Physical Sciences and Engineering

Hydrogen-Powered Aircraft Design Optimization
Eytan Adler, Aerospace Engineering

Aviation’s climate-warming emissions are among the hardest to eliminate. We found that from an aircraft design perspective, liquid hydrogen-fueled aircraft are a promising solution. However, they introduce new challenges: 1) large hydrogen storage volumes may incentivize new aircraft configurations, 2) liquid hydrogen storage temperatures demand careful tank insulation and management, and 3) hydrogen fuel cells have major thermal management challenges. Using state-of-the-art numerical and optimization methods, we created software tools to characterize and alleviate these problems. We implemented aircraft configuration models and measured the hydrogen storage penalty for each; we found that long-range conventional configurations are more viable than previously thought. We developed a new liquid hydrogen tank model to characterize and optimize the tank insulation and pressure, collaborating with a leading hydrogen aircraft company to validate it. Finally, we are building tools that use gradient-based optimization to minimize the weight and drag of fuel cell thermal management systems.

Closing the Loop on Safety-Critical Perception, Planning, and Control of Robotic and Cyber-Physical Systems
Devansh Agrawal, Aerospace Engineering

As robotic systems become more capable, they must operate in increasingly complex and dynamic environments with limited information. Rigorous mathematical tools are needed to optimize information collected, and to analyze the impact uncertainties have on the closed-loop system. This thesis takes a bottom-up, control-theoretic approach to constructing novel algorithms for the perception-planning-control modules of a system. First, we study safety critical controllers, identifying the assumptions which guarantee satisfaction of safety specifications. These assumptions however rarely hold in unstructured environments. We therefore propose a novel and simple trajectory filtering algorithm to make it suitable in such scenarios. Next, we propose new certifiable visual odometry and mapping algorithms, derive rigorous error bounds, and show that these bounds are compatible with the proposed planners and controllers. These ideas are demonstrated experimentally on aerial and mobile robots. Finally, we propose the notion of perceivability, which maximizes information gain by casting it as an optimal control problem.

Integrative Approaches for Exploration of Non-Coding RNA Structural Dynamics: Implications for Function and Beyond
Grace Arhin, Biophysics

The study of non-coding RNAs (ncRNAs) has become increasingly important due to the recognition that these molecules play crucial roles in various cellular processes. Central to understanding how ncRNAs function is our ability to identify dynamic transitions within the system. These dynamics provides a comprehensive understanding of their functions, interactions, and behavior. Within the vast array of ncRNAs, micro-RNAs (miRNAs) emerge as crucial regulators of gene regulation. Dysregulation in miRNA maturation can be drivers of multiple human diseases such as cancers, neurodegenerative disorders, and cardiovascular diseases. Structured around three (3) specific aims, my dissertation endeavors to study the effects of structural dynamics on miRNA maturation, identify drug-like molecules that bind and stabilize a regulatory element in a human miRNA cluster and finally develop a tool for prediction of chemical shifts from 3-D RNA models for dynamic studies using nuclear magnetic resonance (NMR) spectroscopy, chemical probing, molecular modeling, and biochemical assays.

Integrating Data Collection and Modeling to Provide Next-Generation Utility for Environmental Infectious Disease Surveillance
Peter Arts, Environmental Engineering

In the early months of 2020, public health responses to COVID-19 were hampered by an inability to quickly establish and mitigate routes of disease transmission, and to monitor community level disease burden. As the pandemic progressed, the application and improvement of environmental surveillance (i.e., measuring infectious agents in the environment) aided a more informed and effective response to outbreaks. Despite rapid advances in environmental surveillance, interpreting the resulting data is limited by several knowledge gaps related to interactions between humans, infectious agents, and environmental media. This dissertation focuses on improving the utility of environmental infectious disease surveillance by generating and contextualizing individual shedding data, and implementing these results in models of environmental transfer and persistence. Chapters in the dissertation span from small scale environmental surveillance in of air and fomite samples in classrooms to large scale surveillance of municipal wastewater.

Applications of Low-Cost Robots in Post-Stroke Upper Extremity Rehabilitation
Thomas Augenstein, Robotics

Reduced upper extremity strength and coordination are commonplace following a stroke, but survivors often lack sufficient healthcare coverage to fully address their deficits. Rehabilitation robots could bridge this gap, but most devices use large motors, making them bulky and expensive and therefore ill-suited for standard care. To address this shortcoming, we developed a new class of rehabilitation devices, passive and semi-passive robots, that use passive (e.g., cables) and semi-passive (e.g., brakes) actuators instead of motors. The objective of my dissertation is to assess the clinical potential of these devices. Specifically, I will examine two custom-designed devices that address shoulder/elbow function and hand function, respectively. I will examine the clinical potential of each device in short-term interventions and their combined potential in a long-term intervention case study. The results will build knowledge for a promising new generation of rehabilitation technology.

Interactions Between Black Holes and their Environments in X-rays, From Stellar to Supermassive
Mayura Balakrishnan, Astronomy and Astrophysics

Black holes are only observable in the electromagnetic spectrum because they interact with their environments, primarily through the material they draw in, which commonly radiates in the x-ray. There are two main components in this thesis: the study of swiftly changing stellar-mass black holes and the investigation of the faint, consistent emission from our central supermassive black hole, Sgr A*. I use x-ray observations from Swift, Chandra, NICER, and NuSTAR to study stellar-mass black holes GRO J1655-40, GRS 1915+105, and Swift J1728. Moving to the crowded Galactic Center, I use Chandra observations to study the spectra and images of Sgr A* and its surroundings to learn about how the supermassive black hole is interacting with and influencing its environment.

