L'objet de cette thèse est l'étude de phénomènes critiques hors-équilibre. Pour décrire ces systèmes, l'utilisation d'équations de Langevin est souvent incontournable car elles permettent une description heuristique relativement simple du phénomène, construite en ajoutant un terme de bruit à la dynamique macroscopique. J'ai montré qu'il est toutefois possible, dans le cas des processus de réaction-diffusion, d'aller au delà de cette approche et de dériver une équation de Langevin exacte qui décrit la dynamique au niveau microscopique. Une seconde partie de ma thèse est consacrée à l'étude de modèles spécifiques de phénomènes critiques hors-équilibre à l'aide du groupe de renormalisation non-perturbatif (NPRG), une version moderne des blocs de spins de Wilson et Kadanoff. À l'équilibre, cet outil tire son succès de sa capacité à contrôler les fluctuations au voisinage de la transition grâce à un régulateur. Hors équilibre, les fluctuations temporelles doivent être traitées de la même façon, et j'ai donc conçu un régulateur qui contrôle à la fois les fluctuations spatiales et temporelles. Enfin, j'ai appliqué le NPRG à un modèle d'érosion. En effet, l'apparition générique de lois d'échelles dans les paysages suggère l'existence d'un mécanisme sous-jacent qui conduit ces systèmes à leur point critique. L'équation de Kardar-Parisi-Zhang modélise l'érosion à grande échelle (>2 km), mais ne s'accorde pas aux observations à plus petite échelle. Un modèle différent, tenant compte de l'anisotropie (la pente d'une montagne), fut donc suggéré. À l'aide du NPRG, je montre que ce modèle possède une ligne de points fixes qui correspond à un domaine continu d'exposants d'échelle. / This manuscript is focused on the study of critical phenomena taking place out-of-equilibrium. In the description of such phenomena, Langevin equations are ubiquitous and are usually derived in a phenomenological way by adding a noise term to a deterministic mean-field equation. However, I show that for reaction-diffusion processes it is in fact possible to derive an exact Langevin equation from the microscopic process. A second part of my thesis work has been devoted to the study of specific nonequilibrium critical phenomena using the nonperturbative renormalization group (NPRG), which is a modern implementation of Wilson and Kadanoff's block-spin idea. This tool, very powerful in an equilibrium context, takes care of the growing spatial fluctuations that arise near criticality through the use of a regulator. In a nonequilibrium context, the temporal fluctuations also have to be controlled. I have therefore designed a regulator that tackles both spatial and temporal fluctuations. Finally, I have applied the NPRG techniques to a model of landscape erosion: indeed, the generic scaling behaviour that appear in erosional landscapes suggests the existence of an underlying mechanism naturally fine-tuned to be critical. The Kardar-Parisi-Zhang equation seems to give a correct model for landscape erosion at large length scale (>2 km), but fails to predict the scaling observed at smaller scale. A different model was thus suggested which takes into account the intrinsic anisotropy at smaller length scale (the slope of the mountain). Using NPRG techniques, I show that this model possesses a line of fixed points associated with a continuous range of scaling exponents.

Scale invariance, collective behaviours and structural reorganization are crucial for portfolio management (portfolio composition, hedging, alternative definition of risk, etc.). This lack of any characteristic scale and such elaborated behaviours find their origin in the theory of complex systems. There are several mechanisms which generate scale invariance but maximum entropy models are able to explain both scale invariance and collective behaviours.<p>The study of the structure and collective modes of financial markets attracts more and more attention. It has been shown that some agent based models are able to reproduce some stylized facts. Despite their partial success, there is still the problem of rules design. In this work, we used a statistical inverse approach to model the structure and co-movements in financial markets. Inverse models restrict the number of assumptions. We found that a pairwise maximum entropy model is consistent with the data and is able to describe the complex structure of financial systems. We considered the existence of a critical state which is linked to how the market processes information, how it responds to exogenous inputs and how its structure changes. The considered data sets did not reveal a persistent critical state but rather oscillations between order and disorder.<p>In this framework, we also showed that the collective modes are mostly dominated by pairwise co-movements and that univariate models are not good candidates to model crashes. The analysis also suggests a genuine adaptive process since both the maximum variance of the log-likelihood and the accuracy of the predictive scheme vary through time. This approach may provide some clue to crash precursors and may provide highlights on how a shock spreads in a financial network and if it will lead to a crash. The natural continuation of the present work could be the study of such a mechanism. / Doctorat en Sciences économiques et de gestion / info:eu-repo/semantics/nonPublished

