京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第21582号 / 理博第4489号 / 新制||理||1644(附属図書館) / 京都大学大学院理学研究科地球惑星科学専攻 / (主査)准教授 竹見 哲也, 准教授 石岡 圭一, 教授 秋友 和典 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

Disertační práce se zabývá modelováním dispersních materiálů metodou konečných prvků v časové oblasti. V práci jsem navrhl matematický model zahrnující dispersní modely a aplikoval jsem na něj metodu konečných prvků. Pozornost jsem věnoval třem nejvíce používaným dispersním modelům. Jedná se o Debyeho model, Lorentzův model a Drudův model. Dále jsem se zabýval technikami implementace těchto dispersních modelů. Techniky jsem zabudoval do metody konečných prvků. Navíc jsem vyvinul novou metodu založenou na technice digitální filtrace. Funkčnost vyvinutých metod a technik jsem ověřil na řadě testovacích příkladů. Diskutoval jsem dosažené výsledky a navrhuji jejich možná vylepšení.

Abstact: In order to address the wind-industry's need for a new generation of more advanced wake models, which accurately quantify the mean flow characteristics within a reasonably CPU-time, the two-dimensional analytical approach by Belcher et al. (2003) has been extended to a three-dimensional wake model. Hereby, the boundary-layer approximation of the Navier-Stokes equations has been linearized around an undisturbed baseflow, assuming that the wind turbines provoke a small perturbation of the velocity field. The conducted linearization of the well established actuator-disc theory brought valuable additional insights that could be used to understand the behavior (as well as the limitations) of a model based on linear methods. Hereby, one of the results was that an adjustment of the thrust coecient is necessary in order to get the same wake-velocity field within the used linear framework. In this thesis, two different datasets from experiments conducted in two different wind-tunnel facilities were used in order to validate the proposed model against wind-farm and single-turbine cases. The developed model is, in contrary to current engineering wake models, able to account for effects occurring in the upstream flow region. The measurement, as well as the simulations, show that the presence of a wind farm affects the approaching flow even far upstream of the first turbine row, which is not considered in current industrial guidelines. Despite the model assumptions, several velocity statistics above wind farms have been properly estimated, providing insight about the transfer of momentum inside the turbine rows. Overall, a promising preliminary version of a wake model is introduced, which can be extended arbitrarily depending on the regarded purpose.

