Thermodiffusion is one of the mechanisms in transport phenomena in which molecules are transported in a multicomponent mixture driven by temperature gradients. Thermodiffusion in associating mixtures presents a larger degree of complexity than non-associating mixtures, since the direction of flow in associating mixtures may change with variations in composition and temperature. In this study a new activation energy model is proposed for predicting the ratio of evaporation energy to activation energy. The new model has been implemented for prediction of thermodiffusion for acetone-water, ethanol-water and isopropanol-water mixtures. In particular, a sign change in the thermodiffusion factor for associating mixtures has been predicted, which is a major step forward in modeling of thermodiffusion for associating mixtures. In addition, a new model for the prediction of thermodiffusion coefficients for linear chain hydrocarbon binary mixtures is proposed using the theory of irreversible thermodynamics and a kinetics approach. The model predicts the net amount of heat transported based on an available volume for each molecule. This model has been found to be the most reliable and represents a significant improvement over the earlier models. Also a new approach to predicting the Soret coefficient in binary mixtures of linear chain and aromatic hydrocarbons using the thermodynamics of irreversible processes is presented. This approach is based on a free volume theory which explains the diffusivity in diffusion-limited systems. The proposed model combined with the Shukla and Firoozabadi model has been applied to predict the Soret coefficient for binary mixtures of toluene and n-hexane, and benzene and n-heptane. Comparisons of theoretical results with experimental data show a good agreement. The proposed model has also been applied to estimate thermodiffusion coefficients of binary mixtures of n-butane & carbon dioxide and n-dodecane & carbon dioxide at different temperature. The results have also been incorporated into CFD software FLUENT for 3-dimensional simulations of thermodiffusion and convection in porous media. The predictions show the thermodiffuison phenomenon is dominant at low permeabilities (0.0001 to 0.01), but as the permeability increases convection plays an important role in establishing a concentration distribution. Finally, the activation energy in Eyring’s viscosity theory is examined for associating mixtures. Several methods are used to estimate the activation energy of pure components and then extended to mixtures of linear hydrocarbon chains. The activation energy model based on alternative forms of Eyring’s viscosity theory is implemented to estimate the thermodiffusion coefficient for hydrocarbon binary mixtures. Comparisons of theoretical results with the available thermodiffusion coefficient data have shown a good performance of the activation energy model.

Thermodiffusion is one of the mechanisms in transport phenomena in which molecules are transported in a multicomponent mixture driven by temperature gradients. Thermodiffusion in associating mixtures presents a larger degree of complexity than non-associating mixtures, since the direction of flow in associating mixtures may change with variations in composition and temperature. In this study a new activation energy model is proposed for predicting the ratio of evaporation energy to activation energy. The new model has been implemented for prediction of thermodiffusion for acetone-water, ethanol-water and isopropanol-water mixtures. In particular, a sign change in the thermodiffusion factor for associating mixtures has been predicted, which is a major step forward in modeling of thermodiffusion for associating mixtures. In addition, a new model for the prediction of thermodiffusion coefficients for linear chain hydrocarbon binary mixtures is proposed using the theory of irreversible thermodynamics and a kinetics approach. The model predicts the net amount of heat transported based on an available volume for each molecule. This model has been found to be the most reliable and represents a significant improvement over the earlier models. Also a new approach to predicting the Soret coefficient in binary mixtures of linear chain and aromatic hydrocarbons using the thermodynamics of irreversible processes is presented. This approach is based on a free volume theory which explains the diffusivity in diffusion-limited systems. The proposed model combined with the Shukla and Firoozabadi model has been applied to predict the Soret coefficient for binary mixtures of toluene and n-hexane, and benzene and n-heptane. Comparisons of theoretical results with experimental data show a good agreement. The proposed model has also been applied to estimate thermodiffusion coefficients of binary mixtures of n-butane & carbon dioxide and n-dodecane & carbon dioxide at different temperature. The results have also been incorporated into CFD software FLUENT for 3-dimensional simulations of thermodiffusion and convection in porous media. The predictions show the thermodiffuison phenomenon is dominant at low permeabilities (0.0001 to 0.01), but as the permeability increases convection plays an important role in establishing a concentration distribution. Finally, the activation energy in Eyring’s viscosity theory is examined for associating mixtures. Several methods are used to estimate the activation energy of pure components and then extended to mixtures of linear hydrocarbon chains. The activation energy model based on alternative forms of Eyring’s viscosity theory is implemented to estimate the thermodiffusion coefficient for hydrocarbon binary mixtures. Comparisons of theoretical results with the available thermodiffusion coefficient data have shown a good performance of the activation energy model.