<div>We propose an order parameter theory of the quantum Hall nematic in high fractional Landau levels in terms of an Ising description. This new model solves a couple of extant problems in the literature: (1) The low-temperature behavior of the measured resistivity anisotropy is captured better by our model than previous theoretical treatments based on the electron nematic having XY symmetry. (2) Our model allows for the development of true long-range order at low temperature, consistent with the observation of anisotropic low-temperature transport.<br></div><div><div> We furthermore propose new experimental tests based on hysteresis that can distinguish whether any two-dimensional electron nematic is in the XY universality class (as previously proposed in high fractional Landau levels), or in the Ising universality class (as we propose). Given the growing interest in electron nematics in many materials, we expect our proposed test of universality class to be of broad interest.</div><div> </div><div> Whereas the XY model in two dimensions does not have a long-range ordered phase, the addition of uniaxial random field disorder induces a long-range ordered phase in which the spontaneous magnetization points perpendicular to the random field direction, via an order-by-disorder transition. We have shown that this spontaneous magnetization is robust against a rotating driving field, up to a critical driving field amplitude. Thus we have found evidence for a new non-equilibrium phase transition that was unknown before in this model. Moreover, we have discovered an incredible anomaly at this nonequilibrium phase transition: the critical region is accompanied by a cascade of period multiplication events. This physics is reminiscent of the period bifurcation cascade signaling the transition to chaos in nonlinear systems, and of the approach to the irreversibility transition in models of yield in amorphous solids~\cite{reichhardt-dahmen,leishangthem_yielding_2017}. This period multiplication cascade is surprising to be present in a statistical mechanics model, and suggests that the non-equilibrium transition as a function of driving field amplitude is part of a larger class of transitions in dynamical systems.</div><div> Moreover, we show that this multi-period behavior represents a new emergent classical discrete time-crystal, since the new period is robust against changes to initial conditions and low-temperature fluctuations over hundreds of driving period cycles.</div><div><br></div><div> We expect this work to be of broad interest, further encouraging cross-fertilization between the rapidly growing field of time-crystals with the well-established fields of nonequilibrium phase transitions and dynamical systems.</div><div> </div><div> Geometrical configurations gave us a better understanding of the multi-period behavior of the limit-cycles.</div><div> Moreover, surface probes are continually evolving and generating vast amounts of spatially resolved data of quantum materials, which reveal a lot of detail about the microscopic and macroscopic properties of the system. <br></div><div> Materials undergoing a transition between two distinct states, phase separate.</div><div> These phase-separated regions form intricate patterns on the observable surface, which can encode model-specific information, including interaction, dimensionality, and disorder. </div><div> While there are rigorous methods for understanding these patterns, they turn out to be time-consuming as well as requiring expertise. </div><div> We show that a well-tuned machine learning framework can decipher this information with minimal effort from the user.</div><div> We expect this to be widely used by the scientific community to fast-track comprehension of the underlying physics in these materials.</div><div><br></div></div>

