterialcomposition post-fire, with a reduction in alpha- and beta-diversity and no significant recovery of the soil microbiome 1-year post-fire. We also show that certain bacterial clades are clear indicators of fire, with heat-tolerant taxa increasingly dominant post-fire within our study plots, while other clades are indicative of a system without fire. These findings show how forest fires in landscapes adapted to infrequent fire regimes — such as moist, montane communities — may contain soil microbiomes that are less likely to recover post-fire, a concern as climate change alters many regions globally and large-scale forest fires become increasingly common in areas that have not historically experienced them.

    In our last research chapter, we experimentally investigate the effects of temperature, genetic diversity, nutrient levels, and competition on body size (M) and density (N) in Tetrahymena thermophila, based on a unified differential equation model. Our findings highlight the crucial role of environmental conditions in shaping the body size and density of T. thermophila, emphasizing the intricate effects of environmental change on ecosystems. Our model analysis further reveals the specific parameters influenced by temperature, genetic diversity, competition, and resource availability, providing insights into the underlying mechanisms driving population dynamics. Our study sheds light on the complex interplay between body size, environmental factors, and ecological dynamics, contributing to a better understanding of these vital microbial ecosystems.

  • ItemEmbargo
    I. Development of sulfonyl piperazine LpxH inhibitors against multi-drug resistant Gram-negative bacteria II. Exploration towards new TRPM8 agonists for dry eye disease
    (2023) Ennis, Amanda

    The discovery of new drug scaffolds is a time-consuming and costly process. As new technologies are developed and implemented into the drug discovery process, these costs are being lessened. Of these methods, computer-aided drug discovery is the most widely implemented as it reduces the number of compounds that need to be explored synthetically. Computer-aided drug discovery has also become a powerful tool as it envelops a wide range of computational methods that can be used to identify or optimize lead compounds for various disease states or drug targets.The first part of the dissertation presents the design, synthesis, and biological evaluation of sulfonyl piperazine LpxH inhibitors. The emergence of widespread antibiotic resistance among Gram-negative pathogens has led to an urgent need for a new class of antibiotics to fight multidrug-resistant Gram-negative bacteria. Lipid A is a critical component of lipopolysaccharides present on the outer membrane of Gram-negative bacteria that prevents penetration of the membrane by external detergents and antibiotics. The biosynthesis of lipid A occurs through the Raetz pathway via 9 distinct enzymes, one of which is known as LpxH. The dual mechanism of cell killing due to the inhibition of LpxH and its presence in the majority of Gram-negative bacteria makes it an attractive target for novel antibiotics. Based on the structure of AZ1, a small molecule inhibitor of LpxH identified by AstraZeneca, we have synthesized and evaluated a series of sulfonyl piperazine LpxH inhibitors. Our study allowed for the establishment of a comprehensive structure-activity relationship of the various components of AZ1. We also obtained the first crystal structure of Klebsiella pneumoniae LpxH in complex with a sulfonyl piperazine LpxH inhibitor which illuminated how this class of LpxH inhibitors fits into the binding pocket of LpxH and guided the design of inhibitors with increased potency. Our findings will be instrumental in the discovery of new antibiotics against multidrug-resistant Gram-negative pathogens. The second part of the dissertation describes the modification of existing TRPM8 ligands and the search for alternative scaffolds towards the development of a new TRPM8 agonist for the treatment of dry eye disease. As the use of technology continues to increase, so does the incidence of dry eye disease, especially in younger individuals. Modulation of the transient receptor potential cation channel TRPM8 has been implicated as a route for the treatment of dry eye disease as TRPM8 plays a critical role in basal tear production, eye blinking, and is the sensor for eye dryness. Current well-known modulators of TRPM8 such as menthol, icilin, and WS-12, are not suitable for ocular use as they have undesirable physical properties and have off-target effects on other transient receptor potential channels. The recently discovered small molecule cryosim-3 which was identified as a TRPM8 specific agonist provides a good starting point for the development of a new agonist suitable for ocular studies. Through the use of cryo-electron microscopy, molecular docking, and virtual screening, we have begun to understand the different binding orientations of various TRPM8 ligands. This information has aided us in probing new chemical space for the identification of a TRPM8 agonist with improved potency, selectivity, and physical properties over existing agonists.