Two-Electron Chemistry at a Molecular Tricopper Cluster: Informing Mechanisms in CO2 and NOx Reduction Catalysis via Robust Structure/Function Relationships
Andrew Beamer, Chemical Engineering

Anthropogenic climate change is a leading global problem that demands mitigation in the next decade. Even though extensive empirical research has been conducted to discover catalysts capable of forming value-added products from simple oxygenated nitrogen and carbon feedstocks, mechanistic understanding—and therefore rational catalyst design—is still limited. Herein, we describe the synthesis of atomically-precise and structurally modular trimetallic copper clusters that serve as high fidelity models of copper surfaces during electrochemical transformations of carbon dioxide (CO2) and nitrogen oxides (NOx). This work addresses knowledge gaps inhibiting wider-spread implementation of the base-metal copper for these critical small molecule transformations. Combined, these fundamental studies leverage the unique ability of multimetallic constructs—multipoint small-molecule binding sites, delocalized electronic structure, and cooperative substrate activation—to recapitulate surface-relevant properties and catalysis in readily studied, structurally precise cluster compounds.

Mechanism-driven Development of Group 10 Catalyzed Reactions for the Synthesis of Bioactive Scaffolds
Alexander Bunnell, Chemical Engineering

Transition-metal catalyzed reactions are widely employed in the development and production of biologically active molecules relevant to pharmaceuticals, agrochemicals, and medical imaging agents. The utility of these reactions is due, in large part, to the ability to study intermediates that are catalytically relevant to the desired transformation, thus providing detailed mechanistic information that informs reaction optimization. This dissertation contains three projects that utilize this mechanism-driven approach to developing group 10-catalyzed reactions for the construction of biologically relevant scaffolds. The first involves the palladium-catalyzed decarbonylative fluoroalkylation of (hetero)aryl boronate esters to produce difluorobenzylated products. The second explores nickel-catalyzed cyanation to form 13CN-labeled products. This transformation is also applied to the synthesis of 11CN-containing positron emission tomography (PET) tracers. The final project is a collaboration with Merck that explores the palladium-catalyzed C-H functionalization of bicyclo[1.1.1]pentane and related polycyclic C(sp3) bioisosteres.

Design and Synthesis of Porous Polymers for Energy Applications
Cassidy Carey, Macromolecular Science and Engineering

Access to clean water and energy is one of the greatest challenges facing humanity. The water and energy sectors are intertwined in such a way that if one is under stress the other is also affected. This phenomenon, also known as the water-energy nexus, drives the need to develop more efficient purification and energy storage materials to meet increasing demands. Recently, classes of porous polymeric materials have emerged due to their exceptionally high surface areas and unique ability to selectively adsorb and store a target chemical species in both the liquid and gas phases. This work seeks to develop structure-property relationships by synthesizing a variety of porous polymers and evaluating their utility within applications ranging from water purification, battery technology, and gas storage. Completion of this work will result in the development of next-generation materials to offset global demands.

Realizing AI for Data Management: Towards Machine-Actionable Data Stores Guided by FAIR Data Principles
Tianji Cong, Computer Science and Engineering

The resurgence of artificial intelligence (AI) over the last decade has sparked exciting applications in various fields, including data management, which is itself foundational to AI applications. With the exponential growth in data volume, complexity, and creation speed outpacing human intervention capacities, the imperative arises to ensure datasets are readily discoverable and usable by users. My research contributes to the design and implementation of a semantic layer over data stores, guided by the FAIR data principles—aiming to enhance data Findability, Accessibility, Interoperability, and Reusability. While initially proposed to advance open science, we posit that the FAIR principles can greatly benefit any data store or digital repository. However, despite the valuable guidelines provided by these principles, it remains challenging to achieve FAIR data for human users, let alone for machine systems. My dissertation details how we exploit AI techniques from representation learning to probabilistic graphical models to build FAIR data stores.

Towards High Performance, Power Efficient Brain-Machine Interfaces
Joseph Costello, Electrical and Computer Engineering

Brain-machine interfaces (BMIs) are a promising solution for restoring mobility and communication to people who suffer from sensorimotor impairments, including spinal cord injury, stroke, and neurodegenerative diseases. Intracortical BMIs consist of an electrode array implanted in the brain, an algorithm to decode the user’s intentions, and an output device. While BMIs have successfully controlled computer cursors and robotic arms, they are still limited in the accuracy of the decoding algorithm and consume too much power for an implantable system. In this work, I show how recurrent neural network decoders can accurately predict movement from intracortical neural activity in real-time, outperforming the previous state of the art, and I demonstrate methods for reducing BMI power consumption. These include an efficient communication scheme for wirelessly transmitting neural data, a technique for compressing decoders, and a method for reducing the number of active electrodes, to reduce power by up to a factor of 10.

Adaptive Transcriptomics Enabled by Artificial Intelligence and Laboratory Automation
Benjamin David, Biomedical Engineering

Bacteria live in complex environments where they must respond to many signals and stressors simultaneously by altering their patterns of gene expression, i.e. their transcriptional profile. However, transcriptional profiling studies are usually limited to studying the effect of a single stimulus at a time, such as adding a drug to a growth medium. There is a need for a method that can navigate the high dimensional complexity of gene regulation. This project develops an experimental platform for measuring and predicting transcriptional responses in combinations of environmental conditions. My approach, called adaptive transcriptomics, combines artificial intelligence (AI) and laboratory automation to select only the most informative conditions to test. Adaptive transcriptomics searches for experiments without human intervention: AI techniques are used to select experiments that maximize information gained. This automated, data-driven approach to transcriptomic profiling leverages statistical models and laboratory robotics to efficiently learn transcriptional responses.

