UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
Materials Jetting Fabrication and Modeling of Printed Multiferroic Composites
No full textThe full text of this item is not available at this time because the author has placed this item under an embargo until December 11, 2030.This work develops additive manufacturing approaches for magnetoelectric composites based on cobalt ferrite–barium titanate systems, addressing fabrication challenges through integrated process and material design. The research encompasses finite element modeling of magnetoelectric coupling, nanoparticle synthesis and ink formulation, inkjet printing process development, and dielectric characterization of printed materials.
Finite element analysis predicts magnetoelectric coupling in printed nanostructures, incorporating nanoscale effects and processing constraints. Model results show that piezoelectric phase properties dominate magnetoelectric performance, with coupling coefficients ranging from 2 to 900 mV/(cm·Oe) depending on achievable material properties. Magnetic field directed assembly demonstrates control over nanoparticle alignment during printing. Piezoresponse force microscopy indicates local electromechanical activity in printed composites despite cubic symmetry observed in powder X-ray diffraction.
Dielectric characterization of inkjet-printed barium titanate nanoparticles (50 nm and 200 nm) quantifies dielectric permittivity reduction from nanoparticle size effects, surface disorder layers, and porosity inherent to 0-3 connectivity. Printed films achieve = 34–73, a 60–120× reduction versus bulk ceramics.
Analysis of dielectric loss in printed polyimide reveals incomplete imidization undetectable by conventional FTIR spectroscopy of sample surface. Temperature-dependent measurements show optimized thermal cure profile is essential for low-loss performance. Polyimide: composites achieve 2–3× dielectric permittivity enhancement over pure polyimide with tan < 0.003. Characterization from −100°C to room temperature shows less than 5% capacitance variation, qualifying these materials for extreme environment applications, such as aerospace and cryogenic electronics.
The research establishes structure-property-processing relationships for printed dielectrics and identifies limitations imposed by 0-3 connectivity and thermal constraints. This work has resulted in 5 peer-reviewed publications in textit{Journal of Composites Science}, textit{Ferroelectrics}, and textit{Sensors and Actuators A: Physical}.Electrical Engineerin
COPNet: Compositional Orthogonal Polynomials Network
Get PDFApproximating highly oscillatory and nonlinear functions remains a core challenge in scientific computing and machine learning. Conventional neural architectures such as multilayer perceptrons and residual networks often require excessive depth or parameter counts to capture fine-scale variations, leading to inefficient training, redundant representations, and limited accuracy in highly dynamic regimes. This dissertation introduces the Compositional Orthogonal Polynomial Network (COPNet), a recursive neural architecture constructed through a modified three-term recurrence of orthogonal polynomials. COPNet incorporates trainable transformations between recursive stages, enabling adaptive feature extraction while preserving the stability and orthogonality intrinsic to polynomial bases. The resulting compositional design achieves accurate function approximation with compact models that maintain near-orthogonal feature representations across depth. To quantitatively assess orthogonality, a complementary framework Depth-Differential Gram Matrix Analysis (DD-GMA) is developed. Unlike post-training similarity metrics such as SVCCA or CKA, DD-GMA analyzes correlations among layer-wise differentials rather than activations, providing a scale-invariant measure of how each layer contributes new, independent information. Empirical studies demonstrate COPNet's advantages in two domains. In nonlinear function approximation, COPNet achieves smoother convergence and higher accuracy than MLP, ResNet, and SIREN baselines while using significantly fewer parameters. Within physics-informed neural networks, COPNet accurately solves nonlinear partial differential equations including the viscous Burgers' and Allen-Cahn equations, attaining relative L 2 errors on the order of O(10^-3) with reduced computational cost. Together, COPNet and DD-GMA establish a principled and interpretable framework for efficient, stable, and orthogonality-preserving neural modeling of nonlinear and physics-governed systems.Electrical and Computer Engineerin
Journeying Through the Magnetopause: Statistical Study of the Kelvin-Helmholtz Instabilities as Seen by the MMS Mission
Get PDFPlasma in space is frozen-in to magnetic fields, and the continuous outflow of solar wind from the Sun carries magnetic fields that directly interact with Earth’s magnetosphere. On the dayside, magnetic reconnection between oppositely directed fields is the dominant mechanism of energy transfer. On the flanks, where the solar wind flows around the magnetosphere, a velocity shear develops between the fast magnetosheath plasma and the stagnant magnetospheric plasma. This shear can excite the Kelvin–Helmholtz instability (KHI), which grows from small perturbations into large vortices that mix plasma and transport mass, momentum, and energy across the magnetopause.