京都大学 / 新制・課程博士 / 博士(工学) / 甲第23388号 / 工博第4880号 / 新制||工||1763(附属図書館) / 京都大学大学院工学研究科材料化学専攻 / (主査)教授 三浦 清貴, 教授 田中 勝久, 教授 藤田 晃司 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DGAM

Esta dissertação apresenta um estudo experimental sobre o transporte de massa de nanopartículas magnéticas induzido por um gradiente de temperatura, denominado termodifusão. A técnica de Varredura-Z é utilizada para gerar o aumento de temperatura na região irradiada pelo laser Gaussiano e formar o gradiente de temperatura. A sequente migração de nanopartículas é caracterizada pelo gradiente de concentração gerado no estado estacionário do fluxo de partículas, definindo o coeficiente Soret ST. O objetivo deste trabalho é verificar a variação de ST com o tamanho médio d0 das nanopartículas de ferrofluidos eletrostaticamente carregados em solução ácida. A dependência de ST com d0 surge do coeficiente de difusão de massa, explicando a dependência linear encontrada experimentalmente. Nestes materiais, a migração de nanopartículas ocorre para a região quente da amostra. Mostramos que este comportamento ocorre pela diminuição da carga superficial da nanopartícula na parte mais quente da amostra, diminuindo a repulsão eletrostática. A influência dos íons presentes na solução é obtida através da mudança na amplitude de ST com a diminuição do pH na amostra. Uma previsão teórica, baseada na eletrostática da dupla camada elétrica, concorda com estes dados considerando alta blindagem eletrostática das nanopartículas e a diminuição da carga superficial com o aumento da temperatura. / This dissertation presents an experimental study about the mass transport of magnetic nanoparticles induced by a temperature gradient, called thermodiffusion. The Z-scan technique is used to generate the temperature increasing in the region irradiated by the Gaussian laser beam and create the temperature gradient. The following nanoparticles migration is characterized by the concentration gradient of the stationary particles flux, defining the Soret coefficient ST. The aim of this work is to obtain the variation of ST with the average size d0 of electrostatically charged ferrofluid nanoparticles in acidic solution. The ST dependence with d0 comes from the mass diffusion coefficient, in agreement with the linear dependence found experimentally. In these materials the nanoparticles migration occurs to the hot region of the sample. We show that this behavior is owing to the reduction of the nanoparticle´s surface charge in the hottest region of the sample, decreasing electrostatic repulsion. The influence of the ions from solution is obtained through the change in ST amplitude with reduction of the samples pH. A theoretical prediction, based in the electrostatic of the double layer, agrees with this data considering a high screening of the nanoparticles and decreasing of the surface charge with temperature increasing.

