The main objective of this thesis work was to gain understanding of the layer properties and polymer structures that were able to aid lubrication in aqueous media. To this end, three types of polyelectrolytes: a diblock copolymer, a train-of-brushes and two brush-with-anchor mucins have been utilized. Their lubrication ability in the boundary lubrication regime has been examined by Atomic Force Microscopy with colloidal probe. The interfacial behavior of the thermoresponsive diblock copolymer, PIPOZ60-b-PAMPTAM17,on silica was studied in the temperature interval 25-50 ˚C. The main finding is that adsorption hysteresis, due to the presence of trapped states, is important when the adsorbed layers are in contact with a dilute polymer solution. The importance of trapped states was also demonstrated in the measured friction forces, where significantly lower friction forces, at a given temperature, were encountered on cooling than on the preceding heating stage, which was attributed to increased adsorbed amount. On the heating stage the friction force decreased with increasing temperature despite the worsening of the solvent condition, and the opposite trend was observed when using pre-adsorbed layers (constant adsorbed amount) as a consequence of increased segment-segment attraction. The second part of the studies was devoted to the interfacial properties of mucins on PMMA. The strong affinity provided by the anchoring group of C-PSLex and C-P55 together with their more extended layer structure contribute to the superior lubrication of PMMA compared to BSM up to pressures of 8-9 MPa. This is a result of minor bridging and lateral motion of molecules along the surface during shearing. We further studied the influence of glycosylation on interfacial properties of mucin by utilizing the highly purified mucins, C-P55 and C-PSLex. Our data suggest that the longer and more branched carbohydrate side chains on C-PSLex provide lower interpenetration and better hydration lubrication at low loads compared to the shorter carbohydrate chains on C-P55. However, the longer carbohydrates appear to counteract disentanglement less efficiently, giving rise to a higher friction force at high loads. / <p>QC 20170407</p>

In these diploma thesis magnetically ordered materials are studied with the prospect of their application in spintronics. Specifically, we investigated metallic alloy FeRh, which undergoes a magnetic phase transition from antife- romagnetic phase to feromagnetic one around 100◦ C. This phenomenon can be readily used in memory devices. Laser spectroscopy is used as a nondestructive method without need of any electrical contacts. Magnetic properties of FeRh are studied by magnetooptical effects including quadratic magnetic linear dichroism. The measured polarization rotations are of the order of miliradians, therefore, the detection is realized by an optical bridge. At first, we concentrated on discrimina- ting of various magnetooptical effects from each other. The second part is focused on the phase transition induced by different means. Firstly, by heating the whole sample, secondly by illuminating the sample locally by continuous laser.

Most of the water fed into a paper machine is removedmechanically in the forming and press sections. One of thefactor which has an important influence on mechanicaldewatering, i.e. in both forming and pressing, is thestress-strain behaviour of the fibre network. The focus of this thesis is on the development of improvedmathematical descriptions of the stress-strain behaviourexhibited by fibre networks in the forming and press sections.The first part of the thesis presents a physically based modelof the forming and densification of fibre mats in twin-wireformers. The model can calculate the ecect of the applicationof a varied load through the forming section. It was developedfrom mass and momentum balances of the fibre and liquid phases,the fibre mat stress-porosity relation and an expression forthe permeability as a function of the porosity. The fibre-matstress-porosity relation used is rate-independent and presentshysteresis. Simulations have been conducted to study theeffects of roll pressure, blade pulses, wire tension andbeating. The effect of sequential blade pressure pulses afterthe forming roll on the dewatering and the concentrationgradients could be characterised. The simulations alsoexhibited rewetting by expansion when the fibre mats left theforming roll. Increasing wire tension resulted in increaseddewatering, but the rate of increase diminished rapidly withincreasing tension. The simulation results also indicated thatbeating has a large influence on dewatering. The second part of the thesis presents two models of therate-dependent stress-strain behaviour of the fibre networkthat is observed in wet pressing. The first model was based onthe approach pioneered by Perzyna (1966) for strain-ratedependent plasticity and was quite satisfactory for calculatingthe stress-strain behaviour of the fibre network in singlepress nips. It was successfully applied for studyingdensification and dewatering in both normal wet pressing andhigh temperature wet pressing. However, the