Multiscale assessment of swelling and compressibility of fine grained geomaterials

The Danish subsoil is rich with Paleogene Clay Formations, to an extent that most structures and infrastructures constructed in Denmark lay on this type of highly plastic, fissured, fine grained soil. According to the geological history of the North Sea Basin, the shallower Tertiary layers underwent a series of glaciotectonic deformation that sheared them, modifying their structure and folding them. The deeper strata instead were not affected by the glacial movements that occurred during the Quaternary, so that their intact form was not altered. As numerous Danish infrastructures were built on the shallower, folded layers, most of the studies presented in literature focus on the mechanical characterization of these clay strata, while very little attention has been given to that of the deeper, intact layers, nor a systematic comparison of the two types of clays has been performed so far. Moreover, numerous studies claim the existence of an influence of the Paleogene Clay Formations particles arrangement (fabric) and bonging on their compressibility and swelling behaviour, even though no microscopic investigation has been performed on these soils. Indeed the fabric arrangement and porosity distribution has been shown by numerous studies to have an influence on the clay mechanical behaviour, with a dependency on the sample stiffness, geological history and mineralogy. Specifically, it was observed in sedimentation experiments and Mercury Intrusion Porosimetry tests that exposing kaolinitic or smectitic clays to different saline and acidic environments, their aggregation patterns can vary significantly depending on the solution electrical conductivity. Moreover, the mechanical behaviour has been associated with the fabric orientation of natural, stiff clays has been qualitatively studied by means of Scanning Electron Microscope micrographs by several authors. However, due to the lack of updated, powerful quantification tools capable of analysing these microscope images, no quantitative correlation linking the particles orientation of a clay sample to its mechanical indexes has been found. For the same reason, the disturbance experienced by clay samples having different stiffness and mineralogy during the sample preparation that precedes the Scanning Electron Microscope investigation has not been systematically clarified yet. Moreover, despite the numerous X-Ray Diffraction quantification and sample prepratation methods available in literature, their effect on the determination of the mineralogical composition of a mixed sample is not clear yet.
For this study, a Computer Vision based code called MiCA (Microstructural Clay Analyser) capable of giving a high quality characterization of the soil fabric and porosity based on Scanning Electron Microscope micrographs has been developed in order to perform meaningful quantifications of the soil microstructure. MiCA was then validated on a series of reference patterns and on clay images obtained from literature, showing high accuracy and reliability. Moreover, the influence of three Scanning Electron Microscopy sample preparation techniques (i.e. Freeze, Oven and Air Drying), together with that of three imaging methods (i.e. SEM, ESEM and CryoSEM) on the particles orientation of samples having different mineralogy, stiffness and geological history has been studied. The results showed that 1:1 and 2:1 layered minerals show different degreese of disturbance when prepared and imaged in theri soft dispersed state, while a lack of significant microstructural changes was observed for both soft flocculated samples and stiff clays. Finally, in order to gain an increased accuracy in the mineralogical composition of the soil mixtures, the effect of three sample preparation methods and five quantification approaches, two of which were developed by the Author, were investigated. Specific recommendations were formulated in this regard, discouraging the use of Drop prepared samples regardless of the utilized quantification approach.
The tools developed, together with the knowledge on the sample preparations gained were used to investigate the relationship between the 1D compression behaviour and the microstructure of clays having different geological history, fabric arrangements and mineralogy. Specifically, a series of Incremental Load Oedometer tests were performed on folded and intact samples of a highly plastic Danish Paleogene Clay called Røsnæs Clay in its reconstituted and undisturbed state. Moreover, sedimented samples of commercial kaolinite and natural Røsnæs Clay were produced. In particular, the deposition was performed in environments having different pH, to control the particles aggregation process. After characterizing the clays mineralogy by means of X-Ray Diffraction, their particles orientation was quantified, together with their porosity by means of Scanning Electron Microscopy micrographs investigation and Mercury Intrusion Porosimetry tests respectively. Focusing on the mechanical tests performed on the undisturbed and reconstituted samples of Røsnæs Clay, these showed larger compressibility, swelling potential and lower oedometric stiffness for the samples obtained from the folded and the shallower intact layers. Moreover, a preconsolidation stress congruent with that expected from the geological history of the area was observed for all the intact samples. Despite no clear correlation was found between the soil mineralogy and its 1D compression behaviour, the smectite content seems to strongly influence the detected liquid limit with a second degree polynomial relationship. Moving to the micrographs analysis, a good correlation between the sample orientation and the compressibility and swelling index was observed, with different patterns before and after the preconsolidation stress.
The study of the mechanical behaviour of sedimented Røsnæs Clay and commercial kaolinite showed a correlation between the final sedimentation height and the solution pH, with the former soil type showing a direct link, while an inversed relation was found for the latter. Moreover, despite the fact that samples having different pH show a significantly similar 1D compression behaviour, differences in the fabric were observed and successfully linked to the soil compression and consolidation indexes.