Esta dissertação apresenta um estudo experimental sobre o transporte de massa de nanopartículas magnéticas induzido por um gradiente de temperatura, denominado termodifusão. A técnica de Varredura-Z é utilizada para gerar o aumento de temperatura na região irradiada pelo laser Gaussiano e formar o gradiente de temperatura. A sequente migração de nanopartículas é caracterizada pelo gradiente de concentração gerado no estado estacionário do fluxo de partículas, definindo o coeficiente Soret ST. O objetivo deste trabalho é verificar a variação de ST com o tamanho médio d0 das nanopartículas de ferrofluidos eletrostaticamente carregados em solução ácida. A dependência de ST com d0 surge do coeficiente de difusão de massa, explicando a dependência linear encontrada experimentalmente. Nestes materiais, a migração de nanopartículas ocorre para a região quente da amostra. Mostramos que este comportamento ocorre pela diminuição da carga superficial da nanopartícula na parte mais quente da amostra, diminuindo a repulsão eletrostática. A influência dos íons presentes na solução é obtida através da mudança na amplitude de ST com a diminuição do pH na amostra. Uma previsão teórica, baseada na eletrostática da dupla camada elétrica, concorda com estes dados considerando alta blindagem eletrostática das nanopartículas e a diminuição da carga superficial com o aumento da temperatura. / This dissertation presents an experimental study about the mass transport of magnetic nanoparticles induced by a temperature gradient, called thermodiffusion. The Z-scan technique is used to generate the temperature increasing in the region irradiated by the Gaussian laser beam and create the temperature gradient. The following nanoparticles migration is characterized by the concentration gradient of the stationary particles flux, defining the Soret coefficient ST. The aim of this work is to obtain the variation of ST with the average size d0 of electrostatically charged ferrofluid nanoparticles in acidic solution. The ST dependence with d0 comes from the mass diffusion coefficient, in agreement with the linear dependence found experimentally. In these materials the nanoparticles migration occurs to the hot region of the sample. We show that this behavior is owing to the reduction of the nanoparticle´s surface charge in the hottest region of the sample, decreasing electrostatic repulsion. The influence of the ions from solution is obtained through the change in ST amplitude with reduction of the samples pH. A theoretical prediction, based in the electrostatic of the double layer, agrees with this data considering a high screening of the nanoparticles and decreasing of the surface charge with temperature increasing.

No description available.

On étudie l'évolution d'une solution initialement homogène constituée de deux espèces soumises à un gradient thermique qui génère un transfert de matière, ce qui peut conduire à la séparation des espèces du fluide binaire. La configuration choisie pour étudier la séparation est une cellule rectangulaire (ou parallélépipédique), horizontale et placée dans le champ de pesanteur. La présence d'une source piézo-électrique, sur l'une des parois verticales de la cavité, permet de générer un écoulement stationnaire à grande échelle. L'écoulement est induit par la propagation d'ondes ultrasonores au sein du fluide visqueux : la dissipation de l'énergie acoustique de l'onde au sein du fluide porte le nom d'Eckart streaming. On cherche à optimiser la séparation en combinant gradient thermique et source acoustique. La première partie consiste en l'étude de l'écoulement isotherme généré par l'onde ultrasonore dans un fluide mono-constituant. Après avoir calculé le champ d'intensité acoustique avec l'intégrale de Rayleigh, le profil est implémenté dans un code aux éléments finis Comsol Multiphysics. Les résultats numériques sont comparés avec des résultats expérimentaux antérieurs. Dans une seconde partie, on considère une cavité contenant un fluide binaire. On détermine analytiquement, à l'aide du logiciel Maple, la séparation (différence de fraction massique entre les deux extrémités de la cellule) en fonction des paramètres de contrôle du problème. Des simulations numériques 2D et 3D ont montré un bon accord entre les résultats analytiques et numériques, pour un paramètre acoustique constant et un chauffage par le bas ou par le haut de la cellule. Le problème considéré dépend alors de huit paramètres adimensionnels. Trois d'entre eux sont propres à la nature du fluide binaire : le nombre de Lewis Le, de Prandtl Pr et le facteur de séparation ψ. Il y a ensuite deux paramètres de contrôle, le nombre de Rayleigh thermique Ra et la force acoustique adimensionnelle A. Enfin, les autres paramètres adimensionnels sont les deux rapports d'aspect de la cavité, ainsi que l'épaisseur relative du faisceau acoustique / The evolution of an initially homogeneous solution consisting of two species subjected to a thermal gradient which generates a mass transfer, which can lead to the separation species from the binary fluid, is studied. The configuration chosen to study the separation is a rectangular (or parallelepipedic) cell, horizontal and placed in the gravitational field. The presence of a piezoelectric source on one of the vertical walls of the cavity makes it possible to generate a stationary flow on a large scale. The flow is induced by the propagation of ultra-sonic waves within the viscous fluid : the dissipation of the acoustic energy of the wave within the fluid is called Eckart streaming. We seek to optimize the separation by combining thermal gradient and acoustic source. The first part consists of the study of the isothermal flow generated by the ultrasonic wave in a monoconstituent fluid. After calculating the acoustic intensity field with the Rayleigh integral, the profile is implemented in a Comsol Multiphysics finite element code. The numerical results are compared with previous experimental results. In a second part, we consider a cavity containing a binary fluid. A configuration heated from the top is privilegied to allow the insertion of an acoustic source. The separation (difference of mass fraction between the two ends of the cell) is determined analytically using the Maple software as a function of the control parameters of the problem. 2D and 3D numerical simulations showed a good agreement between the analytical and numerical results, for a constant acoustic parameter. The problem considered depends on eight dimensionless parameters. Three of them are specific to the nature of the binary fluid : the Lewis number Le, the Prandtl number Pr and the separation factor ψ. There are then two control parameters, the thermal Rayleigh number Ra and the adimensional acoustic force A. Finally, the other dimensionless parameters are the two aspect ratios of the cavity, as well as the relative thickness of the beam.