This thesis presents a comprehensive experimental and simulation-based study of KHI at Earth’s flanks. Using MMS observations, we constructed a statistically robust event list under quasi-constant interplanetary magnetic field (IMF) conditions, ensuring that solar wind fluctuations do not mask KHI development. Ion composition analysis revealed distinct inner and outer low-latitude boundary layers (LLBLs), consistent with a two-layer structure. We find that KHI commonly initiates at the inner LLBL–magnetosphere interface, where the field geometry reduces stabilizing effects, allowing waves to grow and, in some cases, roll up fully at this interface—an observational stage not widely reported before. Once vortices grow to scales comparable to the boundary layer thickness, the two-layer structure is erased, and plasma is efficiently mixed into the magnetosphere.
Complementary 2D magnetohydrodynamic (MHD) simulations demonstrate that parallel magnetic fields in the magnetosheath suppress instability at the outer LLBL, shifting growth to the inner edge, consistent with observations. Energy budget analysis indicates KH waves can channel ~5 GW in the linear stage and up to ~50 GW in the nonlinear stage, with turbulence-driven cascades providing the most plausible ion heating mechanism. Heavy ion analysis revealed that oxygen is largely absent in our event list, limiting assessment of its role in KHI evolution.
Together, these results highlight KHI as a robust and persistent pathway for solar wind entry and plasma heating at Earth’s flanks, complementing dayside reconnection in driving global magnetospheric dynamics.Physics and Astronom
UNDERSTANDING HOW BACTERIAL COMMUNITY STRUCTURE AND ORGANIC MATTER COMPOSITION ARE ALTERED IN INTERMITTENT VERSUS PERENNIAL REACHES OF A SEMI-ARID, URBAN WATERSHED
Get PDFIntermittent rivers and ephemeral streams (IRES) comprise up to 60% of the total length of all river networks on the globe and are predicted to become more widespread in the future. IRES have been shown to have “hot spots” or “hot moments” of biogeochemical activity due to the rewetting of streambed sediments after drought, yet important knowledge gaps remain. For example, few studies have examined how hydrological intermittency shapes sediment OM chemistry and microbial community structure at a molecular level, specifically in semi-arid systems This study aims to fill those gaps by investigating the spatial differences in stream water and sediment chemistry between intermittent and perennial streams in the San Antonio River (SAR) watershed. In this study, we sampled 14 sites across the San Antonio River watershed during a dry and wet season which spanned from August to September of 2023 and January to February of 2024. At each site, we assessed sediment chemistry, microbial diversity, and organic matter composition. Results showed that the primary driver of microbial community composition was flow conditions, in contrast seasonality had a stronger influence on biogeochemical processes, more specifically in carbon and nitrogen cycling. Perennial streams contained more thermodynamically favorable Organic Matter (OM), enriched in protein-like and lipid-like compound classes, and supported higher bacterial alpha diversity and more stable community composition. In contrast, intermittent streams exhibited less thermodynamically favorable OM, greater variability in organic matter compound classes, suppressed sediment respiration rates, and reduced microbial diversity, with drying and rewetting driving community restructuring rather than recovery. Seasonal effects were strongest in nutrient cycling, as dissolved organic carbon (DOC) and total nitrogen (TN) declined substantially from the dry to wet season in intermittent streams. These findings demonstrate that IRES are not static reservoirs of complex OM. Contrary to the original hypothesis, that prolonged desiccation would increase OM recalcitrance, our results show that, intermittent streams function as dynamic systems undergoing continuous transformation through drying and rewetting. Perennial streams, by contrast, function as more stable reservoirs of both OM and microbial diversity. With increasing hydrologic variability, understanding how flow conditions regulate the energetic quality of OM and microbial community stability will be critical for predicting freshwater carbon and nutrient cycling.Environmental Scienc
The Role of Virtual Care on Access to Mental Health Counseling or Therapy to Reduce Health Inequities in the United States: Considerations on Age, Race, and Education
Get PDFAim: To examine an association between virtual care and mental health care use and whether there is any different pattern in the association across key demographic characteristics, specifically age, race/ethnicity, and educational attainment.
