UC Merced

Catalysis lies at the heart of modern chemical synthesis, underpinning the efficient production of pharmaceuticals, agrochemicals, and advanced materials. As the global scientific community increasingly prioritizes sustainability, there is a growing necessity to move away from catalytic systems that rely on precious metals such as iridium, rhodium, and palladium, elements that are not only scarce and expensive but also associated with significant environmental and supply-chain concerns. Homogeneous and organocatalytic approaches offer a compelling alternative for reducing the ecological footprint of chemical processes. In this context, Earth-Abundant transition metals and main-group elements have emerged as powerful platforms for sustainable catalysis. Despite their enormous potential, their widespread adoption remains hampered by several unresolved challenges, including susceptibility to side reactions, air- and moisture-sensitivity requirements, high catalyst loadings, and harsh reaction conditions. Moreover, the mechanistic landscapes governing these catalytic cycles remain considerably underexplored compared to their precious metal counterparts, limiting the rational design of improved systems. Against this backdrop, this dissertation explores sustainable homogeneous and organocatalytic strategies employing Earth-Abundant elements (manganese, iron and silicon) for three synthetically valuable transformations: borylation, hydroboration, and hydrodefluorination. Borylation and hydroboration target boronate esters, high-value synthetic intermediates widely utilized in pharmaceutical synthesis, agrochemistry, and cross-coupling reactions. Hydrodefluorination, on the other hand, enables the selective synthesis of fluorinated arenes, privileged structural motifs in drug discovery and medicinal chemistry. Beyond their synthetic value, hydrodefluorination reactions carry broader environmental significance, as selective C–F bond cleavage offers a promising strategy toward the remediation of per- and polyfluoroalkyl substances (PFAS). The first chapter of this dissertation presents the first manganese-catalyzed chemoselective C(sp)–H borylation of terminal alkynes, employing [Mn(SiNSi)Cl2] (SiNSi = 2,6-[EtNSi(NtBu)2CPh]2C5H3N), as the precatalyst. Mechanistic studies reveal that the choice of activator critically governs the reaction pathway, dictating chemoselectivity between C–H activation and alkyne insertion. Uniquely, HBPin serves a dual role as both the borylating reagent and an in situ activator, directing the catalyst toward the C–H activation pathway and the selective formation of alkynylboronate esters. This chapter thus underscores the pivotal importance of mechanistic understanding in rationalizing and controlling catalytic outcomes. The second chapter describes the first use of silicon as an organocatalyst for C–B bond formation, employing commercially available triethoxysilane for the chemoselective single and sequential hydroboration of terminal alkynes. This transformation affords alkyl gem-diboronate esters, attractive precursors in drugs containing 3D non-flat motifs. Notably, while silicon-catalyzed hydrofunctionalization reactions reported in literature require the silicon center to be activated as a strong Lewis acid, this work demonstrates that a non-Lewis acidic silicon source is fully competent for catalysis, representing a significant conceptual advance in main-group organocatalysis. Mechanistic insights implicate non-covalent interactions between the multiple bonds of the substrate and the silicon catalyst as key contributors to reactivity. This mechanistic insight directly guided catalyst improvement, wherein a fluorinated analogue of triethoxysilane was identified as a superior catalyst, delivering nearly twice the efficiency of the parent system. The third chapter extends the application of triethoxysilane as a universal catalyst for the hydroboration of a broad range of unsaturated substrates, including alkenes, nitriles, ketones, and esters. Unlike the alkyne hydroboration described in the preceding chapter, mechanistic studies reveal that the electronic and steric properties of heteroatom-containing aromatic substrates exert distinct effects on catalyst efficiency, highlighting substrate-dependent mechanistic divergence. Notably, a fundamentally different mechanistic pathway is operative for the hydroboration of nitriles, involving a catalytically active species distinct from triethoxysilane itself. The final chapter presents the first bench-stable iron catalyst, [Fe(tBubipy)(OTf)₂] (tBubipy = 4,4'-di-tert-butyl-2,2'-bipyridine; OTf = CF₃SO₃⁻), for the hydrodefluorination of fluoroarenes. The catalyst demonstrates remarkable efficiency, achieving a turnover number (TON) of 31.6 for 2,3,5,6-tetrafluoroanisole at room temperature. The extended catalytic lifetime is attributed to the robust five-membered chelate formed by the tBubipy ligand, which stabilizes the catalyst resting state and prevents deactivation through ligand decoordination. At elevated temperatures, the system exhibits enhanced performance, facilitating up to five sequential hydrodefluorination cycles for hexafluorobenzene, underscoring its potential for the exhaustive defluorination of highly fluorinated substrates, a capability of direct relevance to sustainable strategies for PFAS remediation.

