Contributors: Greco, Johnny Paul, Greene, Jenny E, Astrophysical Sciences Department
... Low surface brightness (LSB) galaxies are a common outcome of galaxy formation. Yet, our census of this population is severely incomplete because of the challenges of detecting and studying such diffuse systems. In this thesis, I explore the detection and characterization of these elusive galaxies in the era of deep, wide, and high-resolution optical surveys. I develop software and image-processing algorithms to carry out an automated search for LSB galaxies with the ongoing Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP). Using the first ~15% of HSC-SSP's survey area, I uncover a diverse sample of ~800 LSB galaxies, ranging from dwarf ellipticals in nearby groups to isolated blue ultra-diffuse galaxies to giant LSB spirals. I constrain the redshift distribution of these galaxies using a combination of follow-up spectroscopy and a statistical method that exploits the clustering of galaxies. Their typical distances range from ~30-100 Mpc, with an implied effective radius range of ~0.5-10 kpc. I identify very isolated ultra-diffuse dwarfs, with rotation curve measurements for one such object that suggest it occupies a typical dwarf-like dark matter halo. Using image simulations, I measure the completeness of our survey as a function of galaxy properties, which I combine with the statistical redshift distribution to make a tentative measurement of the mean space density of LSB galaxies with surface brightnesses ~1 mag/arcsec^2 fainter than all previous measurements. The work presented in this thesis will lay the foundation for a more comprehensive search for LSB galaxies with the HSC-SSP, as well as with future wide-field surveys such as the Large Synoptic Survey Telescope. Pushing the surface-brightness limits of these surveys will be necessary to form a more complete census of the galaxy population, which will ultimately provide one of the strongest tests of our standard cosmological framework.
Contributors: Chen, Duyu, Torquato, Salvatore, Chemistry Department
... In this dissertation, the theoretical machinery of statistical mechanics is applied to investigate hyperuniform materials and particle packings. In the first part of the dissertation (Chapters 2-6), we develop novel methods to generate/construct hyperuniform materials in silico, and study their effective physical properties. Specifically, in Chapters 2 and 3, we develop techniques to design and construct hyperuniform two-phase materials and networks, which can be readily realized by 3D printing and lithographic technologies. We also investigate the effective transport and mechanical properties, and wave-propagation characteristics of these materials. In Chapter 4, we computationally explore the use of the self-assembly process of binary mixtures of charged colloids in suspension in order to guide experimentalists to fabricate large samples of effectively disordered hyperuniform materials in two dimensions. In Chapter 5, we employ Lloyd's centroidal Voronoi diagram algorithm to solve the Quantizer problem, and obtain universal disordered hyperuniform final states associated with deep local energy minima when starting from random initial conditions. In Chapter 6, we devise techniques to design experimentally realizable spherical colloidal particles with optimized ``patchy'' anisotropic interactions for a wide class of 2D target low-coordinated hyperuniform structures such as square, honeycomb, kagom\'e, and parallelogrammic crystals that are defect-free. In the second part of the dissertation (Chapters 7-10), we study how particle shape, size distribution, and container can be used as tuning parameters to achieve a rich diversity of emergent properties of particle packings, and develop packing models for real biological systems. In Chapter 7, we ascertain with high precision the stable phases of congruent Archimedean truncated tetrahedra over the entire range of possible densities up to the maximal nearly space-filling density. We also determine the density of jammed (mechanically stable) states with maximal disorder of this system, which in some sense can be regarded to be a prototypical glass. In Chapter 8, we examine disordered jammed binary sphere packings that are confined between two parallel hard planes, which possess packing characteristics that are substantially different from their bulk analogs. In Chapter 9, we employ various sensitive correlation functions to quantitatively characterize structural features of evolving packings of epithelial cells across length scales in mouse skin, and construct a statistical-mechanical model of packings of epithelial cells at late developmental stage. In Chapter 10, we use our knowledge of particle packings to devise a predictive computational model to probe the conditions surrounding tumor dormancy and the ``switch'' to malignant states.