first model onlyincludes rate dependence in the compression phase of thecompressionexpansion cycle; the expansion phase is treated asbeing rate independent The second model of the stress-strain behaviour of the fibrenetwork treats both compression and expansion as being ratedependent, according to experimental observations. It is basedon the idea that the wet fibre web may be conceived as alayered network of restricted swelling gels. A swollen fibre isa restricted gel, the inner swelling pressure in a swollenfibre wall being balanced by the stresses in the fibre wallstructure. The observed rate dependence of wet webs in bothcompression and expansion phases was attributed to the flow ofwater out of and into the fibre walls. The second model gavepredictions that are in good agreement with results fromuniaxial experiments using pressure pulses of arbitrary shapefor both a single pulse and a sequence of pulses. It maytherefore be used as a general model for the rheologicalbehaviour of the wet fibre network in wet pressing, providedthe model parameters are estimated from experimental data withsmall experimental error. KEYWORDS:Paper, modelling, dewatering, forming, wetpressing, fibre network stress, rheology, hysteresis,intra-fibre water, compressibility, structural stress,stress-strain, restricted gels, swelling. / <p>QC 20161026</p>

En dépit de leur importance pratique considérable, et bien que de nombreuses expériences établissent une corrélation certaine entre les hétérogénéités d’interaction de surface (rugosité ou inhomogénéités chimiques) et les propriétés de friction des surfaces, le rôle de ces interactions sur la friction n’est encore pas bien décrit par les modèles et les expériences existants. Dans ce travail de thèse, nous nous sommes intéressés à l’identification des mécanismes moléculaires de la friction aux interfaces polymères souples. Dans ce contexte, nous avons réalisé deux études complémentaires. La première partie du travail concerne le mouvement d’une ligne triple solide-liquide-vapeur qui se déplace sur une surface solide sous l’effet de différentes forces (gravité, forces capillaires et tensions interfaciales), et en particulier le lien entre le piégeage et le dépiégeage de la ligne triple et l’hystérèse de l’angle de contact. Cette méthode permet de mesurer des angles de contact d’avancée et de reculée avec une précision sans précédent (0,1°)et s’avère être particulièrement sensible aux mécanismes qui tendent à ancrer la ligne triple. Ceci en fait un outil de choix pour étudier la friction liquide/solide. Dans la seconde partie du travail, nous avons cherché à comprendre comment des chaînes de polymère flexibles, fortement ancrées sur une surface solide, dans le régime des fortes densités de greffage affectent la friction entre une telle surface et un élastomère réticulé constitué du même polymère. Nous avons montré que le comportement en friction de cette couche confinée suit exactement le comportement rhéofluidifiant observé pour des couches de fondu de masses molaires équivalentes mais avec un temps de relaxation beaucoup plus long que celui des chaînes en fondu,la reptation n’étant pas permise pour les chaînes ancrées. Enfin, en comparant les résultats obtenus pour des couches greffées chimiquement à une extrémité et des couches fortement adsorbées, ayant par ailleurs les mêmes caractéristiques moléculaires (masse molaire des chaînes et épaisseur de la couche ancrée), nous avons mis en évidence que la friction est remarquablement sensible à l’organisation moléculaire au sein de la couche ancrée. / Different experiments show a correlation between interaction heterogeneities (roughnessor chemical inhomogeneities) and friction properties of surfaces but the exact role of theseinteractions is still not clearly established. In this thesis, we try to identify molecular mechanism offriction at soft polymer interfaces. In this context, we have performed two complementary studies.The first one concerns the movement of a triple line solid-liquid-gas which moves on a solidsubstrate and more particularly the link between trapping and entrapping of the triple line andcontact angle hysteresis. This method allows measurement of dynamic contact angles with a verygood precision (0,1°), and is particularly sensitive to mechanisms which tend to anchor a tripleline. This technique is then a very efficient tool to study solid/liquid friction.In the second part of this thesis, we try to understand how flexible polymer chains stronglyanchored on a solid substrate affect friction at the interface between such surface and crosslinkedelastomer made of the same polymer. We show that the friction is dominated by the shearthinning of the grafted layer confined between the elastomer and the substrate, and responding tothe shear solicitation like a melt, with very long relaxation times. We also show that the frictionstress appears highly sensitive to the molecular organization inside the surface anchored polymerlayer, comparing end grafted and strongly adsorbed layers having otherwise the same molecularcharacteristics (molecular weight of the chains, and thickness of the surface anchored layer).