Growth mixture models (GMMs) constitute a versatile statistical framework for capturing population heterogeneity in longitudinal data by identifying latent classes of growth trajectories. Since their introduction in 1999, the models have been extended and disseminated through sustained methodological advancements, yet uncharted gaps remain in their implementation. This dissertation develops and evaluates advanced Bayesian methods to address two such gaps, predictor selection and missing data handling, across two studies. In Study 1, I focused on predictor selection in conditional GMMs. Beyond the usefulness of GMMs in describing heterogeneous growth patterns, predicting these patterns offers an enhanced understanding of substantive phenomena. However, the question of how best to select important predictors has remained unaddressed in the GMM literature. As a principled approach to this issue, I proposed Bayesian variable selection via shrinkage priors and evaluated seven priors (ridge, lasso, hyperlasso, elastic net, horseshoe, regularized horseshoe, and spike-and-slab) through a Monte Carlo simulation. Results showed that the horseshoe, regularized horseshoe, and spike-and-slab offered the best balance between detecting true nonzero predictors, shrinking small effects, and controlling false selections. In addition, the regularized horseshoe, spike-and-slab, and horseshoe achieved superior predictive accuracy, though the regularized horseshoe and spike-and-slab came at the cost of lower convergence rates. The ridge prior showed limited effectiveness in selecting important predictors and high false selection rates. Practical recommendations are provided for choosing shrinkage priors, with caveats and other implementation considerations. In Study 2, I addressed missing data in GMMs under attrition scenarios common in longitudinal research. Existing missing data techniques either discard incomplete cases, require the number of latent classes to be pre-specified as part of the missing data treatment, or ignore the population heterogeneity. To overcome these limitations, I developed a Bayesian nonparametric multiple imputation approach using the Chinese restaurant process (MI-CRP). This new approach retains existing values, imputes missing values, and accounts for population heterogeneity without requiring the number of classes to be predetermined. A Monte Carlo simulation compared MI-CRP with four existing methods, including full information maximum likelihood, Bayesian estimation, multiple imputation via joint modeling, and multiple imputation via fully conditional specification. Results showed that MI-CRP adequately recovered the growth factor means and avoided extreme covariance bias. MI-CRP also achieved higher convergence and valid replication rates at low class separation with smaller sample sizes, and higher classification accuracy with less variability under challenging conditions. Overall, the advanced Bayesian methods developed and evaluated in this dissertation fill unresolved methodological gaps in GMMs and equip applied and methodological researchers with principled tools for implementing these models.

Accurate characterization of fine particulate matter (PM2.5) exposure remains a critical challenge in air quality science, particularly in regions with complex emission sources and limited regulatory monitoring coverage. Traditional regulatory monitoring networks may fail to capture local variability due to their limited number and distribution of monitors because of their higher cost. This dissertation addresses these limitations by integrating information from community-based low-cost air quality monitoring, advanced spatiotemporal statistical modeling, considering socioeconomic status demographics, and global fusion model evaluation to improve understanding of PM2.5 exposure patterns in California’s San Joaquin Valley. Chapter 2 evaluates a dense network of low-cost air quality sensors deployed across Fresno County by comparing observations from community monitors with regulatory monitors. Using spatial interpolation techniques, this analysis identifies periods of elevated PM2.5 concentrations that were not detected by nearby regulatory monitors, demonstrating that low-cost sensors can capture fine-scale variability in air pollution. Chapter 3 applies Bayesian hierarchical spatiotemporal models to examine the relationship between PM2.5 concentrations and socioeconomic and demographic characteristics across the San Joaquin Valley from 2021 to 2023. Results indicate that higher PM2.5 levels are disproportionately concentrated in census tracts with greater socioeconomic disadvantage and higher proportions of residents of color. Comparisons with communities designated under California’s Assembly Bill 617 reveal substantial overlap with identified pollution hotspots, while also identifying additional high-burden areas outside current boundaries used to define high-risk communities. These findings underscore the importance of high-resolution monitoring and modeling approaches for informing equitable air quality management and environmental justice policy. Chapter 4 investigates how low-cost sensor data compares to exposure modeling frameworks, influences estimates of PM2.5 concentrations, and the resulting Air Quality Index (AQI) across California. By comparing model outputs with community monitoring data, this study demonstrates that low-cost sensors improve the spatial representation of pollution. These improvements have important implications for public health communication, as more accurate AQI estimates can better inform individual and community-level decisions regarding air quality exposure. Together, these three studies demonstrate that community-based monitoring networks can serve as both a scientific tool and resource for impacted residents, enhancing exposure assessment, improving model performance, and describing environmental inequities.

