Disertační práce se zabývá asymptotickou stabilitou zpožděných diferenciálních rovnic a jejich diskretizací. V práci jsou uvažovány lineární zpožděné diferenciální rovnice s~konstantním i neohraničeným zpožděním. Jsou odvozeny nutné a postačující podmínky popisující oblast asymptotické stability jak pro exaktní, tak i diskretizovanou lineární neutrální diferenciální rovnici s konstantním zpožděním. Pomocí těchto podmínek jsou porovnány oblasti asymptotické stability odpovídajících exaktních a diskretizovaných rovnic a vyvozeny některé vlastnosti diskrétních oblastí stability vzhledem k měnícímu se kroku použité diskretizace. Dále se zabýváme lineární zpožděnou diferenciální rovnicí s neohraničeným zpožděním. Je uveden popis jejích exaktních a diskrétních oblastí asymptotické stability spolu s asymptotickým odhadem jejich řešení. V závěru uvažujeme lineární diferenciální rovnici s více neohraničenými zpožděními.

A time-delayed response of individual living organisms to information exchanged within flocks or swarms leads to the emergence of complex collective behaviors. A recent experimental setup by (Khadka et al 2018 Nat. Commun. 9 3864), employing synthetic microswimmers, allows to emulate and study such behavior in a controlled way, in the lab. Motivated by these experiments, we study a system of N Brownian particles interacting via a retarded harmonic interaction. For N 3 , we characterize its collective behavior analytically, by solving the pertinent stochastic delay-differential equations, and for N>3 by Brownian dynamics simulations. The particles form molecule-like non-equilibrium structures which become unstable with increasing number of particles, delay time, and interaction strength. We evaluate the entropy and information fluxes maintaining these structures and, to quantitatively characterize their stability, develop an approximate time-dependent transition-state theory to characterize transitions between different isomers of the molecules. For completeness, we include a comprehensive discussion of the analytical solution procedure for systems of linear stochastic delay differential equations in finite dimension, and new results for covariance and time-correlation matrices

Cette thèse est consacrée à l'étude d'une classe d'équations différentielles à retard dépendant de l'état -- ces équations provenant d'un modèle structuré en taille. La principale motivation de cette thèse provient de la volonté d'ajuster les paramètres du système d'équations étudiées vis-à-vis des données générées par un simulateur de forêts, appelé SORTIE. Deux types de forêts sont étudiés ici: d'une part une forêt ne comportant qu'une seule espèce d'arbre, et d'autre part une forêt comportant deux espèces d'arbres (au chapitre 2). Les simulations numériques du système d'équations correspondent relativement bien aux données générées par SORTIE, ce qui montre que le système considéré peut être utilisé afin d'écrire la dynamique de populations d'une forêt. De plus, un modèle plus étendu prenant en compte la position spatiale des arbres est proposé dans le chapitre 2, dans le cas de forêts possédant deux espèces d'arbres. Les simulations numériques de ce modèle permettent de visualiser la propagation spatiale des forêts. Les chapitres 3 et 4 se concentrent sur l'analyse mathématique des équations différentielles à retard considérées. Les propriétés du semi-flot associé au système sont étudiées au chapitre 3, où l'on démontre en particulier que ce semi-flot n'est pas continu en temps. Le caractère dissipatif et borné du semi-flot, pour des modèles de forêts comportant une ou deux espèces d'arbres, est étudié dans le chapitre 4. En outre, afin d'étudier la dynamique de population d'une forêt (d'une seule espèce d'arbre) après l'introduction d'un parasite, nous construisons dans le chapitre 5 un système proie-prédateur dont la proie (à savoir la forêt) est modélisée par le système d'équations différentielles à retard dépendant de l'état étudié auparavant, et dont le prédateur (à savoir le parasite) est modélisé par une équation différentielle ordinaire. De nombreuses simulations numériques associées à différents scénarios sont faites, afin d'explorer le comportement complexe des solutions du au couplage proie-prédateur et les équations à retard dépendant de l'état. / This thesis is devoted to the studies of a class of state-dependent delay differential equations. This class of equations is derived from a size-structured model.The motivation comes from the parameter fittings of this system to a forest simulator called SORTIE. Cases of both single species forest and two-species forest are considered in Chapter 2. The numerical simulations of the system correspond relatively very well to the forest data generated by SORTIE, which shows that this system is able to be used to describe the population dynamics of forests. Moreover, an extended model considering the spatial positions of trees is also proposed in Chapter 2 for the two-species forest case. From the numerical simulations of this spatial model one can see the diffusion of forests in space. Chapter 3 and 4 focus