La valorisation de la biomasse est une alternative intéressante aux ressources fossiles, et s'effectue fréquemment en catalyse hétérogène. L'élaboration de nouveaux catalyseurs est une tâche ardue qui peut être considérablement accélérée in silico. Cependant les molécules de la biomasse sont souvent complexes et hautement oxygénées, rendant ainsi les calculs plus difficiles et couteux en temps. Parmi ces composés, les polyols sont particulièrement importants. Nous avons développé des relations du type Brønsted-Evans-Polanyi (BEP) à partir d'une étude DFT menée sur une famille de monoalcools concernant les dissociations des liaisons C-H et O-H sur des catalyseurs métalliques (Co, Ni, Ru, Rh, Pd, Ir, Pt). Ces relations ont pour but de prédire l’énergie d’activation d’une étape élémentaire à partir de son énergie de réaction. La précision obtenue par ces modèles linéaires est supérieure à 0.10 eV pour l'échantillon considéré. Ces relations ont ensuite étaient appliquées aux étapes élémentaires de la déshydrogénation du glycérol, choisi comme polyol prototype. On observe une erreur moyenne inférieure à 0.10 eV et une erreur systématique de l'ordre de ± 0.10 eV sur Rh. Etant donné que la principale différence entre les monoalcools et le glycérol, vient des liaisons H intramoléculaires présentes dans celui-ci, nous avons mis en place des relations linéaires pour prédire la déshydrogénation des monoalcools assistée par l'eau. Ces nouvelles relations nous ont permis d'améliorer la prédiction sur le glycérol et même d'éliminer la déviation systématique dans le cas de la rupture de la liaison OH. Même si dans cette étude nous nous sommes focalisés sur la déshydrogénation du glycérol, des méthodes similaires pourraient être appliquées à d'autres polyols avec d'autres réactions chimiques, accélérant ainsi considérablement la recherche in silico de catalyseurs solides. Ce travail pave la route pour le développement de nouvelles techniques numériques pour aborder la question de la conversion de la biomasse. / Biomass valorization is an interesting alternative to fossil resources, which is frequently performed via heterogeneous catalysis. Designing new catalysts is a challenging task that can be significantly accelerated in silico. However, biomass molecules are often complex and highly oxygenated, hence rendering calculations more difficult and time consuming. Among these compounds, polyols are particularly important. We developed linear relations of the Brønsted-Evans-Polanyi (BEP) type from the DFT study of C-H or O-H bond dissociation elementary steps for a family of monoalcohol molecules on metallic catalysts (Co, Ni, Ru, Rh, Pd, Ir, Pt). Such relations aim at predicting activation energies from reaction energies. The accuracy of the obtained linear energy models is better than 0.10 eV on the sampling set. Then, the relations were applied for the prediction of the dehydrogenation elementary steps of glycerol, chosen as a prototype of polyalcohols, with an accuracy better than 0.10 eV and with a systematic error around ±0.10 eV for Rh. Keeping in mind that the main difference between glycerol and monoalcohols comes from intramolecular H-bonds present in the former, we designed linear relations for water-assisted dehydrogenation of monoalcohols. These new relations allowed us to improve the prediction on glycerol and to eliminate the systematic deviation in the case of OH bond breaking. Even if in this study we focused on glycerol dehydrogenation, similar methods may be applied to other polyols with other chemical reactions, and considerably speed up the computational design of solid catalysts. This work paves the way for the development of novel numerical techniques to address the issue of biomass conversion.

Two separate mechanistic kinetic models have been developed for the hydrocracking of complex feedstocks. The first model is targeted for the hydrocracking of vacuum gas oil. The second one addresses specifically the hydrocracking of long-chain paraffins, but at a more fundamental level as compared to the first one. Both models are based on an exhaustive computer generated reaction network of elementary steps. In the first model, the dehydrogenation/hydrogenation steps occurring on the metal sites to generate/consume the reactive olefinic intermediates are assumed to be very fast so that the acid site steps are considered as the rate determining steps. The frequency factors for acid site steps are modeled using the single-event concept and the activation energies based on the nature of the reactant and product carbenium ions. This model utilizes a detailed composition of the vacuum gas oil characterized by 16 different molecular classes up to carbon number 40. These classes are divided into 45 subclasses by distinguishing the isomers of a class according to the number of methyl branches. The kinetic model is plugged into an adiabatic multi-bed trickle flow reactor model. The model contains 33 feedstock and temperature independent parameters which have been estimated from the experimental data. The model has been used to study the effect of the operating conditions on the yield and composition of various products. A sensitivity analysis of the distribution of isomers of a class among its different subclasses has been performed showing that the total conversion increases when the content of isomers with a higher degree of branching is increased in the feed. In the second model, the dehydrogenation/hydrogenation steps on the metal sites are also assumed to be rate determining. The rate coefficients for the dehydrogenation steps are modeled depending on the nature of the carbon atoms forming the double bond. The frequency factors for the acid site steps are modeled using the single-event concept. A more rigorous approach has been selected to model the activation energies of the acid site steps by implementing the Evans-Polanyi relationship. The 14 model parameters, which are independent of the temperature and feedstock composition, have been estimated from the experimental data. The model elucidates the effect of the relative metal/acid activity of the catalyst on the isomerization/cracking selectivities and on the carbon number distribution of the products.