Engineering Open Materials for On-Demand Colloidal Matter
Tobias Dwyer, Chemical Engineering

Open colloidal nanomaterials can be self-assembled from simpler, colloidal building blocks that arrange into structurally complex patterns. Their porous structure makes these materials particularly important for water filtration, catalysis, and photonics. Unfortunately, they often serve only a single function, limiting their usefulness. The next grand challenge in engineering open nanomaterials is creating structures with multiple functions. One approach to achieve multifunctionality is to design building blocks that can rearrange in situ to support different functions on demand. Despite the interaction complexity between colloids, clever model development and high performance, open source simulation tools have proven essential in understanding colloidal self-assembly. In this dissertation, I develop bespoke models to engineer open materials that perform differently depending on small changes in experimental parameters. In collaboration with experimentalists, I combine theory, simulations, and experimental validation to create several classes of tunable open nanomaterials whose structures can be reconfigured by following simple design rules.

Highly Charged Ion-Exchange Membranes for the Treatment of Brines Via Electrodialysis
Carolina Espinoza, Chemical Engineering

Brine generation is the Achilles heel of water desalination by reverse osmosis. Electrodialysis (ED) is a promising technology for brine treatment. However, the implementation of ED for this challenging application requires ion-exchange membranes (IEMs) that are highly permeable and highly selective when contacted by brines. To guide the development of IEMs for brine concentration, a better understanding of the interconnection between membrane structure and transport properties is required. My dissertation undertakes a systematic study on ion and water transport in IEMs to address knowledge gaps in the fundamental understanding of transport phenomena. Additionally, this work aims to better understand how the membrane properties translate to system performance by application in a bench-sale ED setup. The results of this work will help to rationally design membranes with increased performance for the treatment of brines via ED.

Exploring Uncertainty and Variation in Seismic Hazard Estimates Using a Novel Relative Amplitude Approach to Reassess Earthquake Magnitude
Sydney Gable, Earth and Environmental Sciences

Probabilistic seismic hazard assessment relies significantly on our understanding of earthquake recurrence which is expressed through statistical parameters including magnitude-frequency distribution (MFD) and b-value. Unfortunately, discrepancies in reported earthquake magnitudes persist due to the diverse methods employed by various earthquake reporting agencies. This leads to uncertainty in MFD and b-value which result in inaccurate assessment of seismic hazard and obscures our understanding of seismic processes. My dissertation research focuses on improving the quality of magnitude estimates for small earthquakes and establishing uniform methods for magnitude calculation and parameterization based on the relative amplitudes of earthquake waveforms. Using the framework I have created, I apply my new magnitude calculation methods to improve estimates of MFD and b-value for induced and tectonic sequences. My research deepens our understanding of the statistical variations in hazard level and physical implications for these statistical parameters.

Decoding Quantum Materials: Novel Phase Transitions Revealed by Nonlinear Optics
Xiaoyu Guo, Physics

Many novel phase transitions are accompanied by spontaneous symmetry breaking governed by the Landau theory of phase transitions. This dissertation focuses on investigating symmetry evolutions across phase transitions into novel quantum states that are previously challenging to detect. We made them possible by developing second and third harmonic generation (S/THG) techniques which are especially sensitive to point symmetries. We investigate three unique quantum phases and transitions: ferro-rotational phase that is pictured as a head-to-tail loop arrangement of electric dipoles, whose domains and domain walls are imaged by our electric quadrupole SHG in NiTiO3; “extraordinary” phase transition that describes the bulk phase transitions with lower critical temperatures than their surface counterparts, whose presence is revealed for the first time by our interference SHG in CrSBr; and two-dimensional magnetism that is subjected to strong fluctuations, whose ordering and fluctuation are closely tracked by our gated photon counting S/THG in NiPS3 films.

Palladium-Catalyzed C-N Bond Forming Alkene Difunctionalization
Jessica Hatt, Chemical Engineering

Carbocycles and heterocycles are highly valuable synthetic building blocks, especially when they incorporate amines. Palladium-catalyzed alkene difunctionalization is a mild and versatile method for the construction of these sought after motifs. This thesis will primarily describe the development of this class of reactions for the synthesis of amine-substituted methylene cyclobutanes and protoberbine alkaloids/analogs. In chapter one, the regiodivergent 4-exocyclization/functionalization for the synthesis of methylene cyclobutanes will be explored including the reaction optimization, substrate scope, and mechanistic investigations. Chapter two will describe the development of intramolecular palladium-catalyzed carboamination to form protoberbine derivatives, which not only offer efficient access to biologically active heterocycles, but are also among the first examples of carboamination reactions with aliphatic amines. Additional mechanistic investigations into the synthesis of malonate-functionalized methylene cyclobutanes and methodology development for an organometallic/biocatalytic cascade process conducted at AbbVie will also be described within this thesis.