Contributors: Peng, Peng, Bagley, Robert W, Art and Archaeology Department
... The present dissertation adopts a comprehensive, thorough, and systematic approach toward early Chinese lost-wax castings. It first verifies that the lost-wax technique truly existed in Bronze Age China, and analyzes how it came to be used in each casting for which it is confirmed. After investigating the chronology and social context of the early lost-wax cast bronzes, it finally looks for the technological origin and artistic driving force of the lost-wax casting in Bronze Age China. As this dissertation shows, the dominant belief that lost-wax is the optimal method for metal casting, and its arrival was an earthshaking development, deserves more rigorous examination. In a broader sense, the dissertation provides a study on the concept of “norms,” which many historians and art historians constantly rely on, but seldom question or clarify. If we accept the lost-wax process as the norm in casting, early Chinese metalworkers must be regarded as favoring an inferior technology. By revealing the picture that Chinese founders often chose not to use the lost-wax process they had clearly mastered, the dissertation refutes the idea that lost-wax is the only “right way” to cast bronzes. This research aims to show that a “norm” is in many ways a misconception that twists our comprehension of art, technology, civilization, and history.
Contributors: Pacheco Pinedo, Diego Armando, Murthy, Mala, Neuroscience Department
... Many animals use sound for communication, from insects to mammals, but we do not yet understand how the brain encodes auditory information in order to extract the information salient for behavioral decisions (e.g. to flee when hearing an alarm call or to approach when hearing a courtship song). One of the main challenges, in any system, has been mapping sensory information past the periphery, and identifying the patterns of neural activity throughout the entire brain involved in representing sensory information (of any modality) - in addition, we lack tools to compare this activity across animals. Here I present new methods for monitoring auditory activity throughout the entire central brain of the vinegar fly, Drosophila melanogaster, at cellular resolution, along with methods for registering activity across trials and brains for direct comparisons. I utilize these new methods to map and characterize auditory representations past the periphery, and investigate how tuning for courtship song-features arises, in addition to examining the reliability of responses across stimulus presentations and individuals. The fly is an excellent model system for addressing questions related to auditory coding because of its compact nervous system, extensive genetic tools, and robust acoustic communication-based behaviors. During courtship, males chase females while producing a dynamic courtship song via unilateral wing vibration. Females extract information from these songs, which informs their mating decisions, while males eavesdrop on the songs of potential competitors. D. melanogaster song comprises only three modes, two types of brief sound impulses termed ‘pulse song’ and one hum-like ‘sine song’. Using a brain-wide approach to study auditory coding in Drosophila, I record from 18,765 stimulus-modulated regions-of-interest across 31 male and female flies. I uncover that auditory activity is diverse and spatially widespread, finding dense auditory activity in several brain regions not previously known to be auditory, including several neuropils of the olfactory system. I find that stimulus-selectivity and temporal properties of auditory responses differ between neuropils, with activity profiles becoming more diverse from primary auditory areas to more downstream and multimodal neuropils. I also demonstrate that auditory responses are more stereotyped across trials and animals in early auditory areas (neuropils that are one to two synapses downstream of the auditory periphery). I find that only one neuropil, the gnathal ganglia, shows strong dimorphic responses between males and females. Finally, I probe stimulus selectivity in more detail in four neuropils, and find that auditory activity can be divided into three distinct classes - neurons preferring sustained pure tones, pulses, or mixtures of the two. In addition, using a novel class of stimuli, I find that several neurons belonging to these classes are sensitive to the envelope modulation of the stimulus, providing new information on how sine versus pulse selectivity emerges in the auditory pathway. These results highlight the power of studying brain-wide sensory-driven activity to provide a systems-level understanding of how sensory stimuli are encoded.
Contributors: Strouse, Daniel, Schwab, David J, Bialek, William, Physics Department
... This thesis explores three applications of information theory in machine learning, all involving the optimization of information flow in some learning problem. In Chapter 2, we introduce a method for extracting the most informative bits that one signal contains about another. Our method, the deterministic information bottleneck (DIB), is an alternative formulation of the information bottleneck (IB). In Chapter 3, we adapt the DIB to the problem of finding the most informative clusterings of geometric data. We also introduce an approach to model selection that naturally emerges within the (D)IB framework. In Chapter 4 we introduce an approach to encourage / discourage agents in a multi-agent reinforcement learning setting to share information with one another. We conclude in Chapter 5 by discussing ongoing and future work in these directions.
