21090 research outputs found
Sort by
Electroluminescence in the Classical and Quantum Regime in Undoped GaAs/AlGaAs Heterostructures
No full textQuantum information processing holds the promise to radically change the way we perform computations and transmit information. In the realm of quantum computing, there has been enormous progress in the last few decades in a huge variety of quantum systems and it is unclear which platform will be the leading system to execute quantum computations. Conversely, photons have always remained the front-runner for the long distance transfer of quantum information since photons travel at the speed of light and have limited mechanisms of decoherence (as compared to other carriers of quantum information) when traveling over long distances. The method used to generate single photons remains the pertinent open question.
Current state-of-the-art single-photon sources (SPSs) are optically-active quantum dots driven by an external laser source. For laboratory-scale experiments, they have proven fruitful in order to demonstrate key components of a quantum network, as well as performing fundamental tests on the nature of quantum mechanics. However, one challenge associated with these optically active quantum dots is two-qubit interactions since the quantum dots are usually spatially isolated. Conversely, two-qubit interactions for spin qubits in gate-defined quantum dots is routinely achieved via the Heisenberg exchange interaction. Thus, it would be highly desirable to have a way to convert the quantum information of the spin state of gate-defined quantum dots to photon polarization. Furthermore, for the prospects scaling of the technology, it would be highly desirable for this quantum information transfer to be all-electrical in order to leverage conventional multiplexing techniques.
In the first part of this thesis, we outline our proposal for an all-electrical SPS where single-photon emission is driven by electroluminescence (EL) at the single-charge to single-photon level. In order to control carriers at the single-charge level, we propose using non-adiabatic single-electron pumps (SEPs) previously investigated as quantized current sources for metrology. We have also previously developed a lateral p—n junction whose geometry allows direct integration with a SEP. We compare our proposed SPS to existing electrically-driven SPS in the literature, highlighting anticipated strengths of our proposed device, including a fabrication process compatible with standard semiconductor fabrication techniques.
Given the key role SEPs play in our proposed SPS, we describe the established theory underpinning the high fidelity operation of SEPs. We also highlight practical considerations for the operation of SEPs, including device fabrication challenges faced during the course of this research, and demonstrate how to measure and characterize a SEP.
A secondary focus of this thesis has been investigating EL from lateral p—n junctions in regimes where there was no attempt to control carriers at the single-charge level. While measuring lateral p—n junctions, we noticed an unconventional form of EL that did not require a forward bias to be applied. By swapping the polarity of the top gate voltage of our ambipolar induced devices, existing carriers recombine radiatively with incoming carriers of the opposite charge. Due to the flow of carriers in and out of the device, we called this form of luminescence the tidal effect. We develop a model to explain the non-monotonic frequency-dependent EL intensity and perform temperature-dependent measurements to identify the species responsible for the observed EL.
We also further investigate a similar phenomenon when two adjacent top gates are periodically swapped with a phase difference between the two signals. We demonstrate that this form of EL is more efficient over larger areas than the tidal effect, and therefore may be more suitable for general illumination purposes.
Lastly, we also performed the first EL measurements from lateral p—n junctions in single heterojunction interfaces. Despite the lack of a bottom barrier in these devices, our measurements suggest that carrier recombination is occurring near the interface. We characterize the EL spectra and observed the so-called H-band, a type of space-indirect exciton created in proximity to a populated single heterojunction interface, which has only previously been observed in photoluminescence experiments. Time-resolved EL experiments suggest reduced dimensionality of neutral excitons. We show that the lifetime of the H-band can be tuned electrically. We also demonstrate that the tidal effect can also be observed in these single heterojunction interfaces
Hybridizable discontinuous Galerkin methods for coupled flow and transport systems
No full textIn this thesis, we propose and analyze hybridizable discontinuous Galerkin methods for coupled flow and transport systems. Such systems may be used to model real-world scenarios in which a fluid contaminant travels through another medium. Common applications include environmental engineering problems and biochemical transport.