Metabolic dysfunction-associated steatotic liver disease (MASLD) affects more than 30% of the global population and primarily afflicts individuals with metabolic syndrome (MetS) due to their overlapping risk factors. MetS is defined as the presence of three or more of the following risk factors: (1) abdominal obesity, (2) atherogenic dyslipidemia, (3) hyperglycemia with or without insulin resistance, (4) hypertension, (5) proinflammatory state, and (6) prothrombotic state. Currently, there is no FDA-approved treatment for MASLD; therefore, the first line of non-pharmaceutical defense is caloric restriction (CR). However, failure to maintain a caloric restriction diet among obese individuals is extremely high, resulting in body mass regain (BMr). BMr following CR has been shown to exacerbate impaired lipid metabolism by inducing excess hepatic fat accumulation and causing an imbalance between lipid availability and its utilization. This impairment in hepatic lipid metabolism may induce MASLD. Subsequently, the accumulation of excess lipids in the liver promotes the production of reactive oxygen species (ROS) through a disequilibrium in redox signaling. This shift catalyzes the transition between MASLD to a progressive onset of metabolic dysfunction-associated steatohepatitis (MASH). The benefits of CR on MetS and MASLD are well established; however, the literature lacks focused studies on the pathways associated with lipid metabolism and redox signaling after BMr. In this dissertation, the Otsuka Long Evans Tokushima Fatty (OLETF) rat was used as a model of MetS due to its similarities to the human conditions. This model was used to investigate the effects of partial and complete BM regain following caloric restriction on hepatic lipid metabolism, redox balance, and thyroid hormone signaling during MetS. We demonstrated that: (1) partial BMr induced hepatic fat accumulation via elevated triglycerides with saturated fatty acids, downregulated fatty acid oxidation proteins, and reversed the benefits induced by CR suggesting that even partial regain of BM (PR) can promote MASLD, (2) PR impairs TH signaling, uncoupling TH signaling and its regulation of hepatic lipid metabolism via a reduction in TH receptor beta (THRβ), associated with changes in the hepatic lipidome (ACars, CLs, LPCs, PCs, and PIs), all of which induce the progression of MASLD to MASH by increasing hepatic injury, and (3) complete BM regain during early Mets increased 4-hydroxenenol (4HNE) despite elevated catalase (CAT) activity demonstrating that complete regain induced oxidative injury associated with impaired hepatic lipid metabolism. Collectively, this dissertation provides insight into the detriments of BMr (complete or partial) following CR on hepatic lipid metabolism, oxidative injury, and TH signaling during MetS.

Rational design of molecules for chemical transformations and optoelectronic devices requires an understanding of excited state electronic transitions and the pathways connecting stable conformers on potential energy surfaces (PES). In this work, we demonstrate how computational methods like PES scans can be used to elucidate and design new reagents for chemical transformations. We also show how to use photophysical characterization of compounds, like charge transfer (CT), oscillator strengths, and absorption profiles can be used to design and chemically tune optical properties. We then show how Δ-SCF methods can be used to model CT states qualitatively using difference densities and quantitatively with CT distance metrics against methods like TD-DFT. Lastly, a new method for transition state (TS) optimization is developed and evaluated on complex topologies like the Müller-Brown and Schlegel surfaces.

