- James Akinola
- Anil Alan
- Olamide Animasahun
- Anna Argento
- Srinivasan Arunachalam
- Margaret Brunette
- Zhongzhu Chen
- Shihao Cheng
- Natasha Dacic
- Hrishikesh Danawe
- Alisher Duspayev
- Easton Farrell
- Katherine Harrison
- Qiana Hunt
- Sarah Katz
- Cecelia Kinane
- Matthew Lasky
- Siyul Lee
- Ju Won Lim
- Ting Lin
- Zhenkun Lin
- Yan Long
- Subha Maity
- Isaac Malsky
- Danielle Maxwell
- Merjem Mededovic
- Jia Mi
- Jonathan Michaux
- Hossein Moghimianavval
- Alexandra Moy
- Aditya Varma Muppala
- Kate Napier
- Ishtiaque Ahmed Navid
- Ilia Nekrasov
- April Nellis
- Eunjae Shim
- Ramakrishnan Sundara Raman
- Franco Tavella
- Rodrigo Tinoco Figueroa
- Irene Vargas-Salazar
- Emily Wearing
- Andrew Wintenberg
- Sangmin Yoo
- Shuqing Zhang
- Oleksii Zhelavskyi
Understanding Aqueous-Phase Adsorption for Predictive Catalyst Design in the Electrocatalytic Hydrogenation of Bio-Oil Model Compounds
James Akinola, Chemical Engineering
Bio-oils from waste biomass can reduce greenhouse gas emissions by replacing fossil fuels, but they need aqueous-phase catalytic or electrocatalytic hydrogenation to upgrade them. Using bio-oil has not been economically feasible because the state-of-the-art Pt group metals used as hydrogenation catalysts still have low activity and are very expensive. This dissertation focuses on understanding descriptors of hydrogenation activity, such as aqueous-phase adsorption energies, to design novel active and inexpensive hydrogenation catalysts. We showed that the adsorption of relevant bio-oil organics on metal surfaces in the aqueous phase requires solvent displacement, which weakens their adsorption energies compared to the gas phase. We then developed a bond additivity model to capture this water displacement and organic solvation to aid accurate and low-cost calculation of adsorption energies and prediction of new catalyst materials. Finally, by controlling organic and hydrogen adsorption energies, we synthesized inexpensive PtxCoy alloy catalysts to improve the hydrogenation rate.
A Framework for Nonconservative Robust Safety with Control Barrier Functions
Anil Alan, Mechanical Engineering
The most critical requirement in designing today’s autonomous systems is safe operation, even if they are surrounded with environmental or operational uncertainties. The safety-critical solution for these systems should not be intrusive in the sense that a nominal operation based on a certain task can be maintained as long as possible. While the well-known control barrier functions provide us practical ways to calculate such safety conditions, their mathematical guarantees may deteriorate in the presence of uncertainties. Uncertainties are typically overcome by considering worst-case scenarios affecting the system. However, this preemptive approach often yields undesirable performances. For an autonomous vehicle, for example, it allows other vehicles to cut in the large headway space left by an unnecessarily overcautious autonomous driver. This dissertation targets improving the control barrier function framework by introducing more flexible compensation for uncertainties.
Genome-Based Spatially Resolved Multi-Omics Data Integration Reveals the Liver’s Cellular Architecture and Dissects the Molecular Underpinnings of Metabolic-Associated Liver Disease
Olamide Animasahun, Chemical Engineering
Globally, metabolic-associated liver diseases are responsible for more than a million deaths every year, and a major obstacle to understanding these diseases is mapping out the spatial heterogeneity and dissecting the molecular milieu of the hepatic-microenvironment in the diseased liver tissue. Here, we developed a work flow that integrates orthogonal multi-omics data to decipher the spatial localization of hepatocytes and other tissue-resident cells in the diseased and healthy liver. This framework also incorporates the reconstructed genome-scale model of human metabolism to dissect and compare the prevailing metabolic pathways in the diseased and healthy tissues. Also, we will probe the nature of the paracrine cellular signaling that exist between the hepatocytes and their neighboring cells. Findings from this work will inspire the identification of biomarkers for early disease detection and pinpointing targetable small molecules for metabolic-associated disease treatment and prevention.
The Spatiotemporal Organization of Brain Tumors: From Oncostreams to Liquid Crystals
Anna Argento, Biomedical Engineering
High grade gliomas (HGG) are the most frequent, most aggressive, and most heterogeneous of all malignant brain tumors. Tumor heterogeneity can be detected at the histological, molecular, cellular, and supracellular level. Heterogeneity makes gliomas very hard to study and to treat. One particular tumor pathology is classified as mesenchymal differentiation and is characterized by cell elongation and cell motility. We recently discovered distinctive fascicles of elongated, aligned, mesenchymal-like cells throughout mouse and human tumors, which we defined as oncostreams. Using time-lapse confocal microscopy in ex vivo slices, and in in vivo two photon imaging, we determined that cells within oncostreams are highly motile. Oncostream motility was classified based on cellular orientation into: streams (cells moving back and forth), flocks (cells moving in one direction), and swarms (cells moving randomly). The molecular characteristics of oncostreams were determined by laser capture microdissection followed by RNA-sequencing, and bioinformatics. Forty-three genes were differentially expressed in oncostreams; COL1A1 was overexpressed in oncostreams. Inhibition of COL1A1 in gliomas, using the Sleeping Beauty transposon model, eliminated oncostreams, reduced tumor aggressiveness, inhibited proliferation, reduced tumor vasculature and collective glioma invasion. Thus, oncostreams are morphologically and molecularly distinct, and contribute to the mesenchymal phenotype and tumor malignity. Further, we recently determined that glioma cell growth in vitro also displays domains of nematic orientation (oncostreams), and topological defects, two essential characteristics of 2-D liquid crystals. Liquid crystals are a mesophase of matter, between crystals and liquids. It is increasingly recognized that biological matter can behave as liquid crystals. Topological defects are singularities of local molecular orientation, i.e., regions where orientation cannot be identified. Topological defects play a critical role in the spatiotemporal organization of active liquid crystals and have been exploited to manipulate liquid crystal behavior. We hypothesize that manipulating brain tumor liquid crystalline behavior will be of therapeutic importance.