This thesis focuses on the Stokes/Darcy-transport and Navier--Stokes/Darcy-transport systems. We consider a two-way coupling between each flow and transport problem: the solution to the flow problem is directly involved in the transport problem, and the solution to the transport problem appears in the flow problem through a parameter function. In each of our considered systems, the flow problem is stationary while the transport problem is time-dependent. The resulting coupled flow and transport systems are quasi-stationary in the sense that the evolution of solutions to the flow problems over time is driven by the transport problem.
Our numerical schemes use a time-lagging method in which the flow and transport problems are decoupled and solved sequentially using hybridizable discontinuous Galerkin methods. This decoupling allows us to establish separate conditions on the discrete flow problem and on the discrete transport problem such that solutions to the combined scheme converge at optimal rates. Moreover, we show how existing results on related discrete flow problems and on the discrete transport problem may be exploited for efficient analysis of the coupled systems. We present this approach in a general setting, and illustrate its use through the specific examples of the Stokes/Darcy-transport and Navier--Stokes/Darcy-transport systems.
For all schemes, we establish the existence of unique numerical solutions over a considered time interval. We prove optimal rates of convergence in space and time, and provide numerical examples to support the theory
Semantic Segmentation of LiDAR Point Clouds for 3D Mapping of Underground Space
No full textUnderground space is among the most challenging environments for 3D mapping because the Global Navigation Satellite Systems (GNSS) signals are often inaccessible. This thesis investigates the use of the LiDAR-based Simultaneous Localization and Mapping (SLAM) technology to map such underground space. Underground parking lots, as an example, offer valuable solutions to the challenges posed by growing populations and urbanization, such as limited surface area, traffic congestion, and environmental concerns. They are GNSS-denied, geometrically repetitive, highly occluded by vehicles and pillars, and contain large, low-texture and specular surfaces that degrade sensing and registration. To support rigorous evaluation under these conditions, this thesis contributes three site-specific underground parking datasets captured using a hand-held LiDAR device, GeoSLAM. Each dataset provides clean point clouds and semantic labels for the core structural and operational classes: wall, pillar, vehicle, and ground, enabling controlled benchmarking.
Since low-cost LiDAR scans yield sparse, non-uniform point distributions that omit fine structural features, the first study of the thesis addresses point cloud upsampling, an essential step for creating high-definition maps that preserve fine structural details while ensuring uniform data distribution for downstream tasks. Five deep learning upsampling models including PU-Net, PU-GAN, PU-GCN, PU-Transformer, and RepKPU are trained and tested in a unified pipeline and evaluated with Chamfer Distance for average surface fidelity, Hausdorff Distance for worst-case deviation, and inference time for deployability. RepKPU consistently delivers the best accuracy–latency trade-off in underground setting.
Since accurate semantic understanding is crucial for structure-aware mapping and autonomous navigation in complex indoor environments, the second and third studies target semantic segmentation for underground parking spaces, first using Transformer-based backbones and then extending the evaluation to Mamba-based architectures. For Transformer-based methods (PT, PCT, and 3DGTN), the generalization across the three different parking lots is assessed using overall accuracy (OA), mean Intersection over Union (mIoU), and F1-score. The results establish 3DGTN as the most accurate and stable Transformer framework across all three sites. Complementing the Transformer study, Mamba-based methods (PointMamba, PoinTramba, and 3D-UMamba) are compared on the same datasets with 3D-UMamba offering the best overall performance
Vertex models for the product of a permuted-basement Demazure atom and a Schur polynomial
No full textThis thesis is about a manifestly positive combinatorial rule for the expansion of the product of two polynomials: Schur polynomials and permuted-basement Demazure atoms. Special cases of the latter polynomials include Demazure atoms and characters; there are known tableau formulas for their expansions when multiplied by a Schur polynomial, due to Haglund, Luoto, Mason and van Willigenburg (2011). We find a vertex model formula, giving a new rule even in these special cases, extending a technique introduced by Zinn-Justin (2009) for calculating Littlewood–Richardson coefficients.