Este trabalho discute o efeito do tamanho de grão, frequência de excitação e resistividade elétrica nas perdas magnéticas, destacando-se a perda anômala. Também é proposto um método de sobreposição de histereses onde é revelada a região de ocorrência da perda anômala durante o ciclo de magnetização e desmagnetização, e posteriormente a curva de histerese da perda anômala é construída. Para tal, três ligas de aço de grão não orientado, com teor de silício de 2,05%, 2,45%, 3,3%, foram tratadas termicamente para aumento do tamanho de grão por crescimento. Cada liga foi tratada sobre a mesma sequência de temperatura, em recozimento contínuo. Os tamanhos de grão das amostras foram medidos pelo método de interceptos. A caracterização das propriedades magnéticas foi realizada por meio do quadro de Epstein. As amostras, no total de 21 conjuntos, foram ensaiadas em regime de frequência de 50, 60, 100, 150 e 200 Hz e regime quase estático (5mHz), ambos a 1 e 1,5 T. Possibilitando desta forma, a construção da histerese da perda histerética e total. Por meio dos resultados da perda total, histerética e parasita, a perda anômala pode ser calculada. Os resultados mostraram o comportamento da perda anômala (Pa) em função do tamanho de grão (l) do tipo Pa &#8733; l0,34 e em frequência (f) a relação de Pa &#8733; f1,65. A respeito do comportamento da perda anômala em função da resistividade elétrica (&#961;), os resultados encontrados são inconclusivos. Constatou-se empiricamente que, o tamanho de grão ótimo é função da frequência e da indução. A partir de tratamento matemático, a histerese da perda parasita mais histerética, foi construída e sobreposta à histerese da perda total. Os espaços vazios entre as curvas corresponder à perda anômala. É visto que a perda anômala ocorre em três regiões distintas da curva de histerese. Tais regiões podem supostamente ser associadas à fenômenos de dissipação de energia como; movimento de parede de domínio, nucleação de domínios e aniquilação de domínio. Como as histereses da perda total, histerética e parasita possuem a mesma indução, a histerese da perda anômala pode ser construída. O campo associado à perda anômala foi obtido como resultado da soma entre campo histerético mais parasita menos o campo magnético da perda total. / This paper discusses about the effect of grain size, excitation frequency and electrical resistivity in magnetic losses, especially the anomalous loss. Also is proposed a method of overlapping hysteresis where the region of occurrence of the anomalous loss during magnetization and demagnetization cycle is disclosed, and thereafter the hysteresis curve of anomalous loss is constructed. For this purpose, three non-oriented steel alloys, with silicon content of 2.05%, 2.45%, 3.3%, were heat treated to increase the grain size by grain growth. Each alloy was treated in the same sequence of temperature in continuous annealing. The grain sizes of the samples were measured by an intercept method. Characterization of magnetic properties was performed using the Epstein frame. The samples, in total 21 sets, were tested under a frequency of 50, 60, 100, 150 and 200 Hz and quasi-static regime (5mHz), both at 1 and 1.5 T. Thus allowing the construction of the hysteresis of quasi-static and total loss. Through the results of the total loss, quasi-static loss and parasite loss, the anomalous loss can be calculated. The results show the behavior of the anomalous loss (Pa) as a function of grain size (l) as Pa &#8733; l0.34 and frequency (f) the relationship of Pa &#8733; f1.65. Regarding the behavior of the anomalous loss due to resistivity (&#961;), the results are inconclusive. It was found empirically that the optimum grain size is a function of frequency and induction. From a mathematical treatment, the hysteresis of parasite loss plus quasi-static loss was constructed and overlapped to the hysteresis of total loss. The areas between the curves correspond to anomalous loss. It can be seen that anomalous loss occurs in three distinct regions of the hysteresis curve. Such regions may be associated with energy dissipation phenomena, such as: domain wall motion, nucleation and annihilation of domain. Since the hysteresis of total loss, quasi-static loss and parasite loss have the same induction, anomalous hysteresis loss can be constructed. The field associated with anomalous loss was obtained as a result of the sum of quasi-static field plus parasite field minus the magnetic field of the total loss.

