- Shivam Barwey
- Markus Borsch
- Gillen Brown
- Sieun Chae
- Caroline Crockett
- Matthew Day
- Peter Dillery
- Nikhil Divekar
- Xinyang Dong
- Katherine Dowdell
- Francisco Holguin
- Aleksander Horawa
- Brian Iezzi
- Johnathon Jordan
- Daniel Kessler
- Gaang Lee
- Daniel Matera
- Christiana Mavroyiakoumou
- Samar Minallah
- Agnit Mukhopadhyay
- Sajedeh Nasr Esfahani
- Ian Nessler
- Subhankar Pal
- Menglou Rao
- Amin Reihani
- Federica Ricci
- Maria Alejandra Rodriguez Mustafa
- Tara Safavi
- Alexander Shane
- Wenhao Shao
- Sanal Shivaprasad
- Prithvi Thakur
- Nicolas Trueba
- Daniel Vallejo
- Matthew Vedrin
- Hannah Vonesh
- Brett Wagner
- Kelly Wang
- Anna White
- Kevin Wu
- Yichao Yan
- Benjamin Yang
- Xubo Yue
- Xin Zan
A Multiscale Modeling Strategy for Compressible Reacting Flow Simulations Using Multiple Feature-Based Refined Meshes
Shivam Barwey, Aerospace Engineering
A modeling strategy inspired by heterogeneous multiscale and multigrid methods is applied to compressible reacting flow simulations. The central idea is to augment a single global mesh, termed the “macroscale” coarse mesh, with one or more “microscale” meshes intended to cover features of interest at high resolution, e.g., shockwaves, detonation waves, or other complex coherent structures. Analogous to adaptive mesh refinement, the goal is to leverage this high accuracy and resolution for each feature through their individual meshes without sacrificing significant computational cost for the simulation as a whole. The principle advantage over traditional mesh refinement is that each microscale mesh can be assigned unique models that satisfy the physical properties of that feature alone, such that simulations corresponding to these features are more accurate, and large-scale quantities of interest communicated back to the macroscale mesh that depend on small-scale interactions are more robustly captured.
Light-Driven Quantum Electronics
Markus Borsch, Electrical and Computer Engineering
Recent breakthroughs in quantum computing, communication, and sensing show that the qunatum information science and technology revolution is well on its way. Light-drive quantum electronics strives to encode quantum information into electrons in solids and modifies it with strong light fields. This thesis presents the full theory alongside computational models to precisely describe the light-driven processes in quantum materials and devices together with a multitude of applications thereof. By connecting the electronic states with measurements, we demonstrate the first experimental flip of a pseudo-vally spin in a transitionmetal dichalcogenide monolayer. Furthermore, we find a fundamental connection between the nonlinear response of quantum materials with the band structure leading to the development of a new technique to measure the band structure of quantum materials in ambient with superresolution precision. Finally, a comparison with spatially resolved experiments shows that our theory and computational model can simulate all aspects of quantum devices.
Chemical Evolution of Milky Way Sized Galaxies
Gillen Brown, Astronomy and Astrophysics
Observations of galaxies have shown that their chemical properties provide unique information about how galaxies formed, but it remains unclear exactly what parts of the galaxy formation process shaped the properties we see. By running high resolution numerical simulations of Milky Way sized galaxies, I will investigate the physical processes that drive the chemical evolution of galaxies as a whole and the star clusters within galaxies. These results will aid the interpretation of observations, describe what causes the metal content of individual galaxies to evolve over time, and show how the star clusters of the Milky Way can reveal its history, all combining to advance our understanding of galaxy formation.
Theoretical Discovery and Experimental Synthesis of Ultra-Wide-Band-Gap Semiconductors with Ambipolar Doping for Power Electronic Applications
Sieun Chae, Materials Science and Engineering
Power-electronics seek to enhance energy efficiency by utilizing ultra-wide-band-gap (Eg > 3.4 eV, UWBG) semiconductors. The state-of-the-art materials (e.g., AlGaN/AlN, diamond, β-Ga2O3) are suffering from doping asymmetry and/or thermal management, which motivates alternative UWBG semiconductors. Through a high-throughput survey and first-principles calculation, I discover that materials having small cation radius, densely-packed crystal structure, and s-orbital conduction/valence bands tend to have wide Eg but small effective mass that enables semiconductivity. This principle led to the discovery of promising semiconductors with Eg up to 11.6 eV, which challenges the conventional gap-based criterion to distinguish semiconductors from insulators. Among the materials, rutile-GeO2 is identified as an alternative UWBG (4.68 eV) semiconductor with predicted ambipolar doping and high thermal conductivity (51 W∙m-1∙K-1). I demonstrate the first synthesis of single crystalline rutile-GeO2 thin films using molecular beam epitaxy. My dissertation research provides opportunities to realize promising UWBG semiconductors to overcome the current challenges in power-electronics.