Performance-Based Wind Engineering Through Stochastic Simulation and High-Fidelity Computational Modeling
Srinivasan Arunachalam, Civil Engineering
With the burgeoning growth of high-rise building construction and increased awareness for the creation of sustainable urban habitats, solutions for performance-oriented, and economical building systems are in great need. To enable efficient quantification of uncertainty in the structural performance, stratified-sampling-based uncertainty propagation techniques were developed. Within the context of performance-based wind engineering, a probabilistic framework was proposed for the investigation of the nonlinear behavior of wind-excited steel structures, with a particular focus on the estimation of the annual rate of inelastic excursions (system-level yielding) and reliability against component and system-level failure modes. For the capture of hurricane-induced aerodynamic effects associated with time-varying wind speed and direction, experimental validation of a non-stationary pressure simulation model was carried out. A novel methodology was introduced for the rapid estimation of nonlinear responses under uncertainty. Finally, the influence of hurricane hazard characterization on the risk assessment of wind-excited nonlinear structures was investigated.
Designing an Immune-Isolating Poly-Ethylene Glycol-Based Capsule for Implantation of Human Ovarian Cortex
Margaret Brunette, Biomedical Engineering
The increase in pediatric cancer survivorship over the last 30 years has concurrently resulted in the increase of young women with premature ovarian insufficiency, resulting in the loss of ovarian endocrine function and inability to undergo physiological puberty. A cell-based therapy that delivers the entire array of hormones, at physiological levels, rates, and reciprocal feedback with the rest of the body could promote normal pubertal development. Previous work has demonstrated that a poly(ethylene glycol) (PEG)-based capsule with a degradable core and a non-degradable shell is effective for mouse tissue. Here I describe updating this PEG capsule to function for human ovarian tissue by evaluating graft health and growth, as well as capsule diffusion and its ability to prevent immune cell infiltration. This technology has shown its effectiveness, and next steps should be to study capsule efficacy in larger animal models and eventually translate to the clinic.
Tackling the Maximum-Entropy Sampling Problem: Upper-Bounding and Lower-Bounding Techniques
Zhongzhu Chen, Industrial and Operations Engineering
The maximum-entropy sampling problem (MESP) is a fundamental and challenging combinatorial-optimization problem, at the intersection of information theory, data science, and optimization. It finds application in spatial statistics (e.g., environmental and geological monitoring), finance, and machine learning. It asks to find a maximum (differential) entropy subset of s random variables, from a universe of n (correlated) Gaussian random variables, which is a means of choosing the s-subset with maximum information. One line of research concentrates on exact algorithms using a branch-and-bound framework, where the problem’s upper and lower bounds play a significant role in the performance. However, this framework fails on large-size instances of MESP, due to its NP-hardness. We propose new upper- and lower-bounding techniques to accelerate the running of the branch-and-bound framework on MESP, thus enabling the tractability of large-size instances. Furthermore, we build a public MATLAB package that enables people to solve MESP by calling user-friendly API.
Controlling a Powered Prosthetic Leg During Continuous and Automatic Transitions Between Multiple Activities
Shihao Cheng, Robotics
Powered prostheses can provide mechanical work to reduce the compensations and effort for lower-limb amputees to walk. Challenges in control of these complex devices over different activities currently prevent this technology from reaching widespread clinical adoption. This dissertation focuses on developing a novel control paradigm for a powered knee-ankle prosthetic leg to perform different locomotion activities and transitions between them, including walking, stair ascending, and stair descending. Using convex optimization, steady-state kinematics are modeled as continuous functions of speed, incline, and gait phase, whereas transition kinematics are modeled by linear interpolation between these steady states. A high-level classifier recognizes the user’s intent with high accuracy so that the mid-level controller based on these kinematics models can control the positions of knee and ankle joints over varying activities. Kinematics results from amputee participants demonstrate similarity with able-bodied data, effectively restoring normative gait biomechanics for amputee users.
How Humans Influence the Environment and Inform Decision-Making
Natasha Dacic, Climate and Space Sciences and Engineering
Anthropogenic greenhouse gas emissions continue to be a major environmental issue. Nitrous oxide (N2O), one of the most potent greenhouse gases, remains one of the most dominant human-emitted stratospheric ozone-depleting substances of the 21st century, slowing the recovery of the Ozone Hole. The primary source of N2O emissions is agricultural soil, specifically from applications of synthetic nitrogen fertilizer. Current emissions of N2O are unquantified, making modeling emissions a challenge; long-term large-scale measurements are scarce, and there are limitations of the geographical spread of measurements to inform models. Thus, these limitations capture the importance of understanding N2O and its effects as a greenhouse gas for preventing further damage to the ozone layer and climate. My thesis aims to use atmospheric observations to quantify modern N2O emissions from the U.S Corn Belt, improve our understanding of processes that control emissions on varying scales, and inform relevant policy applications.