Unraveling Records of Time and Environment in Microbial Ecosystems from the Archean to Today
Cecilia Howard, Earth and Environmental Sciences

Microbial ecosystems have both shaped and been shaped by Earth’s environments since the origin of life. I investigated how environmental variations in energy, atmospheric CO2, and dissolved oxygen are recorded in three microbial ecosystems using geochemical analyses, 16S sequencing for microbial community profiling, and 3-D imaging with x-ray μCT scanning. I used μCT scanning to identify the impacts of stronger tides on Archean (3.48 Ga) microbially induced sedimentary structures. Integrating μCT scanning, field stratigraphy, and elemental mapping, I investigated how short- and long-term environmental changes during a peak in atmospheric CO2 are recorded in 50 Ma lacustrine stromatolites. Finally, I explored how variation in annual ice cover drove changes in microbial community composition and sediment chemistry in a modern anoxic sinkhole in Lake Huron. Jointly, these projects examine how changing environmental conditions impact microbial communities and their preservation.

Privacy-Preserved Sensing for Long-Term Home Activity Monitoring
Yasha Iravantchi, Computer Science and Engineering

Computing technologies have gained significant capabilities, approaching a future where devices anticipate needs and detect health issues. However, privacy concerns, particularly in sensitive areas like bathrooms, have impeded computers’ widespread adoption in crucial contexts such as in-home health monitoring. My prior work demonstrates that novel privacy-preserved sensors can effectively maintain user privacy while achieving comparable health-monitoring performance (e.g., urinary detection, fall detection) to their privacy-invasive counterparts, such as traditional microphones and cameras. Leveraging these privacy-preserved sensors, this research aims to elevate the standard of remote health monitoring where long-term adoption can only be achieved through privacy guarantees; I will work with individuals with multiple sclerosis (MS) and their clinicians to automate tracking Activities of Daily Living, an important measure for chronic conditions. The outcomes promise unprecedented datastreams for MS researchers and continuous health monitoring and forecasting for patients, highlighting the real-world impact of privacy-preserving sensing in the home.

Systematic Management of Signalized Intersections with Vehicle Trajectory Data
Zachary Jerome, Civil Engineering

Traffic congestion at signalized intersections brings roughly $22.9 billion in direct and indirect costs each year. Increasingly available vehicle trajectory data from connected vehicles opens new opportunities for systematic management of signalized intersections, but there are few suitable traffic flow models for such data. This dissertation is among the first to build a stochastic traffic flow model with vehicle trajectories as the only input, enabling more frequent traffic signal parameter optimization. A real-world test of a traffic light optimization system built using this model conducted in Birmingham, Michigan decreased delay and number of stops by 20 percent and 30 percent, respectively. In addition, this dissertation will investigate better signal management during inclement weather conditions and offer suggestions for optimizing intersection road design. Partial results from this dissertation have been published in Nature Communications and licensed by General Motors.

A Computational Study of the Influence of Cardiac Mechanics in Desmoplakin Cardiomyopathy
Javiera Jilberto Vallejos, Biomedical Engineering

Desmoplakin cardiomyopathy is a genetic disease that affects the mechanical linkage between cardiac cells. It is hypothesized that the progression of this disease, as characterized by the apparition of localized fibrosis, can be explained by the regional mechanical loads that the heart bears. To explore this, in my thesis, I build patient-specific heart models to simulate cardiac mechanics at rest and during high cardiovascular activity. Regional mechanical quantities are correlated with the presence of fibrotic tissue. This relationship between high loads and tissue remodeling is further validated by using computational models to augment the observations obtained using engineered cardiac tissues affected by this genetic mutation. This study looks to shed light on the mechanisms underlying desmoplakin cardiomyopathy progression, key knowledge to develop effective therapies, and to show the potential that integrating computational models with in-vitro and clinical data has in the study of disease mechanisms.

Structural Determination and Inhibition Mechanism of the Eukaryotic Fluoride Exporter FEX and the Bacterial Fluoride Exporter CLCF
Chia-Yu Kang, Biophysics

Fluoride (F-) is widely used to prevent tooth decay and also has antimicrobial properties. However, oral pathogens Streptococcus mutans and Candida albicans possess CLCF and FEX proteins, respectively, to extrude fluoride from the cell and avoid fluoride toxicity. The molecular architectures of CLCF and FEX differ from that of the fluoride channels (Flucs) present in beneficial microbes, making CLCF and FEX promising targets for a selective antimicrobial agent. However, no structure of an FEX has been solved, and inhibitors of CLCF and FEX proteins still need to be characterized, leaving a major gap in our understanding of fluoride export by these pathogens. For my dissertation research, I use single-particle cryo-EM to determine the structures of CLCF from S. mutans and FEX from C. albicans and to elucidate the mechanism of inhibition for oral pathogens. These structural insights will support further rational design of antimicrobial drugs for future dental treatment.

Thermodynamic and Kinetic Modeling of Crystal Growth from Unconventional Supersaturated Solutions
Joonsoo Kim, Materials Science and Engineering

Crystals that cannot readily grow in aqueous solutions perplex the understanding of materials manufacturing. Here we modeled the thermodynamics and kinetics of crystal growth for two important materials, leading to novel theories for high-quality crystal growth. First, we reveal the atomistic mechanism underlying the age-old ‘dolomite problem.’ Dolomite, CaMg(CO3)2, is a geologically-abundant mineral that does not grow in supersaturated solutions. By combining first-principles calculations with stochastic simulations, we show a counterintuitive insight that intermittent dissolution is necessary to drive dolomite growth. Periodic dissolution during growth may yield crystals with fewer defects. Second, we derive Pourbaix-inspired phase diagrams for synthesis in supercritical NH3. These new phase diagrams can guide the scalable solution-based bulk synthesis of nitrides—such as the blue LED GaN—for deployment in next-generation microelectronics.