Les conséquences liées à la présence de gradients thermiques sur le transfert de matière en milieu poreux sont encore aujourd’hui mal appréhendées, essentiellement en raison de la complexité induite par la présence de phénomènes couplés (thermodiffusion ou effet Soret). Le but de cette thèse est d’étudier et de comprendre l’influence que peut avoir un gradient thermique sur l’écoulement d’un mélange. L’objectif principal est de déterminer les coefficients effectifs modélisant les transferts de chaleur et de matière en milieux poreux, et en particulier le coefficient de thermodiffusion effectif. En utilisant la technique de changement d’échelle par prise de moyenne volumique nous avons développé un modèle macroscopique de dispersion incluant la thermodiffusion. Nous avons étudié en particulier l'influence du nombre de Péclet et de la conductivité thermique sur la thermodiffusion. Les résultats ont montré que pour de faibles nombres de Péclet, le nombre de Soret effectif en milieu poreux est le même que dans un milieu libre, et ne dépend pas du ratio de la conductivité thermique (solide/liquide). À l'inverse, en régime convectif, le nombre de Soret effectif diminue. Dans ce cas, un changement du ratio de conductivité changera le coefficient de thermodiffusion effectif. Les résultats théoriques ont montré également que, lors de la diffusion pure, même si la conductivité thermique effective dépend de la connectivité de la phase solide, le coefficient effectif de thermodiffusion est toujours constant et indépendant de la connectivité de la phase solide. Le modèle macroscopique obtenu par cette méthode est validé par comparaison avec des simulations numériques directes à l'échelle des pores. Un bon accord est observé entre les prédictions théoriques provenant de l'étude à l’échelle macroscopique et des simulations numériques au niveau de l’échelle de pores. Ceci démontre la validité du modèle théorique proposé. Pour vérifier et consolider ces résultats, un dispositif expérimental a été réalisé pour mesurer les coefficients de transfert en milieu libre et en milieu poreux. Dans cette partie, les nouveaux résultats expérimentaux sont obtenus avec un système du type « Two-Bulb apparatus ». La diffusion et la thermodiffusion des systèmes binaire hélium-azote et hélium-dioxide de carbone, à travers des échantillons cylindriques remplis de billes de différents diamètres et propriétés thermiques, sont mesurées à la pression atmosphérique. La porosité de chaque milieu a été déterminée par la construction d'une image 3D de l'échantillon par tomographie. Les concentrations sont déterminées par l'analyse en continu de la composition du mélange de gaz dans les ampoules à l’aide d’un catharomètre. La détermination des coefficients de diffusion et de thermodiffusion est réalisée par confrontation des relevés temporels des concentrations avec une solution analytique modélisant le transfert de matière entre deux ampoules. Les résultats sont en accord avec les résultats théoriques. Cela permet de conforter l’influence de la porosité des milieux poreux sur les mécanismes de diffusion et de thermodiffusion. / A multicomponent system, under nonisothermal condition, shows mass transfer with cross effects described by the thermodynamics of irreversible processes. The flow dynamics and convective patterns in mixtures are more complex than those of one-component fluids due to interplay between advection and mixing, solute diffusion, and thermal diffusion (or Soret effect). This can modify species concentrations of fluids crossing through a porous medium and leads to local accumulations. There are many important processes in nature and industry where thermal diffusion plays a crucial role. Thermal diffusion has various technical applications, such as isotope separation in liquid and gaseous mixtures, identification and separation of crude oil components, coating of metallic parts, etc. In porous media, the direct resolution of the convection-diffusion equations are practically impossible due to the complexity of the geometry; therefore the equations describing average concentrations, temperatures and velocities must