Ab-initio determination of x-ray absorption near edge structure (xanes) spectra in an ultrasoft and norm conserving pseudopotentials scheme

X-ray absorption spectroscopy (XAS) is a powerful means of investigation of structural and electronic properties in condensed -matter physics. Analysis of the near edge part of the XAS spectrum, the so – called X-ray Absorption Near Edge Structure (XANES), can typically provide the following information on the photoexcited atom: - Oxidation state and coordination environment. - Speciation of transition metal compounds. - Conduction band DOS projected on the excited atomic species (PDOS). Analysis of XANES spectra is greatly aided by simulations; in the most common scheme the multiple scattering framework is used with the muffin tin approximation for the scattering potential and the spectral simulation is based on a hypothetical, reference structure. This approach has the advantage of requiring relatively little computing power but in many cases the assumed structure is quite different from the actual system measured and the muffin tin approximation is not adequate for low symmetry structures or highly directional bonds. It is therefore very interesting and justified to develop alternative methods. In one approach, the spectral simulation is based on atomic coordinates obtained from a DFT (Density Functional Theory) optimized structure. In another approach, which is the object of this thesis, the XANES spectrum is calculated directly based on an ab – initio DFT calculation of the atomic and electronic structure. This method takes full advantage of the real many-electron final wavefunction that can be computed with DFT algorithms that include a core-hole in the absorbing atom to compute the final cross section. To calculate the many-electron final wavefunction the Projector Augmented Wave method (PAW) is used. In this scheme, the absorption cross section is written in function of several contributions as the many-electrons function of the finale state; it is calculated starting from pseudo-wavefunction and performing a reconstruction of the real-wavefunction by using a transform operator which contains some parameters, called partial waves and projector waves. The aim of my thesis is to apply and test the PAW methodology to the calculation of the XANES cross section. I have focused on iron and silicon structures and on some biological molecules target (myoglobin and cytochrome c). Finally other inorganic and biological systems could be taken into account for future applications of this methodology, which could become an important improvement with respect to the multiscattering approach.

Recycling of end of life concrete fines (0 - 4 mm) into silica and cement

The purpose of this work is to find a methodology in order to make possible the recycling of fines (0 - 4 mm) in the Construction and Demolition Waste (CDW) process. At the moment this fraction is a not desired by-product: it has high contaminant content, it has to be separated from the coarse fraction, because of its high water absorption which can affect the properties of the concrete. In fact, in some countries the use of fines recycled aggregates is highly restricted or even banned. This work is placed inside the European project C2CA (from Concrete to Cement and Clean Aggregates) and it has been held in the Faculty of Civil Engineering and Geosciences of the Technical University of Delft, in particular, in the laboratory of Resources And Recycling. This research proposes some procedures in order to close the loop of the entire recycling process. After the classification done by ADR (Advanced Dry Recovery) the two fractions "airknife" and "rotor" (that together constitute the fraction 0 - 4 mm) are inserted in a new machine that works at high temperatures. The temperatures analysed in this research are 600 °C and 750 °C, cause at that temperature it is supposed that the cement bounds become very weak. The final goal is "to clean" the coarse fraction (0,250 - 4 mm) from the cement still attached to the sand and try to concentrate the cement paste in the fraction 0 - 0,250 mm. This new set-up is able to dry the material in very few seconds, divide it into two fractions (the coarse one and the fine one) thanks to the air and increase the amount of fines (0 - 0,250 mm) promoting the attrition between the particles through a vibration device. The coarse fraction is then processed in a ball mill in order to improve the result and reach the final goal. Thanks to the high temperature it is possible to markedly reduce the milling time. The sand 0 - 2 mm, after being heated and milled is used to replace 100% of norm sand in mortar production. The results are very promising: the mortar made with recycled sand reaches an early strength, in fact the increment with respect to the mortar made with norm sand is 20% after three days and 7% after seven days. With this research it has been demonstrated that once the temperature is increased it is possible to obtain a clean coarse fraction (0,250 - 4 mm), free from cement paste that is concentrated in the fine fraction 0 - 0,250 mm. The milling time and the drying time can be largely reduced. The recycled sand shows better performance in terms of mechanical properties with respect to the natural one.

Mitigating non-Lambertian surfaces issues in Stereo Matching with Neural Radiance Fields

Depth estimation from images has long been regarded as a preferable alternative compared to expensive and intrusive active sensors, such as LiDAR and ToF. The topic has attracted the attention of an increasingly wide audience thanks to the great amount of application domains, such as autonomous driving, robotic navigation and 3D reconstruction. Among the various techniques employed for depth estimation, stereo matching is one of the most widespread, owing to its robustness, speed and simplicity in setup. Recent developments has been aided by the abundance of annotated stereo images, which granted to deep learning the opportunity to thrive in a research area where deep networks can reach state-of-the-art sub-pixel precision in most cases. Despite the recent findings, stereo matching still begets many open challenges, two among them being finding pixel correspondences in presence of objects that exhibits a non-Lambertian behaviour and processing high-resolution images. Recently, a novel dataset named Booster, which contains high-resolution stereo pairs featuring a large collection of labeled non-Lambertian objects, has been released. The work shown that training state-of-the-art deep neural network on such data improves the generalization capabilities of these networks also in presence of non-Lambertian surfaces. Regardless being a further step to tackle the aforementioned challenge, Booster includes a rather small number of annotated images, and thus cannot satisfy the intensive training requirements of deep learning. This thesis work aims to investigate novel view synthesis techniques to augment the Booster dataset, with ultimate goal of improving stereo matching reliability in presence of high-resolution images that displays non-Lambertian surfaces.