Les milieux poreux font partie des formations géologiques assez répandue dans la nature et son sujet d'études intensives. L'engouement de ce sujet vient des multiples secteurs d'applications de ces recherches et leur importance dans notre société. Que ce soit de la part des sociétés pétrolières qui souhaitent optimiser leurs moyens de productions, les agences de contrôles environnementaux qui veulent prévenir la contamination de nappe phréatique et la fuite de déchets nucléaires ou encore des industriels avec des problèmes de drainages et de réhabilitation de mines, tous ces acteurs dépendent des recherches faites dans ce domaine. Cependant, un des principaux problèmes de ce sujet est l'inaccessibilité des milieux que nous voulons étudier. Pour palier à cela de nombreuses équipes se sont tournées vers la simulation numérique. Cette thèse s'inscrit dans ce cadre et utilise le module PARADIS du logiciel d'hydrogéologie H2olab pour modéliser le transport de particules dans des milieux poreux fortement hétérogènes. Grâce aux données obtenues et à des comparaisons avec la littérature nous montrerons l'effet du passage au 3D sur la topologie de l'écoulement et les répercussions sur le transport de particules ainsi que l'effet de la diffusion moléculaire sur les coefficients de macro-dispersion. Enfin nous proposerons deux lois de transport reliant macrodispersion, variance du champ de perméabilité et diffusion moléculaire. / Heterogeneous porous media have been intensively studied these last fifty years. The popularity of this subject come from the multiple areas where these researches can be applied and their importance to our society. Whether from the oil companies that want to optimize their methods of production, environmental control agencies who want to prevent contamination of ground water and leakage of nuclear waste or industrial with drainage issues and mine rehabilitation, all these actors depend on research done in this area. However, one of the main problems of this subject is the accessibility of these porous media which are often several hundred meters underground. To overcome this, many teams have turned to computer simulation. This thesis is among them and uses the PARADIS module from the hydrogeology software H2olab to model particle transport in highly heterogeneous porous media. Thanks to the data obtained and comparisons with the literature, we show the impact of switching from a 2D to a 3D porous media on the ow topology and the repercussions on the particle transport. Furthermore, we also investigated the effect of molecular diffusion coefficients on macro-dispersion. Finally, we will propose two empirical functions linking macro-dispersion variance of the permeability field and molecular diffusion.

Ce mémoire porte sur la mise au point de plateforme microfluidique couplée à la microscopie Raman confocale, utilisée dans des conditions d’excitation de la diffusion Raman (diffusion Raman exaltée de surface), dans le but d’obtenir une détection de très haute sensibilité d’espèces moléculaires sous écoulement dans des canaux de dimensions micrométriques. Ce travail a pour ambition de démontrer la faisabilité d’un couplage microscopie Raman/microfluidique en vue de la caractérisation in-situ et locale, des espèces et des réactions mises en jeu dans les fluides en écoulement dans les microcanaux. Nous avons utilisé un microcanal de géométrie T, fabriqué par lithographie douce, dans lequel sont injectées, à vitesse constante, des nanoparticules métalliques d’or ou d’argent dans une des deux branches du canal et une solution de pyridine ou de péfloxacine dans l’autre branche. La laminarité et la stationnarité du processus nous ont permis de cartographier la zone de mélange et de mettre en évidence l’exaltation du signal de diffusion Raman de la pyridine et de la péfloxacine, obtenue grâce aux nanoparticules métalliques, dans cette zone d’interdiffusion. L’enregistrement successif de la bande d’absorption des nanoparticules d’argent (bande plasmon) et du signal de diffusion Raman de la péfloxacine, en écoulement dans un microcanal, nous a permis d’établir un lien entre la morphologie des nanostructures métalliques, et plus précisément l’état d’agrégation des nanoparticules d’argent, et l’exaltation du signal Raman de la péfloxacine observé. Nous avons alors