Methods: A cross-sectional, observation study was conducted among the general population in the United States. This population-based study used data from the 2022 National Health Interview Survey. Main effect and interaction models using binary logistic regression were performed.
Results: Of a total of 708 males and 1,238 females with symptoms of major depressive disorder (MDD), virtual care was associated with an increase in mental counseling or therapy use (males: adjusted odds ratio [aOR]: 2.75, 95% CI: 1.82–4.16, p ≤ 0.001; females: aOR: 3.38, 95% CI: 2.42–4.73, p ≤ 0.001). Females from a Hispanic background (aOR: 7.17, 95% CI: 2.97–17.33, p ≤ 0.001), aged 18–29 (aOR: 7.18, 95% CI: 3.51–14.70, p ≤ 0.001), and with less than high school graduation (aOR: 5.55, 95% CI: 1.97–15.62, p = 0.001) or college degrees (aOR: 5.52, 95% CI: 2.91–10.47, p ≤ 0.001) had notable increases in mental counseling or therapy use with virtual care.
Conclusions: Virtual care is associated with a significant increase in accessing mental health care. Individuals who historically experience challenges related to in-person settings, including those with low educational attainment and Hispanics, may significantly benefit from virtual options.Public Healt
Reinforcement Learning From Human Guidance
Get PDFReinforcement Learning (RL) is a powerful machine learning framework capable of learning behavioral policies autonomously to accomplish desired tasks or make decisions through interaction with an environment, rather than relying on pre-collected datasets as in traditional deep learning approaches. RL has demonstrated remarkable success in simulated environments such as Atari and MuJoCo. However, bridging the gap between virtual environments and real-world applications remains challenging, primarily due to the lack of well-defined reward functions in real-world settings. Reward functions must be task-specific, as RL policies are optimized based on cumulative rewards, making their performance highly sensitive to reward design. In practice, most real-world tasks do not possess inherent reward signals, making it difficult to construct appropriate reward functions, particularly in complex environments.
This dissertation addresses this challenge by proposing methods to learn reward functions or directly shape RL policies through human guidance, including human ratings, demonstrations, and even failed experiences. These approaches aim to enable RL agents to learn effectively when explicit reward signals are unavailable or hard to define.
First, two methods are proposed that utilize human ratings. The first, Rating-based Reinforcement Learning (RbRL), learns a reward model from human evaluations of agent trajectories and explicitly shapes the policy by pushing it away from lower-rated behaviors. The second method learns a reward model through a hybrid multi-class classification and regression approach, improving the accuracy and smoothness of reward estimation from human ratings.
Second, two frameworks are developed based on human demonstrations. The first introduces a learning from failure approach, where policies are shaped by contrasting successful and failed actions sampled from pre-collected failure experiences. The second leverages a Generative Adversarial Network (GAN)-based reward densification model to construct continuous and informative reward signals from limited demonstration data.
Extensive experiments across diverse continuous-control and sparse-reward environments demonstrate that these human-guided methods significantly enhance policy learning efficiency, stability, and generalization compared to standard RL baselines. Collectively, this dissertation provides a unified perspective on reinforcement learning from human guidance, advancing the integration of human knowledge into autonomous decision-making systems.Electrical and Computer Engineerin
Optimal Economic Dispatch and Load Following Control in Nuclear Integrated Energy Systems Using Deep Reinforcement Learning
Get PDFNuclear integrated energy systems represent a transformative approach in addressing the evolving challenges of the modern energy landscape. By coupling nuclear power plants with renewable energy resources, energy storage technologies, and advanced grid management solutions, Nuclear integrated energy systems enable efficient, reliable, and sustainable energy production, distribution, and utilization across multiple sectors. This integrated architecture effectively mitigates the intermittency of renewable generation, thereby enhancing grid stability and operational flexibility. By integrating energy storage and variable renewables, nuclear integrated energy systems are designed to flexibly supply both electrical and thermal demands across diverse end-use applications.. Moreover, these systems generate economic value by participating in wholesale electricity and ancillary services markets, in addition to leveraging revenue opportunities from commodity markets associated with industrial byproducts from tightly coupled