  • ItemEmbargo
    The Afterlives of Archaic Greek Kouroi: An Archaeology of a Greek Sculptural Type
    (2023) Gkiokas, Nikos

    My dissertation “The Afterlives of Archaic Greek Kouroi: An Archaeology of a Greek Sculptural Type” examines the reuses of several kouros fragments after the discard of their initial setting. The focus is centered on their uses at their find spots. I present the reuses chronologically and by type of reuse and find spot. Also, I discuss the find spot context of four fragments found reused in walls on Despotiko and several others found built into walls or deposited in terraces.My thesis has three main objectives. First, I try to classify the reuses of the fragments. This task currently remains a gap in the scholarship. Second, I intend to enrich and broaden the discussion on kouroi beyond their initial display context and their creation phase. Third, I highlight the importance of archaeological context for such study. I argue that the reuse of the four fragments from Despotiko at doors is a structured, deliberate, and conscious event which attributes meaning to the fragments that extends beyond their practical function. Beyond Despotiko, I show that all reuses have a meaningful setting through analysis of other reused fragments at other sites. However, the type and quality of the human engagement in each case varies.

  • ItemEmbargo
    Transfer Learning in Value-based Methods with Successor Features
    (2023) Nemecek, Mark William

    This dissertation investigates the concept of transfer learning in a reinforcement learning (RL) context. Transfer learning is based on the idea that it is possible for an agent to use what it has learned in one task to improve the learning process in another task as compared to learning from scratch. This improvement can take multiple forms, such as reducing the number of samples required to reach a given level of performance or even increasing the best performance achieved. In particular, we examine properties and applications of successor features, which are a useful representation that allows efficient calculation of action-value functions for a given policy in different contexts.

    Our first contribution is a method for incremental construction of a cache of policies for a family of tasks. When a family of tasks share transition dynamics but differ in reward function, successor features allow us to efficiently compute the action-value functions for known policies in new tasks. As the optimal policy for a new task might be the same as or similar to that for a previous task, it is not always necessary for an agent to learn a new policy for each new task it encounters, especially if it is allowed some amount of suboptimality. We present new bounds for the performance of optimal policies in a new task, as well as an approach to use these bounds to decide, when presented with a new task, whether to use cached policies or learn a new policy.

    In our second contribution, we examine the problem of hierarchical reinforcement learning, which involves breaking a task down into smaller subtasks which are easier to solve, through the lens of transfer learning. Within a single task, a subtask may encapsulate a behavior which could be used multiple times for completing the task, but occur in different contexts, such as opening doors while navigating a building. When the reward function changes between tasks, a given subtask may be unaffected, i.e., the optimal behavior within that subtask may remain the same. If so, the behavior may be immediately reused to accelerate training of behaviors for other subtasks. In both of these cases, reusing the learned behavior can be viewed as a transfer learning problem. We introduce a method based on the MAXQ value function decomposition which uses two applications of successor features to facilitate both transfer within a task and transfer between tasks with different reward functions.

    The final contribution of this dissertation introduces a method for transfer using a value-based approach in domains with continuous actions. When an environment's action space is continuous, finding the action which maximizes an action-value function approximator efficiently often requires defining a constrained approximator which results in suboptimal behavior. Recently the RBF-DQN approach was proposed to use deep radial-basis value functions to allow efficient maximization of an action-value approximator over the actions while not losing the universal approximator property of neural networks. We present a method which extends this approach to use successor features in order to allow for effective transfer learning between tasks which differ in reward function.

  • ItemEmbargo
    Control through Constraint
    (2023) Zhang, Boyang

    Recently multi-agent navigation robots have been gaining increasing popularity indiverse applications such as agriculture, package delivery, exploration, search and rescue due to their maneuverability and collaborativity. The control algorithm is the nucleus of such intelligent and autonomous robots performing tasks. In real-world applications, the robots are subjected to nonlinear dynamics, external disturbances, actuator saturation/dynamics, and modeling, estimation, and measurement errors. Furthermore, teams of robots are needed to perform collaboratively while ensuring inter-robot and robot-obstacle collision avoidance.

    To address these needs, a novel control paradigm has been developed for multiagent navigation robots that possesses safety, robustness, resilience, scalability, and computation efficiency. The control rule is fully defined by the current active subset of a superset of inequality constraints, which contrasts methods of minimizing a weighted cost function subject to stability constraints. The advantages of this method are that

    • the constraints (equality, inequality, holonomic, nonholonomic, scleronomic, rheonomic, etc.) can be handled without trying to \look ahead" to a finite time horizon;• the nonlinear control actions are specified by instantly solving a linear matrix equation; • it does not involve a cost function; • it does not involve any dynamics linearization; • the control parameters are physically interpretable; • actuator saturation and actuator dynamics are readily incorporated; and • it is applicable to fully nonlinear, time-varying, and/or arbitrary-order dynamical systems; and • it can simultaneously control the position and orientation of mechanical systems in one unified step.