modifié la géométrie du canal afin d’y introduire une solution d’électrolyte (NaCl et NaNO3) et de modifier localement la charge de surface des colloïdes d’argent en écoulement. Nous avons ainsi confirmé que la modification de l’état d’agrégation des nanoparticules d’argent, induite par l’ajout contrôlé de solutions d’électrolytes, permet d’amplifier le signal SERS de la péfloxacine et d’optimiser la détection en microfluidique. Enfin, nous avons développé une seconde approche qui consistait à mettre en place une structuration métallisée des parois d’un microcanal. Nous avons ainsi démontré que la fonctionnalisation chimique de surface via un organosilane (APTES) permettait de tapisser le canal avec des nanoparticules d’argent et d’amplifier le signal Raman des espèces en écoulement dans ce même microcanal. / This thesis focuses on the development of a microfluidic platform coupled with confocal Raman microscopy, used in excitation conditions of Raman scattering (Surface enhanced Raman scattering, SERS) in order to gain in the detection sensitivity of molecular species flowing in channels of micrometer dimensions. This work aims to demonstrate the feasibility of coupling Raman microscopy / microfluidics for the in situ and local characterization of species and reactions taking place in the fluid flowing in microchannels. We used a T-shaped microchannel, made by soft lithography, in which gold or silver nanoparticles injected at constant speed, in one of the two branches of the channel and a solution of pyridine or pefloxacin in the other one. The laminar flow and the stationarity of the process allowed us to map the mixing zone and highlight the enhancement of the Raman signal of pyridine and pefloxacin, due to the metallic nanoparticles, in the interdiffusion zone. The recording of the both absorption band of the silver nanoparticles (plasmon band) and the Raman signal of pefloxacin, flowing in microchannel, allowed us to establish a link between the shape of the metallic nanostructure, and more precisely the silver nanoparticle aggregation state, and the enhancement of the Raman signal of pefloxacin observed. We then changed the channel geometry to introduce an electrolyte solution (NaCl and NaNO3) and locally modify the surface charge of the colloids. We have put in evidence that the change of the silver nanoparticle aggregation state, induced by the controlled addition of electrolyte solutions, could amplify the SERS signal of pefloxacin and thus optimizing the detection in microfluidics. At last, we established second a approach that consists in the metallic structuring of microchannel walls. This has shown that the surface chemical functionalization through organosilanes (APTES) allowed the pasting of the channel with silver nanoparticles, thus amplifying the Raman signal of the species flowing within the same microchannel.

Canada has the second largest proven oil reserves next to Saudi Arabia which is mostly located in Alberta and Saskatchewan but is unconventional heavy oil and bitumen. The tar sands are found at the surface and are mined, yet 80% of the 173 billion barrels of heavy oil and bitumen exist in-situ according to the Canadian Association of Petroleum Producers (CAPP). Two factors inhibit the economic extraction and processing of Canadian heavy oil; its enormous viscosity ranging from 1000 to over 1 million mPa.s and the asphaltene content (high molecular weight molecules containing heavy metals and sulphur). Heavy oil and bitumen were only included in the reserves estimates through the efforts of Canadian enhanced oil recovery (EOR) research. Viscosity reduction is the one common element of in-situ methods of heavy oil recovery with the exception of cold production. Currently, steam assisted gravity drainage (SAGD) and cyclic steam stimulation (CSS) are being used commercially in the field where the oil’s viscosity is reduced by injecting steam. Thermal methods are energy intensive requiring vast volumes of water such that any improvement would be beneficial. Solvent extraction is one alternative requiring no water, the solvent is recoverable and reusable, and depending on the mode of operation the heavy oil is upgraded in-situ. Vapour Extraction (VAPEX) and enhanced solvent extraction (N-SolvTM) are two such methods. VAPEX and N-Solv reduce the bitumen’s