A Two-Pronged Dissertation: How Students Understand Signals and Systems and Applying Signals and Systems to Image Reconstruction
Caroline Crockett, Electrical and Computer Engineering
Signals and systems (SS) concepts are the theoretical foundation of machine learning and signal processing, cutting-edge fields with real-world applications in many domains. My dissertation considers two aspects of SS. The first part discusses my longitudinal study on which SS concepts undergraduate students understand, what factors predict understanding, and how those factors influence understanding. The results show how instructors and curriculum designers can improve students’ understanding of SS. The second part applies SS concepts to X-Ray Computed Tomography medical image reconstruction, demonstrating by example how students can use SS understanding to benefit society. I propose an image reconstruction method that takes advantage of machine learning to improve image quality, but which is still explainable, improving the likelihood of radiologists trusting and thus adopting the method. The method could ultimately decrease radiation exposure for patients while providing doctors with high-quality images to properly diagnose and treat many diseases.
Simple Diode Frequency Combs to Revolutionize Spectroscopy
Matthew Day, Physics
Frequency combs are lasers which output many thousands of discrete, equally spaced colors. Their development in the late 1990s led to a revolution in precision measurement and the 2005 Nobel Prize. They can connect the widely disparate, but equally important optical and radio frequency domains. Frequency combs enabled the optical atomic clock and undergird an array of precision frequency measurements. Comb-based spectroscopy is the gold standard in rapidity and resolution. These spectroscopies have enormous application potential outside the lab. However, traditional frequency combs are far too bulky to be applicable in the field.
Rigid Inner Forms Over Function Fields
Peter Dillery, Mathematics
We extend the notion of rigid inner forms, introduced by Kaletha, to the new setting of geometric spaces associated to function fields; a function field is all functions defined on a curve carved out by polynomial equations. Rigid inner forms allow one to group together similar spaces, and are thus useful for endoscopy, the study of how the Langlands correspondence—a conjectural connection between the structure of numbers and highly-symmetric colorings of geometric spaces—changes as one varies a geometric space. We surmount the difficulties presented by function fields, including non-smoothness of spaces, using the geometry of an object called a gerbe. Using rigid inner forms, we construct and then analyze transfer factors—organizational tools which interpolate between spaces and are vital for endoscopy—for all spaces associated to function fields. We use these transfer factors to state new conjectures giving a general picture of endoscopy.
Controller Design and Validation of Lower Limb Exoskeletons for Assistance with Neuromuscular Weakness During Multiple Activities
Nikhil Divekar, Robotics
Current exoskeleton technology targets paraplegics and is unsuitable for millions of mobility affected individuals with mere neuromuscular weakness. The latter population requires assistance that is harmonious with their remnant voluntary motion. Via “energy shaping,” this dissertation develops a novel assistance scheme (controller) that aids these individuals with a broad range of daily activities. The controller was implemented on lightweight and highly compliant knee and knee-ankle exoskeletons. Electromyography results from able body validation experiments show holistic reduction in lower limb muscular effort during walking, stair-climbing, sit-stand, and lifting and lowering (L&L) activities. Pending pre-clinical studies on patient populations are hypothesized to show acute improvements in biomechanical metrics. Alleviation of common problems such as “foot-drop” and “knee-buckling,” and a subsequent increase in gait speed are expected in chronic stroke patients. By assisting the fatigued lower limbs during L&L, the biomechanically recommended “squat-lift” is expected to be facilitated for lower-back-injury prone patients.
Numerical Algorithms for Quantum Many-Body Systems
Xinyang Dong, Physics
This thesis encompasses a series of numerical studies of strongly correlated electron systems. In these systems, interesting and emergent properties are brought about by the strong electron-electron interaction, leading to unusual fluctuations and phase transitions.
The first part of the thesis focuses on the two-dimensional single-band Hubbard model, where we develop new methods to study competing fluctuations in the paramagnetic state. We then introduce extensions of the single particle and two particle Green’s function formalism to study the superconducting state.
The second part addresses the issue of effectively solving equations in realistic material calculations. By using mathematical properties of Legendre series expansion, we design a Dyson equation solver with quadratic scaling and unprecedented accuracy. The method can also be used to solve Dyson equations in both imaginary and real time, thereby enabling the simulation of molecules with heavy elements scale.
Factors Shaping Bacterial Opportunistic Pathogens and Free-Living Amoebae in Drinking Water Systems
Katherine Dowdell, Civil and Environmental Engineering
Drinking water opportunistic pathogens (OPs) are a growing public health concern. Effective infection control is limited by a lack of understanding of OP occurrence and of the factors that influence OP concentrations in drinking water. A one year, source-to-tap evaluation of water quality and OPs in a full-scale drinking water system was conducted, including sampling prior to and during the COVID-19 pandemic. OP inactivation and the potential of a novel, real-time microbial monitoring system are investigated in a full-scale ozone disinfection system. The relationship between OPs and free-living amoebae is also being explored. Preliminary results show that the reduced water usage associated with pandemic building closures caused a deterioration of water quality, with lower chlorine residuals and higher bacterial concentrations. This work provides valuable information for drinking water and health professionals as to the factors that influence OPs and provides data useful for microbial risk assessment.
Exploring the Observational Signatures of Cosmic Rays and Radiation Within Galaxy Simulations
Francisco Holguin, Astronomy
The evolution of galaxies from the early universe to the present day has been a long-standing challenge in astrophysics. Observational data from these galaxies is often limited and multiple physical processes within a galaxy produce similar signals. Simulations of galaxies help illuminate the physical processes that shape the dynamical and thermal galactic structure. Producing synthetic results to compare with observational data is challenging due to computational limitations, so it is necessary to make simplifying assumptions in processing a simulation. An understanding of the region of validity of these assumptions, such as the effect of stellar radiation far from the galaxy, is critical. The central objective of my research is to significantly increase the realism of galaxy simulations and their observational predictions, in order to compare with observed signatures to identify key physics that influence galaxy evolution.