    These features are achieved through a novel generalization of Gauss’s Principleof Least Constraint (GPLC). GPLC was originally conceived to incorporate hard equality constraints into second-order dynamical systems. The contribution of this dissertation is to define the control actions from the Lagrange multipliers associated with inequality constraints (e.g., collision avoidance constraints) and to accommodate dynamical systems of any order. Thus, the constrained equations of motion are expressed as a Karush-Kuhn-Tucker (KKT) system (a linear matrix equation), which is solved without iteration at each time step.

    This constraint-based control has been applied to the navigation control of multiagent, multi-swarm systems of double integrators, fully nonlinear quadrotor drones, and nonholonomic, differential drive, wheeled mobile robots subjected to actuator saturation, actuator dynamics, and external disturbances. Two types of constraints are considered for the aforementioned three types of systems: path following and collision avoidance constraints. Each constraint can be formulated based on vector norms or vector components and can be in either equality or inequality format. Thevector-component-based collision constraints lead to a natural byproduct of resolving deadlock in navigating swarms. Furthermore, through a partition of collision avoidance constraints among colliding agents, the control architecture for the navigation swarms can be centralized or decentralized. Numerical studies on swarms of double integrators, nonlinear quadrotor drones, and nonholonomic wheeled mobile robots have demonstrated the effectiveness and efficiency of the proposed approach.

  • ItemEmbargo
    Managing (Unconventional) Water: Essays on Expert Knowledge, Media Framings, and Stakeholder Debates
    (2023) Patel, Ekta

    Ensuring access to adequate quantities of water for basic needs remains a fundamental goal and challenge in many world regions amid the ongoing climate crisis. Yet the uncertainties with existing water supplies coupled with management challenges and rising water demand have shifted attention towards unconventional water sources. Unconventional waters are alternative water sources obtained by using technologies, such as desalination, to transform previously untapped water into drinking water. While desalination is touted as a technical ‘solution’ to alleviate freshwater scarcity by policymakers, businesses, and the public alike, its high financial costs, intensive energy needs, local community impact, and harm to marine life, as well as other yet unseen effects, make it socially, economically, environmentally, and politically contentious. Before the shift towards unconventional waters, and desalination in particular, extends further, it is imperative for informed policymaking to understand how these water options emerge and what constitutes the knowledge base on them.

    This dissertation examines how water management is shaped at the level of international organizations and what information on desalination is shared and debated across two other policy-relevant settings: the global media and a local community. It focuses on these three different settings to capture wide information streams that individually and collectively generate some of the corpus of knowledge on water management and desalination. The first chapter examines UN-Water, the coordinating body on international water policy for the UN system, and how it uses expert knowledge to shape the global water agenda, including the foundations upon which unconventional waters become part of this agenda. The second chapter applies automated content analysis to global news media coverage to examine which framings related to desalination’s adaptive and maladaptive features are most prevalent and in what combinations. The third chapter presents a discourse analysis of stakeholder statements and official deliberations tied to the Huntington Beach desalination facility in southern California, which was ultimately denied, to reveal the key storylines that arise both in favor of and in opposition to desalination during permitting debates. This dissertation’s key findings highlight the significance of expert knowledge in shaping global water policy, biases in the current information landscape of desalination as a climate adaptation strategy, and the opportunities for engaging with diverse stakeholders for collaborative decision-making on water management options like desalination.

  • ItemEmbargo
    Community Dispute Resolution and International Peacebuilding: Competitors or Complementary Actors? Evidence from Liberia
    (2023) Torres, Priscilla

    This dissertation explores the conditions under which international peacebuilding and community dispute resolution (CDR), or the ways in which communities try to address issues, complement or undermine one another. It argues that community dispute resolution can, under some circumstances, address issues before they become larger-scale sources of conflict, thereby enhancing local peace. International peacebuilding in particular, tries to promote a greater reliance on state institutions and broad participation in local affairs. Depending on the CDR characteristics in place within a community, these norms will align or will not be aligned with the CDR practices that the community already has in place. International peacebuilding will either complement, supplement or undermine CDR efforts at maintaining local peace.

    Through novel data collection using community leader interviews, community histories and the implementation of two surveys in Liberia (one community leader survey and one household survey representative at the neighborhood level), this dissertation develops and tests the argument at hand. The qualitative evidence is used to further develop the theoretical expectations of the project and to probe specific mechanisms by which the arguments operate. The quantitative survey evidence is analyzed as the primary test of the theoretical argument. Item response theory (IRT) models are used to further develop the measures included in the statistical models and several different kinds of statistical models (ordinary least squares regression, logit models, hierarchical linear models and hierarchical ordinal logistic models), depending on the dependent variable of interest, are used to test the observable implications of the theoretical framework.