Contributors: Homilius, Max, Troyanskaya, Olga G, Computer Science Department
... In living organisms, biomolecules interact in complex molecular networks that underlie cell function and whose dysregulation leads to disease. These networks thus provide a key lens for understanding the molecular basis of human disease as well as treatment development. In this thesis, I develop three network-based computational approaches for human disease research and gaining insight into the mode of action of drugs. At their core, these methods employ functional interaction networks, which provide a genome-wide view of biochemical and pathway-level interactions and summarize essential functional information derived from diverse and heterogeneous functional genomics experiments. First, I propose a network-based method that can detect critical genes and pathways targeted by a drug treatment from gene expression data even in the absence of large-scale expression differences. This approach enables the analysis of low-dose drug screens, ranking potential targets and drug-perturbed biological processes with higher accuracy than prior network-based methods or gene-expression data alone. Furthermore, I present a method that by inferring and comparing genome-wide profiles for human diseases and animal model phenotypes identifies analogous disease models with high accuracy and more robustly than prior methods relying on shared gene content. This method allows to aggregate the wealth of existing model organism knowledge across multiple species and to identify related phenotypes and novel homologous genes of human diseases for experimental follow-up. Lastly, by constructing a joint tissue-specific classifier for human disease genes, we can significantly improve the prediction of associated genes for rare human diseases. This neural network-based approach makes use of a functional network embedding leveraging tissue-specific expression data and model organism phenotype information in a multi-label classification setting. Overall, the methods I developed provide data-driven, molecular-level solutions to major biological challenges relevant to human health.
Contributors: Shi, Xiao, Selloni, Annabella, Bernasek, Steven, Chemistry Department
... Nickel cobaltite, NiCo2O4, and nickel ferrite, NiFe2O4, are spinel oxides with interesting catalytic properties. Nickel cobaltite oxidizes carbon monoxide and methane, while nickel ferrite is an electrocatalyst for water oxidation. These materials have been recently the focus of intense research aimed at modifying their activities and improving their performances. This thesis describes our theoretical studies of the structural and electronic properties of nickel cobaltite and nickel ferrite, their surfaces, and their interactions with probe molecules. The inverse spinel nickel cobaltite is a promising technological material with complex electronic and magnetic properties. Understanding these properties is important for the development of novel electronic devices and as a basis for the study of their surface and catalytic properties. We have investigated the bulk electronic and magnetic properties of nickel cobaltite using Density Functional Theory (DFT) calculations augmented with on-site Hubbard U repulsion on 3d electrons (DFT+U). Starting from an analysis of nickel doped cobalt oxides, we found that nickel acts as a p-type dopant in Co3O4. NiCo2O4 has a ferrimagnetic half-metallic ground state with fractional valence on Ni and Co cations at tetrahedral sites (Td), caused by the partial occupancy of Ni and Co(Td)’s eg states. We also determined the formation energies of two relevant defects, namely NiCo(Td) exchanges and oxygen vacancies, as a function of the values of the U terms. Facile NiCo(Td) exchange, as observed experimentally, was obtained using U values that are significantly smaller than those predicted by linear response theory. Our computed bulk O-vacancy formation energies suggest that NiCo2O4 is an active oxidant similar to Co3O4. We next extend our study to NiCo2O4 (NCO) surfaces, focusing on the structure, defects and reactivity of (001) surfaces. Our results suggest that the formation of surface oxygen vacancies (VO) on the NCO (001) surface is strongly affected by the neighboring cation in the 3rd layer. In particular, Ni in the 3rd layer significantly reduces the VO formation energy. As a result, VO formation is generally much easier on NCO (001) than on Co3O4 (001) surfaces, suggesting that NCO may be a better catalyst than Co3O4 for oxidation reactions based on the Mars Van Krevelen mechanism. VOs on reduced NCO surfaces can be healed through dissociative water adsorption at room temperature. In contrast, adsorption of molecular oxygen at VOs is energetically unfavorable under ambient conditions, suggesting that O2 adsorption may represent the thermodynamic limiting step for oxidation reactions on NCO (001) surfaces. We again use DFT+U calculations to investigate the mechanism of the low temperature CO oxidation reaction (COOR) on Co3O4(110)/(001) and NiCo2O4(001) as well as methane oxidation on NiCo2O4(001). Our results indicate that the COOR is controlled by the thermodynamics of surface re-oxidation on (001) surfaces and by the kinetic barrier for CO2 formation on the on Co3O4 (110) surface. The COOR is inhibited by water adsorption at surface oxygen vacancies. For methane oxidation, the computed barrier of the first C-H bond agrees well with experimental observations. Nickel ferrite, NiFe2O4, is another spinel oxide with interesting properties and applications, particularly as a catalyst for water oxidation. We have used DFT+U calculations to study the structure, electronic properties, and energetics of the NiFe2O4(001) surface and its interaction with water both in the absence and in the presence of surface oxygen vacancies. In a humid environment, water adsorbs dissociatively on the surface oxygen vacancies leading to the formation of surface hydroxyls. At high temperature, water desorbs leaving a surface containing oxygen vacancies. These defects could represent useful reactive sites for various catalytic reactions. CO and methane oxidation on NiFe2O4 are slightly less favorable in comparison to NiCo2O4, even though the reaction pathways are similar.