In dieser Arbeit entwickeln und untersuchen wir ein numerisches Schema für die eindimensionale Wellengleichung mit Hysterese für unterschiedliche Arten von Randbedingungen. Diese Gleichung ist ein Modell für die Longitudinal- oder Torsionsschwingungen eines homogenen Stabes unter dem Einfluß einer uniaxialen äußeren Kraftdichte, wobei wir ein elastoplastisches Materialgesetz annehmen. Hysterese-Operatoren sind ratenunabhängige Volterra-Operatoren, die Zeitfunktionen in Zeitfunktionen abbilden. Mit ihnen lassen sich Gedächtniseffekte modellieren, wie sie zum Beispiel in der Elastoplastizität oder im Ferromagnetismus auftauchen. Zunächst führen wir Hysterese-Operatoren allgemein ein und analysieren dann eine spezielle Klasse von Hysterese-Operatoren, die Prandtl-Ishlinskii-Operatoren. Wir untersuchen ihre Gedächtnisstruktur und erklären, wie sich die Operatoren numerisch auswerten lassen. Dazu stellen wir zwei verschiedene Approximationsansätze vor. Wir führen aus, wie sich die approximierenden Operatoren implementieren lassen und leiten lineare und quadratische Fehlerabschätzungen her. Zur numerischen Lösung des gekoppelten Systems aus der Wellengleichung mit einem Hysterese-Operator führen wir ein implizites Differenzenschema mit Gedächtnis ein. Für eine Klasse von Hysterese-Operatoren zeigen wir die Existenz und Eindeutigkeit der Lösung des numerischen Schemas, beweisen mit Hilfe von Kompaktheitsschlüssen und einem Monotonieargument die Konvergenz des Verfahrens und leiten eine Fehlerabschätzung der Ordnung 1/2 her. Wir diskutieren, wie das vorgestellte Verfahren auf die Prandtl-Ishlinskii-Operatoren angewendet werden kann. / In this thesis we develop and investigate a numerical scheme for the one-dimensional wave equation with hysteresis for different kinds of boundary conditions. This equation can be regarded as a model for the longitudinal or torsional oscillations of a homogeneous bar under the influence of an uniaxial external force density assuming an elastoplastic material law. Hysteresis operators are rate-independent Volterra operators mapping time functions to time functions. This kind of operator can be used to model memory effects as they appear in elastoplasticity or ferromagnetism, for example. We first give an introduction to the general concept of hysteresis operators before we analyze a special class of hysteresis operators called Prandtl-Ishlinskii operators. We investigate their memory structure and explain how the operators can be evaluated numerically. To that end we present two different kinds of approximation schemes. We point out how the approximating operators can be implemented and we derive linear and quadratic error estimates. For the numerical solution of the coupled system of the wave equation with a hysteresis operator we introduce an implicit difference scheme with memory. For a class of hysteresis operators we show the existence and uniqueness of the numerical solution. We prove the convergence of the scheme by compactness and monotonicity arguments. We derive an error estimate of order 1/2. We discuss the application of the method presented to Prandtl-Ishlinskii operators.

In Brokate-Sprekels 1996 wurde ein Darstellungsresultat für Hysterese-Operatoren, die auf skalaren, stetigen, stückweise monotonen Funktionen definiert sind, hergeleitet. Dieses erlaubt es, die Operatoren eindeutig aus Funktionalen auf Strings reeller Zahlen, d.h. auf Tupeln aus reellen Zahlen mit beliebiger Länge, bei denen die Vorzeichen der Differenzen aufeinander folgender Paare alternieren, zu gewinnen. In dieser Habilitation wird nun ein Ansatz vorgestellt, um auch für Hysterese-Operatoren mit vektoriellen Input-Funktionen eine Darstellung durch Abbildungen auf einer String-Menge zu ermöglichen. Dabei wird die Rolle der monotonen Funktionen von den neu eingeführten sogenannten monotaffinen Funktionen übernommen, die man erhalten kann, indem man die Ausgabe einer monotonen Funktion auf den reellen Zahlen in eine affine Funktion auf den reellen Zahlen einsetzt. Die Rolle der alternierender Strings wird von den sogenannten Konvexitätstripel-freien Strings mit Elementen aus Vektorraum übernommen, d.h. von den Tupel aus Elementen aus dem Vektorraum, so dass sich kein Eintrag als Konvexkombination seines Vorgängers und seines Nachfolgers schreiben lässt. Das Darstellungsresultat erlaubt es, die Hysterese-Operatoren auf den stetigen, stückweise monotaffinen Input-Funktionen eindeutig aus den Abbildungen auf den Konvexitätstripel-freien Strings zu gewinnen. Damit können dann Eigenschaften des Hysterese-Operators durch Untersuchung der entsprechenden Abbildung bestimmt werden. Es wird ein weiteres Darstellungsresultat vorgestellt, bei dem die Hysterese-Operatoren für Funktionen definiert sind, die endlich viele Sprungstellen haben. Die betrachteten Strings sind jetzt Tupel von Quintupeln. In diesen Quintupeln werden für jede Sprung- und Richtungswechselstelle der Funktion der Funktionswert, der links- und der rechtsseitigen Grenzwert und Informationen über das Verhalten der Funktion in der Nähe dieser Stelle gespeichert. / In Brokate-Sprekels 1996 a representation result for hysteresis operators acting on scalar-valued continuous piecewise monotone