processes. This study investigates the economic dispatch performance of a highly integrated NIES configuration by evaluating the efficacy of several conventional reinforcement learning and deep reinforcement learning algorithms in managing the real-time coordination of generation, storage, and demand under both operational and market constraints. Comparative analysis with traditional optimization techniques reveals that deep reinforcement learning approaches not only achieve competitive cost minimization outcomes but also demonstrate enhanced computational efficiency.The results highlight the suitability of deep reinforcement learning-based dispatch strategies for enabling real-time, coordinated control of tightly coupled nuclear integrated energy systems, ensuring operational feasibility and scalability under dynamic system and market conditions.Electrical and Computer Engineerin
SoK: Blockchain Consensus in the Quantum Age
Get PDFConsensus protocols are an important building block in blockchain and blockchain-based systems. The recent focus in developing practical quantum computers reinforces the importance of designing quantum-resistant cryptographic protocols, and in the context of this paper quantum-resistant consensus protocols. In this paper, we systematically review the extant literature on quantum-resistant consensus protocols published between 2019 and 2024. As part of the review, we identify a number of limitations and challenges and discuss potential future research directions.Information Systems and Cyber Securit
Sensing the Black femme: Spit's obsessions and pleasures in Aunt Dicy Tales
Get PDFHow should we read the Black femme body? What practices must we take on to explore the desires and survival written on, and expelled from, her body? To approach with renewed attention the sensory regimes, perhaps in a Wynterian sense, in which the Black femme body is situated? These questions inform how this article thinks within Black feminisms’ interest in sensation as knowing wrapped in flesh. I turn to the U.S. South and how it makes racialization ordinary, clarifying where and how Black femme sensation intervenes. Using African American folklorist J. Mason Brewer's 1956 text, the Aunt Dicy Tales as a case study, folklore is revealed as a racial project where history becomes fiction and fiction history. Folklore shapes what emerges as a mode of Black social thought in Texas, but in the erasure and absence of Black femmes, it remained a tool for state and Black male folklorists to create the Black femme body. By turning to Aunt Dicy's obsession with spitting, I argue that what Brewer intends as an exemplar of Black femme humiliation and masculinization in the immediate afterlives of enslavement, instead, exposes the fissures of Black Texan folklore and the modes of survival Black femme bodies sensate toward.Race, Ethnicity, Gender, and Sexuality Studies (REGS
Interband cascade laser absorption sensor for sensitive measurement of hydrogen chloride in smoke-laden gases using wavelength modulation spectroscopy
No full textThis is the accepted manuscript of a paper that has been published:
K.L. Fetter, L. Munera, M.A. Watts & D.I. Pineda. Interband cascade laser absorption sensor for sensitive measurement of hydrogen chloride in smoke-laden gases using wavelength modulation spectroscopy. Applied Optics, 63(33), pp.8517-8525, 2024. https://doi.org/10.1364/AO.540760A tunable interband cascade laser sensor, based on wavelength modulation absorption spectroscopy near 3.73 µm, was developed to measure hydrogen chloride gas concentration in smoke-laden environments associated with the overhaul stages of firefighting. Wavelength selection near 2678 cm^-1 targets the P(0,9) transition within the fundamental vibrational band of HCl, chosen for its absorption strength and isolation from CO2, H2O, and CH4, as well as proximity to absorption features of other toxicant gases of interest in firefighting applications. Both scanned-wavelength direct absorption with a Voigt lineshape-fitting routine and a wavelength modulation spectroscopy absorption method are employed to recover species concentration. The laser sensor is paired with a compact commercial off-the-shelf 1 m multipass optical gas cell modified to use polished Alloy 20 steel mirrors for increased corrosion resistance against humid and acidic gases, and it is tested by sampling effluent gases from pyrolyzing and burning solid samples of polyvinyl chloride under a radiant heating apparatus in a laboratory fume hood. The wavelength modulation spectroscopy method is demonstrated to enable measurement at the near-ppm-level within a compact form-factor and to provide insights into the thermochemical pyrolysis processes that lead to the formation of hydrogen chloride when polyvinyl chloride is exposed to radiant heating.Federal Emergency Management Agency (EMW-2021FP-00199); National Science Foundation (2135789, 2339502); National Aeronautics and Space Administration (80NSSC19M0194, 80NSSC23PB521)Mechanical Engineerin