DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Design and fabrication of integrated on-chip silicon slow-light structures for optical delay line and eo modulator applications
No full textAugust 2018School of EngineeringThe emerging silicon photonics technology has enabled numerous compact and low-cost on-chip optical systems with various functionalities. In this thesis, the research focuses on silicon on insulator (SOI) based slow-light devices which have the optical signals propagating at a fraction of vacuum speed of light. Slow-light on-chip Bragg grating waveguides on Si/SiO2 are explored both theoretically and experimentally for applications in true-time delay lines and slow-light modulators. The theoretical study of this work focuses on Bragg grating waveguide for its power transmission, group delay, optical bandwidth and dispersion characteristics using the coupled-mode theory (CMT) and finite-difference time-domain (FDTD) methods. The strongest slow light effect is found at the wavelengths near the photonic band edge, while the strong reflection at the slow-light Bragg grating waveguide with regular Si waveguide interface gives rise to unwanted optical power oscillations. Two approaches have studied to suppress the oscillations: the cascaded apodized gratings and the mode-transition gratings. Apodized Bragg gratings have enabled smooth transition at the grating/waveguide interface by gradually modulating the effective index perturbation along the entire grating waveguide. It has been demonstrated that the optical power oscillations can be substantially suppressed using apodization method, thus increasing the delay-bandwidth product of the delay line. Mode-transition gratings, on the other hand, achieves impedance match at the grating/waveguide interface by inserting a short grating taper segment that has a group index smaller than the main section of the slow-light Bragg grating waveguide. By optimizing the taper structure, the spectral oscillations are successfully suppressed. It’s been also demonstrated that the mode-transition gratings exhibit higher group index than apodized gratings. Based on the mode-transition gratings, a dispersion cancellation scheme is designed. The group velocity dispersion (GVD) is negative at the right band edge of a Bragg grating and positive in the left band edge. By cascading two Bragg gratings with opposite signs in GVD, dispersion compensation will lead to much increased delay-bandwidth product. The slow-light grating waveguide has been incorporated in optical electro-optic (EO) modulators for analog photonic applications. Depletion-type EO modulators make use of the free carrier plasma effect to achieve intensity modulation. With light traveling at a slower group velocity, the plasma effect of the waveguide is enhanced, and thus the required modulator length is greatly reduced. The linearity of the modulators is also studied for analog applications. The nonlinear transfer function of modulators gives rise to distortion terms such as 3rd-order intermodulations (IMD3). By taking the transfer function for its third derivative, an optimal biasing point can be found, which gives the maximum spur-free dynamic range (SFDR) and improves the signal-to-noise ratio of the photonic link. In addition, EO crosstalk in densely integrated Mach-Zehnder modulators (MZMs) is studied and characterized, which provides a guideline to the trade-off between system integration and performance optimization.Ph
Real time electricity price time series forecasting models based on deep learning
No full textJuly 2022School of EngineeringDue to its nature of high-noise, high-nonlinearity and high-uncertainty (3H), predicting and modeling the real-time, spot electricity price is of the utmost difficulty. Poor forecasting brings risks and challenges for multiple power system tasks such as strategic bidding and generation rescheduling. Failure to predict price spikes due to climate change events will result in huge social and property loss, on the scale that was witnessed during the 2021 Texas power outage caused by extreme cold weather and the 2019 California power curtailments caused by wildfires. Recently, deep learning (DL) has been actively attracting researchers' attention as a potential solution to the problems posed by 3H. However, there are still huge research gaps. Specifically, most of current existing models are focused on hourly or larger time intervals and only a few models are based on high-frequency (less than 5 minutes) forecasting. Moreover, traditional prediction models cannot effectively model the temporal dependencies among the 3H time series. For example, the spikes caused by commercial forecast loss or looped power flow are a big challenge to most forecasters. This work focus on the development of electricity price forecasting via deep learning based methods. Firstly, a novel multi-branch Gated Recurrent Unit (GRU) architecture (HFnet) is developed for the real-time prediction task, which uses date index features. The strong performance of this model is the basis for several following model development efforts. Secondly, a new multi-branch model with a novel parallel convolution neural network (CNN) and time series statistical features has been developed to enhance the forecasting accuracy (GHTnet). Thirdly, a temporal transfer learning model (GRU-TL) is proposed to measure feature-reuse among hybrid subzone datasets, which contributes to performance improvement and model robustness. Fourthly, a CNN-based Autoencoder (QCAE) is designed to capture features expressed both temporally and spatially. Finally, an explainable model with attention mechanism (ATTnet) is proposed to improve prediction quality and interpretability in real-time electricity price forecasting.Ph