Nous nous intéressons à des problèmes de transmission entre diélectriques et métamatériaux, milieux présentant des propriétés électromagnétiques inhabituelles comme des caractéristiques effectives négatives à certaines fréquences. Par exemple, ces milieux peuvent être construits comme des assemblages périodiques de microstructures résonantes et dans ce cas la théorie de l'homogénéisation permet de justifier mathématiquement ces propriétés effectives. En régime harmonique et dans des géométries à variables séparables, des calculs analytiques peuvent être menés. Ils révèlent dans des cas dits critiques des difficultés mathématiques: les solutions n'ont pas la régularité standard, voire le problème peut être mal posé.La première partie étudie ces problèmes de transmission en régime temporel pour lequel les métamatériaux sont modélisés par des modèles dispersifs (modèle de Drude ou de Lorentz). Les difficultés résident dans le choix d'un schéma de discrétisation mais surtout dans la construction de conditions absorbantes. La méthode retenue ici est celle des Perfectly Matched Layers (PMLs). Comme les PMLs classiques sont instables pour ces modèles du fait de la présence d'ondes inverses, nous proposons une nouvelle classe de PMLs pour lesquelles nous menons une analyse de stabilité. Cette dernière permet de construire des PMLs stables. Elles sont ensuite utilisées pour simuler le comportement en temps long d'un problème de transmission; nous illustrons alors le fait que le principe d'amplitude limite peut être mis en défaut en raison de résonances d'interface.La deuxième partie vise à pallier ces phénomènes d'interface en régime harmonique en revenant sur le processus d'homogénéisation classique, pour un milieu dissipatif. Pour des problèmes de transmission, il est connu que les modèles issus de cette méthode perdent en précision du fait de la présence de couches limites à l'interface. Nous proposons un modèle enrichi au niveau de l'interface. En combinant la méthode d'homogénéisation double-échelle et celle des développements asymptotiques raccordés, nous construisons des conditions de transmission non standards faisant intervenir des opérateurs différentiels le long de l'interface. Le calcul de ces conditions nécessite la résolution de problèmes de cellule et de problèmes non standards posés dans des bandes périodiques infinies. Une analyse d'erreur confirme l'amélioration de la précision du modèle. Des simulations numériques illustrent l'efficacité de ces nouvelles conditions. Enfin, cette démarche est reproduite formellement dans le cas des matériaux à fort contraste se comportant comme des métamatériaux. Nous montrons alors que ces nouvelles conditions permettent de régulariser le problème de transmission dans les cas critiques. / We are interested in transmission problems between dielectrics and metamaterials, that is to say media with unusual electromagnetic properties such as negative constants at some frequencies. These media are often made of periodic assemblies of resonant micro-structures and in this case the homogenization theory can justify mathematically these effective properties. A preliminary part deals with these problems in the harmonic domain and in geometry with separation of variables.Analytical computations are done and reveal in the so-called critical cases some mathematical diffculties: the solutions do not have the standard regularity and the problem can even be ill-posed.The first part examines these transmission problems in the time domain for which metamaterials are modelled by dispersive models (Drude model or Lorentz model for instance). The diffculties reside in the choice of a discretization scheme but especially in the construction of absorbing conditions. The method used here is the use of Perfectly Matched Layers (PMLs). Since classical PMLs are unstable for these models due to the presence of backward waves, we propose a new class of PMLs for which we conduct a stability analysis. The latter allows us to build stable PMLs. They are then used to simulate the long-time behaviour of a transmission problem; we illustrate the fact that the limit amplitude principle can be faulted because of interface resonances.The second part aims to overcome these phenomena by coming back to the classical homogenization in the harmonic domain, for dissipative media. For transmission problems, it is known that models resulting from this method lose accuracy due to the presence of boundary layers at the interface. We propose an enriched model at the interface: by combining the method of two-scale homogenization and that of matched asymptotic expansions, we build non-standard transmission conditions involving tangential derivatives along the interface (Laplace-Beltrami operators). This requires to solve cell problems and non-standardproblems in infinite periodic bands. An error analysis confirms the improvement of the accuracy of the model and numerical simulations show the effectiveness of these new conditions. Finally, this approach is formally reproduced in the case of high contrast materials which behave like metamaterials. We show that these new conditions regularise the transmission problem in the critical cases.

This master thesis studies the behaviour of two aluminium alloys in ultrasonic cleaning at two different intensities, as well as the effect of a cleaning solution, Formula 815 GD-NF on the same surface have been studied with respect to surface roughness and material composition. Methods like Light Optical Microscopy (LOM), Scanning Electron Microscopy (SEM), Energy Dispersive Spectroscopy (EDS) and 3D topography using white light interferometry (VSI) was used to study the surface and material composition. It was found that both the ultrasonic cleaning, as well as the solution itself both increase surface roughness of the samples. When crossing a threshold in ultrasonic intensity micro jet cavitation dislodge precipitates, or areas weakened by precipitate to form a large pit. And the following heat from the jet causes the surface to oxidise, becoming more brittle, and be broken up by other cavitation phenomena to cause rapid acceleration of surface roughness in an area originating from the pit.

With the recent advancements of wireless networks to satisfy new requirements, the investigation of novel transmission schemes to improve the link level performance is of major importance. A very common technique utilized in nowadays systems is the Orthogonal frequency division multiplexing (OFDM) waveform, which has been adopted by several standards, including WiFi, LTE, and more recently 5G, due to its simple equalization process. Despite its success, this dissertation shows that OFDM is a sub-optimal scheme under frequency-selective channel (FSC), when channel state information (CSI) is available at the receiver only. Based on the coded modulation capacity approach, this work demonstrates that the data symbols should experience the same channel gain in order to achieve the best performance, leading to the equal gain criterion (EGC). However, this comes at a cost in terms of losing orthogonality among data symbols. The result is valid for linear modulation matrices under the assumptions of CSI at only at the receiver with perfect feedback equalization. In order to attain the EGC for doubly-dispersive channels, the block multiplexing (BM) waveform is proposed in this thesis, where the data symbols are spread in frequency and time. For instance, the recently conceived orthogonal time frequency space (OTFS) is shown to be a particular case of BM with the classical single-carrier (SC). Regarding the equalization for the robust waveforms, it is shown that the minimum mean squared error with parallel interference cancellation (MMSE-PIC) employed together with convolutional encoder and soft decoder can completely remove the inter-symbol interference (ISI), where a low-complexity implementation is designed. In addition, a waveform with decreased complexity based on the sparse Walsh-Hadamard (SWH) is proposed for two reasons, i) sparse spreading requires a transform with lower size, ii) the Walsh-Hadamard transform is implemented with 1s and −1s, which requires less complexity than fast Fourier transform (FFT) based waveforms. Furthermore, the problem of estimating the time varying channel is considered, where a unique word (UW) or (pilot block) based approach is studied. In this regard, another main contribution of this dissertation is to develop an optimization framework, where the combination of channel estimation plus Doppler spread error is minimized. In particular, the composite error minimization is achieved by properly setting the FFT size of the system, for a fixed data length. Lastly, cyclic prefix (CP)-free system is considered such that the transmission time is decreased, and therefore provides a better channel estimation. Naturally, the CP-free system has undesirable interference, which is resolved by an iterative CP-Restoration algorithm. In this case, we extend the EGC to equal reliability criterion (ERC), i.e., the data symbols should be equally reliable and not only have equal gain. As a consequence, the BM with orthogonal chirp division multiplexing (OCDM) waveform has the best performance due to equal time and frequency spreading. In conclusion, the coded modulation capacity approach of this dissertation provides new insights and solutions to improve the performance of wireless systems.