    There are several findings that emerge from the project as a whole. First, and perhaps most important, peacekeeping exposure at the micro-level has a long-lasting influence on micro-level peace dynamics, however this influence is conditional on the CDR structures in place at the neighborhood level. Second, CDR contributes to and undermines local peace depending on the CDR characteristic at hand. Third, the influence of CDR and peacekeeping exposure are not the same for different components of local peace: i.e. physical security versus perceptions of peace versus dynamics within and between communities. Fourth, peacekeeping exposure does have the potential to undermine, complement or supplement CDR efforts, depending on the CDR characteristic of interest. Lastly, in ``post-conflict'' environments with large-scale peacekeeping operations, peacekeepers often promote a greater reliance on the state. However, doing so can lead to complicated dynamics that are worthy of more time and policy attention. While these efforts can be beneficial to communities if they exhibit norms that are consistent with the goals of peacekeeping operations, they can also lead to adverse effects such as forum shopping, undermining community leader authority and exacerbating local cleavages. CDR can help to ``keep the peace'' once peacekeepers have departed, however, they should not necessarily be considered as substitutes for the state. Instead, this dissertation suggests that they are an additional actor in a complex web of institutions tasked with maintaining order and peace.

  • ItemEmbargo
    Marginalized Voices and Nontraditional Pathways in Higher Education in the Late Roman Empire
    (2023) Küppers, Sinja

    This study analyzes marginalized voices and nontraditional pathways of higher education in the late Roman Empire and diversifies our notion of who was part of “the” educated elite in ancient higher education. I focus on upper-class learners who did not have access to the family’s wealth or faced difficulty with pursuing the discussed traditional paths of schooling designed for young men from wealthy families. The discussed marginalized voices include fatherless students, women, late learners, autodidacts, and disabled students. Since most sources on Roman education were authored by elite men who mention marginalized voices in passing, I piece together the experiences of nontraditional learners and marginalized members of Roman education from an array of literary and epigraphic sources, including letters from teachers to students and families, church historians and Christians commenting on women, orations, tomb stones and legal documents. Most sources discussed are dated to the fourth century C.E., highlighting a period in which girls and women from the upper-class gained a voice in ascetic communities, as educational leaders and philanthropes and in which educational mobility across the Roman Empire flourished. Using Bourdieu’s theory of capital, I analyze how diverse family and educational backgrounds impacted the educational paths of students, discuss the student voices often overlooked in scholarship and bring attention to the challenges that nontraditional and marginalized students have experienced in higher education.

  • ItemEmbargo
    Charge Transfer and Chiral Imprinting on Nano Particles
    (2023) Zhang, Zhendian

    Quantum dots (QDs), being semiconductor particles small enough to exhibit quantum mechanical properties, are leveraged in numerous applications including energy harvesting, quantum computing, and biomedical imaging. Our research theoretically examined two functional QD systems: (1) first one undergoes the electron transfer (ET) between CdSe QDs facilitated by a solvent and linker molecule, and (2) a second performs charge transfer (CT) and triplet energy transfer (TET) between CdSe/CdS core/shell QDs and ligand acceptor. These processes were scrutinized based on the electronic coupling tunneling through the shell, reaction free energy change, and reorganization energy. Furthermore, we explored chiral imprinting mechanism in perovskite nanoplatelet and the ET dynamics of organic molecules with pathway interferences.We devised a discrete variable representation (DVR) method to simulate ET between CdSe QDs, mediated by a solvent and a linker molecule. Employing the effective mass approximation (EMA), we characterized the QDs, ligand, and solvent, studying the distance dependence of donor-acceptor coupling via the energy splitting method. We found that the ET coupling decreases exponentially with the interdot distance. The decay constant is dictated by the size of the linker and the tunneling barrier through the solvent and ligand. When the donor and acceptor sizes significantly iv surpass the diameter of the linker, such as in large QDs with an alkane chain linker, the ET is predominantly regulated by through-solvent tunneling. The DVR method was also applied to simulate the CdSe/CdS core/shell QD system. Notably, experiments observe a subtle TET rate decay with an increase in shell thickness. Simulating QDs of varying shell thickness, we found a large TET coupling decay constant. Marcus analysis finds that the QD TET operates in a deeply inverted regime, where an increase in shell thickness reduces the driving force, leading to a significant increase of the Franck-Condon factor. This in turn offsets the exponential decrease of the electronic coupling with shell thickness. Further, our findings demonstrated that variations in shell thickness could further decrease the TET rate decay constant. Applying density functional theory (DFT) calculations and a charge-perturbed- particle-in-box model, we investigated chiral imprinting of perovskite nanoplatelet by chiral ligands. We found that the imprinted CD signal is sensitive to the orientation of the chiral ligand. As the proportion of chiral surface ligands grows, our model calculations find that the intensity of the CD signal from the lowest energy exciton transition saturates. v

    We also examined the effects of light polarization on the ET yield in the coherent limit, using a model Zinc porphyrin as the ET donor due to its near degenerate excited states and orthogonal transition dipole moments. These two donor excited states, coupling to the acceptor state, produce pathway interference that strongly impact the ET. Introducing dissipation due to system-environment interaction via the Lindblad equation, we found that the ET yield is influenced by the initial light polarization. In the DA system, linearly polarized light (LPL) is predicted to induce an ET yield difference of up to 85% with a 100 fs dephasing time, while the yield difference elicited by R- circularly polarized light (R-CPL) and L-circularly polarized light (L-CPL) was insignificant. The model Hamiltonian was subsequently simulated and the dynamics was predicted by our collaborators with a trapped ion qutrit system.