Advancing Anaerobic Biotechnologies for Medium Chain Carboxylic Acid Recovery from Acid Whey and Food Waste
Dianna Kitt, Environmental Engineering

Globally, there is a growing need for sustainable bioprocesses capable of generating platform chemicals while simultaneously treating organic waste streams. Lactate-based chain elongation (LCE) emerges as a promising anaerobic biotechnology addressing these needs by producing medium chain carboxylic acids (MCCAs) from organic waste streams like acid whey and food waste. This dissertation introduces a novel anaerobic dynamic membrane bioreactor linked with a liquid-liquid extraction system for efficient MCCA recovery from acid whey and digested food waste. The research demonstrates that controlling waste stream composition promotes LCE, maximizing MCCA production by adding LCE microorganisms and MCCA precursors (e.g., high lactate concentration relative to lactose concentration). High LCE efficiency was achieved with an MCCA production rate of 88.3 mmoles carbon per day and a carbon conversion efficiency of 28.9 percent. This work will also provide a deeper understanding of the LCE microbiome and inform engineering decisions and commercialization of this technology.

Scalable Active Acoustic Metamaterials
Dylan Kovacevich, Mechanical Engineering

The ability of engineers to fabricate structures and devices is constrained by the physical properties of the materials available to them. In the field of acoustics, engineered materials with periodic architectures known as metamaterials can be designed to expand the range of accessible acoustic properties, opening new possibilities in sound wave manipulation. In the past couple of decades, the primary acoustic metamaterials research thrust has been in passive composites, but several obstacles limit their practical application. To overcome them, we have developed an alternative active approach, where metamaterial unit cells consisting of microphone-speaker pairs are electronically programmed for desired effective properties. We developed a fundamental model for the programming of these active acoustic metamaterials and experimentally demonstrated the precise control over the effective properties necessary to realize highly sought-after acoustic devices. Some promising areas of application include noise management, medical imaging, and underwater sensing.

Investigation of the N-N Coupling Reaction in Flavodiiron Nitric Oxide Reductases Using Synthetic Model Complexes
Michael Lengel, Chemical Engineering

Flavodiiron nitric oxide reductases (FNORs) represent a class of enzymes found in pathogens. Pathogens equipped with FNORs exhibit resistance toward nitric oxide (NO), a crucial component of the human immune defense, by reducing NO to nitrous oxide (N2O) using a non-heme diiron active site, ultimately affording microbial proliferation and pathogenesis. Understanding how these pathogens break down NO is paramount for developing new therapeutic strategies against resistant strains. The key proposed intermediate(s) (by computation) for N-N coupling of the two NO units to form N2O are iron-hyponitrite species. This reaction is potentially supported by a second coordination sphere (SCS) tyrosine group. The exact coordination chemistry of non-heme iron centers with hyponitrite remains largely unknown and underexplored. Herein, I have used biomimetic synthetic model complexes to gain insights into the coordination chemistry and reactivity of non-heme hyponitrite complexes, and the role of SCS hydrogen-bonding groups for N-N coupling in FNORs.

Phase Behavior and Dynamics of Polyelectrolyte Complexes
Huiling Li, Macromolecular Science and Engineering

Polyelectrolytes are polymers that contain ionizable groups, making them either positively charged (e.g., chitosan) or negatively charged (e.g., DNA and poly (acrylic acid)). Polyelectrolyte complexes (PECs) are usually liquid-like or gel-like materials formed by mixing two oppositely charged polyelectrolyte solutions and have a wide range of practical applications, including use in the food industry, personal products, packaging, water treatment, tissue engineering, drug delivery, and others. This dissertation focuses on understanding the phase behavior of weak PECs and the dynamics of strong PECs. Firstly, to investigate how weak PECs are formed through liquid-liquid phase separation, an experimental phase diagram is constructed, and a thermodynamic model is adapted to understand and explain the effect of pH and salt concentration on formation. Secondly, a combination of rheological measurements and Brownian dynamics simulation is applied to study the viscoelasticity of strong PECs and understand the structure formed through electrostatic interaction.

Experimental Investigation of Intragranular Deformation and Grain Growth Mechanisms in Polycrystalline Materials Using Pointed-Focused High-Energy Diffraction Microscopy
Wenxi Li, Mechanical Engineering

The ability to measure three-dimensional intragranular microstructural changes in polycrystalline materials is essential for advancing micromechanical modeling and developing the next generation of advanced materials. In this dissertation, novel in-situ experimental x-ray diffraction methods are used to investigate three advanced materials: shape memory alloys (SMAs), titanium, and transparent alumina ceramics. Firstly, an innovative analysis approach is developed, building upon established high-energy diffraction microscopy (HEDM) techniques, to study functional fatigue in SMAs. Secondly, an emerging technique, point-focused HEDM (pf-HEDM), is introduced. It features a novel algorithm for reconstructing intragranular strain and orientation evolution with 100 times better spatial resolution than existing HEDM methods. Finally, the results from in-situ pf-HEDM experiments on plastic deformation in titanium and sintering behavior in fine-grained transparent ceramics are reported. These studies offer insights into non-Schmid behavior in hexagonal close-packed metals and microstructural changes in SMAs and fine-grained ceramics, guiding the design of advanced materials.