Learning from event sequences
No full textAugust 2023School of ScienceEvent sequences are fundamental in various domains, encompassing customer transactions, electronic health records, and other scenarios involving actions over time. Understanding the dynamics of event transitions enables prediction of future events, examination of event influence, and quantification of their effects. This thesis focuses on learning and inference problems in event sequences using state-of-the-art machine learning and deep learning approaches.Temporal point processes are widely employed for modeling discrete events in continuous time. In this thesis, we propose an innovative self-supervised learning paradigm coupled with a contrastive module which leverages transformer encoders to predict the subsequent event given history. What sets our approach apart is its ability to capture the continuous time dynamics, distinguishing it from traditional self-supervision techniques used in time series and NLP. In addition, we tackle the intricate task of multi-label prediction in event streams, where multiple event types can co-occur within the continuous time frame- work. To address this challenge, we introduce the Transformer-based Conditional Mixture of Bernoulli Network, enabling the capture of complex temporal and probabilistic interdependencies among concurrent event labels. Through comprehensive experimentation, our proposed methodologies exhibit compelling empirical performance when compared to exist- ing methods, establishing their efficacy across various synthetic and real-world benchmark datasets.
To identify influential events related to specific events of interest, we introduce the influence-aware attention for multivariate temporal point process model. This model lever- ages transformer attention mechanisms and variational inference techniques to capture the temporal dynamics and aggregate instance-instance interactions, thereby revealing type-wise influence between different event types. We also explore scenarios without timestamps and propose a probabilistic attention-to-influence neural model to discover influential event types in timestamp-free datasets. Our models have demonstrated superior performance around influencing set identification and prediction tasks for a particular event of interest.
Moreover, we delve into the realm of causal inference in temporal point processes by extending Rubin’s framework for average treatment effect and propensity scores to accommodate multivariate point processes. This extension enables us to perform causal inference between event variables in recurrent event streams. In the context of datasets without timestamps, we enhance autoregressive transformer models by incorporating pairwise causal knowledge derived from causal knowledge graphs. This incorporation guides the training of transformer models, ultimately bolstering the reliability and trustworthiness of the model. This line of research has proven to be highly effective in our investigations.
Extensive experiments validate the effectiveness of our models and algorithms, outperforming existing methods on synthetic and real-world benchmarks. Finally, we identify future directions for research, including the exploration of multi-modal self-supervision and sequential decision-making within the context of temporal point processes.Ph
Enhancing monte carlo modeling workflows with a metamodel-driven approach for nuclear reactor analysis
No full textMay 2023School of EngineeringNuclear simulation programs are diverse and copious; one can easily find a myriad of programs for modeling applications ranging from neutronics to thermal-hydraulics to species transport and corrosion. Furthermore, it is becoming increasingly common to see such programs, or codes, bundled together to tackle the multiphysics analyses required for advanced reactor concepts. As the ensemble of nuclear simulation programs continues to grow and individual programs become more expansive and complex, it is necessary to consider how to efficiently interact with such programs. Thus, the individual programs, tools, and operations required as part of an analysis problem must be considered together, or as a workflow. A workflow encompasses each piece of the modeling process such as input preparation, simulation execution, and even results processing. Components of a workflow will often be automated or assisted as much as possible to mitigate the need for one to manually manage each individual step or program. The user-facing implementation can take many forms, such as a programming library or a user-friendly graphic interface, but should ultimately alleviate usability