Online harassment, hate speech, and misinformation pose persistent challenges for digital platforms. As centralized moderation faces well-documented limits in scale and accuracy, platforms increasingly turn to community-based systems that mobilize ordinary users to identify and correct harmful content. Yet what makes such systems effective remains poorly understood. This dissertation examines community-based content moderation across three levels of analysis: platform, individual, and collective. It draws on archival data from X’s Community Notes (1.4 million notes posted between 2021 and 2025) and a controlled experiment on EatSnap.Love, a simulated social media platform (N = 137).Chapter 3 investigates how platform design parameters shape bystander intervention. Drawing on bystander intervention theory, it theorizes how disapproval channel availability and reporting effort jointly structure participation, and tests this framework using a 2 × 3 × 2 × 2 between-subjects experimental design. Contrary to predictions from offline bystander theory, higher procedural effort increased rather than reduced reporting, suggesting that effort functions as a commitment filter rather than a uniform deterrent. Disapproval channel availability broadened engagement, and rules reminders moderated the selective effect of effort. Chapter 4 examines personality differences using the Big Five framework. Integrating LIWC-22 linguistic analyses of 2,197 Community Notes contributors with self-reported personality data from the experimental sample, the chapter identifies a systematic engagement–quality gap: Openness predicted who actively intervened, but Conscientiousness predicted whose contributions achieved cross-partisan helpful status. Extraversion and Neuroticism were negatively associated with quality, indicating that the traits motivating intervention differ from those producing effective corrections. Chapter 5 analyzes the linguistic properties of corrections through sentiment analysis, LDA topic modelling, and fractional logit regression on 8,941 tweet-note pairs. Notes adopted a verification-oriented vocabulary and were consistently more negative than the tweets they addressed. Crucially, the degree of shared agreement among note authors that the original tweet was misleading dominated all linguistic predictors of helpfulness, suggesting that effective collective correction depends on shared evaluative consensus rather than stylistic features of the correction itself. Together, these findings advance a design-contingent account of community-based moderation in which platform design, individual differences, and shared evaluative judgment operate as selective conditions determining who recognizes harm, who acts effectively, and when collective correction succeeds.

Understanding the relationship between enzyme structure and function is a central goal in structural biology, with important implications for enzyme design, drug discovery, and biotechnology. In this dissertation, we investigate the role of structural flexibility and dynamics in enzyme catalysis using engineered enzymes as model systems and advanced X-ray crystallography techniques.Our results demonstrate that optimal catalytic function arises from a balance between rigidity and flexibility. In relatively simple, single-step reactions, increased rigidity can enhance catalysis by preorganizing the active site to closely match the transition state geometry, thereby reducing the need for conformational rearrangement. In contrast, more complex, multi-step reactions require enzymes to sample multiple conformational states to accommodate substrates, intermediates, and products throughout the catalytic cycle. In these cases, conformational flexibility becomes essential for efficient function. Through structural and comparative analyses of evolved enzyme variants, we show how directed evolution modulates the conformational landscape of enzymes, enabling expansion or reshaping of the active site and selectively enriching catalytically productive states. Additionally, time-resolved and room-temperature crystallographic approaches provide insight into non-equilibrium conformations and transient states that are not accessible through conventional cryogenic methods. Together, these findings provide a unified framework for understanding how enzymes balance rigidity and flexibility to achieve efficient catalysis, and highlight the importance of dynamics in shaping enzyme function.

Green hydrogen is widely recognized as a promising energy carrier that enables deep decarbonization and integrates variable renewable energy sources. However, its large-scale deployment remains constrained by high production costs and operational challenges associated with intermittent renewable electricity.This thesis presents an hourly techno-economic analysis of green hydrogen production in off-grid hybrid solar–wind systems directly coupled to electrolyzers, without battery storage or grid interaction. To determine the drivers of cost, the study focuses on the impact of renewable resource complementarity and electrolyzer sizing on hydrogen production cost. An hourly modeling framework (8760 h) was developed to simulate renewable electricity generation, electrolyzer operation, hydrogen production, and the Levelized Cost of Hydrogen (LCOH). Renewable generation profiles are obtained using the System Advisor Model (SAM), while a custom model evaluates electrolyzer performance under operational constraints and variable electrolyzer-to-renewable capacity ratios (R). The framework is applied across a diverse matrix including four representative U.S. locations spanning solar-dominant, wind-dominant, low-resource, and highly complementary resource conditions; five renewable energy configurations ranging from PV-only to wind-only systems with hybrid solar–wind combinations; three electrolyzer technologies (AWE, PEM, and SOEC); and electrolyzer-to-renewable sizing ratios ranging from 0.2 to 1.The results reveal that LCOH exhibits a convex relationship with R, confirming the existence of an optimal sizing point driven by the trade-off between renewable curtailment and electrolyzer utilization. It was found that the electrolyzerto-renewable sizing ratio strongly influences electrolyzer capacity factor, curtailed energy, hydrogen production, and overall system economics. Among the evaluated locations, Amarillo–Texas achieves the lowest LCOH (5.61 USD/kg) for a PEM electrolyzer coupled to a hybrid PV45–WD45 configuration, highlighting the benefits of solar–wind complementarity. While hybrid systems generally improve electrolyzer utilization and reduce LCOH in locations with balanced solar-wind resources, while PV-only systems remain optimal in solar-dominant regions such as Phoenix–Arizona. Among the technologies studied, PEM consistently provides the lowest LCOH among the evaluated electrolyzer technologies.Sensitivity analyses indicate that renewable energy and electrolyzer capital costs are the dominant drivers of LCOH reduction. Under cumulative cost reduction scenarios, substantial decreases in both renewable and electrolyzer CAPEX are required to approach low-cost hydrogen targets, while renewable resource quality remains a fundamental determinant of system economics. Overall, the results demonstrate that the economic viability of green hydrogen depends not only on cost reductions, but also on optimal system design, electrolyzer sizing, and renewable resource complementarity. This work highlights the importance of integrated hourly techno-economic modeling for identifying cost-effective renewable hydrogen production under variable renewable energy conditions. 