Rationality in Coherent Cohomology of Shimura Varieties
Aleksander Horawa, Mathematics
My thesis studies rationality in coherent cohomology of Shimura varieties in irregular cases. In sufficiently regular cases, this has been well-understood since the 1990s and has had many applications to special values of L-functions (including generalizations of the famous Birch and Swinnerton-Dyer conjecture). There, an automorphic representation contributes to the cohomology of a vector bundle over a Shimura variety just once. We focus instead on the more difficult, degenerate cases where an automorphic representation contributes to multiple cohomological degrees, multiple times. We propose a motivic action which simultaneously explains this spectral degeneracy and answers the question of rationality. This extends the conjectures of Akshay Venkatesh, Kartik Prasanna, and Michael Harris. It has applications to special values of L-functions and p-adic L-functions.
Directed Evolution of Photonic Crystals via Additive Nanomanufacturing and Physics-Guided Reinforcement Learning Algorithms
Brian Iezzi, Materials Science and Engineering
Photonic crystals (PCs) are materials capable of controlling the phase and intensity of light propagation through the choice of interior feature size, symmetry, and refractive index. They offer a means of dramatically improving figures of merit in many application areas, from telecommunications to biological sensing. However, for all but the simplest versions, there is a lack of rapid and scalable manufacturing capability, particularly for 3-dimensional photonic crystals in the visible and near infrared spectrum. Electrohydrodynamic jet (e-jet) printing is an emerging, high-resolution, additive manufacturing platform that can produce nanoscale structures in 3D, using a variety of “inks” with tunable refractive indices. In this dissertation, I combine an e-jet printing apparatus with an in-situ photonic measurement system and 3D electromagnetic simulations to provide learning data sets to a physics-guided, reinforcement learning algorithm which evolves an optimally manufactured crystal using a fraction of the time and materials that would be needed otherwise.
Physics at Short Baseline Neutrino Experiments
Johnathon Jordan, Physics
Short baseline neutrino experiments feature neutrino detectors which are relatively close to a high intensity, accelerator-produced neutrino source. As a result of this proximity, these experiments offer a rich program of physics measurements. Although principally built to search for sterile neutrinos, short baseline neutrino experiments also enable unprecedented neutrino cross section measurements and searches for new physics beyond the Standard Model. This thesis presents work done in pursuit of the physics goals of three different short baseline neutrino experiments: LSND, MiniBooNE, and JSNS2. First, novel constraints on new physics based on data collected by LSND and MiniBooNE are presented alongside projected constraints from JSNS2. Next, a neutrino interaction measurement using monoenergetic neutrinos from kaon decay-at-rest (KDAR) at MiniBooNE is described. Finally, an improved, high statistics cross section measurement using KDAR neutrinos at JSNS2 is shown.
Statistical Methods for Networks with Applications to Neuroscience
Daniel Kessler, Statistics
Networks are an increasingly popular data structure, especially in neuroscience applications, yet conventional statistical approaches do not readily apply. This thesis considers the setting of a sample of networks along with unit-level covariates. We present a tool for scalar-on-weighted-network regression with optional node covariates that exploits network community structure to perform grouped feature selection. We apply this method to human neuroimaging data and demonstrate that its predictive performance is competitive with benchmarks while offering enhanced interpretability. Next, we consider a generalization of regression that simultaneously considers multiple unit-level covariates. We present an approach for performing inference, including confidence intervals, on the component weights obtained from Canonical Correlation Analysis associated with unit-level covariates. We apply the method to behavioral and brain imaging data to explore the degree to which various behavioral measures have distinct versus common representation in brain features.
Wearable Biosensing to Monitor and Advance the Quality of Interactions Between Humans and the Built Environment
Gaang Lee, Construction Engineering and Management
Wearable biosensing has the potential to understand and advance the quality of interactions between people and the built environment by monitoring human psychophysiological responses in daily life. However, several technical challenges hinder the field application of wearable biosensors. This study aims to develop a field-applicable wearable-based urban psychophysiological sensing framework that addresses the challenges. This framework first denoises biosignals collected from fields by applying reference-based adaptive denoising. A specially designed deep neural network then detects abnormality in a person’s psychophysiological responses (e.g., stress) from the denoised biosignals in a field-applicable subject- and context-independent manner. Finally, environmental features having abnormal interactions with humans are geographically visualized by statistically identifying hotspots where multiple people’s abnormal responses are significantly concentrated. The proposed field-applicable urban sensing framework can contribute to advancing one’s quality of experience in the built environment by providing rich understandings of their psychophysiological responses to the built environment.