challenges of its constituent parts. Given the complexity of nuclear reactor modeling and analysis tools, one would expect smooth and coherent workflows to be intrinsic to the field. Therefore, it is counterintuitive that Monte Carlo neutronics codes, a cornerstone of reactor physics modeling and simulation, have antiquated workflows for both standalone and coupled operations. With their continuous energy simulation capabilities, Monte Carlo codes are irreplaceable sources for both high-fidelity reference solutions and multigroup cross-section generation for deterministic solvers. For Monte Carlo codes such as MCNP, Serpent, and KENO, their most glaring usability obstacle is their expansive and intricate input syntax. When modeling a reactor or other system in such codes, one must specify everything from geometry, to physical conditions, to their simulation quantities of interest in a terse text-based input format. Consequently, users must understand an expansive input syntax that appears arcane to the uninitiated and still remains tedious for veterans. This directly impedes manual input development as navigating the syntactic obstacles is time-consuming and error-prone. At the same time, there are limited capabilities to allow automated execution, model transformation, or support for user-defined logic and abstractions which would circumvent manual operations or permit workflow enhancements. Thus, developing Monte Carlo simulation inputs is a belabored process with limited workflow integration. In this work, these impediments to Monte Carlo modeling workflows are addressed through a model-driven development approach to provide modernization and unification across workflows. Following this strategy, a comprehensive ``model of the model'', or metamodel, is created for codes of interest, including MCNP, which fully describes all syntactic and semantic elements of their input formats. From the metamodel representation, editor-services (syntax highlighting, error-checking, reference finding, etc.) and Application Program Interface (API) capabilities are established. By being derived from the same underlying model, both of these avenues for input development become inherently interoperable with significantly less development effort than independent solutions. These functions are leveraged to create modern editing environments for Monte Carlo codes and more importantly, to support full-featured Python APIs. Through the developed APIs, all input features described by a code's metamodel can be managed programmatically. This enables advanced operations such as transforming and translating input files between Monte Carlo codes. These capabilities are demonstrated on applications including criticality searches, processing models for 3D viewing, modifying cross-section libraries, and iteratively translating between input formats. Collectively, these applications demonstrate the viability of a metamodel-driven approach towards workflow unification and modernization for Monte Carlo codes.Ph
Stochastic first order methods for distributed composite optimization with differential privacy
No full textAugust 2023School of ScienceABSTRACTDistributed optimization has gained much attention over recent years as the world generates
ever more data. At the same time machine learning methods are able to solve many new
problems. As the desire to train models on large data sets becomes of great concern to more
fields, it is more important than ever to have fast and secure methods to solve distributed
optimization problems. In this work we address three fundamental concerns of distributed
optimization. In Chapter 1, we provide motivation and an overview of recent works on distributedoptimization. In Chapter 2, we introduce the Async-Parallel Adaptive stochastic gradient Method (APAM) algorithm. The APAM method is designed to solve non-convex, smooth optimization problems. This method allows the use of asynchronous updates to the model during training, thus avoiding delays caused by lagging compute nodes. We first give a brief overview of important works on adaptive stochastic gradient methods, then prove convergence of APAM in both convex and non-convex settings. Also, empirical experiments are given to show improvement of APAM on practical problems. In Chapter 3, we discuss the Federated Learning framework for training a model in common among several disparate parties. In this setting different clients hold their own data privately and only model parameters, rather than data or gradients, can be communicated. We introduce the Federated Proximal Gradient Method (FedPGM) algorithm for solving non-convex and non-differentiable problems in the Federated setting. We prove convergence of FedPGM for both full and partial client participation cases. In Chapter 4, we review key ideas of Differential Privacy, a popular method for ensuring privacy while training a model. Differential Privacy gives rigorous guarantees of privacy for those individuals included in the training set when the trained model will be released publicly. We give a new analysis of the Gaussian Mechanism, a popular method in the machine learning community. Our analysis shows that one can use dynamic noise schedules in training, rather than a fixed level of additive noise in each iteration. We also show the effectiveness of our technique with experiments on several data sets. Finally we present our concluding remarks in Chapter 5, where we give a brief summaryof the contents of this dissertation.Ph
