12023 research outputs found
Sort by
3D in situ Chemical Synthesis: Additive Manufacturing of Functional Polymeric Materials via Vat Photo-polymerization
Get PDFAs additively manufacturing gains popularity in rapid-prototyping, manufacturing and customized production, there is a continuous demand in seeking for new materials with advanced functionalities to satisfy the wide range of applications in aerospace, construction, optics, actuation, dentistry, biomedical practices and even food industry. Vat photopolymerization (VP), a light-enabled AM technique, is particularly promising due to its ability to achieve good surface quality, high resolution, and large volumetric throughput. The vast majority of materials obtained by VP are covalently-crosslinked thermosets with nondegradable carbon backbones. This highly crosslinked molecular structure gives rise to stiff and brittle materials, limiting the structural functionality in desired applications.
This thesis explores a variety of molecular structures for new VP photopolymers: a) dynamically-crosslinked compliant polymer, b) interpenetrating network (IPN) hydrogel, and c) covalently-crosslinked polymer with labile group (ex. ester) insertion to polymer backbone. With the dynamic crosslinking system, we demonstrate tunable mechanical behaviors of the metal-coordinated supramolecular polymers. These materials display a range of failure strain of 450% - 940% and ultimate tensile strength of 12.4 - 2.2 MPa with varying resin compositions. To incorporate multifunctionality, we design thermoresponsive IPN hydrogels by fabricating a hydrophilic host polymer network via VP and a subsequent formation a thermoresponsive 2nd network (poly(N-Isopropylacrylamide)). The architected IPNs consistently display strong polymer-liquid phase separation behavior and a tunable water release behavior with volumetric shrinkage between 30% and 70% upon heating at 50oC. Finally, to promote the degradability of the acylate-based photoresin, we demonstrated successful incorporation for ester functional groups into the polymer backbone via radical ring opening polymerization of cyclic ketene acetals. The obtained polymer undergoes 84% mass loss within 7 hours under hydrolytic degradation condition. Overall, we demonstrated VP as a powerful technique to achieve one-pot synthesis and fabrication of functional materials. Our explorations on the development of degradable photopolymers, thermoresponsive double-network hydrogels, and metal-coordinated supramolecular polymers provide valuable insights into the impact of resin formulation on mechanical properties. From analyzing the molecular weight of 3DP materials to finetuning of phase separation behavior and degradability, we demonstrate that VP provides a new platform to inspire advanced photoresin design strategies for desirable mechanical performance.</p
Unveiling Incipient Reactivity via Tandem Hydrosilylation Reaction Cascades and the Progress Toward the Total Synthesis of (–)-Cylindrocyclophane A
Get PDFThe two pillars of synthetic organic chemistry, reaction methodology development and total synthesis of complex natural products, has remained the focus of chemical research for synthetic chemists since their fundamental inception. In particular, harnessing the reactivity of unstable, but useful, chemical intermediates through telescoping reaction conditions is emerging as an attractive approach to rapidly access complex molecular architecture from readily available building blocks. Herein is described two unique reaction methodologies relying on tandem hydrosilylation reaction cascades to synthesis saturated N-heterocyclic products in a stereoselective manner. We have developed a diastereoselective Mannich reaction combining α-substituted-γ-lactam pronucleophiles with N-silyl imine electrophiles generated in situ via catalytic hydrosilylation of aryl nitriles. Additionally, we have developed a tandem hydrosilylation, enantioselective allylic alkylation reaction of substituted pyridines to yield chiral tetrahydropyridine products. This serves as the first example of using hydrosilylation of pyridines to generate enamine nucleophiles that can undergo an asymmetric allylic alkylation reaction. The final portion of this thesis describes the progress toward a total synthesis of (–)-cylindrocyclophane using C–H functionalization logic. We were able to access the necessary [7.7]-paracyclophane core in 8 steps from a feedstock aryl diazoacetate compound and n-hexene. Through functional group manipulations, we were able to advance this paracyclophane core to an intermediate possessing the exact stereocenters and carbon framework in (–)-cylindrocyclophane A. We are currently modeling the necessary deoxygenation needed to advance this intermediate and complete the total synthesis.</p
Engineering Artificial Systems with Natural Intelligence
Get PDFAlthough Deep neural networks achieve human-like performance on a variety of perceptual and decision-making tasks, they perform poorly when confronted with changing tasks or goals, and broadly fail to match the flexibility and robustness of human intelligence. Additionally, artificial neural networks rely heavily on human-designed, hand-programmed architectures for their remarkable performance. In this thesis, I work towards achieving two goals: (i) development of a set of mathematical frameworks inspired by facets of natural intelligence, to endow artificial networks with flexibility and robustness, two key traits of natural intelligence; and (ii) inspired by the development of the biological vision system, I propose an algorithm that can ‘grow’ a functional, layered neural network from a single initial cell, with the aim of enabling autonomous development of artificial networks akin to living neural networks.