be developed. They might be obtained using an up-scaling method, in which the complicated local situation (transport of energy by convection and diffusion at pore scale) is described at the macroscopic scale. At this level, heat and mass transfers can be characterized by effective tensors. The aim of this thesis is to study and understand the influence that can have a temperature gradient on the flow of a mixture. The main objective is to determine the effective coefficients modelling the heat and mass transfer in porous media, in particular the effective coefficient of thermodiffusion. To achieve this objective, we have used the volume averaging method to obtain the modelling equations that describes diffusion and thermodiffusion processes in a homogeneous porous medium. These results allow characterising the modifications induced by the thermodiffusion on mass transfer and the influence of the porous matrix properties on the thermodiffusion process. The obtained results show that the values of these coefficients in porous media are completely different from the one of the fluid mixture, and should be measured in realistic conditions, or evaluated with the theoretical technique developed in this study. Particularly, for low Péclet number (diffusive regime) the ratios of effective diffusion and thermodiffusion to their molecular coefficients are almost constant and equal to the inverse of the tortuosity coefficient of the porous matrix, while the effective thermal conductivity is varying by changing the solid conductivity. In the opposite, for high Péclet numbers (convective regime), the above mentioned ratios increase following a power law trend, and the effective thermodiffusion coefficient decreases. In this case, changing the solid thermal conductivity also changes the value of the effective thermodiffusion and thermal conductivity coefficients. Theoretical results showed also that, for pure diffusion, even if the effective thermal conductivity depends on the particle-particle contact, the effective thermal diffusion coefficient is always constant and independent of the connectivity of the solid phase. In order to validate the theory developed by the up-scaling technique, we have compared the results obtained from the homogenised model with a direct numerical simulation at the microscopic scale. These two problems have been solved using COMSOL Multiphysics, a commercial finite elements code. The results of comparison for different parameters show an excellent agreement between theoretical and numerical models. In all cases, the structure of the porous medium and the dynamics of the fluid have to be taken into account for the characterization of the mass transfer due to thermodiffusion. This is of great importance in the concentration evaluation in the porous medium, like in oil reservoirs, problems of pollution storages and soil pollution transport. Then to consolidate these theoretical results, new experimental results have been obtained with a two-bulb apparatus are presented. The diffusion and thermal diffusion of a helium-nitrogen and helium-carbon dioxide systems through cylindrical samples filled with spheres of different diameters and thermal properties have been measured at the atmospheric pressure. The porosity of each medium has been determined by construction of a 3D image of the sample made with an X-ray tomograph device. Concentrations are determined by a continuous analysing the gas mixture composition in the bulbs with a katharometer device. A transient-state method for coupled evaluation of thermal diffusion and Fick coefficients in two bulbs system has been proposed. The determination of diffusion and thermal diffusion coefficients is done by comparing the temporal experimental results with an analytical solution modelling the mass transfer between two bulbs. The results are in good agreement with theoretical results and emphasize the porosity of the medium influence on both diffusion and thermal diffusion process. The results also showed that the effective thermal diffusion coefficients are independent from thermal conductivity ratio and particle-particle touching.