Group decision makings from partial preferences
No full textMay 2023School of ScienceGroup decision making is the situation that a group of agents makes collective choices over a set of alternatives. The input of group decision making is the partial preferences of agents, and its output is one (or more) candidates. This thesis focuses on a two-step (\emph{Preference learning} and \emph{rank aggregations}) group decision making framework. A preference learning method learns a statistical ranking model from the users' preferences, which might be partial rankings. In rank aggregations, the group decision is made according to the learned ranking model. Both preference learning and rank aggregations are highly challenging when the number of alternatives becomes large. This dissertation focuses on solving the following questions. How can we design efficient preference learning algorithms when the preferences are partial rankings? How can we design rank aggregation rules with privacy guarantees? The first part of this dissertation focuses on preference learning. We propose two preference learning algorithms, both compatible with partial preferences. Our main theoretical contributions are the complexity analysis of the proposed preference learning algorithms, which guarantees that all our proposed preference learning algorithms can finish in polynomial (or sub-polynomial) time. The experiments confirm that our algorithms are robust, efficient, and practical. Furthermore, this dissertation proposes a group of rank aggregation methods with privacy guarantees. We found Differential privacy (DP), a widely accepted notion of privacy, is unsuitable in many rank aggregation scenarios because it requires external noises. Thus, we proposed a novel privacy notion, smoothed DP, for rank aggregations. The smoothed DP notion can achieve a similar privacy guarantee with DP without requiring external noises. We theoretically proved and experimentally confirmed that most real-world elections are private under the smoothed DP notion. Finally, we apply the private group decision making framework to a downstream application, improving the robustness of interpretation maps. Interpretation maps explain the reason of why deep neural networks output a certain classification and can be used in the application scenarios like objective detection, medical recommendation, and transfer learning. Inspired by rank aggregation, we proposed the first interpretation method with theoretically guaranteed robustness against l_∞-norm attacks. The proposed method is not only ~30% more robust but also surprisingly more accurate than state of the art according to an experiment on real-world datasets.Ph
Design of silicon, iii-v, diamond, and graphene terafet detectors
No full textMay 2023School of EngineeringThe plasmonic terahertz field-effect transistor (TeraFET) has been studied extensively since the 1990s, and has shown promise for a range of applications such as sensing, imaging, biomedical engineering, and wireless communication, owing to its high speed, low noise-equivalent power, and high tunability. Recently, TeraFETs have been identified as ideal candidates for 6G communication systems, as they exhibit excellent detection performance in the 200–500 GHz band. However, to accelerate the development and commercialization of these applications, it is essential to further enhance the sensitivity of TeraFET detectors. This thesis explores the improvement of TeraFET sensitivity using theory and verified hydrodynamic models. We compare and analyze the detection performance of TeraFETs in multiple material systems, including Si, AlGaN/GaN, AlGaAs/InGaAs, Diamond, and graphene. Our results show that p-Diamond TeraFETs have a relatively high continuous-wave (CW) response at the sub-THz band due to their high quality factor, while graphene TeraFETs are promising for pulse detection due to their low fictitious effective mass, which results in a fast response time. Additionally, electrons in graphene TeraFETs exhibit strong viscous features and possess various transport regimes. In addition to material considerations, we discuss improved TeraFET designs based on non-uniform structures. We demonstrate that using a sawtooth gate capacitance configuration can achieve a theoretically 40% increase in CW response, owing to the enhanced boundary asymmetry effects in a non-uniform geometry. Non-uniformity in TeraFETs can also be introduced by drain biasing. In a short-channel device where plasmonic oscillations reach the drain, the voltage response can be negative, indicating the amplification of incoming THz signals. This suggests that current-driven TeraFETs can be used as THz amplifiers under certain conditions. The above results provide preliminary insights into the design and optimization of TeraFETs under different operating conditions, thereby facilitating their application in various fields.Ph