  • ItemOpen Access
    Efficient and Collaborative Methods for Distributed Machine Learning
    (2023) Diao, Enmao

    In recent years, there has been a significant expansion in the scale and complexity of neural networks. This has resulted in significant demand for data, computation, and energy resources. In this light, it is crucial to enhance and optimize the efficiency of these ML models and algorithms. Additionally, the rise in computational capabilities of modern devices has prompted a shift towards distributed systems that enable localized data storage and model training. While this evolution promises substantial potential, it introduces a series of challenges. Such challenges encompass addressing the heterogeneity across systems, data, models, and supervision, balancing the trade-off among communication, computation, and performance, as well as building a community of shared interest to encourage collaboration in the emerging era of Artificial General Intelligence (AGI). In this dissertation, we contribute to the establishment of a theoretically justified, methodologically comprehensive, and universally applicable Efficient and Collaborative Distributed Machine Learning framework. Specifically, in Part I, we contribute to methodologies for Efficient Machine Learning including for both learning and inference. In this direction, we propose a parameter-efficient model, namely Restricted Recurrent Neural Networks (RRNN), that leverage the recurrent structures of RNNs using weight sharing in order to improve learning efficiency. We also introduce an optimal measure of vector sparsity named the PQ Index (PQI), and postulate a hypothesis connecting this sparsity measure and compressibility of neural networks. Based on this, we propose a Sparsity-informed Adaptive Pruning (SAP) algorithm. This algorithm adaptively determines the pruning ratio to enhance inference efficiency. In Part II, we address both efficiency and collaboration in Distributed Machine Learning. We introduce Distributed Recurrent Autoencoders for Scalable Image Compression (DRASIC), a data-driven Distributed Source Coding framework that can compress heterogeneous data in a scalable and distributed manner. We then propose Heterogeneous Federated Learning (HeteroFL), demonstrating the feasibility of training localized heterogeneous models to create a global inference model. Subsequently, we propose a new Federated Learning (FL) framework, namely SemiFL, to tackle Semi-Supervised Federated Learning (SSFL) for clients with unlabeled data. This method performs comparably with state-of-the-art centralized Semi-Supervised Learning (SSL), and fully supervised FL techniques. Finally, we propose Gradient Assisted Learning (GAL) in order to enable collaborations among multiple organizations without sharing data, models, and objective functions. This method significantly outperforms local learning baselines and achieves near-oracle performance. In Part III, we develop collaborative applications for building a community of shared interest. We apply SemiFL to Keyword Spotting (KWS), a technique widely used in virtual assistants. Numerical experiments demonstrate that one can train models from the scratch, or transfer from pre-trained models in order to leverage heterogeneous unlabeled on-device data, using only a small amount of labeled data from the server. Finally, we propose a Decentralized Multi-Target Cross-Domain Recommendation (DMTCDR) which enhances the recommendation performance of decentralized organizations without compromising data privacy or model confidentiality.

  • ItemEmbargo
    Essays on International Tax and Firm Behavior
    (2023) Boller, Lysle Thomas

    This dissertation examines unique tax incentives faced by multinational corporations (MNCs). Although nearly all businesses engage in profit-maximizing behavior, many do not have the ability to set up numerous subsidiaries in multiple jurisdictions. MNCs, on the other hand, often operate in a global context. The largest MNCs have hundreds of related subsidiaries that are distributed worldwide. Firms can utilize this geographic dispersion to take advantage of heterogeneity in tax laws across the jurisdictions in which they operate to minimize their tax burden. In the following chapters, I examine (1) common mechanisms used by some of the largest MNCs that facilitate the movement of income from high-tax jurisdictions to low-tax jurisdictions, (2) the extent to which researchers can trust different sources of data to measure the foreign activity of MNCs, and (3) how regulatory uncertainty concerning tax regimes can distort various aspects of firm behavior and financial reporting.