viscosity via mass transfer and a combination of mass and heat transfer, respectively. A light hydrocarbon solvent (instead of steam) is injected into an upper horizontal well where the solvent mixes with the heavy oil, reduces its viscosity and allows the oil to drain under gravity to a bottom production well. The idea of using solvents for heavy oil extraction has been around since the 1970s and both VAPEX and N-Solv are patented processes. However, there is still much to be learned about how these processes physically work. Research to date has focused on varying system parameters (including model dimensions, permeability, heavy oil viscosity, solvent type and injection rate, etc.) to observe the effect on oil production from laboratory scale models. Based on an early mass balance model by Butler and Mokrys (1989) and an improvement by Das (1995), molecular diffusion alone cannot account for the produced oil rates observed from laboratory models. Until recently, very little progress had been made towards qualifying and quantifying the mass transfer mechanisms with the exception of the diffusivity of light hydrocarbons in heavy oil. Mass transfer can only be by diffusion and convection. Differentiating and quantifying the contribution of each is complex due to the nature and viscosity of the oil. The goal of this thesis is to investigate the mass transfer mechanisms during the solvent recovery of heavy oil. Quantifying the diffusion of light hydrocarbon solvents has been an active topic of research with limited success since the mid 1990’s. The experimental approach presented here focused on capturing the rate of solvent mass transfer into the bitumen by measuring the bitumen swelling and the butane uptake independently. Unlike early pressure decay methods, the pressure is held constant to not violate the assumed equilibrium solvent concentration at the interfacial boundary condition. The high solubility of solvent in heavy oil complicates the physical modeling because simplifying assumptions of a constant diffusion coefficient, constant density and a quiescent liquid should not be used. The model was developed from first principles to predict the bitumen swelling. The form of the concentration dependent diffusivity was assumed and the diffusivity coefficients initially guessed. The swelling (moving boundary) was fixed by defining a new dimensionless space coordinate and the set of partial differential equations solved using the method of lines. Using the non-linear regression (lsqnonlin) function in MATLAB®, optimising for the difference in predicted and experimentally found bitumen heights and independently validating the result using the solvent uptake, the diffusivity of butane in heavy oil (at 25oC) was found to be Dsb = 4.78 x 10-6ωs + 4.91 x 10-6 cm2/s where ωs is the solvent mass fraction. Diffusion alone has proven inadequate in predicting oil recovery rates from laboratory scale models. It is logical to assume that convective mass transfer plays a role at mixing the solvent and bitumen while draining via gravity through the reservoir porous matrix. Solvent extraction experiments were conducted in etched glass micromodels to observe the pore scale phenomena. The pore scale mechanisms were found to differ depending on how the solvent extraction was operated, with non-condensing (VAPEX) or condensing (N-SolvTM) solvent. Observations show increased convective mixing and an increased rate of interface advancement when the solvent condenses on the bitumen surface. Evidence of trapped butane vapour being mobilised with the draining live oil and a technique of observing solvent extraction using UV light confirm that the draining live oil is on average one pore deep. While the interface appears from a distance to be uniform, at the pore scale it is not. Live oil can drain from one to two pores via capillary displacement mechanisms in one section of the interface and via film flow only in another area (James and Chatzis 2004; James et al. 2008). This work also shows the detrimental impact of having a non-condensable gas present during solvent extraction (James and Chatzis 2008). In summary, this work emphasises the mass transfer and drainage displacement mechanisms of non-condensing (VAPEX) and condensing (N-Solv) solvent extraction methods of heavy oil recovery.