For the first goal of endowing networks with flexibility and robustness, I propose a mathematical framework to enable continuous training of neural networks on a range of objectives by constructing path connected sets of networks, resulting in the discovery of a series of networks with equivalent functional performance on a given machine learning task. In this framework, I view the weight space of a neural network as a curved Riemannian manifold and move a network along a functionally invariant path in weight space while searching for networks that satisfy secondary objectives. A path-sampling algorithm trains computer vision and natural language processing networks with millions of weight parameters to learn a series of classification tasks without performance loss while accommodating secondary objectives including network sparsification, incremental task learning, and increased adversarial robustness. Broadly, for achieving this goal, I conceptualize a neural network as a mathematical object that can be iteratively transformed into distinct configurations by the path- sampling algorithm to define a sub-manifold of networks that can be harnessed to achieve user goals.
For the second goal of ‘growing’ artificial neural networks in a manner similar to living neural networks, I develop an approach inspired by the mechanisms employed by the early visual system to wire the retina to the lateral geniculate nucleus (LGN), days before animals open their eyes. I find that the key ingredients for robust self- organization are (a) an emergent spontaneous spatiotemporal activity wave in the first layer and (b) a local learning rule in the second layer that ‘learns’ the underlying activity pattern in the first layer. As the bio-inspired developmental rule is adapt- able to a wide-range of input-layer geometries and robust to malfunctioning units in the first layer, it can be used to successfully grow and self-organize pooling architectures of different pool-sizes and shapes. The algorithm provides a primitive procedure for constructing layered neural networks through growth and self-organization. Finally, I also demonstrate that networks grown from a single unit perform as well as hand-crafted networks on a wide variety of static (MNIST recognition) and dynamic (gesture-recognition) tasks. Broadly, the work in the second section of this thesis shows that biologically inspired developmental algorithms can be applied to autonomously grow functional ‘brains’ in-silico.</p
Co-Option of the piRNA Pathway to Regulate Neural Crest Specification
Get PDFThe piRNA pathway has persisted throughout evolution as an essential regulatory pathway to protect genomic integrity in the metazoan germline. It achieves this through the repression of transposable elements, or “selfish genes,” which would otherwise jump throughout the genome unchecked, causing genomic instability and infertility. While transposable elements are generally deleterious in nature, their persistence in our genome remains an important driver of evolution, both as an agent of mutation and source of raw genetic material. Thus, a delicate balance must be struck to both maintain genomic integrity for the next generation but still enable enough mutation to allow for adaptation. The arms race between the ever-adapting piRNA pathway and its transposon targets provides this balance, and for a long time the piRNA pathway was considered to be germline specific, since that is where both transposon and piRNA pathway activity is highest. It has since become clear that the piRNA pathway is also active in somatic tissue of several invertebrate species, and may even target host genes in some. Whether the piRNA pathway plays a role outside of the germline in vertebrates, however, has remained elusive.