    The second chapter studies the adoption and use of specific forms of tax planning by US multinational corporations (MNCs). Along with my coauthors, I use IRS data to identify hybrid tax planning affiliates (HTPs) that allow MNCs to avoid corporate income taxes by targeting mismatches between US and Irish, Dutch, and Luxembourg tax law. By 2016, we find that more than 35\% of the foreign profits of US MNCs were linked to these HTPs. Difference-in-difference models reveal that, after adoption, corporations intensify behavior commonly linked to profit shifting, significantly increasing related-party loans, book values of foreign intangible assets, and profits held abroad relative to control MNCs. These changes coincide with stark reductions in foreign effective tax rates. MNCs that adopt HTPs also experience large increases in foreign tangible assets and in domestic R\&D, payroll, and investment relative to other types of multinationals. To separate selection and treatment effects, we develop and estimate a model that rationalizes the selection of MNCs into HTPs and the changes that we observe in their reported economic activity.

    The third chapter examines whether data that is commonly used to study MNCs can be trusted by researchers. Recent literature has noted the potential for significant measurement error, unique to MNCs, in tax and accounting data. I examine a key source of measurement error that can affect administrative tax data --- aggregation error. I link data from tax filings and public disclosures to quantify the extent to which commonly-used aggregation techniques may result in double counting of foreign earnings. A comparison of book and tax data reveals that aggregation error has been increasing over time. The matched sample also reveals large inconsistencies in the reporting of corporate income tax across firms' books and tax filings, particularly in extractive and financial industries. I introduce a simple correction for aggregation error and examine the extent to which it harmonizes measurement of foreign income and tax rates across firms' books and tax filings. Applying this correction yields a 30% reduction in the magnitude of foreign earnings as measured in tax data in 2016 and significantly reduces book-tax differences. Furthermore, this correction breaks the systematic relationship between book-tax differences and the size of multinationals' foreign affiliate networks. Unadjusted book-tax differences are increasing over time. After applying the correction, this is no longer true. Finally, I replicate estimates from prior literature that do not correct for aggregation error and discuss robustness.

    The fourth chapter examines how MNCs reacted to a regulatory loophole that emerged after the implementation of transfer pricing regulations passed in 1995. This loophole allowed US-based MNCs to shift labor costs from low-tax foreign affiliates to US parent companies by utilizing stock-based compensation. A 2005 tax court ruling created regulatory uncertainty around enforcement of the loophole, increasing the market value of firms that were in a position to take advantage of it. This period of regulatory uncertainty extended for over a decade. During this period, exposed firms increased their overall usage of stock-based compensation and its relative intensity as a share of their total labor costs. Exposed firms also increased their overall R&D activity and claimed domestic tax credits for larger portions of their overall R&D expense.

  • ItemOpen Access
    Investigating the Influence of Red Blood Cell Interactions on Large-Scale Cancer Cell Transport: Bridging the Gap through Advances in Computational Techniques
    (2023) Roychowdhury, Sayan

    Metastatic cancer, the leading cause of cancer mortality, involves the complex process of circulating tumor cells (CTCs) spreading through the bloodstream and forming secondary tumors far away from the primary mass, which often travel a distance many thousand times its size. The interactions between CTCs and their neighboring red blood cells (RBCs), as well as the local hemodynamics in vessels, play a crucial role in determining these cells' fate; however, the mechanisms guiding their transit are still unclear. Predicting secondary tumor sites remains challenging due to the intricate dynamics of CTC migration. Thus, there remains a need to understand the interplay between the fluid dynamics, intercellular interactions, and vessel topology which can determine the fate of the CTC and subsequent likelihood of cancer progression.

    Investigating CTC transport has involved a range of \textit{in vivo} and \textit{in vitro} studies to unravel the intricate mechanisms that dictate cellular outcomes. However, the process of tracking an individual CTC's trajectory through the massive vascular system is still not possible today \textit{in vivo}. The integration of \textit{in silico} models has proven instrumental, complementing traditional experimental approaches. In this work, we utilize our advanced fluid dynamics solver HARVEY to perform high-fidelity hemodynamic simulations to capture CTC dissemination. We outline several key contributions, including the addition of new physics interactions models and software optimizations, to enable these simulations to better capture biological phenomena and run to completion within a reasonable timeframe.

    Numerical optimizations for \textit{in silico} models are still necessary: the drastic difference in length scales of CTC size versus distance traveled hinders current simulation models. To accurately capture intercellular dynamics, interactions must be modeled with sub-micrometer precision; meanwhile, the characteristic length scale of CTC traversal through the blood stream can be on the order of hundreds of millimeters, over 5 orders of magnitude larger. Numerically modeling hundreds of millions of individual cells at a sub-micron resolution over this timescale would require the entirety of multiple leadership-class supercomputers over the course of several weeks, if not months. Therefore, there still exists a disparity between these two ranges that needs to be addressed to make simulations of CTC transport with the presence of neighboring RBCs tractable.