Canada has the second largest proven oil reserves next to Saudi Arabia which is mostly located in Alberta and Saskatchewan but is unconventional heavy oil and bitumen. The tar sands are found at the surface and are mined, yet 80% of the 173 billion barrels of heavy oil and bitumen exist in-situ according to the Canadian Association of Petroleum Producers (CAPP). Two factors inhibit the economic extraction and processing of Canadian heavy oil; its enormous viscosity ranging from 1000 to over 1 million mPa.s and the asphaltene content (high molecular weight molecules containing heavy metals and sulphur). Heavy oil and bitumen were only included in the reserves estimates through the efforts of Canadian enhanced oil recovery (EOR) research. Viscosity reduction is the one common element of in-situ methods of heavy oil recovery with the exception of cold production. Currently, steam assisted gravity drainage (SAGD) and cyclic steam stimulation (CSS) are being used commercially in the field where the oil’s viscosity is reduced by injecting steam. Thermal methods are energy intensive requiring vast volumes of water such that any improvement would be beneficial. Solvent extraction is one alternative requiring no water, the solvent is recoverable and reusable, and depending on the mode of operation the heavy oil is upgraded in-situ. Vapour Extraction (VAPEX) and enhanced solvent extraction (N-SolvTM) are two such methods. VAPEX and N-Solv reduce the bitumen’s viscosity via mass transfer and a combination of mass and heat transfer, respectively. A light hydrocarbon solvent (instead of steam) is injected into an upper horizontal well where the solvent mixes with the heavy oil, reduces its viscosity and allows the oil to drain under gravity to a bottom production well. The idea of using solvents for heavy oil extraction has been around since the 1970s and both VAPEX and N-Solv are patented processes. However, there is still much to be learned about how these processes physically work. Research to date has focused on varying system parameters (including model dimensions, permeability, heavy oil viscosity, solvent type and injection rate, etc.) to observe the effect on oil production from laboratory scale models. Based on an early mass balance model by Butler and Mokrys (1989) and an improvement by Das (1995), molecular diffusion alone cannot account for the produced oil rates observed from laboratory models. Until recently, very little progress had been made towards qualifying and quantifying the mass transfer mechanisms with the exception of the diffusivity of light hydrocarbons in heavy oil. Mass transfer can only be by diffusion and convection. Differentiating and quantifying the contribution of each is complex due to the nature and viscosity of the oil. The goal of this thesis is to investigate the mass transfer mechanisms during the solvent recovery of heavy oil. Quantifying the diffusion of light hydrocarbon solvents has been an active topic of research with limited success since the mid 1990’s. The experimental approach presented here focused on capturing the rate of solvent mass transfer into the bitumen by measuring the bitumen swelling and the butane uptake independently. Unlike early pressure decay methods, the pressure is held constant to not violate the assumed equilibrium solvent concentration at the interfacial boundary condition. The high solubility of solvent in heavy oil complicates the physical modeling because simplifying assumptions of a constant diffusion coefficient, constant density and a quiescent liquid should not be used. The model was developed from first principles to predict the bitumen swelling. The form of the concentration dependent diffusivity was assumed and the diffusivity coefficients initially guessed. The swelling (moving boundary) was fixed by defining a new dimensionless space coordinate and the set of partial differential equations solved using the method of lines. Using the non-linear regression (lsqnonlin) function in MATLAB®, optimising for the difference in predicted and experimentally found bitumen heights and independently validating the result using the solvent uptake, the diffusivity of butane in heavy oil (at 25oC) was found to be Dsb = 4.78 x 10-6ωs + 4.91 x 10-6 cm2/s where ωs is the solvent mass fraction. Diffusion alone has proven inadequate in predicting oil recovery rates from laboratory scale models. It is logical to assume that convective mass transfer plays a role at mixing the solvent and bitumen while draining via gravity through the reservoir porous matrix. Solvent extraction experiments were conducted in etched glass micromodels to observe the pore scale phenomena. The pore scale mechanisms were found to differ depending on how the solvent extraction was operated, with non-condensing (VAPEX) or condensing (N-SolvTM) solvent. Observations show increased convective mixing and an increased rate of interface advancement when the solvent condenses on the bitumen surface. Evidence of trapped butane vapour being mobilised with the draining live oil and a technique of observing solvent extraction using UV light confirm that the draining live oil is on average one pore deep. While the interface appears from a distance to be uniform, at the pore scale it is not. Live oil can drain from one to two pores via capillary displacement mechanisms in one section of the interface and via film flow only in another area (James and Chatzis 2004; James et al. 2008). This work also shows the detrimental impact of having a non-condensable gas present during solvent extraction (James and Chatzis 2008). In summary, this work emphasises the mass transfer and drainage displacement mechanisms of non-condensing (VAPEX) and condensing (N-Solv) solvent extraction methods of heavy oil recovery.