Exploring the Role of the Physical Microenvironment During Fibrosis Development
Daniel Matera, Chemical Engineering
Fibrosis is attributed to nearly 45% of all deaths in the developed world and is characterized by excessive scar-like extracellular matrix (ECM) deposition within structural organ spaces. While mechanisms of disease onset are similar across organs, pulmonary fibrosis in the lung is rapidly progressive and incurable; a lack of effective treatment options suggest inadequate understanding of disease mechanism. In fact, while myofibroblast driven ECM deposition and tissue stiffening has been linked to hindered lung function, there is currently limited understanding as to how these changes impact signaling at the cellular scale. Hence, the objective of this dissertation is to determine how ECM structure and remodeling effect the phenotype of myofibroblasts and vascular endothelial cells in the lung. Using a combination of cell biology techniques, materials engineering, and tissue-on-chip models, the information gleaned from this thesis work will advance biomimetic model development and enhance pathophysiological understanding.
Membrane Flutter in Inviscid Fluid Flow
Christiana Mavroyiakoumou, Applied and Interdisciplinary Mathematics
We study the large-amplitude flutter of membranes (of zero bending rigidity) with vortex sheet wakes in two-dimensional inviscid fluid flows. We apply small initial deflections and track their exponential decay or growth and subsequent large-amplitude dynamics in the space of three dimensionless parameters: membrane pretension, mass density, and stretching modulus. We also study the instability of a thin membrane to out-of-plane deflections by solving the nonlinear eigenvalue problem iteratively with large ensembles of initial guesses, for three canonical boundary conditions—both ends fixed, one end fixed and one free, and both free. Over several orders of magnitude of membrane mass density, we find instability by divergence or flutter (particularly at large mass density, or with one or both ends free). We find good quantitative agreement with unsteady time-stepping simulations at small amplitude, but only qualitative similarities with the eventual steady-state large-amplitude motions.
From the Laurentian Great Lakes to the Himalaya-Karakoram Glaciers: A Study on the Governing Atmosphere-Cryosphere-Hydrosphere Processes for Regional Freshwater Resources
Samar Minallah, Climate and Space Sciences and Engineering
Freshwater resources perpetually change and respond to large-scale climatic patterns. Therefore, assessment of their evolution under future climate projections first require understanding of the governing processes behind their seasonal fluctuations and long-term variability. This work focuses on process-based assessment of the atmospheric, hydrologic, and cryospheric systems that influence freshwater resources by using data analysis and modeling tools together with application of fundamental physical concepts. The first project focuses on large inland water bodies; specifically, the Laurentian Great Lakes, the African Great Lakes, and Lake Baikal, and how they influence the regional hydroclimates. An in-depth analysis of the atmospheric and hydrological processes in the Laurentian watershed is conducted, and their future changes are evaluated based on climate model projections. The second project evaluates the mass balance sensitivity and dynamics of the Himalaya-Karakoram-Hindukush glaciers under changing climate and the processes that are important for the formation and growth of glacial lakes.
Sources of Ionospheric Conductance – Balance and Impacts
Agnit Mukhopadhyay, Climate and Space Sciences and Engineering
The interaction of the solar wind with Earth’s magnetic field produces a myriad of magnetospheric currents. Since several of these currents circulate through the ionosphere, the ionospheric conductance becomes a crucial factor for predictive investigations of the near-Earth space environment during space weather events. While several investigations have attempted to estimate it, the exact factors determining the conductance remain unclear. In this dissertation, we describe the development of a numerical model that computes each individual source of conductance to address the following questions:
- How do the different sources of conductance impact the overall auroral pattern?
- How do these diverse sources contribute to space weather threats?
- How do the driving mechanisms of these sources vary during extreme events?
We seek to address these questions by conducting systematic numerical experiments with extensive validation studies. Our ultimate goal is to gain an improved understanding of the magnetosphere-ionosphere coupled system through a predictive model proficient for space weather forecasting.
A Simple Method for Human Primordial Germ Cell Development in a Synthetic Embryonic Niche
Sajedeh Nasr Esfahani, Mechanical Engineering
Primordial germ cells (PGCs), the embryonic precursors of sperm and eggs, are established in postimplantation, pre-gastrulation stage embryos in mammals. PGCs transmit genetic and epigenetic information across generations and are the origin of new individuals and the driving force for genetic diversity and evolution. Understanding the origin and specification of human PGCs (hPGCs) have important implications for advancing regenerative medicine and reproductive medicine and contributing to knowledge and models of human germ cell pathologies. Here, in this thesis, we report the induction of hPGC-like cells (hPGCLCs) directly from primed human pluripotent stem cells (hPSCs) in a synthetic embryonic environment recapitulating the in vivo niche of the posterior end of the primate embryonic sac. We demonstrated that amniotic ectoderm-like cells (AMLCs) trigger the specification of PGCLCs from primed hPSCs. Importantly, using this synthetic system, we, for the first time, elucidate the important role of ACTIVIN-NODAL signaling in hPGCLC specification.
Molecular Engineering of Protein-Drug Conjugates for Solid Tumor Immunotherapy
Ian Nessler, Chemical Engineering
Antibody drug conjugates (ADCs) are an emerging class of hybrid drugs that attempt to reduce side effects while maintaining efficacy by combining a small molecule cytotoxin with a cancer specific antibody. Although there has been some clinical success, ADCs encounter multiple delivery challenges such as limited tumor penetration. In addition, their inherent complexity leads to multiple mechanisms of action that are not fully understood but may provide insight to rational design based on driving mechanisms. In this work, the connections between ADC design and efficacy are studied through a systematic modification of ADC design elements, development of a quantitative fluorescent method, and the use of a syngeneic mouse model with a customized ADC. These studies provide guidance to improve ADC function through rational design based on the key factors in ADC efficacy.