Role of Phononic Crystals in Elastic Wave Focusing, Subwavelength Imaging, and Sensing Applications
Hrishikesh Danawe, Mechanical Engineering
In the last decade, metamaterials have given new dimension to mechanical wave propagation research due to their unconventional ability to control elastic/acoustic waves. In this dissertation work, we study novel design concepts for a particular class of metamaterials called phononic crystals (PCs) which are artificially engineered periodic structures. In the first part of the thesis, we work on generalizing the existing concept of planar gradient index phononic crystal (GRIN-PC) lenses to curved structures and develop a theoretical framework for design and analysis of non-planar elastic lenses that would make elastic wave focusing feasible over a broad range of structures. In the second part, we design a negative refraction-based flat PC lens to experimentally demonstrate subwavelength imaging of flexural elastic waves, which can benefit many applications such as achieving super-resolution in ultrasonic imaging. Finally, we are working on a compact sensor to determine acoustic properties of unknown liquid samples using defect modes in phononic crystals.
Precision Measurements and Quantum Sensing Using Cold Atoms
Alisher Duspayev, Physics
I present experimental and numerical investigations on several fundamental properties of laser-cooled atoms in different quantum states and how these may translate into applications in emerging technologies. First, I describe precision measurements of the dynamic polarizability and photo-ionization cross-section at 1064-nm light of the rubidium 5D state, which is of interest for portable atomic clocks. I demonstrate a cold-ion-beam source based on the photo-ionization of cold rubidium 5D atoms trapped in a tightly focused optical-cavity mode. The source employs a novel method to monitor the electric micro-fields of the ions exploiting the Stark effect of highly excited (Rydberg) atoms. In related work, I investigate a new type of Rydberg-atom-ion molecule bound by long-range multipolar forces and examine the influence of nonadiabatic dynamics on its stability. Lastly, I present a novel technique for atom interferometry that employs uninterrupted three-dimensional confinement and manipulation, including sensitivity estimations and initial experimental work.
Piezo1-Driven Synovial Fibrosis Through Yap1-Mediated Canonical Wnt Activation in Post-Traumatic Osteoarthritis
Easton Farrell, Biomedical Engineering
Osteoarthritis is a debilitating condition that can be induced by joint trauma and gives rise to painful bone growths, cartilage depletion, and severe inflammation of synovium. Rampant matrix deposition by activated synovial fibroblasts raises tissue stiffness and stimulates exacerbatory stromal-immune crosstalk. Mechanoreceptive synovial fibroblast activity driving synovial fibrosis is uncharacterized. Canonical Wnt/β-catenin signaling (cWnt) is implicated in matrix mediation and clinical cases of osteoarthritis. Piezo1 is a mechanosensitive cation channel that activates under direct mechanical deformations of the cell membrane and has been shown to drive fibrosis in other tissues. The role of Piezo1 in synovial fibroblasts, however, is not understood. My objective is to characterize Piezo1-mediated mechanotransduction driving osteoarthritis pathology. The guiding hypothesis is that Piezo1 mediates cWnt signaling and drives the pro-inflammatory and pro-fibrotic synovial fibroblast phenotypes. Given the complete absence of disease-modifying osteoarthritis treatments, this work seeks to uncover novel therapeutic avenues for pursuit.
A Mechanistic Approach to Wastewater Monitoring for Viruses
Katherine Harrison, Environmental Engineering
Viral pathogens such as SARS-CoV-2, influenza, smallpox, Ebola, and HIV/AIDS are a major public health concern resulting in increased hospitalization and death. Since viruses are excreted in stool and urine, monitoring wastewater for viruses is a promising method to conduct epidemiological surveillance without the high cost of testing or bias towards symptomatic cases. However, a mechanistic understanding of the diverse structural components of viruses is necessary to fully understand how viruses interact with the complex wastewater environment and therefore how these measurements are interpreted for public health. My dissertation investigates viruses in untreated wastewater to understand how viruses decay and partition, and ultimately how current wastewater monitoring can be translated into successful public health interventions. My work demonstrates that through mechanistic understanding of viruses in wastewater, engineers and public health officials can utilize wastewater monitoring to implement accurate and effective public health interventions and therefore limit viral infections.
X-ray Binaries and Ultraluminous X-ray Sources in Nearby Galaxies
Qiana Hunt, Astronomy and Astrophysics
X-ray binaries dominate the X-ray emissions of most galaxies and are a crucial element in population synthesis and gravitational wave models. Due to their different evolutionary timescales, XRBs with donor stars of different masses can be used as proxies for different properties of their host galaxies, such as total stellar mass and star formation rate. However, since low- and high-mass XRBs have similar spectral shapes and luminosities in the X-rays, a clear way of differentiating between these classifications in galaxies that host a mix of them has been elusive. By combining Chandra X-ray data with HST optical imaging, we are able to directly identify low-mass and high-mass XRBs within nearby, star-forming galaxies. This method uniquely allows us to identify the distinct XRB populations within star-forming galaxies, determine their correlations with environmental properties, and understand the role of stellar clusters in the formation of XRBs.
Andean Interglacial Climate and Hydrology Over the Last 650,000 Years
Sarah Katz, Earth and Environmental Sciences
Due to anthropogenic global warming, terrestrial landscapes may experience enhanced climatic impacts like warming temperatures and/or hydrologic intensification (e.g. drought or floods). However, long climate records from the low-latitudes are scarce, so it is uncertain how these understudied regions, where 40% of the world’s population lives, have responded to global change in the past. Records of global warm periods (i.e. interglacials) from Lake Junín, Peru, could provide insights into Andean climate during recent interglacials, but traditional techniques generate ambiguous climate interpretations of the Lake Junín record. My dissertation uses novel geochemical techniques to overcome this ambiguity. I reconstruct continent-scale rainfall patterns, temperature, and regional water stress during the last two interglacials (the Holocene and “penultimate” interglacial) and an ancient interglacial 650,000 years ago. I find that the central Andes experienced acute water stress during past interglacials. It is likely that ongoing climate change will threaten water security in the Andes.