functions was derived. Thanks to this result, the operators can be derived in a unique way from the functionals on alternating strings, i.e. on the tuple of real numbers of arbitrary length, such that the sign of the differences between the elements alternates. In this habilitation an ansatz will be presented that allows also to represent hysteresis operators with vector valued input functions by a mapping defined on a set of strings. Here,the monotone functions are replaced by the monotaffine functions, intruded by the author. One can describe these functions by considering the composition of a monotone function on the real numbers with an affine function from the real number to the vector space such that the monotone function is applied first. Instead of alternating strings of real numbers, the so called convexity triple free string are considered. These strings are tuple of elements of the vector space such that no element in the tuple can be written as the convex combination of its predecessor and its successor. Thanks to representation result, one can generate uniquely hysteresis operator on continuous piecewise monotaffine input functions from mappings on the set of all convexity triple free string. This allows to investigate properties of operators by investigating properties of the corresponding mapping. Moreover, a further representation result is presented for hysteresis operators with input functions having a finite number of jumps. The corresponding strings are now tuple of quintuples. In the quintuple for each position of a change of the direction of the input function and for each jump, the corresponding value of the function, its limits from the right and from the left and information about the behavior of the function near to this position are stored.

Le CyberKnifeTM est composé d’un accélérateur linéaire de 6 MV monté sur un bras robotisé, avec 6 axes de rotation et d’un système d’imagerie permettant de guider le faisceau d’irradiation sur la cible à traiter. Le but est d'améliorer la précision du traitement et la réduction de l’irradiation des organes critiques environnants. Le traitement est réalisé par la convergence « isotrope » d’une centaine d’orientation pour créer jusqu’à 1200 mini faisceaux dirigés sur la cible avec une précision sub-millimétrique. Cet ensemble est complété par une table de traitement montée sur un bras robotisé qui offre 6 degrés de liberté supplémentaires, permettant d’améliorer encore la précision du traitement et de lever d’éventuelles limitations. Grâce à son sous système SynchronyTM, le CyberKnifeTM est capable de traiter les tumeurs abdo-thoraciques, qui bougent avec la respiration en déplaçant dynamiquement le LINAC afin de compenser le mouvement respiro-induit. La forte dose utilisée dans ce genre de traitement hypofractionné, rend toute erreur même minimale inacceptable et impose une très grande précision géométrique, tout en assurant une précision dosimétrique maximale.Notre travail est consacrée à évaluer la qualité de traitement en termes des précisions géométrique et dosimétrique, pour les différents modes de suivi disponibles dans le système en statique, et en dynamique avec suivi respiratoire. Dans cette étude, nous avons utilisé différents types de détecteurs, et trois plateformes afin de simuler des simples mouvements respiratoires, des mouvements réels provenant des patients traités et enfin des mouvements complexes avec hystérésis / The CyberKnifeTM consists of 6MV LINAC mounted on a robotic arm, with six degree of freedom and is coupled to an image guiding system, allowing us to guide the irradiation beams toward the target. The aim is to improve the treatment accuracy and to reduce the irradiation of critical surrounding organs. The treatment is realized by the isotropic convergence of hundreds of orientations for creating up to 1200 mini-beams, which are orientated to the target with submillimetric accuracy. This group is completed by a treatment couch, which is also mounted on a robotized arm, that offers 6 additional degrees of freedom, allowing an additional improvement of accuracy, and eliminates the possible limitations. Using its subsystem SynchronyTM, the CyberKnifeTM is capable of treating the abdo -thoracic tumors, which move with respiration, by moving dynamically the LINAC to compensate the respiratory motion of the tumors. The high dose level, which is used in this kind of hypofractionated treatment, makes the smallest error unacceptable, and needs a very high geometric accuracy with keeping a maximal dosimetric accuracy. Our work is dedicated to evaluate the quality of treatment, in the terms of dosimetric and geometric accuracies. For the different modes of tracking, which are available in the system in static mode, and dynamic mode with respiratory motion tracking. By using different kinds of detectors (ionization chambers, radiochromic films) and three different platforms, which allow simulating simple respiratory motion, real respiratory motion coming from real treated patients, and finally complex motion with hysteresis