<p>  </p> <p>Atmospheric particles originate from all over the globe with wildly different sources such as sea spray aerosols of the ocean, mineral dust from deserts, biogenic emissions from forests, anthropogenic emissions of urban and industrial areas, volcanic eruptions, and many more. All of these particles can then be transported during which aging can occur where the external and internal chemical composition of particles and drastically be altered in which their physiochemical properties change or new particles as a whole are formed. Understanding what can cause this aging and correctly identifying how these particles change is vital for assessing climate in local areas. </p> <p>Chapter 3 focuses on dry intrusion (DI) and non-DI periods where vertical mixing of air occurs and allows for long range transport of particles. DI periods introduces populations of aged particles from far away sources into local regions. Identification and chemical characterization is performed for both of these periods to highlight the changes the DI period introduces in regards to particle morphology, chemical composition and lifetime. Analysis was performed via computer controlled scanning electron microscopy (CCSEM) for external information of the particles and scanning transmission x-ray microscopy (STXM) was used for internal information. The combination of these two techniques allows for a complete and thorough understanding of the particles during the two periods. </p> <p>Chapter 4 covers the first experiment done on a newly constructed cryogenic cooling cart which was created in the hopes to identify individual ice nucleating events of particles <em>in situ</em> when mimicking real world atmospheric conditions through temperature and humidity control. </p>

To design high pressure fuel system components that resist wear and scuffing failure when operated in low viscosity fuels, a comprehensive study on the tribological performance of various existing coating materials is necessary. This thesis aims to provide the relative performance of a variety of coating materials across different fuel environments by testing them in conditions that model those experienced in fuel pumps. The relative performance of these coatings are then indexed across a variety of material properties, including hardness, elastic modulus, wettability, and the interaction between the surface and the various types of fuel molecules.

Pressure swirl atomizers are commonly used in IC, aero-engines, and liquid propellant rocket combustion. Understanding the atomization process is important in order to enhance vaporization, mitigate soot formation, design of combustion chambers, and improve overall combustion efficiency. This work utilizes non-invasive techniques such as ultra -speed imaging, and Phase Doppler Particle Anemometry (PDPA) in order to investigate the cascade atomization process of pressure-swirl atomizers by examining swirling liquid film dynamics and the localized droplet characteristics of the resulting hollow cone spray. Specifically, experiments were conducted to examine these effects for three different nozzles with orifice diameters .3mm, .5mm, and .97mm. The ultra-speed imaging allowed for both visualization and interface tracking of the swirling conical film which emanated from each nozzle. Moreover, this allowed for the measurement of the radial fluctuations, film length, cone angle and maximum wavelength. Radial fluctuations are found to be maximum near the breakup or rupture of a swirling film. Film length decreases as Reynolds number increases. Cone angle increases until a critical Reynolds number is reached, beyond which it remains constant. A new approach to analyze the temporally unstable waves was developed and compared with the measured maximum wavelengths. The new approach incorporates the attenuation of a film thickness, as the radius of a conical film expands, with the classical dispersion relationship for an inviscid moving liquid film. This approach produces a new long wave solution which accurately matches the measured maximum wavelength swirling conical films generated from nozzles with the smallest orifice diameter. For the nozzle with the largest orifice diameter, the new long wave solution provides the upper bound limit, while the long wave solution for a constant film thickness provides the lower bound limit. These results indicate that temporal instability is the dominating mechanism which generates long Kelvin Helmholtz waves on the surface of a swirling liquid film. The PDPA was used to measure droplet size and velocity in both the near field and far field of the spray. For a constant Reynolds number, an increase in orifice diameter is shown to increase the overall diameter distribution of the spray. In addition, it was found that the probability of breakup, near the axis, decreases for the largest orifice diameter. This is in agreement with the cascading nature of atomization.