We derive a coloured vertex model for permuted-basement Demazure atoms. This model is inspired by Brubaker, Buciumas, Bump and Gustafsson's model for Demazure atoms (2021) and Borodin and Wheeler's model for permuted-basement nonsymmetric Macdonald polynomials (2022). We make this model compatible with an uncoloured vertex model for Schur polynomials, putting them in a single framework. Unlike previous work on structure coefficients via vertex models, a remarkable feature of our construction is that it relies on a Yang–Baxter equation that only holds for certain boundary conditions. However, this restricted Yang–Baxter equation is sufficient to show our result
Interface Shear Behaviour and Modelling of Ultra-High Performance Concrete
No full textThe use of ultra-high-performance concrete (UHPC) in bridge construction has expanded considerably in recent years due to its exceptional mechanical properties, particularly its high compressive strength and superior post-cracking tensile resistance and deformation capacity. One particularly common and critical application of UHPC is its use in the connections of prefabricated structural components. Therefore, understanding UHPC shear resisting performance is essential.
To investigate the broadly varying and complex shear stress conditions that can develop in connection regions of modern concrete structures, an experimental program involving UHPC push-off specimens subjected to combined shear and lateral loading was performed. Comparisons of existing shear strength estimation procedures and the applicability of classical concrete failure criteria to unreinforced UHPC interfaces was examined. The findings provide insight into the shear transfer mechanisms of unreinforced UHPC interfaces under varied stress conditions, clarify the influence of external loading on UHPC interface shear strength, and provide a basis for refining design models for UHPC structural connections. Dog-bone direct tension tests were also conducted to investigate the tensile behaviour of UHPC, given the substantial influence of tensile properties on interface shear response. The results demonstrated that UHPC casting volume/size have a significant effect on the measured tensile properties, and therefore would influence the interface shear behaviour.
Finite element modelling of UHPC interfaces was performed to validate a proposed smeared-crack modelling approach in which UHPC was incorporated using an adapted tensile behaviour model originally developed for steel fibre-reinforced concrete (SFRC). Comparisons between experimental and numerical results were conducted to validate the accuracy of the proposed modelling approach for UHPC. The findings demonstrate that incorporating a user-defined UHPC, DEM-calibrated, tensile constitutive model calibrated enhanced the predictive capability for UHPC interface behaviour under shear-dominated multiaxial loading
Efficient Learning for Large Language Models
No full textArtificial Intelligence (AI) systems have become indispensable across domains such as healthcare, finance, robotics, and scientific discovery. At the heart of this revolution, Large Language Models (LLMs) have emerged as the central paradigm, demonstrating remarkable reasoning, generalization, and multi-domain adaptability. However, their exponential growth in scale introduces severe computational bottlenecks in training, fine-tuning, and inference, limiting accessibility, sustainability, and real-world deployment.
This dissertation advances the efficiency of LLMs across all lifecycle stages by introducing a suite of five frameworks that significantly reduce compute, memory, and latency costs with minimal or no loss in accuracy. First, Quantized Dynamic Low-Rank Adaptation (QDyLoRA) enables memory-efficient fine-tuning across multiple LoRA ranks in a single training pass, achieving competitive performance to QLoRA while reducing GPU memory usage by up to 65% and supporting flexible rank selection at inference time. Second, Sorted-LoRA introduces a stochastic depth–aware fine-tuning framework that co-trains multiple sub-models of varying depths within a single cycle. On LLaMA2–7B, it produces submodels up to 40% smaller that retain over 98% task accuracy, with the largest variant even surpassing the base model by +0.34%. Third, LoRA-Drop accelerates autoregressive inference by dynamically substituting computationally redundant layers with lightweight low-rank modules during decoding. It delivers up to 2.6× faster decoding and a 50% reduction in KV-cache memory with less than 0.5% degradation in accuracy, offering latency-aware adaptability for real-world deployment. Fourth, EchoAtt exploits redundancy in attention maps by sharing attention matrices among similar layers. On TinyLLaMA–1.1B, it achieves 15% faster inference, 25% faster training, and a 4% parameter reduction while improving zero-shot accuracy, highlighting that structural compression can enhance rather than degrade model generalization. Finally, ECHO-LLaMA introduces cross-layer Key–Value (KV) and Query–Key (QK) sharing to reduce redundant attention computation. This approach achieves up to 77% higher token-per-second throughput during training, 16% higher Model FLOPs Utilization (MFU), and 7% higher test-time throughput, while preserving language modeling performance. On the mechanical-domain RoboEval benchmark, ECHO-CodeLLaMA-7B boosts average accuracy from 62.15% to 63.01% with only 50% KV sharing, confirming its robustness in domain adaptation.