Step-by-step: defining the catalytic properties of heterodimeric kinesin-2 motors
No full textMay 2019School of ScienceKinesin is a class of MT-based molecular motors that is involved in vesicle transport, signal transduction, microtubule cytoskeletal remodeling, and cell division for proper organismal physiology and development. The kinesin-2 family, for example, is well-known for its transport roles because it is highly processive, meaning it can take multiple 8-nm steps along the microtubule track before it detaches. The expression of four genes, namely KIF3A, KIF3B, KIF3C, and KIF17 can result in mammalian kinesin-2s heterodimeric KIF3AB and KIF3AC as well as homodimeric KIF17. KIF3AB, which is associated with a cargo adaptor protein called kinesin accessory protein (KAP), is essential for intraflagellar transport for ciliary assembly and can act as a scaffold for hedgehog signaling. Much more is known about KIF3AB/KAP than KIF3AC, mainly because there is no KIF3C orthologue in other model organisms. Unlike KIF3AB/KAP, heterodimeric KIF3AC is primarily expressed in neurons. One of the longstanding questions in the field has been why mammalian kinesin-2 is preferentially expressed as a heterodimer. One role for heterodimerization may be to specify adaptor and cargo binding. However, we hypothesize that heterodimeric kinesins may have also evolved to tune the catalytic properties of the heterodimer for its transport roles. KIF3AC serves as an ideal kinesin for testing this hypothesis because the intrinsic properties of KIF3A within engineered KIF3AA are significantly faster than KIF3C within engineered KIF3CC. For example, the single-molecule velocity of KIF3AA is 240 nm/s whereas the velocity of KIF3CC is 7.5 nm/s. However, KIF3AC achieves a velocity of 186 nm/s which is intermediate of KIF3AA and KIF3CC. This leads to the question of how the catalytic properties of KIF3A and KIF3C within KIF3AC differ from their intrinsic properties within homodimeric KIF3AA and KIF3CC. We addressed this question using stopped-flow presteady-state ADP release kinetics experiments and computational modeling of the KIF3AC stepping cycle. The modeling predicted that KIF3A and KIF3C collide with the microtubule with similar rates. However, once KIF3AC is on the microtubule, KIF3A and KIF3C retain their relative intrinsic catalytic properties. To better understand the mechanism of KIF3AC, we also modeled the stepping cycle of KIF3AB. The modeling predicted that heterodimerization alters the microtubule association properties of KIF3A and KIF3B but once on the microtubule, each head steps with equivalent fast rates of ~40 s-1. To confirm these results, both presteady-state phosphate release and dissociation kinetics experiments were conducted. Mathematical modeling of the data from the phosphate release kinetics experiments, which capture the steps from ATP association through coupled phosphate release and dissociation, is currently in progress. However, the experiment results demonstrate that KIF3A likely dominates the fast initial exponential rate of phosphate release and dissociation of KIF3AC. Together, these results suggest that heterodimerization serves as a mechanism for regulating the motility of heterodimeric KIF3AC.Ph
Electrical resistivity size effect in compound conductors
No full textJune 2023School of EngineeringThe downscaling of modern integrated circuits is facing a major challenge posed by the resistivity size effect in interconnect lines which is mainly due to electron scattering at surfaces and grain boundaries. The resistivity increase and associated signal delay can be mitigated if a suitable material can be found that has a small resistivity size effect and therefore conducts better than Cu at reduced dimensions. Thus, a lot of experimental and theoretical efforts have been made to reduce electron scattering at small dimensions or decrease the thickness of barrier layer. Nevertheless, there is still no clear winner on replacing Cu in the back-end-of-line process and resistivity scaling of many materials are still unknow. Therefore, in this thesis, I perform a series of experiments dedicated to quantifying the resistivity size effect of various compound conductors to pave the road for the prospective barrier-free interconnect integration.The resistivity of metal can greatly increase once its dimension reaches below the phonon-electron scattering mean free path λ of 39 nm in Cu. The increase resistivity including surface scattering and grain boundary scattering is proportional to the ρoλ product where ρo is the bulk resistivity of the material. Therefore, we are going to search for material with lower ρoλ to achieve smaller resistivity increase at small metal pitch. ρoλ is a intrinsic material’s property determined by its electronic structure and changes inversely related to the effective electronic states near to the Fermi level. In principle, the smaller ρoλ would result in a better conductivity at small dimensions. However, we have to make sure the resistivity of the new material is comparable with the know metals, based on which we are mainly searching for material with smaller electron mean free path to reduce the electron scattering contribution from both surface and grain boundaries.