on the mathematical analysis of the state-dependent delay differential equations. The properties of semiflow generated by this system are studied in Chapter 3, where we find that this semiflow is not time-continuous. The boundedness and dissipativity of the semiflow for both single species model and multi-species model are studied in Chapter 4. Furthermore, in order to study the population dynamics after the introduction of parasites into a forest, a predator-prey system consisting of the above state-dependent delay differential equation (describing the forest) and an ordinary differential equation (describing the parasites) is constructed in Chapter 5 (only the single species forest is considered here). Numerical simulations in several scenarios and cases are operated to display the complex behaviours of solutions appearing in this system with the predator-prey relation and the state-dependent delay.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2018-11-12T14:27:26Z No. of bitstreams: 1 raffaelbechararameh.pdf: 1503042 bytes, checksum: 87e66fa77937ca9a85aac3231b27ac84 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2018-11-23T13:13:22Z (GMT) No. of bitstreams: 1 raffaelbechararameh.pdf: 1503042 bytes, checksum: 87e66fa77937ca9a85aac3231b27ac84 (MD5) / Made available in DSpace on 2018-11-23T13:13:22Z (GMT). No. of bitstreams: 1 raffaelbechararameh.pdf: 1503042 bytes, checksum: 87e66fa77937ca9a85aac3231b27ac84 (MD5) Previous issue date: 2018-08-31 / Para representar diferentes fenômenos e sistemas modelos matemáticos são largamente utilizados. Muitos deles são fundamentados em sistemas de equações diferenciais ordinárias (EDOs), isto é, baseiam-se em conjuntos de igualdades que envolvem variáveis dependentes, suas derivadas de primeira ordem e a variável independente. Neste trabalho, estudamos a modelagem da geração do potencial de ação em células excitáveis, como os neurônios. Existem dois modelos tradicionais e pioneiros que se destacam nessa área: Hodgkin-Huxley e FitzHugh-Nagumo. O objetivo desta dissertação é avaliar a possibilidade de modelar a geração do potencial de ação via uma única equação diferencial com atraso. Equações diferenciais com atraso são importantes por sua capacidade em reproduzir uma grande diversidade de fenômenos. Porém, seu uso na modelagem do potencial de ação de células excitáveis é ainda incipiente. Nesta dissertação, o método usado para alcançar este objetivo se baseou no desenvolvimento, inicialmente, de uma equação integro-diferencial que aproxima o sistema de EDOs. Em seguida, desenvolvemos uma aproximação para as integrais que usa termos tanto no instante atual quanto em instante anteriores, i.e., atrasados no tempo. Dessa forma, mostramos que é possível aproximar cada um dos sistemas de EDOS dos modelos de Hodgkin-Huxley e FitzHugh-Nagumo por uma única equação diferencial com atraso. Por fim, estes novos modelos são comparados com os originais, e são apontadas direções para a continuidade desta pesquisa. / To represent different phenomena and systems mathematical models are widely used. Many of them are based on systems of ordinary differential equations (ODEs), that is, they are based on sets of equalities involving dependent variables, their derivatives of first order and the independent variable. In this work, we study the modeling of action potential generation in excitable cells, such as neurons. There are two traditional and pioneering models that stand out in this area: Hodgkin-Huxley and FitzHugh-Nagumo. The objective of this dissertation is to evaluate the possibility of modeling the generation of the action potential via a single differential equation with delay. Differential equations with delay are important because of their capacity to reproduce a great diversity of phenomena. However, its use in modeling the action potential of excitable cells is still incipient. In this dissertation, the method used to achieve this goal was based on the development, initially, of an integral-differential equation that approximates the ODE system. Next, we develop an approximation for integrals that uses terms at both the current instant and the previous instant, i.e., time delayed. Thus, we show that it is possible to approximate each of the ODEs systems of the Hodgkin-Huxley and FitzHugh-Nagumo models by a single differential equation with delay. Finally, these new models are compared with the original ones, and directions are indicated for future works.

Disertační práce formuluje asymptotické odhady řešení tzv. sublineárních a superlineárních diferenciálních rovnic se zpožděním. V těchto odhadech vystupuje řešení pomocných funkcionálních rovnic a nerovností. Dále práce pojednává o kvalitativních vlastnostech diferenčních rovnic se zpožděním, které vznikly diskretizací studovaných diferenciálních rovnic. Pozornost je věnována souvislostem asympotického chování řešení rovnic ve spojitém a diskrétním tvaru, a to v obecném i v konkrétních případech. Studována je rovněž stabilita numerické diskretizace vycházející z $\theta$-metody. Práce obsahuje několik příkladů ilustrujících dosažené výsledky.