Biomass valorization is an interesting alternative to fossil resources, which is frequently performed via heterogeneous catalysis. Designing new catalysts is a challenging task that can be significantly accelerated in silico. However, biomass molecules are often complex and highly oxygenated, hence rendering calculations more difficult and time consuming. Among these compounds, polyols are particularly important. We developed linear relations of the Brønsted-Evans-Polanyi (BEP) type from the DFT study of C-H or O-H bond dissociation elementary steps for a family of monoalcohol molecules on metallic catalysts (Co, Ni, Ru, Rh, Pd, Ir, Pt). Such relations aim at predicting activation energies from reaction energies. The accuracy of the obtained linear energy models is better than 0.10 eV on the sampling set. Then, the relations were applied for the prediction of the dehydrogenation elementary steps of glycerol, chosen as a prototype of polyalcohols, with an accuracy better than 0.10 eV and with a systematic error around ±0.10 eV for Rh. Keeping in mind that the main difference between glycerol and monoalcohols comes from intramolecular H-bonds present in the former, we designed linear relations for water-assisted dehydrogenation of monoalcohols. These new relations allowed us to improve the prediction on glycerol and to eliminate the systematic deviation in the case of OH bond breaking. Even if in this study we focused on glycerol dehydrogenation, similar methods may be applied to other polyols with other chemical reactions, and considerably speed up the computational design of solid catalysts. This work paves the way for the development of novel numerical techniques to address the issue of biomass conversion.

Ce travail de recherche portant sur la compréhension des premières étapes de synthèse des zéolithes s'inscrit dans le cadre général de développements efficients de nouveaux catalyseurs performants. De nombreuses incertitudes demeurent quant aux mécanismes moléculaires de leur formation, en particulier lors de la nucléation en conditions de synthèse hydrothermale. Dans ce contexte, une méthodologie globale a été proposée afin d'évaluer l'impact des variables de synthèse sur la formation des espèces siliciques. La réalisation de ce travail a nécessité l'usage de différentes méthodes théoriques, combinant des calculs ab initio DFT et des simulations kMC. Un nouveau module kMC nommé Réacdiff a ainsi été développé et validé avec le modèle Lodka. L'extension des chemins réactionnels vers des espèces plus complexes a été effectuée et le recours à des modèles théoriques du type Brønsted-Evans-Polanyi (BEP) a été indispensable afin de les estimer rapidement. Un nouveau modèle des étapes réactionnelles d'oligomérisation a été proposé et a permis de tester l'effet crucial des agents organiques structurants (AS). Ces travaux ont permis de donner un nouveau regard sur la compréhension de la condensation des espèces siliciques durant les premières étapes de la synthèse des zéolithes. / This research work on the understanding of the early stages of zeolites synthesis fits withinthe general framework of the technological lock on the lower cost of new efficient catalysts synthesisoptimization. Many uncertainties remain with respect to the molecular mechanisms of their formation,in particular during nucleation in hydrothermal synthesis conditions. In this context, a comprehensivemethodology was proposed to assess the impact of synthesis variables in the formation of silicicspecies. This objective has required the use of different theoretical methods, combining ab initio DFTcalculations and kMC simulations. The creation of a new kMC module named Reacdiff has beendeveloped and validated with the Lodka model. The extension of reaction pathways to more complexspecies was carried out with the essential use of theoretical models like Brønsted-Evans-Polanyi(BEP) to quickly estimate. A new anionic model of the reaction steps of oligomerization has beenproposed and has allowed testing the crucial effect of organic templates. These thesis works gave anew understanding on the condensation of silicic species in early stages of zeolite synthesis.