In this thesis, we demonstrate that the piRNA pathway is active in a vertebrate somatic cell type, the chick neural crest, where it has been co-opted into the gene regulatory network to repress the transposon-derived gene, ERNI. ERNI, in turn, supresses Sox2 when piRNA pathway protein Piwil1 is downregulated upon neural crest specification. Thus, the piRNA pathway functions to maintain Sox2 expression in the neural plate border stem cell niche, protecting its proliferative abilities and setting the timing of neural crest specification. We also provide preliminary evidence that the neural crest piRNA pathway might be conserved in other vertebrate species, and that two highly conserved transcription factors regulate its expression in the chick neural crest. Our work provides mechanistic insight into a novel function of the piRNA pathway as a regulator of somatic development in vertebrates, and raises the possibility that this ancient pathway may play a more significant role in evolution and transposon co-option by host genomes than previously thought.</p
Electronic Correlations and Topology in Graphene Moiré Multilayers and InAs/GaSb-Derivative Systems
Get PDFTwisted bilayer graphene (TBG) near the magic angle exhibits a wide variety of correlated and topological phases such as superconductivity, correlated insulators, and orbital ferromagnetism. We show using electrical transport measurements that adding a layer of tungsten diselenide in proximity to twisted bilayer graphene stabilizes superconductivity to twist angles significantly below the magic angle despite the disappearance of correlated insulators and insulators at full moiré filling. These findings--along with our report of a relationship between superconductivity and symmetry breaking Fermi surface reconstruction--suggest constraints on theories of the origin of superconductivity in TBG. In the context of this TBG-tungsten diselenide system, we study how the correlated phases evolve over a wide twist angle range and classify them into a hierarchy based on where they occur relative to the magic angle (or where bands have been maximally flattened). While effects such as orbital ferromagnetism near one electron per moiré unit cell and gapped correlated insulators only exist in close proximity to the magic angle, superconductivity and high-temperature cascade transitions survive in a wider twist angle range.
We also analyze the structures of twisted trilayer, quadrilayer, and pentalayer graphene (and all proximitized to tungsten diselenide) near their respective theoretical magic angles, revealing robust electron- and hole-side superconductivity in each heterostructure. We additionally find previously unreported insulating states in twisted trilayer and quadrilayer graphene along with an enlarged filling range of superconductivity in pentalayer. Our studies on twisted graphene multilayers beyond two layers allow us to generalize the correlated physics found in TBG and consider the role of the additional bands introduced.
In the last part of this thesis, we measure the two-dimensional topological insulator candidate system InAs/GaSb with added stoichiometric impurities. Previous studies in pure InAs/GaSb structures have revealed low bulk resistivity and edge states that arise from trivial effects which can be easily mistaken for topological effects. Due, in part, to the strain effects of Indium impurities added to GaSb, our results show high bulk resistivity. We also, due to the wide gate-tunability in our devices, are able to measure the expected spin-orbit-split valence band structure. Our development of highly tunable InAs/GaSb-derivative structures paves the way for another look at two-dimensional topological insulator behavior in these systems and for their integration into superconducting devices.</p
Understanding the Origins of Photoexcited XUV Spectra
Get PDFA full measurement of photoexcited dynamics, from excitation to recombination, is required to understand the photochemical processes at the heart of solar energy materials and devices. Measuring these complete dynamics is often unachievable with a single experimental tool. Transient X-ray spectroscopies, however, have proven to be powerful techniques as they can separately measure electron and hole dynamics, as well as vibrational and structural modes, all with elemental specificity. The interpretation of these measurements is still challenging, as the core-hole created following a core-level transition distorts the measured spectrum. This thesis aims to develop complementary experimental and computational techniques to measure and interpret transient X-ray spectra. Initially, the measured photoexcited dynamics of ZnTe and CuFeO₂, which reveal polaron formation and lattice coupling, as well as electron and hole kinetics and band gap dynamics, are presented. Following this experimental work, we develop an ab initio computational method for modeling transient X-ray and extreme ultraviolet (XUV) spectra. The ab initio method is a Bethe-Salpeter equation (BSE) approach based on the previously developed Obtaining Core Excitations from Ab initio electronic structure and the NIST BSE solver (OCEAN) code. Building on the foundations of the OCEAN code, we incorporate photoexcited states for a range of transition metal oxides and demonstrate the method’s ability to simulate the effects of state filling, isotropic thermal expansion and polaron states on XUV absorption spectra. Importantly, our method is also able to fully decompose the calculated spectra into the constituent components of the X-ray transition Hamiltonian, providing further insight into the origins and nature of spectral features. The XUV absorption spectra for the ground, photoexcited, and polaron states of α-Fe₂O₃, as well as for the ground, photoexcited, and thermally expanded states of other first row transition metal oxides – TiO₂, α-Cr₂O₃, β-MnO₂, Co₃O₄, NiO, CuO, and ZnO – are calculated to demonstrate the accuracy of our approach. This method is easily generalized to K, L, M, and N edges to provide a general approach for analyzing transient X-ray absorption or reflection data.</p