Functional Polymer Networks for Self-Powered Microrobots
Cecelia Kinane, Macromolecular Science and Engineering
Mobile microscale robots are a promising technology to address complex problems in drug delivery and medical treatments, environmental contamination, and microfabrication. Presently, traditional motor components cannot be used at small scales due to challenging manufacturing size limitations. Therefore, in order to push the boundaries of small-scale devices and enable new miniaturized designs in robotics, there is a need for new materials and methods for powering and controlling microscale robots. This work seeks to understand the fundamental mechanisms in microrobot locomotion and to develop new functional polymer network materials to power microrobots, with the overarching goal of creating efficient, effective, adaptable, and biocompatible microrobots that can navigate in biologically relevant environments. This work will bring functional, real-world application of these materials and microrobotic devices closer to reality by overcoming short lifetimes, poor performance, autonomy, and toxicity problems.
Development of Novel Photochemical Methods for the Construction of Aryl Carbon–Nitrogen Bonds
Matthew Lasky, Chemistry
Arylamine products are of high value in pharmaceutical, agrochemical, and materials chemistry applications; thus the development of methods to access these motifs are of great importance. In the past decade, light-driven organic chemical transformations have emerged as a powerful new technology by providing access to reactive radical intermediates that afford a variety of novel bond-forming protocols, which are not readily accessible under thermal control. Herein, we detail three projects utilizing photochemical activation for the construction of aryl C–N bonds. First, we describe two closely related methods for SNAr and C–H pyridination of electron-rich arenes via photoredox catalysis. Second, we establish a complementary method for photochemical C(sp2)–H pyridination that eliminates the requirement of an exogenous photocatalyst by leveraging the photoactivation of electron donor–acceptor complexes. Finally, we describe the discovery and development of a versatile, photocatalytic acridine–Lewis acid system for C(sp2)–H amination of arenes.
Anatomy of Black Hole in Five-Dimensional Anti-de-Sitter Spacetime
Siyul Lee, Physics
Black hole entropy is the pinnacle of modern physics where quantum, gravitational, and thermal laws of nature intertwine. In 2018, the entropy of supersymmetric black holes in five-dimensional Anti-de-Sitter (AdS) spacetime was accounted for by number of microstates in the dual four-dimensional Super-Yang-Mills theory. This thesis spells out some details of the matching. We solve Klein-Gordon equation in 5D supersymmetric AdS black hole background and demonstrate superradiance. We present a class of D3-brane geometry that preserves the same amount of supersymmetry as the black hole. We quantize their configuration space to define and count microstates of giant gravitons. We comment on their relation to black hole microstates. In somewhat orthogonal direction, we present explicit expressions for operators in Super-Yang-Mills theories with small gauge groups SU(2) and SU(3) that are not of graviton type and thus are candidates for actual microstates of smallest quantum black holes.
Study on Thermal Energy Manipulation for Renewable Energy Application
Ju Won Lim, Materials Science and Engineering
Manipulation and harvesting of thermal energy play an important role as the demand for non-carbon energy resources has increased. Recent experimental work has verified that the total radiative thermal transfer between two objects can exceed the black body limit when one or more dimensions of the objects are smaller than the dominant thermal wavelength, even in the far-field. Here, we first demonstrate a far-field thermal transistor consisting of two coplanar SiN nano-membranes combined with a phase transition material-coated modulator as a third body to manipulate the radiative thermal energy. By only adjusting the modulator (gate) temperature, we are able to modulate heat energy between the emitter (source) and the receiver (drain) by factors of three in the far field. Our new design of the thermal transistor will pave the way for a wide range of applications, such as thermal management, energy harvesting, and related thermal energy industries.
Elucidating the Interfacial Molecular Interaction Mechanisms of Silicone Adhesive, Polymer Degradation, and Polymer Bio-applications Using Advanced Spectroscopy
Ting Lin, Macromolecular Science and Engineering
Interfacial properties and interfacial interaction of polymer are important to multiple applications including adhesion, bio-interaction of polymer, and more. To better research and design desirable interfacial properties, learning how molecular structure changes at interface is extremely important. However, the lack of techniques to probe the molecular structure directly and nondestructively at surface or buried interface makes it difficult to perform systematic analysis on such systems. In this thesis, an interface sensitive techniques-sum frequency generation (SFG) vibrational spectroscopy and other characterization techniques are used to elucidate the interfacial structure—properties correlations of multiple polymeric material systems. The systems of interest include silicone adhesive, bio-interaction between polymer and biological molecules, and polymer interfacial degradation.
Reconfigurable Metasurfaces for Low-Frequency Acoustic/Elastic Wavefront Control
Zhenkun Lin, Mechanical Engineering
Emerged as a new kind of phase-modulated structure in the last decade, metasurfaces offer compact and lightweight solutions for wave control, which are especially advantageous for wavefront control in the low-frequency regime. This study explores different mechanisms to control the acoustic/elastic wavefront and achieve various phenomena, including wave deflecting, wave focusing, and non-paraxial propagation. We endow metasurfaces with reconfigurable capabilities by exploiting different mechanisms, including easily tuned local resonators, origami-inspired folding mechanisms, and piezoelectric shunts. Furthermore, to enhance wave control opportunities beyond conventional linear metasurfaces, we devise a nonlinear acoustic metasurface formed by curved beams to effectively control higher-harmonic generation. This study uncovers unconventional wave control via reconfigurable material systems adaptable to different environments, e.g., operating frequencies, geometric scales, and vibration patterns. The finding in this work can be used in a broad range of engineering applications, e.g., energy harvesting, signal demultiplexing, and imaging with enhanced resolution.