End-to-End Transportation Network Analysis, Planning, and Operation
Zhichen Liu, Civil Engineering

For over seventy years, transportation network equilibrium models have been foundational in transportation planning, illustrating traveler competition on congested networks to reach an equilibrium state, where no traveler benefits from changing routes. Originating in the 1950s, these models faced limitations due to scarce travel data and simplified behavioral assumptions. Today, the emergence of vehicle-to-everything data collection technology offers an exciting opportunity to transform transportation network analysis.
My dissertation seeks to revolutionize transportation network equilibrium modeling by leveraging multi-source data and machine learning to enhance decision-making in connected transportation systems. Specifically, I establish a deep learning-based “end-to-end” network equilibrium framework that (i) models network-level traveler interaction from data, (ii) integrates modeling with subsequent decision-making to prescribe tailored policies, and (iii) offers theoretical guidelines for mathematical robustness. This work pioneers a paradigm shift in transportation network management, evolving from traditional rule-based models to an advanced artificial intelligence-enabled approach.

Effective and Guaranteed Computational Methods in Robust Machine Learning
Jianhao Ma, Industrial and Operations Engineering

This dissertation focuses on developing efficient computational methods for robust machine learning (ML) tasks under “atypical”, adversarial data generation processes, where the extreme behavior of the data and noise is likely to occur. To solve this class of problems, the overarching idea of this dissertation is to resort to a general nonconvex optimization framework with robust l1-loss. Our framework exhibits excellent statistical guarantees and computational efficiencies and has been successfully applied for diverse tasks like robust mean estimation, linear regression, and matrix recovery. Notably, the nonconvex models boast strong statistical guarantees due to their inherent adaptability to problem complexity, automatically finding the sparsest solutions without explicit regularization. Moreover, a simple sub-gradient method demonstrably converges efficiently even for nonconvex, nonsmooth objective functions. This combination of robustness, efficiency, and strong statistical guarantees opens new avenues for tackling challenging problems in modern ML.

Investigating Electrocatalyst and Electrocatalyst/Semiconductor Interface Design Principles for High Performing Photoelectrochemical Water Splitting Systems
Aarti Mathur, Chemical Engineering

Photoelectrochemical water splitting is a promising solution for converting solar energy directly into hydrogen fuel. High-efficiency devices consist of a tandem system with two semiconductor (SC) light absorbers coupled to electrocatalysts (EC) to perform the hydrogen and oxygen evolution half reactions. However, these systems remain poorly understood, especially for nanoparticle metal ECs, as the SC/EC interface is complex on atomic scales, dynamic under reaction conditions, and difficult to probe directly using experimental methods. My thesis aims to shed light on rational design principles for reducing efficiency losses at the SC/EC interface using a combination of electrochemical testing, atomic characterization, and analytical modeling. Additionally, we propose a method for identifying and testing highly active and stable metal alloys for OER that reduce dependence on precious metals such as iridium while maintaining kinetic activity.

Quantifying Electrochemical Charge and Discharge in Individual Battery Particles
Jinhong Min, Materials Science and Engineering

This dissertation challenges the prevailing Newman model in lithium-ion battery research, presenting groundbreaking insights into the electrochemical behavior of cathode materials. Using a novel multi-electrode array for a battery, originally used in neuroscience, the study enables individual cycling of cathode particles, specifically focusing on Li(Ni0.5Mn0.3Co0.2)O2 (NMC) used in electric vehicles. Contrary to the Newman model, findings reveal a linear relationship between particle size and exchange current density, j0, with larger particles exhibiting higher j0. Cross-sectional analysis using plasma-focused ion beams uncovers particle cracking, leading to a new hypothesis that electrolyte penetration through cracks enhances reaction surface area, irrespective of particle size. This contradicts the current trend towards single crystal cathode particles, as polycrystalline particles with cracks demonstrate superior discharge performance. This research significantly impacts battery design, highlighting the necessity of cracks for faster charging, and urges a reevaluation of electrochemical parameters for accurate battery modeling.

Control and Reconfigurability of Polarization in Epitaxially Grown Wide Bandgap III-Nitrides: From Materials to Devices
Shubham Mondal, Electrical and Computer Engineering

Moving beyond the silicon era in semiconductors, III-nitrides have emerged as frontrunners in next-generation electronics, characterized by their wide and tunable direct bandgap combined with robust, spontaneous, and piezoelectric polarizations. This unique polarization property makes III-nitrides highly desirable for ultraviolet (UV) optoelectronics and power devices. Additionally, the controllable reconfigurability of this polarization imparts ferroelectricity to III-nitrides, a functionality that opens new doors in the field of data-centric computing applications. My research focuses on effective strategies for the molecular beam epitaxy of polarity-controlled wide-bandgap III-nitrides on large area commercially available silicon carbide substrates. Moreover, I also investigate the ferroelectric functionality of emerging rare-earth alloyed III-nitrides in the context of optoelectronics and ferroelectric memories. My work showcases the potential of polarization engineering in addressing some of the most pressing issues facing optoelectronic devices (UV-lasers, LEDs), high frequency devices (HEMTs), and also promises to drive new possibilities by integrating ferroelectrics for emerging in-memory computing and edge-intelligence.