Contributors: Fasoranti, Oluseyi Oyebode, Koel, Bruce E, Chemistry Department
... Liquid metal plasma facing components (LM-PFCs) such as lithium (Li) and tin (Sn) have been proposed as potential solutions to first wall and divertor challenges in tokamak fusion reactors. These liquid metals are of interest due to their regenerative and heat transfer properties which would allow them to withstand high heat and particle fluxes emanating from the plasma. Fundamental investigations of these metals using a surface science approach will yield insights into how they can be optimized for use in the various engineering configurations proposed for a fully functional reactor. Ultrathin (up to 10 monolayers, ML) pure Li and Sn films in the solid and liquid state were deposited on polycrystalline substrates of molybdenum (Mo), molybdenum alloy (titanium zirconium molybdenum, TZM), tungsten (W), and single-crystal Mo(100). The thermal behavior, film structure, composition, oxidation characteristics and deuterium uptake capabilities of these deposited films were studied under controlled ultrahigh vacuum (UHV) conditions with Auger electron spectroscopy (AES), X-ray photoelectron spectroscopy (XPS), low energy ion scattering spectroscopy (LEIS) and temperature programmed desorption (TPD). Generally, for Li films deposited on Mo, TZM and W, the monolayer of Li in contact with the substrate is bound much stronger than in bulk Li films, and thermally desorbs at much higher temperatures. Interfacial Li on Mo(poly) has a higher thermal stability than that on TZM(poly), where the limiting values for the desorption activation energies, Ed, are 3.56 and 2.84 eV, respectively, in the low coverage, high temperature desorption tail. LEIS indicates some clustering or interdiffusion of the Li films on the TZM substrate at 500 K. No appreciable irreversible absorption of Li occurs on Mo, TZM or W under the conditions of these experiments. The Li films grown on the TZM substrate showed non-ideal layering at the monolayer level. For post-oxidized Li films on TZM, no Li desorption occurred until temperatures above 620 K, and then Li desorbed from the surface via three desorption peaks at temperatures of 860, 990 and 1220 K. The formation of lithium oxide (Li2O) and peroxide (Li2O2), respectively was observed after post-oxidation. The peroxide converts to oxide after heating to 680 K with no Li desorption, and then this film decomposes to liberate Li into the gas phase while leaving oxygen at the TZM surface. Heating the LiOx films to 1070 K caused oxidation of the substrate to form MoO2 and MoO3, and also a condensed binary lithium molybdenum oxide (LixMoOy) phase. The LiOx film at 310 K initially wetted the TZM substrate well and no de-wetting of the LiOx film occurred prior to evaporation of Li above 680 K. Li deposition on an oxygen-containing TZM surface formed a Li-O-Mo interfacial oxide. This was most clearly seen for the thinnest, sub-monolayer Li films studied. Li desorption from multilayer Li films on oxygen-containing TZM surfaces occurred in a metallic Li multilayer peak and three other oxide-derived peaks at 812, 934 and 1157 K. Sn films deposited on Mo, TZM and W demonstrated stability up to temperatures of 900 K, and thereafter begin to desorb. Multilayer desorption of these films was observed at temperatures of 1187-1270 K. The Sn monolayer films are stable until very high temperatures of about 1800-1900 K on these substrates. These films formed islands after film deposition at 310 K for all substrates studied, with agglomeration of liquid droplets to form larger clusters occurring after annealing to temperatures greater than 500 K. A complex clustering behavior was seen for the Sn films studied on TZM. The electronic properties of sub-monolayer Sn on Mo was modified as demonstrated by a -0.4 eV binding energy shift for the Sn 3d core level indicative of strong Sn-Mo electronic interactions. Sn films were found to oxidize rapidly at 800 K to form SnO2 due to facile O and Sn interdiffusion. Deuterium uptake on Sn films at 310-750 K from irradiation using 700 eV D2+ ions showed lower uptake by liquid Sn compared to solid Sn films. Irradiation of oxidized Sn films by a low flux of 700 eV D2+ caused reduction of the film to metallic Sn.