Modeling and Mitigating Side Channels in Optical and Embedded Sensing Systems
Yan Long, Electrical and Computer Engineering
The Internet of Things and mobile devices depend on sensors to make life-critical decisions ranging from steering an autonomous vehicle to defibrillating a patient’s heart. It’s extremely important to ensure trustworthy and confidential data from sensors. However, a substantial gap persists between what application engineers expect from sensor semiconductors and what sensors actually provide in terms of trustworthiness for protecting confidentiality of sensitive information, leading to sensor side channel problems. By investigating representative examples of such problems with embedded camera sensors, we seek to measure and close this gap by (1) discovering and modeling the consequence and root cause of sensor side channels, (2) inventing techniques that can undo or prevent threats posed by sensors against existing and future systems, and (3) complementing existing side channel literature with an analytical framework for sensor side channels.
A Study of Transfer Learning Problems and Their Connections to Algorithmic Fairness
Subha Maity, Statistics
This dissertation contributes to several problems in the area of responsible data-science related to transfer learning (TL) and algorithmic fairness. The first chapter investigates the theoretical limitations of the label shift problem in non-parametric classification settings in terms of sample sizes and several model parameters. The second chapter develops a linear adjustment based model and method for the posterior drift problem, performs a minimax study on the proposed method, and showcases its utility in mortality prediction for a demographic minority group in a UKBiobank dataset. The third chapter designs a statistical framework for TL in classification settings in the absence of labeled samples from the target domain, which makes the problems considerably challenging. The fourth chapter analyzes the efficacy of group fairness techniques in mitigating the bias caused by a subpopulation shift model, while the fifth studies the effect of “stereotyping’” in unsupervised representation learning techniques for an under-represented group.
Modeling the Three-Dimensional Atmospheric Structure of Hot Gaseous Planets
Isaac Malsky, Astronomy and Astrophysics
By studying exoplanets with numerical simulations, we can further our understanding of their formation, evolution, and dynamics. I am expanding upon the Rauscher-Menou General Circulation Model to study hot Jupiters—massive gaseous planets with orbital semi-major axes less than 0.05 au. I have added to the capabilities of the GCM in order to study inhomogeneous cloud formation, hazes, varying planet inclinations, and wavelength-dependent energy transport. Numerical simulations that can simulate multiple processes in concert are critical, as they allow us to understand the mutual feedback that these mechanisms induce. Furthermore, I have created a pipeline to simulate telescope observations from our models and shown how changes in planetary structure manifest in detectable ways. Throughout my Ph.D., I have used these models to interpret observations of a number of exoplanets. With further additions to the GCM in my final year, I will expand our understanding of exoplanet physics.
Navigating the Hidden Curriculum of Academia: A Critical Approach to Understanding the Racial and Science Identity Formation of Latine Undergraduate Students at Hispanic-Serving Institutions
Danielle Maxwell, Chemistry
The institutional histories, disciplinary practices, and norms in science constrain the representation and advancement of minoritized students by deciding what knowledge is acceptable, influencing the ways of knowing and identities that students bring to the classroom. With science identity being one proposed construct for predicting the persistence of minoritized students in science, it is imperative to understand how racialized experiences contribute to science identity formation. Centering the voices of Latine undergraduate students and drawing from critical epistemological perspectives, I examine how racialized experiences in science may influence the science identity development of Latine undergraduate students. My research questions explore how Latine students navigate epistemological border crossing between racial and science identity. Specifically, I use the genre of testimonio to challenge objectivity through the narratives of my research participants. Drawing upon these narratives, my dissertation disrupts dominant narratives and exposes inequitable systems of oppression in science.
Enhanced Cell-Cell Communication Using Gene Delivery Coating for Accelerated Bone Fracture Healing
Merjem Mededovic, Biomedical Engineering
Bone fractures are common—there are 15 million annually, however 22% exhibit delayed healing. Patients experiencing delayed healing require repeat interventions and experience pain for six months or longer. The proposed work demonstrates a holistic approach to accelerating fracture healing by exogenously upregulating gap junction intercellular communication (GJIC), which improves coordination of the mesenchymal stem cells (MSCs) and vasculature. Lentiviral vectors delivering GJA1/Connexin43, a GJIC protein, were used to determine effects on MSC differentiation, migration, and angiogenesis, and immobilized to a polymer deposited to titanium using chemical vapor deposition. GJIC upregulation increases MSC differentiation, but does not affect MSC migration or angiogenesis, and may reduce fracture bridging delay. Paracrine communication regulating bone and vascular regeneration is increased with Connexin43 delivery, indicating extensive effects on cell behavior. This work is a framework for localized gene delivery, which is adaptable and mitigates limitations of viral gene delivery and implies GJIC as a platform to alter cell behavior.
Oscillating Surge Wave Energy Converters: Novel Design, System Integration and Case Study
Jia Mi, Naval Architecture and Marine Engineering
Marine renewable energy has enormous but untapped potential. In the United States, the total technical available wave energy equals to 34% of the yearly electricity generation from all resources. This study aims to investigate two types of oscillating surge wave energy converters (OSWEC) with novel design. Firstly, a nearshore bottom-hinged OSWEC is proposed to directly power reverse osmosis desalination. This pilot study experimentally reveals that ocean wave energy is a promising source to sustainably power desalination. Secondly, to extend the operation of bottom-hinged OSWEC from nearshore to offshore with larger wave energy visibility, a dual-flap out-of-phase floating OSWEC is introduced. The proposed floating OSWEC consists of a floating platform and two pivoting flaps. The distance between the two flaps is around half of the wavelength to achieve out-of-phase motion. Numerical modeling results show the proposed dual-flap design can significantly mitigate the platform’s horizontal motion and the mooring load can be reduced.