Together, these contributions form a coherent research program on the efficiency of large-scale Transformers. They demonstrate that intelligently exploiting representational redundancy—through quantization, low-rank structure, cross-layer sharing, and adaptive computation—can yield substantial compute savings with minimal trade-offs
A Speculative Design Exploration of Voice User Interfaces to Support Storytelling Among Older Adults
No full textReminiscing and more broadly, storytelling is an inherent part of what it means to be human. We reminisce through stories to form connections, share life wisdom, or remember significant events in our lives. Current technologies can act as a catalyst for documenting and sharing various life stories. Older adults have lifetimes of stories to tell, though they have unique and diverse needs in using advanced technologies. This research aims to understand how older adults view current Voice-User Interfaces (VUIs) such as Voice Agents (VAs) and speculate on their future role in supporting storytelling through reminiscence.
Using semi-structured interviews involving speculative design and questions from Dignity Therapy (DT), nineteen older adults aged 70-96 shared their experiences with storytelling, reminiscence, and using technology. Reflexive thematic analysis brought forward overarching themes of autonomy and agency in technology usage, understanding storytelling as a learned skill, and connections to memory, and meaning-making. While most VUI technologies focus on supporting care or assistance, the themes from this work
help reframe voice technologies as a potential tool for narration led by older adults, for other older adults. In speculating on the upper limits of current VUI capabilities and the seemingly endless potential of Artificial Intelligence, participants also call for simplicity and high utility value needed to adopt a new technology into daily life. This offers insights into creating dignified and meaningful interactions around reminiscence in this final stage of life
UringCL: A Lightweight io_uring Convergence Layer for Adoption in Legacy Event Loops
No full textHigh-performance network servers depend on efficient I/O mechanisms to manage thou-
sands of concurrent connections with minimal latency and overhead. While traditional
readiness-based interfaces (e.g., select, poll, epoll) notify applications when I/O oper-
ations can proceed, they still require synchronous system-calls to execute the operations.
This synchronous requirement causes frequent user–kernel transitions, which limits scala-
bility under heavy load. In contrast, the io uring interface offers a fundamentally different
approach by providing a completion-based I/O model that minimizes system-call overhead
and enables true asynchronous data transfer.
Although the performance benefits of io uring are well established in storage systems,
its integration into high-throughput network applications remains limited. This thesis aims
to bridge this integration gap by making the adoption of io uring accessible and provid-
ing a structured vehicle for evaluating its performance in network-bound environments. To
this end, the io uring Convergence Layer (UringCL) is presented to transparently map
the synchronous I/O calls of readiness-driven applications onto asynchronous io uring
operations. The UringCL simplifies initialization, event handling, and data transfer while
preserving the existing control flow of legacy applications, allowing for incremental migra-
tion toward completion-based I/O without major redesign.
The UringCL architecture facilitates the practical integration of io uring into estab-
lished network architectures and provides a consistent framework for measuring its impact
on throughput, latency, and CPU efficiency. Experimental results demonstrate significant
performance advantages over traditional models. In bulk-transfer workloads, the system
delivers up to 40% higher throughput than epoll due to superior batching capabilities.