Combinatorics of Quiver Mutations
Scott Neville, Mathematics

A cluster algebra is a ring defined by combinatorial data: a quiver (directed graph) viewed up to mutation equivalence. Mutations are a particular kind of quiver transformation. A central open question in cluster combinatorics is that of detecting whether two given quivers are mutation equivalent. We approach this problem from two directions. First, we introduce new mutation invariants that yield necessary conditions for mutation equivalence. In many cases, these invariants provide a quick way to establish that two quivers are mutation inequivalent. Second, we construct new arbitrarily long mutation cycles, i.e., sequences of mutations transforming a quiver into itself. These cycles give a barrier to greedy algorithms for deciding mutation equivalence. We then combine these results to construct a heuristic algorithm for the mutation equivalence problem. We conjecture that the algorithm terminates for all four-vertex quivers, and give examples of larger quivers where it does not.

Strategies for Physiological Noise Correction in Oscillating Steady-State Functional MRI and its Impact on Image Quality
Mariama Salifu, Biomedical Engineering

Functional Magnetic Resonance Imaging (fMRI) is a cutting-edge technique to explore the brain’s various functions, including vision, language, and cognition. However, one of the significant challenges in this area of study is maintaining a high signal-to-noise ratio (SNR) to accurately capture the fine details of the brain. Traditional methods to increase SNR, such as high-field MRI machines, can be limited by their high costs and practical feasibility, especially in settings with limited resources. To address this issue, we have developed the oscillating steady-state imaging (OSSI) technique, which offers a remarkable two- to three-fold increase in SNR compared to conventional fMRI methods. However, respiration-induced magnetic field fluctuations can affect the OSSI steady state, reducing the temporal signal-to-noise ratio (tSNR) and functional contrast. This project aims to develop effective techniques for minimizing physiological noise in OSSI, evaluate their impact on data quality, and demonstrate the feasibility of utilizing OSSI for low-field fMRI.

Lunar Stories Unfolded by Volatile Elements: From Water in the Lunar Mantle to Volcanic Eruptions on the Lunar Surface
Xue Su, Earth and Environmental Sciences

Returned lunar samples from Apollo missions deliver valuable information about the Moon, such as its age and composition, and provide unique insights into the formation and evolution history of both the Moon-Earth system and the broader solar system. Volatile elements—elements that are exceptionally sensitive to temperature, such as water—are ideal indicators for tracking temperature-related events during geological processes. This thesis focuses on using novel experimental techniques to analyze various Apollo lunar samples—from both the lunar surface and interior—for volatile elements, and using computational simulations to unravel the stories behind the experimental data. Specifically, this thesis targets enhancing our understanding of 1) the lunar mantle’s composition, with a particular focus on its water content, as proxy for the Moon’s formation, evolution history, and relation to the Earth, and 2) the processes that occurred during ancient lunar volcanic eruptions that shaped the current Moon.

Large-Scale Electronic Structure Calculations at Quantum Accuracy
Vishal Subramanian, Materials Science and Engineering

Electronic structure calculations have been dichotomous between accessible systems sizes and accuracy. This work seeks to break this size-accuracy barrier and presents a computational framework for large-scale quantum accuracy materials simulations. This entails developing algorithms and scalable implementations for fast density function theory (DFT) calculations on large-scale systems involving many tens of thousands of electrons. The accuracy limitation of DFT is tackled by developing an accurate and efficient implementation for “inverse DFT” that provides the exact exchange-correlation or exact correlation potentials corresponding to quantum many-body densities. These are in turn used to improve the model exchange-correlation functionals, which provide a systematic path to realizing the goal of quantum accuracy in DFT calculations.

Certifiable Robot Control and Perception via Convex Geometric Methods on Lie Groups
Sangli Teng, Robotics

Lie groups are ubiquitous modeling tools for studying symmetry and geometric structures in artificial intelligence and robotics, often involving the kinematics and dynamics of sensors or robots composed of 3-D rigid bodies. Because of the nonlinear motion constraints, many highly sought-after problems, such as motion generation, stabilizing control, and state estimation, are generally nonconvex and hard to solve. In this dissertation, I will show that with a proper modeling framework leveraging Lie groups’ rich geometric structure, the globally optimal solutions to these problems are accessible via solving convex optimization problems, leading to certifiable algorithms for robot control and perception. In particular, I will solve the kinodynamic (kinematically and dynamically feasible) motion planning for 3-D rigid body systems globally, optimally, and certifiably; I will then generalize the theory of Kalman filtering from linear systems with Gaussian noise to nonlinear polynomial systems with arbitrary noise distributions with a certificate of optimality.

Leveraging Low-Carbon Diesel Fuel Alternatives to Maximize Life Cycle Greenhouse Gas Emissions Reduction
Courtney Videchak, Mechanical Engineering

Transportation sector emissions must decrease to reach 2050 emissions goals. Electrifying passenger vehicles will help, however, emissions from medium- and heavy-duty vehicles must be addressed. This work decreases the emissions from these vehicles using low-carbon alternatives to traditional fossil-derived diesel fuel, including biodiesel, renewable diesel, and their blend. These fuels are in the marketplace and customers operate their vehicles on them with no modifications, but they are unknowingly neglecting the nuance associated with the specific fuel chemistry. This work aims to reclaim these benefits by creating an optimized fuel-specific engine calibration for maximized engine efficiency and minimized carbon dioxide emissions. This calibration is created by leveraging the effects of the fuel chemistry to modify the input parameters for the engine’s fuel injection strategy and air path management. The conclusions of this work quantify the greenhouse gas emissions benefits for the alternative fuels relative to the conventional fossil diesel fuel.