Safe Robot Autonomy with Structured Neural Implicit Representations
Jonathan Michaux, Robotics
Deployment of robots in human-centric environments is limited due to key challenges such as ensuring safety and real-time performance. My research addresses these challenges by proposing three algorithms called ARMOUR, RADAR, and DART. ARMOUR, the core method, is an optimization-based trajectory planner and tracking controller that operates in real-time, is robust to uncertainty, and is certifiably safe. Importantly, ARMOUR is able to solve complex motion planning tasks on a real robotic arm. Replacing ARMOUR’s constraints with neural networks, RADAR’s fast computations allow ARMOUR’s optimization problem to scale to higher dimensional robots. DART incorporates ARMOUR as a layer in a trainable deep neural network, thus enabling certifiably safe robot learning. The integration of ARMOUR, RADAR, and DART provides a novel framework that synthesizes robot behaviors that are provably correct, computationally tractable, generalizable, have rigorous performance guarantees, and realizable on physical robotic systems.
Synthetic Biology Platforms for Recreating and Engineering Functional Membrane-Membrane Interfaces
Hossein Moghimianavval, Mechanical Engineering
Cell-cell communication and signal transduction through direct cell-cell contact are found ubiquitously in biology, with examples including T-cell activation upon contact formation with antigen-presenting cell in immunological synapses and electrical signal transduction in neuronal synapses. Despite the prevalence of intercellular communication at cell-cell interfaces, it is difficult to engineer and reconstitute functional proteins at membrane-membrane interfaces. A synthetic biology tool that can recapitulate signal transduction exclusively at membrane-membrane interfaces will be useful for designing synthetic intercellular communication pathways or providing a potential replacement for damaged and malfunctional natural cell-cell communication systems. Here, we present a synthetic biology tool for programmable activation of split proteins at the membrane-membrane interface between synthetic cells. We utilize this system to demonstrate reconstitution of a fluorescent protein as well as a bioluminescent protein. We then leverage the bioluminescent membrane-membrane interfaces and show a light-based contact-dependent communication pathway between synthetic and natural cells.
Investigation of the Grain Boundary, Heteroelectrolyte, and Cathodic Interphases and Interfaces of Ceramic Solid Electrolytes for Solid-State Batteries
Alexandra Moy, Materials Science and Engineering
The need for improved battery performance has increased the impetus to replace state-of-the-art carbon-based anodes with Li metal. However, it is generally known that Li metal anodes cannot cycle with state-of-the-art liquid electrolytes. Owing to its high conductivity, Li metal stability, wide electrochemical window, and stiffness, lithium lanthanum zirconium oxide (LLZO) is considered a promising next generation solid-state electrolyte for Li metal batteries. However, its interfaces, especially internal and cathodic, are not well understood and, thus, will be the basis of this thesis. This dissertation is broken into four main sections: effects of doping on LLZO, effects of grain boundaries on electrochemical behavior of LLZO, characterization of the cathode/LLZO interface, and characterization of the sulfide electrolyte (catholyte)/LLZO interface. This research will provide a greater understanding of these crucial but under-investigated areas of solid-state battery research, contributing to solid electrolyte and solid-state battery advancement.
High-Resolution Millimeter-Wave Imaging Radars for a Safe and Equitable Future
Aditya Varma Muppala, Electrical and Computer Engineering
The most fascinating imaging systems are those that make the invisible visible. In the age of black-hole telescopes and molecular microscopes, one such system is the imaging radar. Imaging radars can see through objects over long distances without using harmful radiation. A good example is the scanning chamber that one walks through at an airport to detect concealed objects. The focus of my research is to democratize imaging radars by making them portable, scalable, and cost-effective, thus allowing their widespread use in applications such as: concealed weapons detection at stadiums, schools, and concerts; autonomous navigation; low-cost medical and industrial diagnostics; and search and rescue operations. An inter-disciplinary holistic systems approach is taken to propose new ideas in reflector antennas, polarization beam-steering, array processing algorithms, and millimeter-wave integrated circuits. Using these ideas, we have demonstrated high-resolution high-speed imaging radars that are 100 times smaller and cheaper than the market standard.
Cluster-Lensed Quasars: From Discovery to a Measurement of the Hubble Constant
Kate Napier, Astronomy and Astrophysics
We know that the universe is expanding at an accelerating rate, yet astronomers cannot agree on how fast this expansion is occurring. Different precise methods that rely on either the nearby or far away universe to measure the expansion give different answers. The focus of my dissertation is to use an independent method—using gravitational lensing and lens modeling—that will help answer this important question of how fast the universe is expanding. Using a different method that is sensitive to different systematic uncertainties might shed light on why current state-of-the-art measurements are discrepant with each other. If the methods continue to disagree, we might require a new understanding of physics.
Nanoscale and Alloy Engineering of III-Nitride Semiconductors for High Efficiency Solar Photocatalysis and LEDs
Ishtiaque Ahmed Navid, Electrical and Computer Engineering
GaN based nanostructures are increasingly being used for a broad range of device applications, including LEDs, laser diodes, transistors, and more recently artificial photosynthesis. Artificial photosynthesis processes like photocatalytic and photoelectrochemical solar water splitting have become very promising means of clean fuel production for sustainable ecology and environment. For such applications, III-Nitride semiconductors such as GaN, InN, and their alloys InGaN have drawn substantial attention for their direct and tunable energy bandgaps and structural stability. In this work, we have investigated the design, epitaxy, fabrication, and characterization of highly efficient GaN/InGaN nanowire based artificial photosynthesis devices. We have also studied the material synthesis and optical properties of GaN based deep-nano network structures. This work provides significant insights in advancing the development of high efficiency artificial photosynthesis devices and offers a new path for realizing the next generation nanoscale optoelectronics.