Le comportement d’un fluide confiné dans un milieu poreux peu perméable (micro- and méso-pores) a été étudié en ce qui concerne ses propriétés de diffusion de masse, de conductivité thermique et de thermodiffusion. Pour ce faire des simulations de dynamique moléculaire hors équilibre ont été réalisées sur des mélanges binaires modèles placés dans des conditions thermodynamiques diverses, confinés dans des milieux poreux de géométrie lamellaire de différentes natures (lisse ou atomique, plus ou moins adsorbant) en utilisant l’ensemble __//_ et l’ensemble grand canonique. Les résultats ont montré que les effets du milieu poreux sur les propriétés de transport sont d’autant plus marqués que lataille de pore est petite, que l’adsorption est forte et que la température est basse. Les résultats ont permis d’évaluer quantitativement ces effets. Il a aussi été montré que la rugosité des murs a un impact très important sur le coefficient de diffusion de masse et non négligeable sur celui de thermodiffusion. / The aim of this work was to study how a fluid confined in a low permeability porous medium (micro- and meso-porous) behaves concerning its properties of mass diffusion, thermal conductivity and thermal diffusion. For this purpose, non-equilibrium molecular dynamics simulations have been performed on simple binary mixtures placed in various thermodynamic conditions, confined in a porous medium of lamellar geometry of different types (structure-less or atomistic, more or less adsorbent) in __//_ and grand canonical ensembles. The results show that the effects of porous medium on transport properties are more pronounced when the pore size is small, the adsorption is strong and the temperature is low. The results allowed to evaluate these effects quantitatively. In addition, it has been found that the wall roughness has a major impact on the mass diffusion coefficient and a non negligible one on the thermal diffusion coefficient.

Des questions fondamentales concernant les propriétés de diffusion des systèmes biologiques dans des conditions isothermes et non-isothermes restent en suspens en raison de l’absence de techniques expérimentales capables de visualiser et de mesurer les phénomènes de diffusion avec une très bonne précision. Il existe en conséquence un besoin de développer de nouvelles techniques expérimentales permettant d’approfondir notre compréhension des phénomènes de diffusion. La convection naturelle en cavité tridimensionnelle inclinée est elle-aussi très peu étudiée. Cette inclinaison de la cavité peut correspondre à un léger défaut expérimental ou être imposée volontairement. Dans cette thèse, nous étudions les phénomènes de transport de chaleur et de masse en cavité parallélépipédique, nous intéressant particulièrement à la thermodiffusion en situation sans convection et à la convection naturelle en fluide pur (sans thermodiffusion). La diffusion de masse est étudiée à l’aide d’une nouvelle technique optique, tandis que la convection naturelle est tout d’abord étudiée en détails avec une méthode numérique sophistiquée, puis visualisée expérimentalement à l’aide du même système optique que pour les mesures de diffusion. Nous présentons l’interféromètre optique de haute précision développé pour les mesures de diffusion. Cet interféromètre comprend un interféromètre polarisé de Mach–Zehnder, un polariseur tournant, une caméra CCD et un algorithme de traitement d’images original. Nous proposons aussi une méthode pour déterminer le coefficient de diffusion isotherme en fonction de la concentration. Cette méthode, basée sur une analyse inverse couplée à un calcul numérique, permet de déterminer les coefficients de diffusion à partir des profils de concentration transitoires obtenus par le système optique. Mentionnons de plus que c’est la première fois que la thermodiffusion est visualisée dans des solutions aqueuses de protéines. La méthode optique proposée présente trois avantages principaux par rapport aux autres méthodes similaires : (i) un volume d’échantillon réduit, (ii) un temps de mesure court, (iii) une stabilité hydrodynamique améliorée. Toutes ces méthodes ont été validées par des mesures sur des systèmes de référence. La technique optique est d’abord utilisée pour étudier la diffusion isotherme dans des solutions de protéines : (a) dans des solutions binaires diluées, (b) dans des solutions binaires sur un large domaine de concentration, (c) dans des solutions ternaires diluées. Les résultats montrent que (a) le coefficient de diffusion isotherme dans les systèmes dilués décroit avec la masse moléculaire, comme prédit grossièrement par l’équation de Stokes-Einstein ; (b) la protéine BSA a un comportement diffusif de type sphère dure et la protéine lysozyme de type sphère molle ; (c) l’effet de diffusion croisée est négligeable dans les systèmes