Contributors: Case, Sarah, Dolven, Jeff, English Department
... The controversy over who would take the throne after Elizabeth I’s death arose from her refusal to do what was expected of her as both a monarch and a woman: namely, marry and produce an heir. The great uncertainty over what would come next for the Elizabethans after the death of their monarch, a subject of debate from the time of Elizabeth’s accession in 1558 to her death in 1603, provided an opportunity for poets to explore uncertainty as it related to ideas of continuity in verse. Coupled with an enduring national anxiety about the possibility of civil war after Elizabeth’s death was the Parliamentary gag order, in the form of the 1571 Treasons Act, which banned discussion of the succession. Increase of Issue: Poetry and Succession in Elizabethan England explores the relationship between monarchical succession and poetic innovation during the reign of Elizabeth I, arguing that a scepticism of the ability of the past to make secure the future emerged in English poetry during this period. The first chapter explores the relationship between gender, rhetoric, and debates about the succession in the works of George Puttenham. Analyzing his lesser-known writings, this chapter argues that the succession debates shaped Puttenham’s poetics manual, The Arte of English Poesy. Chapter Two examines how Elizabeth’s demand for silence on the matter of the succession led Sir Philip Sidney to imagine alternate forms of counsel in the pastoral eclogues of his Arcadia. The third chapter focuses on the genealogical cantos of The Faerie Queene, arguing that Edmund Spenser used the stanza and canto structures of this poem to imagine a version of order in reaction to the disrupted succession. Chapter Four concludes the project by considering the implications of the succession debate on Shakespeare’s sonnets. In tracing the multiple chronologies of these poems, this chapter focuses in particular on the fact that reproduction could no longer provide an answer to the uncertain succession in the late period of Elizabeth’s reign. Shakespeare’s sonnet speaker debates how forms of reproduction and increase operate in relation to the desire for political, personal, and poetic continuity.
Contributors: Schneider, Jonathan, Braverman, Mark, Computer Science Department
... Learning algorithms are often analyzed under the assumption their inputs are drawn from stochastic or adversarial sources. Increasingly, these algorithms are being applied in strategic settings, where we can hope for stronger guarantees. This thesis aims to understand the performance of existing learning algorithms in these settings, and to design new algorithms that perform well in these settings. This thesis is divided into three parts. In Part I, we address the question of how agents should learn to bid in repeated non-truthful auctions -- and conversely, how should we design auctions whose participants are learning agents. In Part II, we study the dynamic pricing problem: the question of how should a large retailer learn how to set prices for a sequence of disparate goods over time, based on observing demands for goods at various prices. Previous work has demonstrated how to obtain O(log T) regret for this problem. We show how to achieve regret O(log log T), which is tight. Our algorithm uses ideas from integral geometry (most notably the concept of intrinsic volumes). Finally, in Part III, we study how to learn the ranking of a set of N items from pairwise comparisons that may be strategic or noisy. In particular, we design mechanisms for a variety of settings (choosing the winner of a round-robin tournament, aggregating the top-K items under the strong stochastic transitivity noise model) which outperform the naive rule of ranking items according to the total number of pairwise comparisons won.