Die Dichtefunktionaltheorie (DFT) einschließlich Dispersionkorrekturen (+D) wird mit der Random-Phase-Approximation (RPA) für die Adsorption von Alkanen auf der Pt(111)-Oberfläche verglichen. RPA wird zuerst im Hinblick auf relevante technische Parameter evaluiert und für die Methanadsorption an der Pt(111)-Oberfläche getestet. Im Vergleich zum Perdew-Burke-Ernzerhof-Funktional (PBE) mit Tkatchenkos Many-Body-Dispersionskorrektur (PBE+MBD) liefert RPA gute Ergebnisse. Auch reproduziert RPA experimentelle Adsorptionsenergien bei verschiedenen, physikalisch sinnvollen Beladungsstufen der Pt(111) Oberfläche mit Alkanmolekülen. Für Platin in der hexagonal dichtesten Kugelpackung sagt RPA richtigerweise die Methanadsorption an der hollow-tripod-Stelle voraus, während mit PBE+MBD die Adsorption an einer anderen Stelle bevorzugt wäre. Dies geht aus Schwingungsspektren hervor. Da periodisches RPA sehr rechenaufwändig ist, wird ein QM:QM Hybridansatz (QM=Quantenmechanik) angewendet, wobei periodisches PBE(+D) mithilfe von RPA Rechnungen an Clustern korrigiert wird (RPA:PBE(+D)). In einem Test verschiedener Dispersionskorrekturen schneiden RPA:PBE und RPA:PBE+MBD am besten ab. Diese Arbeit ist wegbereitend für die Anwendung des QM:QM Hybridansatzes zur Beschreibung der Adsorptionsprozesse an Metalloberflächen ‒ bei hoher Genauigkeit und deutlich verringertem Rechenaufwand. Auch Kresses low-scaling RPA Algorithmus wird getestet. Dieser Algorithmus ermöglicht, große Systeme, wie z.B. die Methan-, Ethan-, Propan- und n-Butanadsorption an Pt(111), zu untersuchen. Der Vergleich mit experimentellen Daten zeigt, dass mit RPA stets die beste Übereinstimmung erreicht wird. Dabei wird eine deutliche Verbesserung gegenüber allen untersuchten Dichte-Funktionalen erzielt. Obwohl Bindungen mit RPA etwas zu schwach vorhergesagt werden, ist es die derzeit beste Methode zur Untersuchung der Adsorption an Metalloberflächen und damit der Benchmark für diese Systeme. / Density Functional Theory (DFT) including dispersion (+D) is compared against the Random Phase Approximation (RPA) for the adsorption of alkanes on the Pt(111) surface. RPA is first benchmarked with respect to technical parameters and tested for methane adsorption on Pt(111). It is found to perform well relative to the Perdew–Burke–Ernzerhof (PBE) functional augmented with the many-body dispersion scheme of Tkatchenko (PBE+MBD). It also compares well relative to experimentally derived adsorption energies at physically relevant coverages. RPA correctly assigns the adsorption of methane to the hcp (hexagonal close packed) hollow tripod site, matching vibrational spectra, whereas PBE+MBD found another site. Given the high cost of periodic RPA, a high-level: low-level QM:QM (QM = quantum mechanics) hybrid approach is applied using RPA (RPA:PBE(+D)), which has also been tested with several dispersion corrections, with RPA:PBE and RPA:PBE+MBD performing best. This extends the QM:QM hybrid approach to the study of adsorption on metal surfaces, resulting in high accuracy at significantly reduced cost. Finally we test the performance of the low-scaling RPA algorithm of Kresse and co-workers. This algorithm enables the study of larger systems and is applied to the first four n-alkanes (C1-C4) on the Pt(111) surface. Comparison against experiment indicates that RPA offers the best agreement, consistently better than any studied density functional. RPA underbinds slightly but is still found to be the best method for studying adsorption on metal surfaces and is the current benchmark for such systems.