ternaires dilués. La technique optique est aussi utilisée (d) dans des solutions binaires diluées non-isothermes, révélant que les molécules d’aprotinin (6.5 kDa) et de lysozyme (14.3 kDa) sont, respectivement, thermophiliques et thermo-phobiques, quand elles sont en solutions aqueuses à température ambiante. Enfin, la technique optique est utilisée pour l’étude de la convection de Rayleigh-Bénard en cavité cubique horizontale. Puisque la convection peut aussi être étudiée de façon réaliste en utilisant les équations de Navier-Stokes, une analyse numérique de bifurcation est proposée, permettant une étude approfondie de la convection naturelle dans des cavités tridimensionnelles parallélépipédiques. Pour cela, une méthode de continuation a été développée à partir d’un code aux éléments finis spectraux. La méthode numérique proposée est particulièrement bien adaptée aux études de convection correspondant à des diagrammes de bifurcation complexes. [...] / Fundamental questions concerning the mass diffusion properties of biological systems under isothermal and non-isothermal conditions still remain due to the lack of experimental techniques capable of visualizing and measuring mass diffusion phenomena with a high accuracy. As a consequence, there is a need to develop new experimental techniques that can deepen our understanding of mass diffusion. Moreover, steady natural convection in a tilted three-dimensional rectangular enclosure has not yet been studied. This tilt can be a slight defect of the experimental device or can be imposed on purpose. In this dissertation, heat and mass transfer phenomena in parallelepiped enclosures are studied focusing on convectionless thermodiffusion and on natural convection of pure fluids (without thermodiffusion). Mass diffusion is studied with a novel optical technique, while steady natural convection is first studied in detail with an improved numerical analysis and then with the same optical technique initially developed for diffusion measurements. A construction of a precise optical interferometer to visualize and measure mass diffusion is described. The interferometer comprises a polarizing Mach–Zehnder interferometer, a rotating polariser, a CCD camera, and an original image-processing algorithm. A method to determine the isothermal diffusion coefficient as a function of concentration is proposed. This method uses an inverse analysis coupled with a numerical calculation in order to determine the diffusion coefficients from the transient concentration profiles measured with the optical system. Furthermore, thermodiffusion of protein molecules is visualized for the first time. The proposed method has three main advantages in comparison to similar methods: (i) reduced volume sample, (ii) short measurement time, and (iii) increased hydrodynamic stability of the system. These methods are validated by determining the thermophysical properties of benchmark solutions. The optical technique is first applied to study isothermal diffusion of protein solutions in: (a) dilute binary solutions, (b) binary solutions with a wide concentration range, and (c) dilute ternary solutions. The results show that (a) the isothermal diffusion coefficient in dilute systems decreases with molecular mass, as roughly predicted by the Stokes-Einstein equation; (b) BSA protein has a hard-sphere-like diffusion behaviour and lysozyme protein a soft sphere characteristic; and (c) the cross-term effect between the diffusion species in a dilute ternary system is negligible. The optical technique is then applied to (d) non-isothermal dilute binary solutions, revealing that that the aprotinin (6.5 kDa) and lysozyme (14.3 kDa) molecules are thermophilic and thermophobic, respectively, when using water as solvent at room temperature. Finally, the optical technique is applied to study Rayleigh-Bénard convection in a horizontal cubical cavity. Since natural convection can be studied in more depth by solving the Navier-Stokes equations, a bifurcation analysis is proposed to conduct a thorough study of natural convection in three-dimensional parallelepiped cavities. Here, a continuation method is developed from a three-dimensional spectral finite element code. The proposed numerical method is particularly well suited for the studies involving complex bifurcation diagrams of three-dimensional convection in rectangular parallelepiped cavities. This continuation method allows the calculation of solution branches, the stability analysis of the solutions along these branches, the detection and precise direct calculation of the bifurcation points, and the jump to newly detected stable or unstable branches, all this being managed by a simple continuation algorithm. This can be used to calculate the bifurcation diagrams describing the convection in tilted cavities. [...]