    We also address one of the pillars of the inherent variability in cell transport: the fate of a single CTC can exhibit significant variations due to its interactions with neighboring RBCs in the context of an \textit{in vivo} experiment,. A single simulation of a CTC may not encompass the range of outcomes, necessitating the consideration of many simulations with different RBC distributions. Multiplying the number of simulations required to capture this variability by the computational workload of a single simulation results in a computationally intractable workload, making it essential to optimize the number of simulations required for proper results. The number of potential cell configurations is vast, which makes it essential to identify representative configurations that encompass the full range of possible outcomes while optimizing computational feasibility.

    This dissertation explores the influence of several hemodynamic and geometric parameters, microvasculature interactions, and the impact of RBCs on CTC movement, including the presence of RBC aggregation, RBC volume fraction, microvessel size, and shear rate. Furthermore, it discusses the enhancement of adaptive physics refinement methods to model cellular transport phenomena and highlights the capabilities of fluid-structure interaction models in capturing the dynamics of CTCs and RBCs across the system-scale. The dissertation concludes by discussing the development of a novel framework to account for the range of outcomes in CTC transport due to the variability in neighboring RBCs; it addresses the importance of generating representative configurations using quantitative metrics such as the Jaccard index applied to sets of sphere and RBC data sets. By integrating these advances, we further reduce the gap towards biologically accurate computational models of cancer cell transport, which holds promise for improving our understanding of cancer metastasis and developing effective strategies for cancer treatment.

  • ItemEmbargo
    Compliance Under Pressure: Strategic Bureaucratic Control and Policy Implementation in China
    (2023) Gui, Xiaoshu

    Scholars contend that in weak institutional contexts, political leaders rely on informal networks or ad hoc campaigns to push through policies that challenge powerful local interests. My dissertation challenges this conventional view by arguing that top leaders in authoritarian regimes can effectively enhance local policy compliance by improving and adapting formal institutions to different local conditions, with the support of advanced technology and strong political commitment at the top. With a special focus on the central directives of differentiated work priorities and performance evaluation standards in the context of air pollution control in China, I find that a place-based performance evaluation system enables the central government to exert flexible political control over different regions by assigning them varying work priorities that cater to their specific local conditions and changing environment. In addition to the high-powered promotion incentive, I show that disclosing performance rankings creates an additional incentive for local officials to exert effort on future performance improvement, particularly those who appear at the bottom of the rankings. Finally, I argue that aligning local officials’ interests through formal institutions does not always yield desired outcomes, especially during times of political uncertainty. Informal institutions are, however, more effective in motivating local officials to carry out policy tasks because they can provide protection and assurance to officials. I substantiate this claim by examining how patronage networks shape China's local COVID-19 responses. In sum, a combination of formal and informal institutions remains at work in ensuring policy implementation and bureaucratic compliance in authoritarian countries.

  • ItemEmbargo
    Some Advances in Statistical modeling of Brain Structural Connectomes
    (2023) Dey, Pritam

    It has become routine in neuroscience studies to measure brain networks for different individuals using neuroimaging. These networks are typically expressed as adjacency matrices, with each cell containing a summary of connectivity between a pair of brain regions. There is an emerging statistical literature describing methods for the analysis of such multi-network data in which nodes are common across networks but the edges vary.

    There has been essentially no consideration of the important problem of outlier detection in the structural connectomics literature. In particular, for certain subjects, the neuroimaging data are so poor quality that the network cannot be reliably reconstructed. For such subjects, the resulting adjacency matrix may be mostly zero or exhibit a bizarre pattern not consistent with a functioning brain. These outlying networks may serve as influential points, contaminating subsequent statistical analyses. In chapter 2, we propose a simple Outlier DetectIon for Networks (ODIN) method relying on an influence measure under a hierarchical generalized linear model for the adjacency matrices. An efficient computational algorithm is described, and ODIN is illustrated through simulations and an application to data from the UK Biobank.

    Another problem in statistical modeling of brain networks is that in the aforementioned framework based on dividing the brain into regions, the choice of regions is highly subjective and often analyses are sensitive to the way these regions are chosen. Alternative methods which do not depend on choosing such regions are computationally expensive. In chapter 3, we propose a simple but computationally efficient method based on density estimation using Ensembles of MOndrian processes(EMO). We showed our method has strong theoretical properties and is computationally fast in real world scenarios. In chapter 4, we proposed a simple way of extending the method in chapter 3 to a case where we simutaneously estimate several densities for several subjects.