Elucidating Rotator Cuff Tendon Tear Growth Mechanisms with Full-Field Strains and Data-Driven Multiscale Modeling
Carla Nathaly Villacis Nunez, Mechanical Engineering

Rotator cuff tendon tears can cause aggravating pain and disability. Tear evolution cannot be predicted based on the tear’s original characteristics and when to surgically repair remains controversial. In this dissertation work, we seek to elucidate tear progression mechanisms with an animal model of the intact and torn rotator cuff tendons, using a holistic perspective. In the first part, we employ magnetic resonance imaging to obtain full-field strains of injured tendons, providing quantifiable evidence that high-grade partially torn tendons behave as fiber reinforced materials primarily failing in shear when loaded in the longitudinal direction. In the second part, we examine tearing at the microscale, developing a damage cascading model of collagen fiber networks and characterizing the role of the enthesis in tear evolution. Finally, we integrate the micro- and macro-scales into a fully descriptive model of the rotator cuff tendon, which can predict tear growth, potentially informing treatment strategies.

Total Syntheses of Complex Polycyclic Natural Products
Trenton Vogel, Chemical Engineering

Compounds isolated from natural sources serve as an inspiration for modern drug design and development due to their diverse and beneficial biological activities. Many biologically active natural products feature multiple ring systems, which are often fused, resulting in intricate, highly functionalized core structures. Despite recent advances in developing efficient methods for carbon-carbon bond formation, the selective synthesis of these complex structures remains challenging, thus creating opportunities for the development of new strategies and methods for their synthesis. Research toward my dissertation has focused on two total synthesis projects. First, Cochlearol B, a meroterpenoid isolated from the fruiting bodies of the fungus Ganoderma cochlear. It contains a unique 4/5/6/6/6 polycyclic core structure and has documented antifibrotic activity. Second, Atropurpuran and the arcutines, which were isolated from the roots of a variety of Acotinum plants commonly utilized in traditional Chinese medicine. They bare a unique cage-like framework consisting of a bis-bicyclo[2.2.2]octane system.

Efficient Mapping for Safety-Critical Autonomous Robots in Dynamic Environments
Joey Wilson, Robotics

Autonomous robots require robust world models which succinctly describe their surroundings in order to navigate and plan. Such world models or maps must encompass static intricacies of the environment, such as lane markings, while also accounting for the highly dynamic nature of other vehicles, pedestrians, and animals. Although deep learning has led to major breakthroughs in robotics, temporal modeling of the spatial world remains a challenge due to computational and safety constraints of autonomous robots. In this thesis, I improve the reliability of mapping networks by bridging the gap between probabilistic and deep learning methods. Next, I propose algorithms which leverage and improve deep learning architectures to temporally integrate the static and dynamic world within a unified map. Finally, I combine the two previously mentioned ideas to create a deep neural network which simultaneously models static and dynamic environments with quantifiable uncertainty, tested in challenging real-world scenarios.

Toward a Better Understanding of Arctic Climate Using Novel Techniques
Yan Xie, Climate and Space Sciences and Engineering

The Arctic region is rapidly warming approximately three times faster than the global mean, mainly because the melting surface snow and ice absorb more solar energy, thus reinforcing the warming. Uncertainties remain in current understanding of the Arctic climate and can be reduced by analyzing observations, including satellite and ground measurements. Our research focuses on developing novel techniques to detect and evaluate Arctic rainfall events and surface parameters and to gain insights into their future changes in a warming climate. A satellite remote sensing algorithm has been developed to monitor surface status in the Arctic. Results from ground observations have shown that rainfall can remarkably accelerate the surface snow melting. A machine learning framework is also designed to investigate the vertical structure of rainfall events, which helps to depict the future Arctic climate more accurately.

Design and Control of an Active Mechanical Motion Rectifier Power Take-off for Wave Energy Conversion
Lisheng Yang, Naval Architecture and Marine Engineering

Ocean wave energy has great potential to increase global renewable energy penetration. However, currently wave energy conversion technology faces a significant bottleneck due to the wave’s irregular oscillatory motion. This dissertation presents a new wave energy power take-off design that can perform mechanical motion rectification and advanced control at the same time. A control co-design framework is introduced to optimize the new design’s key parameters for different controllers. The switching nonlinearity of the new power take-off is efficiently handled by a proposed semi-analytical approach to significantly increase the computational speed of co-optimization in the co-design loop. In addition, a compatible control algorithm for the new design is proposed and shown to achieve superior performance. Wave tank experiments validated the functionality of the new design. A novel dry lab hardware-in-the-loop test rig is designed and built to validate the proposed control methods on a small-scale power take-off prototype.

Worker-Centric Adaptations for Scalable Master-Apprentice Relationships Between Human Workers and Robotic Assistants in Field Construction Work
Hongrui Yu, Civil Engineering

This work seeks to transform construction operations and enable an inclusive construction workforce by developing robots capable of construction craft skills and adaptive cooperation with workers. Imitation learning is applied to transfer skills from human workers to robots to streamline learning and include workers of diverse educational backgrounds. The robots emerge skillful and observant of human intent and take over heavy and repetitive physical tasks, improving safety and productivity. Additionally, a generative craft skill metaverse to bridge the chasm between robots and human workers’ intents, skills, and knowledge is developed. The proposed concept is transformative to construction site operations and robot programming schemes through centralization of labor and knowledge resources without temporal and geospatial limitations. The proposed approach enables scalable usage of robot learning data to address robot learning sustainability, human-robot partnership, and ubiquity in robotic construction operations for a leaner industry and an inclusive workforce in the future.