Where the Wild Things Are: Computer Vision for Global-Scale Biodiversity Monitoring
Get PDFWe require a real-time, modular earth observation system that unites efforts across research groups in order to provide the necessary information necessary for global-scale impact in sustainability and conservation in the face of climate change. The development of such systems requires collaborative, interdisciplinary approaches that translate diverse sources of raw information into accessible scientific insight. For example, we need to monitor species in real time and in greater detail to quickly understand which conservation efforts are most effective and take corrective action. Current ecological monitoring systems generate data far faster than researchers can analyze it, making scaling up impossible without automated data processing. However, ecological data collected in the field presents a number of challenges that current methods, like deep learning, are not designed to tackle. These include strong spatiotemporal correlations, imperfect data quality, fine-grained categories, and long-tailed distributions. Our work seeks to overcome these challenges, and this thesis includes methods which can learn from imperfect data, systematic frameworks and benchmarks for measuring and overcoming performance drops due to domain shift, and the development and deployment of efficient human-AI systems that have made significant real-world conservation impact
Engineering and Rapid Prototyping for Biology in Extreme Conditions
Get PDFIn this thesis we show three projects in which biological systems are engineered for increased robustness to environmental stressors such as toxic small molecules. Several lignocellulose-derived growth inhibitors commonly found in industrial feedstocks for fermentation were used to grow a panel of yeast knockouts for several efflux pumps and detoxifying enzymes. Some specific knockout strains showed slowed growth on specific growth inhibitors, while other knockout strains showed the same growth rate as the wild-type. One efflux pump was identified for vanillin, YHK8, and was overexpressed in yeast. The overexpression strain did not show an improved tolerance to vanillin, and grew more slowly than the wild-type. To regulate the expression of the vanillin pump, a sensor for vanillin was created. The starting enzyme was the wild-type qacR transcription factor, and several variants were generated using computational protein design. The designs were synthesized and tested using in vitro transcription-translation (TX-TL) as part of a rapid prototyping process. This rapid prototyping considerably sped up the design-build-test process. Finally, four bacteria, Pseudomonas synxantha 2-79, Pseudomonas chlororaphis PCL1391, Pseudomonas aureofaciens 30-84, and E. coli are tested against the same lignocellulose growth inhibitors. The Pseudomonas spp. show an improved tolerance to the growth inhibitors. We then develop some ability to engineer and prototype in all three species. A panel of promoter parts were integrated into the P. synxantha genome to produce a collection of test strains. These same promoter parts were also used as DNA templates for TX-TL reactions. The in vivo measurements of promoter strength and in vitro measurements show similar relative strengths between the parts, showing the Pseudomonas-based TX-TL systems can be used for design-build-test activities in these non-model organisms. This alternate approach to developing tolerance, starting with a species that already has a working tolerance to the stressor in question, changes the problem to one of building engineering capabilities in the new chassis
Tracer Transport in Three Dimensions: Dispersion of Methane on Mars, Coupled Chemistry and Dynamics on Exoplanets, and Submesoscale Mixing in the Ocean
Get PDFOne-dimensional (1D) modeling from a horizontally averaged perspective can oftentimes greatly simplify problems in atmospheric and oceanic sciences and thus capture leading-order physics. Meanwhile, 1D numerical models have great advantages such as numerical stability and time efficiency, hence they are widely used to gain insights into complex problems. However, oversimplification by 1D models may cause failures in finding solutions, revealing novel phenomena, and discovering scaling laws in the three-dimensional (3D) real world, and those are when 3D thinking proves its value. Also, the rise in computational power has allowed investigations using 3D numerical models. This thesis discusses three examples of how 3D modeling transcends the limitations of 1D modeling and reveals new solutions, phenomena, and scalings in planetary atmospheres and Earth’s ocean.