  • ItemEmbargo
    The Good Old Days: The Concept of the Golden Age in Greek Political Thought
    (2023) Nathan, Charles

    Why are visions of “the good old days” such a prevalent part of political life? This dissertation investigates why the golden age—the idealized utopia of the past—has been employed so frequently by political thinkers and movements throughout history. By examining how ancient political thinkers utilized the concept of the golden age, I identify the unique power and attraction of the idealized past. First, I analyze Hesiod’s myth of the metals to define the golden age as a political concept and identify its core elements. Next, I turn to Plato, who demonstrates how a political thinker can use the golden age to open possibilities for political theorizing and praxis. Lastly, I examine the comedies of Aristophanes, which illustrate how golden age thinking inevitably surfaces in actual politics as a polity endures over time. In the end, I argue that the golden age has the unique ability to blur the line between idealism and realism by placing a normative ideal into the past, raising important questions about the role of fantasy, idealism, and the power of the past in politics today.

  • ItemEmbargo
    Information Intermediation in Opaque Markets: Evidence from Equity Crowdfunding Analyst Reports
    (2023) Burke, Gregory

    Equity crowdfunding (ECF) is a growing market in the U.S., where firms issue unregistered securities to the public over the Internet. This market is characterized as informationally opaque and dominated by unsophisticated retail investors. I provide the first evidence with respect to information intermediation in this market by examining the role of KingsCrowd (KC), the sole ECF information intermediary that provides analyst reports for a selected subset of Regulation Crowdfunding offerings. I find a 136% (298%) increase in average weekly web traffic to the KC report page (platform click-throughs) the week a report is released, consistent with investors acquiring reports and considering an investment on the offering platform. I find a one unit increase in the report recommendation favorableness is associated with a 17-19% increase in average daily investment pledges. Further, this effect is more heavily concentrated in the first three weeks following the release of a report, whereby a one unit increase in report favorableness is associated with a 22-27% increase in average daily investment pledges, and less thereafter. In terms of dollar magnitude, a one unit increase in report favorableness is associated with a $25,076 increase in investment pledges over the nine weeks following the report release. The collective evidence suggests information intermediation by KC in the form of recommendations are utilized by investors when making investment decisions in the opaque ECF market.

  • ItemOpen Access
    Essays on Concentrated Animal Feeding Operations
    (2023) Ma, Yu

    This dissertation is an empirical study of the livestock industry and its environmental impacts on residents. Concentrated animal feeding operations, abbreviated as CAFOs, are livestock production facilities where large numbers of animals are raised in confined spaces. Although the hog and poultry industries provide jobs and economic benefits, they also produce significant air pollution, contaminate waterways, and affect people's quality of life. North Carolina (NC) is currently the third largest hog producing state in the nation and also hosts a high concentration of poultry farms. Most of the animal farms are located in the eastern area of the state, which is also the area where many low-income people and people of color (POC) reside.

    Because of environmental pollution produced by CAFOs, local real estate markets could be affected. Chapter 4 examines how having CAFOs nearby could affect housing price. In this co-authored paper, we utilize housing transaction data from CoreLogic and study the impacts of CAFOs on housing price. We consider co-location of hogs and poultry and separately examine the impacts for houses on private wells and community water systems as water contamination is channeled as an important exposure route. Results show significant housing price reductions for nearby housing properties. The costs increase disproportionately for really large CAFO exposure and are even larger for the houses with private wells. We find that being exposed to the highest levels of exposure to hogs could cause housing price decreases ranging from 13% to 50% for houses with private wells, while only a 13% to 27% price decrease for community-water-dependent houses, depending on the distance between CAFOs and the residential property.

    In NC, most of the farms are located in the eastern region, where many communities of color and low-income populations live, and such high concentration raises environmental justice concerns. Chapter 5 explores the relationship between race and income and exposure to CAFOs. In this co-authored paper, we collect information on both hog and poultry farms, use novel micro-data from InfoUSA, and investigate how exposure varies by both income and race. We find POC are more likely to be exposed to both hogs and poultry. Results show strong evidence of high exposure for low-income Hispanic households, compared to white households. Higher income helps reduce the exposure gap for Hispanics, but does not similarly help Black residents, suggesting such uneven exposure patterns are more related to race other than class.

    Climate change brings another challenge to CAFOs. During the past two hurricane events in NC, Hurricane Matthew (2016) and Hurricane Florence (2018), CAFOs caused large damages to local communities and contaminated neighborhood drinking water sources. In my job market paper, I first use individual demographic data from InfoUSA to examine household's out-migration behaviors after floods. Results suggest floods make people move out, especially for those with CAFOs around or with private wells. Besides out-migration behaviors, this study also examines how household race and income composition change after floods. Results show more lower-income and POC households move into flooded areas, especially places near animal farms, after floods. Such migration patterns highlight equity concerns under climate change and in the future hurricane events.