Towards Closing the Programmability-Efficiency Gap Using Software-Defined Hardware
Subhankar Pal, Computer Science and Engineering
The slowdown of transistor scaling as their sizes approach the width of a few atoms have caused a surge in Application-Specific Integrated Circuits (ASIC) based accelerators that deliver significantly better efficiency than general-purpose processors, such as CPUs. However, costs associated with design, fabrication, and testing of new ASICs, coupled with lack of programmability, deter their use for scenarios that involve a broad set of application domains. This work aims to bridge the programmability-efficiency gap and deliver near-ASIC efficiency across diverse applications, while retaining CPU-like programmability. It proposes programmable hardware that supports fast reconfiguration across different modes, thus catering to the requirements of the running application. This is complemented with runtime software that monitors hardware performance counters to predict the next best configuration upon detecting a change in phase. Overall, this work aims for impact to the broader computing community, as it takes a concrete step towards realizing practical, introspective computers.
Miniaturized Antennas and Radiation Measurement Techniques for Extremely Small Electromagnetic Systems
Menglou Rao, Electrical and Computer Engineering
In all areas of electromagnetics, electrically small radiation systems are of great importance both in real-world applications and in fundamental research. At low frequencies, the implementation of small antennas enables the use of compact communication systems. At millimeter-wave (mmWave) frequencies, miniaturized antennas are also highly desirable, especially for 5G smartphones. However, miniaturized antennas always suffer from poor performance. On the measurement side, small radiation systems such as biological cells generate extremely weak signals. Special measurement techniques are required for detecting such signals.
This thesis first focuses on the design of highly efficient miniaturized antennas. An extremely low-profile monopole antenna for low-frequency applications and a miniaturized mmWave array for 5G smartphones are presented. Another key aspect of the thesis is the measurement of radiation from biological samples. Key issues with measuring extremely weak signals are addressed and a very sensitive system is developed for the measurement.
High-Resolution Tools for Probing Energy Transport, Conversion, and Storage at the Nanoscale
Amin Reihani, Mechanical Engineering
In recent years, probing energy dissipation and transport at micro/nano-scale has become increasingly important with a wide range of applications in electronics, energy conversion, and energy storage. In this study, we have developed high-resolution calorimetry and thermometry techniques for probing such phenomena.
First, we present a calorimeter capable of measuring the enthalpy of reactions on nanomaterials with a calorimetric resolution of <3 µW/√Hz. This instrument was utilized to measure the enthalpy of hydrogen absorption on Pd nanoparticles with applications in hydrogen storage and electrochemical energy storage.
Second, we demonstrate a novel optical micro-thermometer which achieves the highest reported thermometry resolution of 65 nK/√Hz at room temperature and micro-scale. The temperature sensing mechanism is based on employing Fabry–Pérot resonance at the Urbach edge of direct band-gap semiconductors.
Third, we present a Scanning Thermal Microscopy technique using custom fabricated probes for quantitative mapping of temperature with nanometer resolution. This technique is utilized for mapping the temperature across the junction of a tunnel diode to detect potential solid-state cooling mechanisms. Experimental demonstration of such phenomena is of significant importance for design of high-efficiency light emitting diodes and optoelectronic devices.
Electronic Quantum Coherence Dynamics in Semiconducting Materials at Room Temperature Using the Two-Photon Time-Resolved Near-Field Scanning Optical Microscopy
Federica Ricci, Chemistry
Emerging technologies in electronics are in continuous need for new, high-performance and low-cost semiconducting materials. Electronic coherence has a significant impact on the properties of these materials, with implication in fields ranging from photovoltaic devices to quantum information processing. My research seeks to study the microscopic characteristics and timescale of electronic coherences in energy and charge transfer processes. In these studies, I use various laser techniques such as the three-pulse photon echo spectroscopy and the two-photon time-resolved near-field scanning optical microscopy in order to reveal the role of coherences as it relates to longer exciton diffusion length. Interestingly, coherent charge transport dynamics have been found in the solid phase for dendric macromolecules. And, long-lived coherences have been measured on a single perovskite nanoparticle with extremely high resolution in both space and time. My overall goal is to pinpoint the processes that make semiconducting materials excellent candidates for optoelectronic applications.
Geochronology and Source of Metals and Fluids in Iron Oxide-Apatite and Iron Oxide-Copper-Gold Mineral Deposits
Maria Alejandra Rodriguez Mustafa, Earth and Environmental Sciences
My dissertation investigates the origin of iron oxide – copper – gold (IOCG) and iron oxide – apatite (IOA) mineral deposits. These deposits are major sources of Fe, Cu, and Au, and are enriched in other elements that are critical for the development of a post-carbon infrastructure. Even though these deposit types have been studied separately, their close spatio-temporal association has led to the suggestion that they have the same origin. I use the geochemistry of the mineral magnetite to find the source of the metals and fluids that form these deposits. I also apply novel geochronological techniques to establish the timing of different mineralization stages to constrain a unifying, conceptual model for mineralization. The results of my dissertation will allow for the development of geophysical and geochemical exploration techniques to find new deposits that will optimize the exploration process and maintain the supply of resources for future generations.