In the field of compound conductors, I first focus on Ti4SiC3(0001), a MAX-phase material with a previously predicted ρoλ= 3.1 × 10 16 Ωm2 that is more than two times smaller than that of Cu and exhibits a 2.5 times larger cohesive energy. Magnetron co-sputtering from three elemental sources at 1000 °C onto 12-nm-thick TiC(111) nucleation layers on Al2O3(0001) substrates yields epitaxial growth with Ti4SiC3(0001) || Al2O3(0001) and Ti4SiC3(101 ̅0) || Al2O3(2(11) ̅0), a low and thickness-independent surface roughness of 0.6 0.2 nm, and a measured stoichiometric composition. The room-temperature resistivity ρ increases slightly with decreasing thickness, from ρ = 35.2 0.4 to 37.5 1.1 cm for d = 92.1 to 5.8 nm, and similarly from 9.5 0.2 to 11.0 0.4 cm at 77 K, indicating only a minor effect of electron surface scattering on ρ. The curve fitting based on semiclassical model yields a low λ = 1.1 0.6 at 293 K and λ = 3.0 2.0 nm at 77 K. This provides a great potential for MAX phase materials to be used for high-conductivity narrow interconnect lines in spite of its relatively high resistivity ρo = 35.1 0.4 cm.
I also conduct some experiments on isotropic intermetallic compounds. The resistivity ρ as a function of thickness d of epitaxial CuTi(001) and CuAl2(001) layers is measured to quantify the resistivity size effect of these compound conductors and evaluate their promise as a replacement material for Cu in highly scaled interconnect lines. The layers are deposited by magnetron co-sputtering onto MgO(001) substrates and their epitaxy is confirmed by x-ray diffraction -2 scans, ω rocking curves and φ-scans. The surface morphology is quantified by x-ray reflectivity and atomic force microscopy, and the composition measured by photoelectron spectroscopy and Rutherford backscattering. Data fitting of the measured ρ vs d yields room-temperature electron mean free paths λ = 12.5 and 15.6 nm for CuTi and CuAl2, respectively, and bulk resistivities ρo = 19.2 0.8 and 7.7 0.4 μΩ·cm. The overall analysis yields ρoλ benchmark values of 24 × 10-16 and 12× 10 16 Ωm2, respectively, indicating that the conductivity advantage of the evaluated compound conductors against Cu can only be realized if their higher cohesive energies and stability can be exploited to achieve liner-free lines.
In my third research thrust, I study the directional conductors with suitably anisotropic Fermi velocity distributions such that they can achieve superior conduction along specific axis. Thus, epitaxial VNi2 layers are deposited onto MgO(001) and their resistivity ρ measured as a function of layer thickness d = 10.5-138 nm to quantify the resistivity size effect. A cube-on-cube epitaxy of the fcc parent structure on MgO(001) leads to two possible layer orientations for orthorhombic VNi2(010) and VNi2(103), resulting in considerable atomic disorder at domain boundaries. In situ ρ vs d measurement yield a bulk resistivity ρo = 46 2 cm and a benchmark quantity ρoλ = (160 10) × 10 16 Ωm2, where λ is the bulk electron mean free path. These values are considerably higher than theoretically predicted, which can be attributed to the two possible atomic arrangements, causing extra electron scattering at domain boundaries. The superior predicted conduction along a specific crystalline direction could not be experimentally verified. The superior anisotropic high conductivity can only be achieved with a single crystalline orientation which necessitates the suppression of any domain boundaries between the two nearly identical orientations.