Diese Arbeit befaßt sich mit der Koppelung von Vertex-Modellen an die planaren $\phi^4$-Zufallsgraphen des Zugangs zur Quantengravitation über dynamische Polygonifizierungen. Das betrachtete System hat eine doppelte Bedeutung, einerseits als die Koppelung einer konformen Feldtheorie mit zentraler Ladung $C=1$ an zweidimensionale Euklidische Quantengravitation, andererseits als Anwendung von geometrischer, "annealed" Unordnung auf ein prototypisches Modell der statistischen Mechanik. Da das Modell mit Hilfe einer großangelegten Reihe von Monte Carlo Simulationen untersucht wird, müssen entsprechende Techniken für die Simulation von dynamischen Quadrangulierungen bzw. die dualen $\phi^4$-Graphen entwickelt werden. Hierzu werden verschiedene Algorithmen und die dazugehörigen Züge vorgeschlagen und hinsichtlich ihrer Ergodizität und Effizienz untersucht. Zum Vergleich mit exakten Ergebnissen werden die Verteilung der Koordinationszahlen bzw. bestimmte Analoga davon konstruiert. Für Simulationen des $F$-Modells auf $\phi^4$-Zufallsgraphen wird ein Ordnungsparameter für den antiferroelektrischen Phasenübergang mit Hilfe einer Plakettenspindarstellung formuliert. Ausführliche "finite-size scaling"-Analysen des Kosterlitz-Thouless-Phasenübergangs des $F$-Modells auf dem Quadratgitter und auf Zufallsgraphen werden vorgestellt und die Positionen der jeweiligen kritischen Punkte sowie die dazugehörigen kritischen Exponenten werden bestimmt. Die Rückreaktion des Vertex-Modells auf die Zufallsgraphen wird in Form der Koordinationszahlverteilung, der Verteilung der "Baby-Universen" und dem daraus resultierenden String-Suszeptibilitäts-Exponenten sowie durch die geometrische Zweipunktfunktion analysiert, die eine Schätzung der intrinsischen Hausdorff-Dimension des gekoppelten Systems liefert. / In this thesis, the coupling of ice-type vertex models to the planar $\phi^4$ random graphs of the dynamical polygonifications approach to quantum gravity is considered. The investigated system has a double significance as a conformal field theory with central charge $C=1$ coupled to two-dimensional Euclidean quantum gravity and as the application of a special type of annealed connectivity disorder to a prototypic model of statistical mechanics. Since the model is analyzed by means of large-scale Monte Carlo simulations, suitable simulation techniques for the case of dynamical quadrangulations and the dual $\phi^4$ random graphs have to be developed. Different algorithms and the associated update moves are proposed and investigated with respect to their ergodicity and performance. For comparison to exact results, the co-ordination number distribution of the dynamical polygonifications model, or certain analogues of it, are constructed. For simulations of the 6-vertex $F$ model on $\phi^4$ random graphs, an order parameter for its anti-ferroelectric phase transitions is constructed in terms of a "plaquette spin" representation. Extensive finite-size scaling analyses of the Kosterlitz-Thouless point of the square-lattice and random graph $F$ models are presented and the locations of the critical points as well as the corresponding critical exponents are determined. The back-reaction of the coupled vertex model on the random graphs is investigated by an analysis of the co-ordination number distribution, the distribution of "baby universes" and the string susceptibility exponent as well as the geometric two-point function, yielding an estimate for the internal Hausdorff dimension of the coupled system.

<p dir="ltr">Complexity has played an increasingly important role in recent years. In this dissertation, we study some notions of complexity in systems that exhibit critical behaviour. Our results show that complexity as it is generally understood in holographic and lattice models of criticality can have several ambiguities. But despite these ambiguities, there are some features that are universally true. On the phase diagram of the system, it is the critical point which has the most complex ground state. States of physical systems with a large complexity tend to be hard to simulate using quantum circuits. Near the critical point, there is a part of complexity which is non-analytic and scales universally, i.e, the scaling is independent of the microscopic details of the Hamiltonian but depends only on the dimensionality of the system, and of the deforming operator. The coefficient of this term is unambiguous, i.e, it is not affected by the various changes in the definition of complexity which plague all the analytic terms near the critical point. We show this in lattice, field-theoretic and holographic calculations. These results were first presented in our earlier studies.</p>

Condensed matter systems undergoing phase transitions rarely allow exact solutions. The presence of disorder renders the situation  even worse but collective Monte Carlo methods and parallel algorithms allow numerical descriptions. This thesis considers classical phase transitions in disordered spin systems in general and in effective models of superfluids with disorder and novel interactions in particular. Quantum phase transitions are considered via a quantum to classical mapping. Central questions are if the presence of defects changes universal properties and what qualitative implications follow for experiments. Common to the cases considered is that the disorder maps out correlated structures. All results are obtained using large-scale Monte Carlo simulations of effective models capturing the relevant degrees of freedom at the transition. Considering a model system for superflow aided by a defect network, we find that the onset properties are significantly altered compared to the $\lambda$-transition in $^{4}$He. This has qualitative implications on expected experimental signatures in a defect supersolid scenario. For the Bose glass to superfluid quantum phase transition in 2D we determine the quantum correlation time by an anisotropic finite size scaling approach. Without a priori assumptions on critical parameters, we find the critical exponent $z=1.8 \pm 0.05$ contradicting the long standing result $z=d$. Using a 3D effective model for multi-band type-1.5 superconductors we find that these systems possibly feature a strong first order vortex-driven phase transition. Despite its short-range nature details of the interaction are shown to play an important role. Phase transitions in disordered spin models exposed to correlated defect structures obtained via rapid quenches of critical loop and spin models are investigated. On long length scales the correlations are shown to decay algebraically. The decay exponents are expressed through known critical exponents of the disorder generating models. For cases where the disorder correlations imply the existence of a new long-range-disorder fixed point we determine the critical exponents of the disordered systems via finite size scaling methods of Monte Carlo data and find good agreement with theoretical expectations. / <p>QC 20150306</p>