Representing and Inferring Relational World Knowledge in Machines
Tara Safavi, Computer Science and Engineering
Endowing machines with human-like relational reasoning skills is a long-standing goal of artificial intelligence. Focusing on world knowledge derived from text, this thesis explores and connects two classes of relational knowledge representation in machines. The first part of the thesis considers representations that have explicit relational structure enforced by an ontology. We characterize the strengths and weaknesses of these representations in the link prediction task, in which a machine learning algorithm is trained to predict missing relationships in a graph. In the second part of the thesis, we expand our scope to consider implicit relational knowledge stored in the parameters of deep neural networks trained to model the language of large encyclopedic text corpora. We show that such representations, which surface relational knowledge via querying or downstream usage rather than from an ontology, are flexible, powerful, and ultimately necessary complements to traditional relational data structures that store world knowledge.
Wave-Particle Interactions with Superthermal Electrons on Dayside Martian Crustal Fields
Alexander Shane, Climate and Space Sciences and Engineering
The study of the transport of superthermal electrons is important in planetary space environments because they are able to efficiently heat and ionize the upper atmosphere, a contributing factor in atmospheric escape. Using data from the MAVEN mission, I showed that the high energy (100-500 eV) pitch angle distributions of electrons are not explainable by adiabatic invariants and collisions alone, and that our understanding of electron transport at Mars is incomplete. Whistler waves were investigated and it was found that the plasma environment of the crustal fields is conducive for whistler waves to interact with 100-500 eV electrons. Modeling of the diffusion equation will be performed to determine if whistler waves can produce the observed distributions. The crustal fields of Mars offer a unique system to study wave-particle interactions. Using data, theory, and numerical models this work is working toward a more complete picture of electron transport at Mars.
Metal-Free Purely Organic Phosphors: From Fundamental Molecular Design to Performance Amplification in Modern Applications
Wenhao Shao, Chemistry
Metal-free purely organic phosphors (POPs) as novel organic semiconducting materials are promising functional components in modern and emerging technologies such as organic light-emitting diodes (OLEDs), solid-state lighting, bio and chemical sensors, and data encryption. However, realizing the full capacities of POPs remains challenging due to the deficient fundamental understanding of their emission mechanism and the dearth of systematic molecular design blueprint. The dissertation research is built upon a fundamental design rule for POPs we proposed emphasizing a dynamic symbiosis of heavy atom effects and orbital angular momentum manipulation, which could greatly enhance the Spin-Orbit Coupling efficiencies in POPs. An advanced molecular simulation method has been developed for the novel POPs. The design rule has been applied to prototypic POP-based OLEDs and functional POPs having Excited-State Intramolecular Proton Transfer properties that could respond to external stimuli. Furthermore, “state-specific” manipulation on triplet excitons was demonstrated in high-efficiency POPs with ultralong emission.
Convergence of Complex Geometric Measures to Non-Archimedean Measures
Sanal Shivaprasad, Mathematics
We study the limits of families of natural measures on complex manifolds. This includes the natural measure on log Calabi-Yau manifolds as well as the Bergman measure on Riemann surfaces. We interpret the limit of these measures as a measure on the Berkovich analytification of the family. The convergence takes place on a Berkovich hybrid space obtained by compactifying the family using the aforementioned analytification. In particular, we show that Bergman measures on Riemann surfaces converge to the Zhang measure on the analytification. We also compute the limit on various other hybrid spaces such as the Boucksom-Jonsson hybrid space and the metrized curve complex hybrid space.
Earthquake Cycle Simulations of Strike-Slip Fault Systems Surrounded by Fault Damage Zones
Prithvi Thakur, Earth and Environmental Sciences
Numerical modeling of earthquake cycles aims to combine seismologic, geodetic, and geologic observations with physical principles to explain the long-term behavior of active faults. We use earthquake cycle simulations to demonstrate that the geometry and mechanical behavior of fault zones can significantly affect the spatial and temporal seismicity distribution. Shallow fault zones can also produce a bimodal depth distribution of earthquakes. We further develop physical models to show that the evolving strength of fault zone structures during the seismic cycle can have pronounced effects on the temporal evolution of the locked and creeping segments of the fault. Lastly, we will discuss the interplay between stress heterogeneities and fault damage zones at various spatial scales through seismic cycles and the role of plastic deformation within the fault damage zone. Our results highlight the importance of mapping the fault zone structures in strike-slip fault systems for seismic hazard assessment.
Probing the Inner Regions Near Accreting Black Holes and Neutron Star with Chandra
Nicolas Trueba, Astronomy and Astrophysics
Accretion disks are ubiquitous in the universe, ranging from the protoplanetary disks responsible for transporting material onto a newly born star after the collapse of a cold gas cloud, to those funneling gas into supermassive black holes in the centers of galaxies. Despite decades of study, many of the key physical mechanisms responsible for accretion lack direct observational evidence. This is especially true in accreting black holes and neutron stars, where questions regarding the primary mechanisms responsible for mediating angular momentum transfer and the nature of the central X-ray corona remain unanswered. In this dissertation, I demonstrate how in-depth high-resolution spectroscopic studies of absorption phenomena in accreting black holes and neutron stars can set meaningful constraints on the physics that underlie the disk. In addition, I develop a new method for constraining the size of the central emitting region of the disk, providing new angles into their physical nature.
