Résumé : Glass-ceramics are of growing interest due to their enhanced properties compared to the base glasses. More specifically, the control of microstructures is a major challenge as the properties of glass-ceramics are the direct consequences of microstructures. Microstructures can be modified by forming specific crystal phases or by using a prior amorphous phase separation before crystallization. The PhD thesis objectives are to demonstrate that the properties of silicate glasses can be enhanced by controlling their microstructure genesis with composition and thermal process parameters. More specifically, two systems were studied and compared: the BaO-TiO2-SiO2 system and the soda-lime silica Na2O-CaO-SiO2 used industrially. Both systems exhibit a large zone of immiscibility allowing the study of the influence of phase separation on crystallization.

The first system BaO-TiO2-SiO2 has gathered interest from the interesting properties of fresnoite (Ba2TiSi2O8): piezo and pyroelectricity, second harmonic generation and blue/white photoluminescence. Many studies on the stoichiometric composition were conducted to understand and improve those promising properties. However, it was recently suggested that the photoluminescence can be improved with composition exhibiting phase separation. This indicates that the photoluminescence intensity can be improved through a microstructural control. The possible role of a prior amorphous phase separation on the subsequent crystallization has been however the topic of vigorous debates over the last decades and has not yet been clarified, especially regarding the role of the interfaces created by the phase separation. In this PhD, the effect of phase separation on fresnoite crystallization was studied. This had to pass through the calculation of the liquid-liquid immiscibility in the phase diagram in order to select suitable compositions to compare in a systematic study. The systematic study concludes to a surface crystallization mechanism for all non- stoichiometric compositions and shows no influence between amorphous droplets and matrix crystallization. This study was also completed with the investigation of the effect of composition (i.e. SiO2-excess), annealing temperature and prior heat treatment, i.e. heating rate, cooling rate or a prior isothermal step before annealing. It is shown that specific microstructures are obtained depending on the process parameters. Finally, selected compositions and heat treatment show how photoluminescence intensity can be improved by a microstructural control. The highest intensity is obtained with a high crystallization fraction and a maximization of the number of interfaces.

The results obtained in the study of the BaO-TiO2-SiO2 system are extended to the soda-lime-silica system in order to study the effect of phase separation on crystallization. It is shown that cristobalite forma- tion from the surface cannot be avoided and that the involved composition shift inhibits phase separation. It is consequently difficult to observe an interplay. Those studies lead to a general discussion about the criteria allowing to observe an interplay between phase separation and crystallization in oxide glasses.