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal.

Made available in DSpace on 2017-07-21T19:26:00Z (GMT). No. of bitstreams: 1 Cristiano Santos Lopes.pdf: 13551286 bytes, checksum: 1ba67329a96370d271e71e39f117ec62 (MD5) Previous issue date: 2010-11-19 / In this work we report on conductivity fluctuation measurements in polycrystalline samples of the Er1xPrxBa2Cu3O7 and Gd1xPrxBa2Cu3O7superconductor. Pr contents are 0.00, 0.05 and 0.10. Samples were prepared by the standard solid-state reaction technique and characterized by SEM, X-ray diffraction and electrical transport experiments. The samples were granular and homogeneous. The results were analyzed in terms of the temperature derivative of the resistivity and of the logarithmic temperature derivative of the conductivity, what allowed identifying power-law divergences of the conductivity. For Er1−xPrxBa2Cu3O7 samples, the results show that the transition proceeds in two stages: pairing and coherence transition. Also, our results, from the critical exponent analysis, show a two-peak splitting at pairing transition, indicating possibly a phase separation. On approaching the zero resistance state, our results show a power-law behavior that corresponds to a phase transition from paracoherent to a coherent state of the granular array. For Gd1−xPrxBa2Cu3O7 samples, it was observed that the critical temperature decreases and that the transition width increases with increasing Pr doping. Systematic measurements of fluctuation conductivity are reported and special attention is taken above the critical temperature, where Gaussian and critical regimes are observed. Below the critical temperature, on approaching the zero resistance state, our results show a power-law behavior consistent with a phase transition from a paracoherent to a coherent state of the granular array. / Neste trabalho foram estudados os efeitos das flutuações térmicas na condutividade elétrica de amostras policristalinas com base nos supercondutores Er1xPrxBa2Cu3O7 e Gd1xPrx Ba2Cu3O7. A quantidade de Pr considerada foi x = 0, 00, 0, 05 e 0, 10. As amostras foram preparadas pela técnica padrão de reação de estado sólido e caracterizadas por microscopia eletrônica de varredura, difração de raios-X e medidas de transporte eletrônico. As amostras são homogêneas e granulares. Para a obtenção dos expoentes críticos, os dados foram analisados em termos da derivada da resistividade em função da temperatura e da derivada logarítmica da resistividade em função da temperatura. Essa análise permitiu identificar regimes em lei de potência na paracondutividade durante a transição normal-supercondutora. Para a amostra de Er1−xPrxBa2Cu3O7, os resultados mostraram que a transição ocorre em dois estágios: transição de pareamento e transição de coerência. Contudo, os resultados obtidos através da análise dos expoentes críticos mostraram o desdobramento em dois picos da transição de pareamento, indicando uma possível separação de fase. Na aproximação do estado de resistência nula, os resultados mostraram um comportamento em lei de potência que corresponde à transição de fase paracoerente-coerente, típica de sistemas granulares. Para a amostra de Gd1−xPrxBa2Cu3O7, foi observado que a temperatura crítica diminuiu e que a largura da transição aumentou com o acréscimo da dopagem de Pr. Medidas sistemáticas na condutividade são apresentadas e é dada atenção especial em temperaturas ligeiramente acima da temperatura crítica, região na qual regimes Gaussianos e críticos são observados. Abaixo da temperatura crítica, na aproximação ao estado de resistência nula, os resultados mostram claramente regimes em lei de potência consistentes com a transição de fase paracoerente-coerente.

Cette thèse présente de manière synthétique mes travaux de recherche dont les deux thématiques principales sont le comportement critique du modèle de Potts en présence de désordre et le vieillissement de modèles de spin lors d'une trempe.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal