12023 research outputs found
Sort by
Harvesting Insights from Advanced Microscope Acquisitions: Techniques and Applications
No full textSince their inception, microscopes have evolved significantly, becoming essential tools across various fields, from pathology diagnosis to biological studies. Morphological information that cannot be otherwise observed has always been regarded as the primary data a microscope could deliver. Yet microscopy data embodies further valuable information worth exploring. This thesis demonstrates extracting three types of information beyond morphology by modifying microscope systems, incorporating physical models, and applying image processing: 1) depth information, 2) object size information, and 3) object developmental information.
The first part of the thesis describes an all-in-focus technique based on Fourier Ptychographic Microscopy (FPM) for depth information extraction. It synthesizes an all-in-focus image and depth map from an FPM-reconstructed multi-focal image stack. This technique benefits thyroid fine needle aspiration samples, relieving pathologists from the need to constantly adjust focal planes, enabling convenient data transfer, and potentially aiding machine learning tasks on cytology specimens.
The second part of the thesis focuses on a non-destructive subvisible particle (SbVPs) analyzer for estimating size and concentrations of SbVPs in drug products. This analyzer aims to estimate the size and concentrations of SbVPs within a drug product while keeping the sample intact. Incorporating a light-sheet microscope with custom housings to compensate for container-induced astigmatism, it uses side-scattered light as a size indicator based on Mie scattering theory. Its functionality is demonstrated on polystyrene beads and biological drug products. Additionally, a new metric named the strip density is discovered from the same microscope images, which could serve as a more precise and robust size indicator beyond scattering light intensity. This new size indicator is used to train a particle detection neural network, verifying its effectiveness through good performance.
For the final part, we focus on an embryo sex classification project, aiming to extract subtle developmental differences between male and female embryos from early development videos taken by Embryoscope. A combined convolutional and recurrent neural network structure is employed. While the prediction accuracy reaches 61%, which is not high, the deep learning model outperforms both human and random predictions, demonstrating its ability to acquire embryo developmental information from the Embryoscope videos to some extent.</p
Strategic Planning and Sensitivity-Enhancing Tactics for Detecting Low-Mass Particle Dark Matter with Phonon-Mediated Detectors
Get PDFA non-baryonic matter beyond the framework of the Standard Model is required to explain a vast set of astrophysical and cosmological phenomena in our universe; it is referred to as dark matter and comprises 85% of all matter. Previously centered on particle candidates in the 1 GeV to 10 TeV mass range, dark matter model building has expanded to masses well beyond that range, with an emphasis toward low-mass particles below 1 GeV. Low-mass dark matter models have invoked new and creative mechanisms for producing the relic abundance of dark matter and in doing so have provided a variety of new laboratory-testable hypotheses about the early universe.
Direct detection experiments seek to directly measure a dark matter particle interaction from the Milky Way dark matter halo with ultra-sensitive detector technologies in low-background environments. As the paradigm has shifted toward lower-mass particle candidates, detector technologies have followed suit: single-charge-sensitive detectors and low-threshold, purely phonon-mediated detectors are among the best detector architectures for probing the most immediately accessible
theoretical models of low-mass dark matter. The Super Cryogenic Dark Matter Search (SuperCDMS) has used and developed detector technologies on both of these fronts. On the axis of single-charge-sensitive detectors, the High Voltage (HV) detector program of the SuperCDMS Collaboration has demonstrated gram-scale, single-charge sensitive detectors known as HVeV detectors. Recent advances in the collaboration’s understanding of single-charge backgrounds have enabled much improved sensitivity to low-mass dark matter parameter space with even these gram-scale detectors. HVeV detectors are a prototype version of the HV kg-scale detectors to be deployed at the flagship SuperCDMS experiment in SNOLAB. HV detectors are projected to test vast regions of unconstrained parameter space for both electron- and nuclear-recoiling dark matter, as shown in Chapter 4 of this thesis among many other SuperCDMS sensitivity projections.
A potentially limiting background for SuperCDMS detectors at SNOLAB is the zero-charge low energy excess, which is characterized by an exponentially rising spectrum of background phonon events below about 100 eV to 1 keV recoil energy. Chapter 5 of this thesis presents a data-driven technique to subtract the zero-charge low energy excess (0QLEE) as observed in HVeV detectors. A search for charge-producing, nuclear-recoiling dark matter is performed with this background-subtraction technique. The resultant exposure-limited constraint on the nucleon-dark-matter cross section is nearly a factor of 10× stronger than the background-limited constraint and is within tens of percent from unconstrained parameter space. The two dominant sources of systematic uncertainties for this search are (1) the uncertainty on the total rate and spectral shape of zero-charge low energy excess events and (2) the completely unknown behavior of the ionization yield function in silicon for nuclear recoils below 100 eV.
On the axis of low-threshold phonon-mediated detectors, the SuperCDMS Collaboration must improve phonon energy thresholds to below 1 eV in order to attain sensitivity to sub-GeV nucleon-coupled dark matter. Presently, SuperCDMS has achieved detector phonon thresholds in the range from 10 eV to 200 eV depending on detector size. In Chapters 6, 7, and 8, we present a radically different phonon sensor architecture that may provide long-term gains in sensitivity: the kinetic inductance detector.
There are two main quantities that constrain the capacity of kinetic inductance detectors to be effective phonon sensors: the detector readout noise and the phonon collection efficiency. Chapter 7 explores the former, detailing the variety of noise sources in kinetic inductance detectors and how they may impact a sensor’s energy resolution using both theoretical calculations and experimental measurements. In general, resolution on energy absorbed in the sensor is presently limited to a range from 1 eV to 5 eV. Chapter 8 then reports on the overall detector energy performance of three different KID-based phonon-mediated (KIPM) detectors, each of which suffers from percent-scale phonon collection efficiencies. An empirical model is then built to parametrize and understand the reasons for the poor phonon collection efficiencies, thereby outlining a path forward to lowering energy thresholds in KIPM detectors.</p
Scalable On-Chip Platforms for Quantum Microwave-Optical Interface with Solid-State Ensembles
No full textSuperconducting quantum circuits based on Josephson junctions are one of the most promising platforms for future quantum information processing. Tens of superconducting quantum bits have been integrated on a single chip with performances exceeding the most advanced classical computers. However, these new quantum machines operate at microwave frequencies, which have enormous thermal noise and photon loss at room temperature. This fundamentally limits the future application of this technology in distributed quantum computing and quantum networks. Conversely, optical photons are an ideal information carrier as the photon loss is extremely small in fibers and the thermal noise is negligible at room temperature. Therefore, a quantum transducer that converts between microwave and optical frequencies at the single-photon level is of great importance.
This thesis is centered on building such chip-scale interfaces with rare-earth ion (REI) doped crystals. First, we focus on developing a theoretical understanding of microwave-to-optical transducers. Based on coupled mode theories, we derive a clean theoretical result of the on-resonance transduction model. This allows us to condense the relevant material properties for transduction into a single parameter, effective χ⁽²⁾, describing the strength of the non-linearities provided by the rare-earth ion materials. Next, we designed, fabricated, and measured the chip under cryogenic temperatures, where percent-level efficiency and single-photon level of added noise referred to the input is achieved. To further demonstrate the unique advantage of atom-based platforms, we perform two transducer interference experiments, showing the scalability and capacity towards transducer-assisted remote entanglement of superconducting quantum bits. Lastly, with large microwave cooperativities achieved, we observe novel quantum electrodynamics enabled by controllable initialization of the excited-state spin system. By initializing the spins into spin-down and spin-up states, we observe collectively induced transparency and periodic superradiant emissions, respectively. Simulations are developed to explain the experimental results.
These results establish REI doped crystals as a highly competitive platform for microwave-optical quantum interfaces and pave the way toward remote transducer-assisted entanglement of superconducting quantum machines.</p
Advanced Nano Manufacturing Enables Probing Fundamental Mechanical Behaviors of Materials
Get PDFThe trend of miniaturization has revolutionized modern technologies, with micro- and nanoscale materials driving transformative advancements in high-tech industries and scientific discovery. Among the various properties and applications enabled at these small scales, nanomechanical properties play a fundamental role, underpinning the integrity and functionality of any structures or systems. However, despite advancements in both conventional and emerging micro- and nano-manufacturing strategies, there has remained a lack of direct “bottom-up” experimental pathways to fabricate and probe the mechanical responses of submicron-sized monolithic nano-specimens with unconventional microstructures and/or 3D nano-architectures with submicron-sized features, particularly for non-carbon materials.
In this work, I will present novel nano-fabrication and manufacturing strategies and their applications in addressing these nanomechanical challenges through three key studies. In Chapter 2, the deformation characteristic of organic ice is studied via cryogenic micro-compression and molecular dynamics simulations, providing insights into a benzene-ring re-orientation-mediated densification deformation route and offering new insights into planetary geology for celestial bodies such as Titan. In Chapter 3, we experimentally unveiled unprecedented two-regime size effects in additively manufactured metallic nanopillars with hierarchical microstructures, revealing a nanocrystallinity-, nanoporosity-mediated plasticity mechanism through atomistic insights. In Chapter 4, we extended this nano-manufacturing approach to explore nanoporosity-driven deformation behaviors in nano-architected metals with in situ experiments and finite element analysis. Together, these studies not only elucidate previously unprobed fundamental small-scale mechanical behaviors but also lay the groundwork for developing an advanced micro-to-nanoscale manufacturing platform, enabling complex systems and functional applications such as energy storage, biomedical microrobots, nanophotonics, and beyond, which I will briefly discuss in Chapter 5 as an outlook with a few examples from metal/oxide nanocomposites to interpenetrated pyrolytic carbon microarchitectures.</p
Localized Catalytic DNA Circuits for Integrated Information Processing in Molecular Machines
Get PDFThis thesis supports the long-term goal of engineering molecular devices with computational complexity akin to cells. Like cells, artificial molecular devices can benefit from integrating multiple computational modalities.
To that end, this thesis advances molecular computing systems in three modalities: dynamic molecular assembly, well-mixed circuits, and spatially-organized cascades. Specifically, it introduces methods to enhance control over DNA structural assembly, well-mixed DNA circuits, and DNA circuits localized to a DNA origami surface.
As DNA structural assembly grows increasingly complex, so too grows the potential for off-target structures. This issue can be addressed through developmental self-assembly, where components join a growing structure in a programmed sequence under controlled kinetics. The scope of developmental self-assembly is here expanded by a method enabling specific pathway selection among multiple encoded options.
Well-mixed DNA circuits require catalytic motifs for signal restoration and amplification. A catalytic motif is presented where two input strands cooperate to control catalysis. This motif could enhance AND gates and thresholding, and could enable adaptive memories and learning behaviors in DNA-based neural networks.
Localized DNA circuits lack cascadable catalytic mechanisms for signal restoration and amplification. Two designs for a localized catalytic mechanism are presented. Each omits any intermediate diffusible species to support nanodevices compatible with uncontrolled environments, as in biomedical contexts. This constraint leads to design lessons; principally, we respond to leak in the first design through geometric constraints in the second design.</p
Bridging Space and Time: Resolving the Temporal Dynamics of the Seminiferous Epithelial Cycle Using Spatial Transcriptomics
Get PDFBiology is inherently spatial, with tissue architecture and cell–cell interactions shaping dynamic developmental and homeostatic processes. In this thesis, we harness high-resolution spatial transcriptomics via RNA seqFISH+ to show how spatial information can be used to resolve temporal information in complex tissues, using adult mouse spermatogenesis as a model. By profiling 2,638 genes in over 216,000 cells, we find that each seminiferous tubule cross-section represents a distinct timepoint of the seminiferous epithelial cycle, and collectively all tubules form a circular topology in gene expression space that precisely aligns with the known 12-stage progression. Intriguingly, Sertoli cells exhibit a robust cyclic transcriptional program synchronized with germ cell differentiation, raising the question of whether this cycle is driven solely by germ cells or whether Sertoli cells display an intrinsic cyclic expression profile. To address this, we ablate differentiating germ cells using a DNA alkylating agent, busulfan. In this model, despite the lack of differentiating germ cells, Sertoli cells maintain much of their cyclic expression suggesting an autonomous cycle that partially dephases without germ cell input. Integrative analyses suggest that the underlying mechanism of this oscillation may involve an innate retinoic acid metabolic cycle and/or an interconnected transcription factor network. Finally, we discuss how these findings broaden our understanding of tissue processes and propose that spatial transcriptomics can be adopted to reconstruct temporal dynamics for many tissues from static snapshots
The Landscape of Stellar Mergers with Time-Domain Surveys
Get PDFStellar mergers result in a wide range of outcomes, from luminous cosmic transients to peculiar variable stars. Mergers provide valuable insights into a broad range of astrophysical phenomena, from stellar evolution to gravitational waves to the origins of the universe's heaviest elements. In this thesis, I systematically explore the outcomes of merging white dwarfs, merging neutron stars, and merging massive stars, using robotic survey telescopes to fill gaps in our understanding of the diverse merger landscape.
In Part I of this thesis, I present the first infrared census of dusty variable stars formed from low-mass white dwarf mergers. This population offers new insights into the binary white dwarfs that will be detected by the upcoming Laser Interferometer Space Antenna (LISA). I also present the least luminous thermonuclear supernova discovered to date, which possibly originated in the merger of two massive white dwarfs.
In Part II of this thesis, I present the data processing pipeline of the novel Wide-field Infrared Transient Explorer (WINTER) surveyor at Palomar Observatory, designed for infrared followup of gravitational wave events from neutron star mergers. I present results from WINTER’s first search for an infrared counterpart to a neutron star merger recently detected by the International Gravitational Wave Network.
In Part III of this thesis, I present the first systematic study of extragalactic transient eruptions from massive stellar mergers and estimate their volumetric rate and luminosity function. I also present the first infrared observations of such mergers with the James Webb Space Telescope, which suggest that stellar mergers could be significant contributors to the cosmic dust budget. Additionally, I present a slow-evolving infrared transient identified by WINTER that originated in a merger involving a giant star primary, revealing a new class of events that have been overlooked by previous optical surveys.
Together, these studies set the stage for more comprehensive explorations of the merger landscape in the future, with i) the Vera Rubin Observatory to study large populations of low luminosity transients from massive stellar mergers and white-dwarf mergers, and ii) the upcoming suite of ground and space-based infrared surveys to discover the dustiest stellar mergers and quantify their contributions to the cosmic dust budget.</p
Computational Complexity and Quantum Gibbs Sampling for Local Hamiltonians
Get PDFOne of the primary motivations for building quantum computers is to simulate quantum many-body systems. While significant progress has been made in simulating quantum dynamics, much less is known about simulating ground states and Gibbs states, an essential task for understanding the static properties of quantum many-body systems. From a computer science perspective, problems on ground states and Gibbs states are quantum analogues of the Boolean satisfiability problem (SAT) and classical Gibbs sampling, which have wide applications in optimization, machine learning, and computational complexity.
This thesis leverages tools from computer science to explore the potential quantum advantage in simulating ground states and Gibbs states, through two complementary approaches: designing new quantum algorithms and evaluating the extent to which classical algorithms remain effective. In particular,
Quantum Gibbs sampling. In the first part, we describe our progress in developing quantum algorithms for preparing quantum Gibbs states. For general Hamiltonians, we develop a quantum analogue of the Metropolis-Hastings algorithm that is both conceptually simple and provably correct, with the Gibbs state as its approximate unique fixed point. Note that generalizing the Metropolis-Hasting algorithm to the quantum setting is non-trivial due to the unclonability of quantum states. Additionally, for a broad class of commuting Hamiltonians, we propose a different approach which constructs efficient quantum Gibbs samplers by leveraging reductions to existing classical sampling algorithms.
Sharpening the understanding of classical algorithms. In the second part, we present new complexity results to deepen our understanding of the capabilities of classical algorithms for ground energy estimation. The potential quantum advantage in solving many-body systems stems from the sign problem in general Hamiltonians, which classical algorithms struggle to handle. We give rigorous evidence to show that under certain conditions, widely used classical methods, such as fixed-node Monte Carlo and tensor network contraction, may overcome this barrier and effectively resolve the sign problem.
</p
Robust Gravitational Wave Analysis at the Catalog Scale
No full textThe rapid improvement in the sensitivity of ground based gravitational wave detectors has produced a huge variety of technical insights, but has also brought new challenges in gravitational wave data analysis. In this dissertation I address two of those challenges: the rapid increase in the number of detected events, and the need for robust astrophysical inferences in the presence of transient detector glitches. To manage the number of gravitational wave transients now regularly detected, I developed infrastructure for the LIGO-Virgo-KAGRA collaboration which monitors and collates the results of many disparate analyses in order to produce the final transient catalog. I implemented physically informed models for scattered light glitches into standard parameter estimation tools, and so that the potential realizations of these glitches can be marginalized over when performing astrophysical inference. This method was used to better understand GW191109, an event from the third observing run with potentially dynamical formation history. These tools were also applied to better understand the behavior of parameter estimation in the presence of glitches, and to search for statistical tests which can identify if parameter estimation is biased by the presence of a glitch
Nanophotonic Engineering of Thermal Emitters
No full textThermal emission is our most ubiquitous light source, as all objects with non-zero temperature emit this type of radiation. Consequently, our ability to shape the spectral and directional properties of thermally emitted and absorbed light by structures is both intriguing at a fundamental level and has practical implications for infrared light sources, radiative cooling, and energy harvesting systems. To impart desired properties to emitted radiation, nanophotonic designs where subwavelength features are patterned into structures have proved effective in preliminary demonstrations of engineered nanoscale control of thermal emission.
In this thesis, we leverage nanophotonic designs to demonstrate new phenomena in the context of thermal emission. We first use a guided-mode structure made of α-Si to resonantly couple to magneto-optically active InAs. The magneto-optic response is a common effect used in nonreciprocal optical elements, which we use here to directly observe a violation of the Kirchhoff thermal radiation law, a strict equality in the spectral, directional absorptivity and emissivity. This demonstration is significant in two ways: first, it opens new avenues to design thermal emitters with distinct spectral, directional emissivity and absorptivity properties, and second, it confirms theoretical predictions which have long lacked experimental confirmation.
We then extend this experimental Kirchhoff violation to a broadband, directive thermal emitter. The nanophotonic design to achieve this is a deeply subwavelength structure of gradient epsilon-near-zero InAs layers that couple to a Berreman mode. The angular selectivity is determined by the stack thickness, while the broadband spectral range of the effect is imparted by the closely spectrally separated epsilon-near-zero wavelengths.
Finally, we theoretically and experimentally lay the groundwork for a thermal lens, where emitted radiation is directed to a focus a given distance above the surface of the structure. Using a combination of coupled dipole approximation, global optimization, and experimental measurements, we realize the necessary collective and local resonance conditions for this effect.</p