Individuals often face decisions in which the consequences of their actions for others are uncertain, and in many cases this uncertainty is actively maintained. A growing literature documents that individuals avoid information about the social consequences of their choices, a behavior commonly referred to as strategic ignorance. This dissertation argues that such ignorance is not merely an individual choice, but a socially embedded phenomenon shaped by interactions with others, internalized norms, and beliefs about the behavior of others. Using a series of experiments based on moral wiggle room environments, I examine how different features of the social context influence both the willingness to acquire information and the behavioral consequences of remaining ignorant. The results reveal a striking asymmetry: while many individuals avoid acquiring information themselves, others are readily willing to provide it. When information is imposed, a substantial share of decision-makers revise their choices in a more prosocial direction, demonstrating that ignorance—and its behavioral consequences—can be constrained by others. At the same time, prompting individuals to consider what constitutes appropriate behavior prior to making a decision increases prosocial actions, consistent with the activation of latent norms, whereas eliciting such beliefs after decisions instead reflects post-hoc justification. In addition, individuals respond to information about the behavior of others by aligning both their information acquisition and allocation choices with perceived norms. Taken together, these findings provide a unified account of moral decision-making in which behavior is shaped by a layered system of enforcement, ranging from external interventions to fully internalized regulation. The results highlight that the key determinant of behavior is not only what individuals know, but how the social environment shapes what they choose not to know.

Modern computing systems increasingly rely on heterogeneous architectures that integrate general-purpose processors with specialized accelerators such as GPUs and deep learning engines. To meet the performance and energy demands of modern workloads, these systems depend heavily on optimization techniques including speculative execution, transfer learning, hardware acceleration, and dynamic power management. While these mechanisms substantially improve efficiency, they also introduce new forms of shared state, resource contention, and externally observable behavior, thereby expanding the attack surface. As a result, security vulnerabilities in modern platforms often emerge not from a single component, but from interactions across multiple layers of the computing stack. This dissertation investigates such cross-layer vulnerabilities in heterogeneous computing systems and develops new insights into how performance-oriented design choices can unintentionally expose sensitive information. It presents three case studies spanning machine learning systems, edge GPU platforms, and CPU microarchitecture. First, this dissertation presents Decepticon, a model extraction attack on large-scale transfer-learned models. Decepticon exploits execution fingerprints inherited from shared pre-trained models to identify the underlying model used by a black-box victim and significantly reduce the effort required to reconstruct a high-fidelity clone. The results demonstrate that even large transformer-based models remain vulnerable when transfer learning and GPU execution characteristics are jointly exploited. Second, this dissertation investigates the power and frequency behavior of edge GPU platforms and their security implications. Through detailed characterization of instruction-level and application-level telemetry on commercial NVIDIA Jetson devices, this work demonstrates that GPU and deep learning accelerator behavior can be leveraged to construct covert channels. It further evaluates mitigation strategies based on telemetry access restriction, resolution reduction, and perturbation, highlighting the trade-offs between observability and security. Third, this dissertation introduces SCPC (Securing Cross-Process Collision-Based Transient Attacks), a low-overhead defense against cross-process collision-based transient execution attacks in modern CPUs. SCPC protects speculative execution structures by selectively virtualizing vulnerable predictor state while preserving much of the performance benefit of shared hardware resources. Evaluation shows that the mechanism significantly improves security with minimal performance overhead and can be generalized to other speculative structures. Together, these contributions show that information leakage in modern heterogeneous systems is often a byproduct of the same mechanisms that enable high performance. This dissertation argues that securing future computing platforms requires a cross-layer perspective in which machine learning frameworks, accelerator behavior, and processor microarchitecture are analyzed jointly rather than in isolation. By exposing these vulnerabilities and exploring corresponding defenses, this work advances the design of more secure and robust heterogeneous computing systems.