Oligomorphic Groups and Geometry of Semi-Infinite Wedge
Ilia Nekrasov, Mathematics
In the first part, we prove the (equivariant) noetherian property for a wide class of varieties generalizing the class of Plucker varieties. It improves previous results of Draisma-Eggermont, who treated the case of bounded Plucker varieties. Our approach is based on new results about the geometric structure of the infinite wedge. These structures include previously known varieties (as hyper-Pfaffians) and completely new equivariant subvarieties of wedge powers. In the second part, we describe smooth representations of oligomorphic groups via a combinatorial model. In most cases these categories are infinitely generated and not semi-simple, so we prove several fundamental finitary properties (e.g., all objects have finite length). Our main examples include the infinite symmetric group, the group of order-preserving diffeomorphisms of real numbers, and the automorphism group of the Rado graph.
Numerical Methods in and Economic Applications of Stochastic Optimal Control Theory
April Nellis, Applied and Interdisciplinary Mathematics
Stochastic optimal control problems can model a rich variety of situations where a controller facing uncertainty wishes to optimize their expected outcomes. By using stochastic differential equations, complicated control problems can be represented in a probabilistic setting. We take advantage of this setting to apply machine learning and Monte Carlo methods to problems that would be extremely difficult to solve analytically or using traditional grid-based numerical methods. Namely, we consider three exigent problems with economic applications. The first is an optimal control problem in pandemic modeling and mitigation with novel macroeconomic elements. In the second, we develop an efficient and accurate machine learning algorithm to solve the high-dimensional optimal switching problem faced by a power plant operator under uncertain production costs and profits. Finally, we combine theoretical models and empirical data to determine the optimal behaviors of small-scale liquidity providers in a decentralized cryptocurrency exchange.
Expanding the Nickel-Catalyzed Cross-Coupling Toolbox with Machine Learning and Quantum Chemistry
Eunjae Shim, Chemistry
Nickel-catalyzed cross-electrophile-coupling is a promising approach that expands the scope of cross-coupling reactions. For these reactions, substrates with seemingly small structural differences commonly require substantially different reaction conditions. This makes prediction of conditions for new substrates challenging, resulting in a resource-intensive reaction space exploration. Machine learning algorithms that utilize prior reaction data and prioritize conditions for the new substrates could facilitate this exploration. However, the number of available datapoints is insufficient to train mainstream machine learning models, necessitating novel algorithms that make better use of small data. Drawing inspiration from expert chemists’ workflows, an active transfer learning algorithm is developed by building a simple initial model and elaborating it as data is collected. The utility of this strategy is prospectively demonstrated, doubling the yield of a challenging nickel-catalyzed cross-electrophile-coupling reaction. In all, the developments of this thesis will enrich building blocks applicable to this reaction, expediting access to useful molecules.
Global, Longitudinal, and Data-Driven Investigation of Network Censorship
Ramakrishnan Sundara Raman, Computer Science and Engineering
Governments, authorities, and private actors increasingly seek to control how Internet citizens access content and communicate online. These information controls typically take the form of blocking of access to certain websites or online services. The community’s understanding of the current state and global scope of such network interference remains limited: Most work has focused on the practices of particular countries at specific points in time, or on the reachability of limited sets of online services. Creating a global, longitudinal, and data-driven view of network interference is an extremely challenging proposition, since such practices are intentionally opaque, and censorship mechanisms may vary. Moreover, advances in network interference technology and recurring instances of censorship events all over the world have necessitated high-quality measurement tools and data that can help researchers, journalists, policymakers, and advocacy groups characterize network interference technology and ensure accountability. This dissertation develops novel measurement methods to investigate the network technology that enables network interference, and proposes frameworks to monitor their deployment around the world. This dissertation also advances methods to rapidly measure large-scale network interference to gain insight on events such as censorship and interception attacks. Finally, this dissertation explores the development of a global, longitudinal network observatory, Censored Planet, and advances the analysis of large-scale censorship measurement data.
Robustness and Tunability of Biological Oscillators
Franco Tavella, Biophysics
Many natural processes, such as heartbeats, cell cycles, and circadian rhythms, are cyclical and powered by a network of biomolecules broadly called biological oscillators. The network structure of many of these oscillators is highly conserved. Yet, which network features determine their performance remains unclear. My dissertation explores the relationship between network architecture and oscillatory function through modeling and experimentation. Computationally, I analyze which network motifs display tunability. Tunable networks can change their period and amplitude in response to environmental signals. Experimentally, I explore the circuit design that allows cell cycles to sustain perturbations in cytoplasmic density and temperature. Previous theoretical work suggested that positive feedback increases cell cycle robustness. However, my results show that antagonistic regulation is the determining factor for robustness and tunability. Overall, the results of this dissertation deepen our understanding of biological oscillations and provide a set of guidelines for designing robust and tunable cyclical behavior.
The Evolution of the Ray-Finned Fish Brain: Novel Data on Brain Anatomy in Fossil and Living Lineages
Rodrigo Tinoco Figueroa, Earth and Environmental Sciences
Ray-finned fishes represent the most diverse lineage of living vertebrates and show a wide range of morphologies occupying equally diverse environments. Since brain anatomy relates to evolutionary relationships and ecomorphology, it is expected that ray-finned fishes would exhibit diverse brain morphologies. However, there is lack of neuroanatomical information for most living ray-finned fishes and absence of direct evidence of brain anatomy from most fossils. In my dissertation I describe novel exceptional preservation of fossil brains and explore the hidden diversity of living ray-finned fish neuroanatomy using high-resolution micro-computed tomography (µCT) to provide a better understanding of patterns of brain evolution deep in the history of the group. The results of my dissertation provide the framework for a new line of research in vertebrate paleontology, integrating three-dimensional soft-tissue data from fossil and living groups in an evolutionary context and demonstrating that data from exceptionally preserved fossils can provide key anatomical information.