Chapter 2 is focused on the dispersion of methane plumes on Mars and how it can reconcile the discrepancy between observations. In the face of ostensibly inconsistent observational results of methane on Mars, we adopt a novel approach—inverse Lagrangian modeling in 3D space—to find the scenarios in which the inconsistency in the observations can be reconciled and locate the methane source. We find that the inconsistency between the results of the near-surface in situ methane measurements and the satellite remote sensing measurements can be reconciled if and only if an active methane emission hot spot is located in the immediate vicinity of the Curiosity rover in northwestern Gale crater, or unknown physical or chemical processes are rapidly removing methane.
Chapter 3 presents a novel phenomenon that could exist on exoplanets—self-sustained photochemical oscillations, which is only produced by 3D atmospheric models. We use a 3D, fully coupled, chemistry-radiation-dynamics model to simulate the ozone-NOx-HOx photochemistry in the atmosphere of a tidally locked Earth-like exoplanet in the circumstellar habitable zone, and calculate the transmission spectra during transits. We find that under certain conditions, biological nitrogen fixation like the one on the Earth can drive large-magnitude, self-sustained photochemical oscillations in the atmospheres of terrestrial exoplanets. The resulting large temporal variability in ozone abundance on exoplanets, if observed, may suggest a strong surface NOx emission source, which could signal extrasolar life participating in the nitrogen cycle on exoplanets. Fully coupled, three-dimensional atmospheric chemistry-radiation-dynamics models can reveal new phenomena that may not exist in one-dimensional models, and hence they are powerful tools for future planetary atmospheric research.
Chapter 4 uses a 3D fluid dynamics model to study the vertical exchange in the upper part of Earth’s ocean that potentially has great implications for the marine ecosystem. We develop scaling laws for the exchange rate between the surface ocean and the ocean interior which is critical to the rate of nutrient supply to phytoplankton near the ocean surface. These scaling laws could substitute the crude 1D parameterizations that are currently widely used in ocean models. We find that submesoscale turbulence energized by baroclinic instability in the ocean mixed layer can induce tracer exchange between the surface ocean and the ocean interior. Various environmental physical parameters affect the exchange rate. The exchange is stronger where the ocean mixed layer is thicker, the Richardson number (defined as the ratio of the squared buoyancy frequency to the squared vertical shear of the horizontal flow) of the thermocline is smaller, and the Richardson number of ocean mixed layer is larger. The associated nutrient supply from the ocean interior to the surface ocean is also expected to be stronger under these conditions.</p
Reliable Learning and Control in Dynamic Environments: Towards Unified Theory and Learned Robotic Agility
Get PDFRecent breathtaking advances in machine learning beckon to their applications in a wide range of real-world autonomous systems. However, for safety-critical settings such as agile robotic control in hazardous environments, we must confront several key challenges before widespread deployment. Most importantly, the learning system must interact with the rest of the autonomous system (e.g., highly nonlinear and non-stationary dynamics) in a way that safeguards against catastrophic failures with formal guarantees. In addition, from both computational and statistical standpoints, the learning system must incorporate prior knowledge for efficiency and generalizability.
This thesis presents progress towards establishing a unified framework that fundamentally connects learning and control. First, Part I motivates the benefit and necessity of such a unified framework by the Neural-Control Family, a family of nonlinear deep-learning-based control methods with not only stability and robustness guarantees but also new capabilities in agile robotic control. Then Part II discusses three unifying interfaces between learning and control: (1) online meta-adaptive control, (2) competitive online optimization and control, and (3) online learning perspectives on model predictive control. All interfaces yield settings that jointly admit both learning-theoretic and control-theoretic guarantees.</p