Constraining the behaviour of small fractions of partial melt in a solid silicate matrix has been the focus of numerous experimental petrology studies over several decades, and is an important factor in constraining upper mantle rheology, melt extraction at mid-ocean ridges and mechanisms of core formation in the early solar system. Deformation of partially molten rock has been observed to change melt geometry, and may enhance permeability and interconnectivity of melt otherwise trapped in a solid silicate matrix, although it is uncertain how applicable results of high strain-rate laboratory experiments are to the real Earth. The addition of deformation precludes attainment of textural equilibrium, complicating textural analysis, which has previously relied on extrapolation of 3D textures from quenched and polished 2D sections for hydrostatically annealed samples. X-ray computed tomography gives the potential to visualise sample textures directly in three dimensions, and is becoming popular as a complementary technique for textural analysis in petrologic studies. The aim of this project has been to develop techniques to improve visualisation of small fractions of partial melt within a solid silicate matrix using X-ray CT, to examine textures of various partially molten systems at high PT in hydrostatic, and dynamically deforming systems. Experiments carried out in the FeS-melt, solid olivine system have examined the potential for deformation-enhanced percolation of core forming melts before the onset of silicate melting. Access to the newly designed rotational Paris-Edinburgh Cell (roPEC/rotoPEC) equipment has allowed us to carry out controlled, torsional deformation experiments under PT conditions applicable to planetary interiors. Experiments conducted at lower strain-rates over longer duration than in previously published studies show that deformation enhances connectivity at low melt fractions, at strain-rates down to 10-6s-1. This is in contrast to earlier work suggesting melt textures are unaffected at strain-rates below 10-5s-1. Quenched melt networks have been fully characterised in 3D using multi-scale CT, with voxel sizes down to 70nm for small sample sub-volumes. Results suggest segregation of metallic melt below the silicate solidus could be an efficient process, and should be taken into account in geochemical models of planetary evolution. Experiments on basaltic melt in a solid silicate matrix were conducted in application to upper mantle melting. A heavy element, hafnium, was added to the basaltic glass starting composition to enhance contrast between the basalt and olivine phases during CT scans. In-house micro-CT equipment was used to visualise post-quench run products of hydrostatic and deformation experiments. The doping technique was successful for long-duration, high temperature hydrostatic experiments. Some issues with undissolved / re-precipitated HfO¬2 crystals complicated tomographic imaging of partial melt textures in a number of experiments, particularly those carried out on the rotoPEC equipment, limiting comparison between samples. The doping technique requires further adjustment, but is shown to be a viable way to improve visibility of basaltic melt without significantly affecting melt texture. The X-ray transparent design and fully rotating top and bottom anvils of the rotoPEC allow X-ray tomography to be carried out in-situ while experiments are in progress, enabling collection of 4D datasets. During this project, the rotoPEC equipment was incorporated into two different synchrotron beamlines, to carry out time-resolved studies of textural development within samples of varying composition. The migration of gold melt along fractures with a BN matrix was imaged using 2D radiography, in combination with repeated 3D tomography to fully characterise the 3D fracture geometry. This allowed melt migration velocity to be estimated directly from in-situ observations. These techniques could be developed further to constrain melt migration processes quantitatively for a number of geological systems in the near future.

La problématique de cette thèse est l’identification des phénomènes physico-chimiques se produisant entre les roches encaissantes du gisement de sel gemme de Lorraine et des saumures, puis l’étude de leurs effets sur le comportement mécanique. Une démarche multi-échelle a été entreprise : du minéral à l’échelle d’une exploitation en passant par celle des essais mécaniques. La confrontation de la géologie locale avec la géométrie des cavités de dissolution de sel (logiciel GOCAD) au travers de ces roches encaissantes rend compte d’un délitage relativement rapide lorsque les argilites sont au contact d’une saumure de cavité. A l’inverse, l’anhydrite et la dolomie peuvent constituer le toit de cavité pendant plusieurs années, avant de se rompre. Suite à une caractérisation minéralogique, et microtexturale des roches encaissantes (Marnes irisées inférieures et moyennes), une expérimentation de type batch a été mise en œuvre afin de comprendre ces différences: les faciès lithologiques qualifiés de majeurs ont été immergés dans des saumures pendant plus d’un an. Des analyses chimiques et microscopiques ont permis de suivre les modifications minéralogiques et microtexturales. Au contact de saumures saturées en chlorure de sodium, les faciès argileux se délitent en raison de l’hydratation en gypse du minéral anhydrite. Au niveau de l’anhydrite massive, cette hydratation n’affecte qu’une frange superficielle des échantillons, en raison d’une porosité connectée quasi-nulle. En présence d’une saumure de cavité, l’hydratation est promue par la présence de potassium et de strontium en solution. Ces résultats permettent d’expliquer qualitativement la dilatation voire la rupture des éprouvettes d’argilite lors d’essais de fluage en saumures. Le comportement en flexion de l’anhydrite massive ne semble pas être affecté par la présence de saumure, en raison d’une porosité trop faible pour permettre l’accès de la saumure au site réactionnel que sont les cristaux d’anhydrite / The aim of this study is the understanding of the physico-chemical interactions between saturated brine and the rocks (Marnes Irisées inférieures) enclosing the underground salt workings in Lorraine (eastern France), and also the study of their effect on the mechanical behaviour. A multi-scale study was undertaken, from the mineral scale to the one of a salt working. Whereas anhydrite-rich argillites flake quickly with the presence of saturated brine at the border of solutions cavities, the dolomudstone and massive anhydrite don’t and can constitute the top of cavities for several years. In order to explain this difference, these three lithologies were analysed in terms of mineralogy, micro-texture and porous media. Then, samples of argillites and massive anhydrite were immersed in saturated brines for more than one year. Chemical and microscopic analysis testified the hydration of anhydrite crystals into gypsum. This transformation occurs in a superficial way on massive anhydrite; to the contrary, it is located inside the anhydrite-rich argilites. As this transformation induces a volume increasing of 63%, the anhydrite crystals swell. In saturated brine, the water activity is low enough to prevent the swelling of clays such as smectites. Thus, anhydrite swelling might be the responsible of the argillites splitting in a saturated brine environment. The superficial anhydrite hydration on massive anhydrite can be explained by the low values of connected porosity (less than 1%) for this lithology. This results can explain, in a qualitative way, the dilatant behaviour of argilites samples during creep tests with brine. The bending behaviour of massive anhydrite don’t seem to be affected by the presence of brine for one year or less, probably because of the too low porosity of this lithology