Lastly, a polycrystalline CuTi is fabricated to study the stability of compound conductors on dielectric SiO2. With similar deposition condition with epitaxial layers, polycrystalline CuTi layers show a preferred out-of-plane orientation of 001 direction and a random in-plane orientation. The resistivity scaling of CuTi is only comparable with elemental metals if it facilitates barrier-free interconnect lines. Accordingly, the stability study includes interdiffusion reliability test, thermal stability test and interfacial adhesion test. The overall results indicate that CuTi has a much better stability on SiO2 than Cu and enhances the potential for compound interconnect to achieve a reduced barrier thickness or even barrier-free interconnect.Ph
The effect of interfacial forces on the performance of wickless heat pipe
No full textDecember 2022School of EngineeringA heat pipe is a passive thermal unit that utilizes a phase change mechanism to dissipate the heat from the hot spot of a device to the outside environment. Evaporation of the liquid at the hot end of the heat pipe absorbs the heat. The vapor then travels to the cold end and condenses back to the liquid phase. The condensation of the vapor releases the latent heat, and heat sinks at the cold end draw the heat out of the heat pipe system. Fluid mechanics, however, allows the cooling process to be continuous in the heat pipe. The surface tension of the liquid creates the capillary pressure that recycles the condensate from the cold end to the hot end. Wick structures are often used on the internal solid surface of the container to improve the efficiency of capillarity. Therefore, the heat transfer cycle can be repeated indefinitely without any external mechanical pumping systems.The engineering of an efficient heat pipe requires an understanding of processes at both the macroscopic and microscopic scales of operation. Among the different factors that can affect a heat pipe’s performance, the physical properties of working fluid in a heat pipe system play a critical role in the phase change capability, heat transfer, and fluid flow behavior. The Constrained Vapor Bubble (CVB) experiment used a transparent and wickless fused silica cuvette to serve as the heat pipe. Thus, the behavior of the working fluid could be analyzed through optical measurements.
In the first section of the study, we examined how varying condenser temperatures affect heat transfer and fluid flow behavior using pentane as the working fluid. The system's performance declined as the condenser temperature dropped. The behavior seen while employing a mixture of 94 vol% pentane and 6 vol% iso-hexane was the opposite of this performance decline when using the pure working fluid. A rise in the apparent strength of Marangoni flows at the heater end of the system was the cause of the performance drop as the condenser temperature was lowered. For experiments where we held the condenser temperature constant while increasing the heater power and for experiments where we held the heater power constant while decreasing the condenser temperature, a fin-type heat transfer model was fit to the experimental data and used to extract an evaporator heat transfer coefficient. One Nusselt number vs. Marangoni number curve was found to represent all the results. In this formulation, it was discovered that the Nusselt number decreased as the Marangoni number was raised to the third power.
In the second section of the study, we analyzed more than 100 nucleation events in a wickless heat pipe under microgravity. The experiment ran for 20 hours with pentane as the working fluid. Peak pressures and vapor temperatures in the device were momentarily increased by bubble nucleation at the heater end. The original vapor bubble collapsed due to increasing pressure at the time of nucleation, and the heater wall temperature considerably dropped due to enhanced evaporation. Heat transfer coefficients near the heater end of the system were determined using a thermal model that was created using the measured temperatures and pressures. Peak heat transfer coefficients during the nucleation event exceeded steady-state values by a factor of three. A linear correlation between the Nusselt number and the Ohnesorge number was found that explains all the heat transfer coefficient data from the nucleation events.
The third section of the study focuses on the microscopic perspective of the heat pipe system. The local heat flux in the wickless heat pipe is governed by the shape of the microscale meniscus. Thus, we employed an interferometry method to calculate the thickness profile of the corner meniscus. This process is quick and simple and does not direct interaction with the liquid surface. Thin Film Interference generates fringe patterns at the meniscus region. Five different image analysis methods are explained to demonstrate the estimation of meniscus thickness from the interferometry images collected in two experiments.
In the last section of the study, we analyzed and questioned the interfacial ideality of the pentane and iso-hexane liquid mixture. The ideal mixture assumes linear behavior of the surface tension and Hamaker constant. These physical characteristics, however, might not always follow a linear relationship with changes in the volume fraction. The non-ideality of the binary mixture was investigated using five different compositions. Through the isothermal experiments employing the CVB system on Earth, the retarded Hamaker constants and surface tension of the liquid mixtures were evaluated.
The experiments, discoveries, and analyses presented in this thesis provide an in-depth inspection of the wickless heat pipe operation under the microgravitational environment. In addition, the non-ideal behavior of the pentane and iso-hexane mixture was revealed using the interferometry images of the corner meniscus and multiple processing tools. In the future, two newly designed CVB apparatuses, linear and looped systems, with pentane/iso-hexane mixture as the working fluid, will be studied under non-isothermal conditions. These experiments will further investigate the impact of the non-linear interfacial behavior of the liquid mixture on the heat transfer and fluid mechanics of the wickless heat pipe. Ph