Development of Ion Mobility-Mass Spectrometry as a High Throughput Assay for Biotherapeutic Characterization
Daniel Vallejo, Chemistry
Thermal stability measurements have, in many cases, become surrogates for protein structure stability following the modulation of their solvation environments, ligand binding states, or their sequences through mutation. Importantly, such stability “fingerprints” are used extensively to assay the fitness of potential protein-based biotherapeutics, the many successful applications of which now drive a multibillion-dollar research and development enterprise.
The technologies currently in place to measure protein stabilities, while well validated, are limited by large amounts of protein or require labeling chemistry not amenable to high-throughput measurements. Our improvements to MS-based methods that measure the stabilities of gas-phase proteins, such as collision induced unfolding (CIU), demonstrate the potential to bridge the gaps above, and rapidly deliver information-rich protein stability assessments in an order of magnitude faster than their solution phase counterparts for complex mixtures for biotherapeutics.
Improving Water Quality in Drinking Water Distribution Systems Through Engineering Action Research
Matthew Vedrin, Environmental Engineering
Most people in the U.S. receive drinking water via centralized water treatment and distribution systems. People expect the water to be safe at all times, which is difficult to achieve given the complex technical and social elements of these water systems. Most water providers are not equipped to improve distributed water quality through modern data science approaches. Conversely, engineering researchers are well positioned to conduct applied research in partnership with utilities to achieve both practical outcomes and generate new knowledge. My dissertation takes an action research approach in partnership with the City of Ann Arbor to develop strategies for improving water quality and related monitoring after treatment. I utilize city records and historical data alongside primary collected data from two separate longitudinal water sampling campaigns. Experiences across middle- and high-income contexts have demonstrated the global relevance of drinking water quality challenges and the need for local practical solutions.
Development and Mechanistic Elucidation of Lewis Acid-Catalyzed Carbonyl–Olefin Metathesis Transformations
Hannah Vonesh, Chemistry
Olefin–olefin metathesis and has drastically changed how olefins are synthesized in materials, agrochemicals, and pharmaceuticals. An important variation of olefin–olefin metathesis is carbonyl–olefin metathesis, which provides an additional approach to access olefins, but has lacked advancements in methodology. Recently, the development of catalytic protocols for carbonyl–olefin metathesis has brought a renewed interest to field. The proposed catalytic cycle for FeCl3-catalyzed carbonyl–olefin metathesis operates through asynchronous, concerted [2+2]-cycloaddition, forming an oxetane, that fragments via retro-[2+2]-cycloreversion to provide the desired metathesis products. The Schindler lab has previously studied the mechanism of intramolecular carbonyl–olefin metathesis, but fundamental questions remain. 13C Natural abundance KIEs are an attractive way to investigate these questions, as they can identify whether the reaction is stepwise or concerted, what the turnover-limiting step is, and characterize the transition state.
Achieving Mainstream Partial Nitritation/Anammox Nitrogen Removal via a Membrane Aerated Biofilm Reactor (MABR)
Brett Wagner, Environmental Engineering
Nitrogen removal from wastewater is energy intensive and accounts for 3% of the total energy load in the United States. This dissertation examines how an innovative, energy-efficient wastewater treatment technology, the membrane aerated biofilm reactor (MABR), can be combined with resource efficient anammox bacteria to remove nitrogen in mainstream wastewater. This research addresses practical questions focusing on optimizing the reactor startup, finding the ideal configuration of the reactor in relation to other treatment processes, and determining the impact of placing the reactor in series. Ultimately, answering these questions will enable practitioners to incorporate the anammox metabolism into MABRs and will hopefully increase their adoption for nitrogen removal in wastewater treatment.
Grain Coalescence in Quasicrystals
Kelly Wang, Macromolecular Science and Engineering
Quasicrystals (QCs) are solids possessing long-range orientational order but lacking translational periodicity possessed by conventional crystals. Due to their unique structures, metallic QCs possess unusual properties, such as high corrosion resistance and thermal, electrical, and frictional anisotropy. These distinct properties have a variety of promising applications, such as in insulation, solar energy, and a Teflon-alternative. QCs challenged the paradigm of crystallography—crystals were once defined by their translational periodicity, or a repeating pattern with a single, distinct unit cell. In 1984, Dan Schechtman won a Nobel Prize for the discovery of these unique crystals. Despite advances made in QC research, fundamental questions about aperiodicity effects on defect and dislocation behavior remain. In this thesis, I examine grain coalescence, or the formation of a single crystal grain from multiple, misoriented grains in QCsusing molecular dynamics simulations. This work paves the way towards fabrication of large-scale, dislocation-free QCs for novel applications.
