This paper proposes a new technique for the 3D identification of clay particle orientations using images obtained from FIB-SEM observations. The method is based on a three-dimensional reconstruction that combines the Focused Ion Beam abrasion technique and Scanning Electron Microscopy, applied to kaolinitic clay selected for this study. The clay was first subjected to one-dimensional compression up to a given stress level, after which microstructural observations were performed using a post-mortem approach. A novel methodology using appropriate image processing was established for this purpose, allowing for a precise treatment of the obtained FIB-SEM images. The proposed methodology first involved removing “curtain effects” and “charging artefact”, which are specific types of noise commonly associated with FIB-SEM images. Two methods were employed to address this issue and were compared to evaluate their effectiveness: the first method was based on Fourier Transformation and Total Variational Reconstruction, while the second used a stochastic approach formulated as a convex optimization problem. Subsequently, a machine learning technique was integrated to enhance the segmentation process of the images. The final stage of the methodology involved creating a 3D model by reconstructing the clay particles in their spatial configuration. This paper aims to demonstrate how the proposed 3D observation method enables the quantification of the structural organization of clay particles in space in relation to mechanical loading.

The importance of physico-chemical processes at the particle scale for the engineering scale behaviour of fine-grained geomaterials is undisputed. Yet, despite great advances in the discipline, experimental evidence that fully resolves the clay micromechanics i.e. linking the evolving microstructure and interparticle actions under loading, is lacking. This paper will discuss the challenges ahead in quantifying the evolving kinematics and interparticle interactions of finegrained geomaterials. As such, the current limitations, and the potential opportunities of experimental methodologies for manipulating, monitoring and (post-mortem) analysing fine-grained materials at the particle scale will be discussed. In addition to the need of integrating multiple experimental techniques that span several length scales and modalities, the critical role of advanced data reduction and analysis is highlighted, as required for a measurement as opposed to qualitative observation. Throughout the paper, the link between experimental clay micromechanics and modelling will be discussed.