Systeme mit Zeitverzögerung sind dadurch charakterisiert, dass deren zukünftige Entwicklung durch den Zustand zum aktuellen Zeitpunkt nicht eindeutig festgelegt ist. Die Historie des Zustands muss in einem Zeitraum bekannt sein, dessen Länge Totzeit genannt wird und die Gedächtnislänge festlegt. In dieser Arbeit werden fundamentale Effekte untersucht, die sich ergeben, wenn die Totzeit zeitlich variiert wird. Im ersten Teil werden zwei Klassen periodischer Totzeitvariationen eingeführt. Da diese von den dynamischen Eigenschaften einer eindimensionalen iterierten Abbildung abgeleitet werden, die über die Totzeit definiert wird, werden die Klassen entsprechend der zugehörigen Dynamik konservativ oder dissipativ genannt. Systeme mit konservativer Totzeit können in Systeme mit konstanter Totzeit transformiert werden und besitzen gleiche charakteristische Eigenschaften. Dagegen weisen Systeme mit dissipativer Totzeit fundamentale Unterschiede z.B. in der Tangentialraumdynamik auf. Im zweiten Teil werden diese Ergebnisse auf Systeme angewendet, deren Totzeit im Vergleich zur internen Relaxationszeit des Systems groß ist. Es zeigt sich, dass ein durch dissipative Totzeitvariationen induzierter Mechanismus, genannt resonanter Dopplereffekt, unter anderem zu neuen Arten chaotischer Dynamik führt. Diese sind im Vergleich zur bekannten chaotischen Dynamik in Systemen mit konstanter Totzeit sehr niedrig-dimensional. Als Spezialfall wird das so genannte laminare Chaos betrachtet, dessen Zeitreihen durch nahezu konstante Phasen periodischer Dauer gekennzeichnet sind, deren Amplitude chaotisch variiert. Im dritten Teil dieser Arbeit wird auf der Basis experimenteller Daten und durch die Analyse einer nichtlinearen retardierten Langevin-Gleichung gezeigt, dass laminares Chaos robust gegenüber Störungen wie zum Beispiel Rauschen ist und experimentell realisiert werden kann. Es werden Methoden zur Zeitreihenanalyse entwickelt, um laminares Chaos in experimentellen Daten ohne Kenntnis des erzeugenden Systems zu detektieren. Mit diesen Methoden ist selbst dann eine Detektion möglich, wenn das Rauschen so stark ist, dass laminares Chaos mit bloßem Auge nur schwer erkennbar ist.:1. Introduction 2. Dissipative and conservative delays in systems with time-varying delay 3. Laminar Chaos and the resonant Doppler effect 4. Laminar Chaos: a robust phenomenon 5. Summary and concluding remarks A. Appendix / In systems with time-delay, the evolution of a system is not uniquely determined by the state at the current time. The history of the state must be known for a time period of finite duration, where the duration is called delay and determines the memory length of the system. In this work, fundamental effects arising from a temporal variation of the time-delay are investigated. In the first part, two classes of periodically time-varying delays are introduced. They are related to a specific dynamics of a one-dimensional iterated map that is defined by the time-varying delay. Referring to the related map dynamics the classes are called conservative or dissipative. Systems with conservative delay can be transformed into systems with constant delay, and thus have the same characteristic properties as constant delay systems. In contrast, there are fundamental differences, for instance, in the tangent space dynamics, between systems with dissipative delay and systems with constant delay. In the second part, these results are applied to systems with a delay that is considered large compared to the internal relaxation time of the system. It is shown that a mechanism induced by dissipative delays leads to new kinds of regular and chaotic dynamics. The dynamics caused by the so-called resonant Doppler effect is fundamentally different from the behavior known from systems with constant delay. For instance, the chaotic attractors in systems with dissipative delay are very low-dimensional compared to typical ones arising in systems with constant delay. An example of this new kind of low-dimensional dynamics is given by the so-called Laminar Chaos. It is characterized by nearly constant laminar phases of periodic duration, where the amplitude varies chaotically. In the third part of this work, it is shown that Laminar Chaos is a robust phenomenon, which survives perturbations such as noise and can be observed experimentally. Therefore experimental data is provided and a nonlinear delayed Langevin equation is analyzed. Using the robust features that characterize Laminar Chaos, methods for time series analysis are developed, which enable us to detect Laminar Chaos without the knowledge of the specific system that has generated the time series. By these methods Laminar Chaos can be detected even for comparably large noise strengths, where the characteristic properties are nearly invisible to the eye.:1. Introduction 2. Dissipative and conservative delays in systems with time-varying delay 3. Laminar Chaos and the resonant Doppler effect 4. Laminar Chaos: a robust phenomenon 5. Summary and concluding remarks A. Appendix