Novel Quantitative Approaches to Vulnerability in the United States
Anna White, Industrial and Operations Engineering
Vulnerable populations are understudied by quantitative researchers, despite a rich field of existing qualitative work. I use predictive and inferential modeling with simulation analysis to detect patterns related to root causes and effects of vulnerability. I apply these methods to study vulnerability in four distinct areas: discriminatory housing policy, human trafficking, opioid misuse treatment, and hazard resilience. I use spatial data and weighted logistic regression to study differences in health, income, education, employment, and racial composition in the context of historic discriminatory housing policy. I use inferential modeling to understand what spatial characteristics and local policies favor illicit massage businesses, one of the primary industries of human trafficking. I am working to improve access to treatment for opioid misuse through transportation solutions. I developed a new framework to model cellphone service resilience during natural hazards. Each of these applications demonstrates a step forward in human- and vulnerability-centered quantitative research.
Molecular Mechanism of ATP-Independent Chaperone-Mediated Protein Folding
Kevin Wu, Biophysics
Being able to correctly fold into a well-defined three-dimensional structure is vital for proteins to be able to function properly but this folding is a challenging task. To protect proteins from misfolding or aggregation and to guide proper protein folding, cells rely on molecular chaperones. ATP-independent chaperones work on the front lines of stress defense. It is generally assumed that these chaperones very tightly hold onto non-native proteins in order to prevent them from aggregating under stress conditions. Here, by combining biophysical and biochemical approaches, we found that two ATP-independent chaperones instead actually allow their client proteins to fold while they are bound to the chaperone. This folding-while-bound mechanism could make ATP-independent chaperones not just efficient aggregation inhibitors but also folding catalysts.
Physical Organic Approach Towards High-Performance Materials for Non-Aqueous Redox Flow Battery
Yichao Yan, Chemistry
Redox flow batteries (RFBs) are a promising technology for grid-scale energy storage of renewable energy resources. Aqueous RFBs have been studied extensively and have resulted in the development of promising commercial batteries. However, aqueous systems suffer from a narrow thermodynamic cell voltage window of 1.5 V. Switching from water to a non-aqueous solvent expands the theoretical cell voltage window from 1.5 V up to 5 V, drastically increasing the energy that can be stored. One of the major research efforts has been focusing on the discovery of proper catholyte and anolyte materials for this application. Molecules that are stable at all redox states, but also electrochemically active and soluble, are required to develop RFBs that take full advantage of the non-aqueous solvent. My research focuses on addressing these challenges in the context of developing new materials for non-aqueous RFBs using basic principles of physical organic chemistry.
Towards Defining Principles of Cellular Plasticity
Benjamin Yang, Biomedical Engineering and Scientific Computing
Cell fate plasticity, a cell’s ability to change its identity and function, is molecularly specified and tightly regulated in all mammalian cells. Elucidating cell fate regulators can reveal therapies that prevent disease and aging by attenuating aberrant cell fate transitions. My dissertation employs a combination of experimental techniques to characterize and manipulate three factors that guard cell fate: 1) chromatin organization and gene expression, 2) nuclear elasticity, and 3) cell-cell signaling. The first project describes how an understudied chromatin modification (H4K20me1) is affected in muscle stem aging, and explores the role of Sestrins, a group of stress-inducible proteins, as metabolic regulators of muscle stem cell regeneration and aging. The second project uses a micro-engineered cell-squeezing device to show how mechanical properties of the nucleus contribute to chromatin organization and cell fate decisions. Finally, the third project investigates the biological significance of cell-cell fusion as a mechanism to promote tumor metastasis.
Bayesian Machine Learning: Theory and Applications in the Modern Engineering System
Xubo Yue, Industrial and Operations Engineering
Bayesian methods such as Gaussian process and Bayesian optimization have become increasingly popular and promising in the modern engineering system. Meanwhile, there are many unsolved challenges such as how to perform stable parameter estimations (also known as inference), how to provide accurate predictions, and most importantly, how to ensure fairness and privacy. The aim of this dissertation is to address those challenges in Gaussian processes, Bayesian optimization, and federated learning. The first part of this dissertation will focus on the theoretical aspect of Gaussian processes. Specifically, I provide a new inference framework for Gaussian processes using Renyi divergence. The second and third part of this work are applying Gaussian processes to the Bayesian optimization and the joint statistical model, respectively. The key advantage is that the proposed models can help industrial engineers make maintenance timely decisions and save economic loss. The last part of this dissertation is devoted to the study of the federated learning. I will develop a distributed learning framework to ensure fairness and privacy. This framework enables collaboration among many institutes without sharing private data.
Scalable Architectures for High Frequency and Very High Frequency Wireless Power Transfer
Xin Zan, Electrical and Computer Engineering
Wireless charging is taking hold while high frequency (HF) and very high frequency (VHF) wireless power transfer (WPT) stands out because of spatial freedom, efficiency, miniaturization, and integration. I work on a completely new scalable architecture for HF-VHF WPT which involves interdisciplinary studies among control, electromagnetics, and circuits with an application range from watts to hundred watts for biomedical and consumer electronics, robots, and drones, breaking the trade-offs among device, power, frequency, and transfer distance. I investigate novel segmented power converters, which consist of several current-mode class D (CMCD) modules. The first 100 MHz inductive WPT is demonstrated using a singleton CMCD system. Then 100 MHz segmentation CMCD WPT demonstration aggregates the magnetic flux and power together from each identical and synchronous module by electrical connection, which physically increases the coil size at HF-VHF and extends the transfer distance and power level, but maintains the efficiency of the optimized singleton system.