Un nouveau matériau d’âme à base de fibres enchevêtrées et réticulées a été précédemment développé dans le but d’améliorer certaines propriétés des structures sandwichs dont l’amortissement vibratoire. Cependant, son comportement mécanique et vibratoire doit être optimisé afin de l’utiliser dans le domaine aérospatial. Plusieurs paramètres morphologiques entrent en jeu lors de sa fabrication. L’objectif de cette thèse est de développer un modèle numérique permettant de mieux comprendre le comportement de ce matériau enchevêtré réticulé. Le comportement d’un volume élémentaire représentatif de fibres de carbone enchevêtrées et non-réticulées est d’abord étudié en compression par éléments finis. La géométrie numérique du réseau de fibres s’appuie sur les données morphologiques du matériau réel. Les simulations numériques permettent de suivre, au cours de la compression confinée, l’évolution des différents paramètres, tels que la distribution des orientations des fibres, la distance entre contacts ou la fraction volumique. Ces résultats constituent une base robuste pour le développement du modèle numérique du matériau enchevêtré et réticulé qui est ensuite utilisé pour modéliser le comportement mécanique en cisaillement, et en particulier pour simuler et expliquer les boucles d’hystérésis observées expérimentalement. A la fin de ce travail, une étude numérique est proposée afin de décrire l’influence des différents paramètres morphologiques sur la rigidité en compression et en cisaillement du matériau enchevêtré réticulé. / A new core material based on entangled and cross-linked fibers has been previously developed in order to improve certain properties of sandwich structures including vibration damping. However, its behavior must be optimized for use in the aerospace field. Several morphological parameters can be modified during the manufacturing process. The aim of this thesis was to develop a numerical model to better understand the behavior of this entangled cross-linked material. The behavior of a representative volume element of entangled carbon fibers without cross-links is first studied in compression using finite element. The numerical geometry of the fiber network relies on the morphological parameters of a real sample. Numerical simulations make it possible to follow, during the confined compression, the evolution of the different parameters, such as the distribution of fiber orientations, the distance between contacts or the volume fraction. These results provide a robust basis for developing the numerical model of the entangled cross-linked material which is then used to model the mechanical behavior in shear, and in particular to simulate and explain the hysteresis loops observed experimentally. At the end of this work, a numerical study is proposed to study the influence of different morphological parameters on the compressive stiffness and shear stiffness of the entangled cross-linked material.

La combustion prévaporisée prémélangée pauvre est une piste de choix pour réduire les émissions polluantes des moteurs d'avions mais peut conduire à l'apparition d'instabilités thermo-acoustiques. Afin d'améliorer la stabilité de telles flammes, l'étagement du combustible consiste à contrôler la distribution spatiale du carburant. Une telle procédure s'accompagne cependant d'une complexité accrue du système pouvant déboucher sur des phénomènes inattendus.Un brûleur à l'échelle de laboratoire alimenté par du dodécane liquide est utilisé dans cette thèse. Le combustible est injecté dans deux étages séparés, permettant ainsi de contrôler sa distribution. Cette particularité permet l'observation de différentes formes de flammes et notamment de points bistables pour lesquels deux flammes différentes peuvent exister malgré des conditions opératoires identiques.L'utilisation de diagnostics optiques à haute cadence (diffusion de Mie des gouttes de combustible et émission spontanée de la flamme) est couplée à des méthodes de post-traitement avancées comme la Décomposition en Modes Dynamiques. Ainsi, des mécanismes pilotant la stabilisation des flammes ainsi que leurs changements de forme sont proposés. Ils mettent notamment en lumière les interactions entre l'écoulement gazeux, les gouttes de combustible et la flamme. / A promising way to reduce jet engines pollutant emissions is the use of lean premixed prevaporized combustion but it tends to trigger thermo-acoustic instabilities. To improve the stability of these flames, a procedure called staging consists in splitting the fuel injection to control its spatial distribution. This however leads to an increased complexity and unexpected phenomena can occur.In the present work, a model gas turbine combustor fed with liquid dodecane is used. It is equipped with two fuel injection stages to control the fuel distribution in the burner. Different flame stabilizations can be observed and a bistable case where two flame shapes can exist for the same operating conditions is highlighted.High-speed optical diagnostics (fuel droplets Mie scatering and chemiluminescence measurements) are coupled with advanced post-processing methods like Dynamic Mode Decomposition. The results enable to propose mechanisms leading to flame stabilization and flame shape transitions. They show a strong interplay between the gaseous flow, the fuel droplets and the flame itself.