In request-response scenarios involving Memcached, the integration achieves higher peak
throughput and maintains significantly lower and more stable tail latency under heavy
load. Furthermore, UringCL achieves these benefits with negligible overhead, proving that
completion-based I/O can be adopted seamlessly to enhance the efficiency of modern net-
work servers
Leak Detection and Localization in Water Distribution Networks
No full textLeaks in water distribution networks remain a significant challenge for utilities, resulting in substantial economic and environmental losses and health risks. However, existing leak detection and localization approaches face several shortcomings, including (i) limited understanding of how algorithms generalize across different networks, (ii) limited adoptability of empirical characterization of dispersive wave behavior in water-filled pipes, and (iii) heavy dependence on cross-correlation methods when performing leak localization, failing if leaks are not located on a direct sensor-to-sensor path.
This thesis addresses these gaps using machine learning-driven leak detection and localization techniques using hydrophone time-series data. First, I introduce structured frameworks for leak detection and leak localization algorithms, which define the key processing stages from signal collection to post-processing. To evaluate the ability of leak detection algorithms to generalize across different networks, I present a novel leak detection dataset collected from three real-world water distribution networks, and propose two evaluation schemes - Cross-Domain F1 Scoring and Multi-Domain F1 Scoring. Using these schemes, over 33,000 leak detection models were evaluated by varying modeling parameters, revealing that certain transformation techniques and low-frequency energy-based features (e.g., 62–124 Hz energy vs. 0–500 Hz centroid) can yield up to a 37% higher mean cross-domain F1 score. Further, I found that when sufficient training data are available, convolutional neural networks generalize better than hand-crafted-feature-based algorithms, achieving a multi-domain F1 score of 0.87 compared to 0.72 for exhaustive feature selection and 0.50 for simple feature selection when eight unique leak scenarios were included in the training data.
Next, I characterize wave propagation in a controlled lab-scale system and experimentally demonstrate dispersive shell-borne surface waves traveling at approximately 291 m/s, waterborne plane waves at 350 m/s, and high-velocity ultrasonic waves traveling at approximately 1,300 m/s. I show that analytical models that predict wave speed can be inaccurate by up to 16%, and that waves traveling along the shell wall exhibit dispersive behavior, which poses problems for traditional cross-correlation-based leak localization methods.
The viscothermal wave equation is implemented using the finite difference method to explore how spectral features correlate with leak proximity. These findings motivate the use of spectral features such as energy and centroid for predicting leak proximity. I then propose a novel leak localization algorithm that produces a heat map describing the probability of leakage along each point in a pipe network. The algorithm achieves reliable leak localization results, even in leak scenarios where conventional cross-correlation cannot be used. Calibration is shown to improve leak proximity regression performance by more than 53%, and the approach reliably localizes leaks within 3.66 m across leak scenarios not included in its training data, even in scenarios where traditional cross-correlation-based methods cannot be used.
Overall, my thesis contributes new datasets, quantitative evaluation methods, numerical modeling, insights into wave behavior, and learning-based algorithms that together advance the development of deployable and generalizable leak detection and localization systems
Duped by Dream Sellers: A Case Study of Student Immobility, Precarity, and Profit in Northern Cyprus
No full textYoung migrants arrive in Northern Cyprus seeking opportunity and safety through international higher education. Instead, they find themselves in a state of continuous precarity as financial benefactors sustaining complex legal liminalities of a de facto state. What happens when systemic marketing of an affordable, internationally recognized education and work opportunities targets individuals from countries in active war, extreme poverty, and political unrest? With safety concerns in their home countries, low-ranking passports, and limited international options, student migrants continue to arrive in Northern Cyprus despite its difficult living conditions.
Drawing on 29 qualitative interviews, this thesis examines how student migration both responds to and sustains the political and economic structures of Northern Cyprus. It shows how education, labour, and legality intertwine to produce a system that depends on students’ presence and their restricted mobility. By situating student migration within the political economy of an unrecognized state, this thesis contributes to empirical research on the governance of mobility and the production of precarity for non-elite, de facto refugee students, facilitated through higher education and its institutions