Utilizing Field Massive Stars to Study the Properties and Evolutionary Consequences of Binaries
Irene Vargas-Salazar, Astronomy and Astrophysics
Almost all massive stars are born in binaries, causing them to exchange mass with their companions, which affects their structure and evolution. Therefore, binaries profoundly affect stellar populations and their feedback effects on galaxy evolution. I study a complete sample of massive field stars in the nearby Small Magellanic Cloud, which are linked to binarity. By using cluster-finding algorithms to search for small clusters, I find that 95% of these massive stars are ejected from their parent clusters by binary interactions. These “field stars” may themselves be ejected as binaries, thus their properties can help us understand the binary statistics and ejection mechanisms. I use multi-epoch spectroscopic observations to obtain radial velocities to extract their binary properties. Combining these with binary population synthesis models that model the number of ejections and their evolution, I examine the implications for the massive star population and their evolutionary consequences.
Development of New Visible-Light-Mediated Synthetic Methods to Access 4-Membered N-Heterocycles
Emily Wearing, Chemistry
The development of new pharmaceuticals relies on access to organic molecules that provide drugs with the desired biological activity. While modern organic synthesis has evolved to provide the wealth of pharmaceutical compounds available today, development of new drugs hinges on a careful balance of drug properties and is therefore limited by which compounds are accessible. Nitrogen-containing heterocycles are essential components of pharmaceuticals, however, despite valuable properties, 4-membered N-heterocycles (azetidines and azetines) are vastly underrepresented due to challenges in their synthesis. Research towards my dissertation has utilized visible-light-mediated triplet energy transfer photocatalysis as a mild method to access excited state compounds and enable the development of previously inaccessible [2+2]-cycloaddition reactions to form azetidines and azetines. The development of these methods has provided access to previously inaccessible azetidines, 1-azetines and 2-azetines, and has provided insight into the mechanisms by which photocatalysis can enable the synthesis of challenging molecules through energy transfer.
Privacy and Utility in Dynamic Systems: Verification and Enforcement
Andrew Wintenberg, Electrical and Computer Engineering
Recently, society has become reliant upon cyber-physical systems that integrate physical processes across computer or cyber networks. Many of these systems, ranging from the smart grid to medical devices, communicate sensitive information which when mishandled can lead to serious harm to both the system and its users. Protecting this information is challenging, as eavesdroppers can leverage their knowledge about how the system evolves over time to learn information in unexpected ways. In my work, I study privacy requirements for dynamic systems using the formal notion of opacity. In the framework of discrete event systems, I develop more efficient methods for verifying that a system is privacy-preserving. Furthermore, when systems cannot be verified, I propose a framework for opacity enforcement. In this framework, the outputs of a system are altered to mislead eavesdroppers and shape their beliefs in a process called obfuscation.
Theoretical Discovery and Experimental Characterization of Memristor Internal Physics and the Physics-Based Neural Network System for the Future of Artificial Intelligence
Sangmin Yoo, Electrical and Computer Engineering
Memristors are a new class of devices that enable efficient computing. The capability of co-locating compute and memory allows memristors to implement multiply-and-accumulate (MAC), the most fundamental operation in neural networks, through simple Kirchhoff’s current law with orders-of-magnitude improvements in throughput and power-efficiency compared with conventional von-Neumann implementations. Furthermore, the internal physics of memristors can be directly used to accelerate neural network learning and spatiotemporal data processing. Efficiently leveraging the native electronic, ionic, and thermal dynamic processes will allow the device to autonomously process temporal inputs and perform rich synaptic and neuron functions. This research direction is however still at its infancy, partly due to difficulties in controlling the internal dynamics in memristors. This dissertation aims to develop theoretical and experimental approaches to control the internal memristor physical processes, demonstrate neural networks empowered by tunable internal dynamics, and produce efficient and reliable computing hardware for AI-related tasks.
Analysis of Polymer Materials for Industrial Applications
Shuqing Zhang, Macromolecular Science and Engineering
Polymer materials are being developed in numerous fields to meet the expanding demands of the industry. Meanwhile, plastic waste accumulated in the environment has created a worldwide concern. This work focuses on analyzing polymer materials for industrial applications from the perspectives of performance improvement and sustainability. In the first part, sum frequency generation (SFG) vibrational spectroscopy was applied as the main tool to investigate the adhesion mechanisms of coatings for sealing applications. The observed interfacial molecular behaviors were correlated to mechanical performances to help design future products with improved properties. In the second part, the environmental weathering process of polyethylene materials for packaging applications was studied using a combination of analytical techniques. Based on the relationship between the manufacturing parameters and degradation rates, the study will assist the industry in developing and modifying plastics that are less harmful to the environment or with accelerated degradation rates after usage.
Diastereoselective and Enantioselective Sustainable Catalysis for Carbohydrate and Heterocyclic Chemistry Application in Batch and Continuous Flow
Oleksii Zhelavskyi, Chemistry
Stereoselective transformations play a significant role in modern organic chemistry synthesis, but still face issues related to low availability and cost of the chiral catalysts, their recovery and recycling. The dissertation describes synthesis and application of immobilized Chiral Phosphoric Acids (CPAs) as sustainable and recyclable heterogenous catalysts for regioselective carbohydrate modifications and efficient asymmetric hydrogen atom transfer (HAT) reduction of heterocyclic molecules in continuous flow. Computational data generated via Quantum Mechanical/Molecular Mechanical studies along with 2D-NMR experiments are used for mechanistic investigations and determining of regioselectivity controlling factors of transformations catalyzed by CPAs. Further studies are focused on exploring reactivity of purely described fused aziridino-quinoxalines as useful precursors for generation of aromatic aziridine ylides salts via thermal and photochemical ring opening process. Asymmetric reactions of in situ generated ylides with nucleophiles, dipolarophiles, and HAT-agents are studied in the thesis.