Dans cette thèse, nous nous intéressons à certains aspects concernant les phénomènes hadroniques à basse énergie sous 1 GeV, en dessous de laquelle la symétrie chirale de la Chromodynamique Quantique (QCD) est spontanément brisée. En dessous de cette échelle d'énergie, le spectre de QCD se réduit à un octet de mésons légers pseudo-scalaires (π, K and η). Mais à cause du confinement, QCD sous 1 GeV devient hautement non perturbative – il n'est donc plus possible de décrire à basse énergie la dynamique de ces mésons en termes de gluons et de quarks (ici seuls les quarks légers u, d et s sont concernés). Deux alternatives principales à cet obstacle majeur existent néanmoins: la QCD sur réseau ainsi que les Théories Effectives des Champs. La QCD sur réseau consiste à calculer de manière numériques les diverses observables hadroniques, alors que les théories effectives permettent de nouveau une approche analytique (et perturbative) adaptée à une échelle d'énergie donnée. Dans le cas de QCD à basse énergie, c'est la Théorie Chirale des Perturbations (ChiPT) qui joue le rôle de théorie effective. Cette théorie peut être construite à partir de deux saveurs de quarks légers (u et s) ou trois (u,d, et s). Il est alors possible d'utiliser certains résultats de calculs sur réseau (ainsi que certains résultats expérimentaux) afin d'extraire des valeurs numériques pour les divers paramètres libres que contient la théorie chirale. Il fut néanmoins observé que le développement en séries chirales de quelques observables hadroniques sont numériquement “malades” dans le cadre de la théorie à trois saveurs. En effet, des travaux antérieurs montrent qu'il pourrait exister une possible compétition numérique entre l'Ordre Dominant (LO) et l'Ordre Sous-Dominant (NLO): en place de la hiérarchie usuelle LO>>NLO, l'équivalence LO~NLO prévalerait. La partie principale de la thèse consiste ainsi à la description et l'utilisation d'une version alternative de ChiPT, nommée Théorie Chirale des Perturbations Ressommée (ReChiPT ). Quelques observables hadroniques de basse energie sont calculées puis étudiées dans ce cadre “ressommé”, puis nous procédons à l'ajustement de certaines données de QCD sur réseau obtenues par des simulations à 2+1 quarks dynamiques sur ces observables exprimées en ReChiPT: les constantes de désintégrations et les masses de l'octet (π, K, η), ainsi que les facteurs de forme Kl3. Nous testons ensuite la validité de notre assertion concernant la possible compétition numérique observée dans les séries chirales. Enfin, dans la dernière partie, nous discutons plusieurs aspects analytiques et numériques concernant certaines quantités topologiques liées de manière intrinsèque à la très complexe structure du vide de QCD, dans le cadre de ChiPT (ressommé), et nous confrontons de nouveau cette étude à des données réseau 2+1. / In this thesis, we focus on some aspects concerning hadronic phenomena at low energy, below 1 GeV, under which the spontaneous breaking of chiral symmetry takes place. Under this scale, the spectrum of Quantum Chromodynamics reduces to an octet of light pseudo-scalar mesons (π, K and η). But because of the confinement property, QCD under 1 GeV is highly non-perturbative, it is thus not possible to describe at low energy the dynamics of these mesons in terms of gluons and quarks (in that case the three light quarks u,d, and s). Two main alternatives exist to circumvent this major obstacle: Lattice QCD and Effective Field Theories. Lattice QCD is concerned with the numerical computations of various hadronic observables, while Effective Field Theories correspond to analytical frameworks adapted to a particular energy scale. In the case of QCD at low energy, this role is devoted to Chiral Perturbation Theory (ChiPT). This theory can be built either from two quark flavours (u and d), or three (u,d, and s). Using the numerical results from Lattice QCD, it is possible to obtain numerical values for the unknown parameters that ChPT contains. It was however observed that the series expansions of hadronic observables stemming from ChiPT calculations do not “behave well” numerically in the three-flavour case. Indeed, previous works shown that there could exists at the numerical level a competition between the Leading and the Next-to- Leading order (LO and NLO); i.e., instead of the usually expected hierarchy LO>>NLO, one would have LO~NLO. The main part of the thesis work consists in the description and the use of a modified version of ChiPT allowing this numerical competition in the chiral series that was called “Resummed ChiPT”. Within this “Resummed” framework, we proceed to fitting data from 2+1 lattice calculations to hadronic observables computed in ChiPT: decay constants and masses of π, K and η, and Kl3 form factors, and check the consistency of our claim about the numerical competition in ChiPT expansions. In the last part, we discuss topological quantities that are intrinsically tied to the very complex structure of the QCD vacuum, in the (resummed) ChiPT framework and in the light of 2+1 lattice data, in their analytical and numerical aspects.

In dieser Arbeit wird die Berechnung der farbführenden Zweischleifen-QCD-Korrekturen für die Top-Quark-Paarproduktion mit einem zusätzlichen Jet an einem Lepton-Collider präsentiert. Das Matrixelement wird in Vektor- und Axial-Vektorströme zerlegt und die Ströme werden weiter in Dirac-Spinorstrukturen und Formfaktoren zerlegt. Die Formfaktoren werden mit Projektoren extrahiert. Die auftretenden Feynmanintegrale werden mittels IBP-Identitäten und Dimensionsverschiebungstransformationen durch eine Basis quasi-finiter Masterintegrale in 6−2ϵ Dimensionen ausgedrückt. Die Mehrheit der Feynmanintegrale gehört zu einer Doppelbox-Integralfamilie. Die Berechnung der Masterintegrale erfolgt durch numerisches Lösen von Differentialgleichungen in kinematischen Invarianten. Asymptotische Reihenentwicklungen der Masterintegrale in der Top-Quarkmasse werden verwendet, um die Anfangsbedingungen für die numerischen Lösungen der Differentialgleichungen zu bestimmen. Die führenden Terme dieser Entwicklung werden mit der Expansion-by-Regions-Methode berechnet. Höhere Reihenkoeffizienten werden durch die Anwendung einer Differentialgleichung auf einen Ansatz für die Reihenentwicklung bestimmt. Die renormierten Formfaktoren und die farbführende Zweischleifenamplitude werden an einem Referenzphasenraumpunkt zu hoher Präzision numerisch ausgewertet. Die Resultate werden mit elektroschwachen Ward-Identitäten und durch numerische Vergleiche der IR-Singularitäten mit der erwarteten Singularitätsstruktur überprüft. / In this work, the calculation of the leading-colour two-loop QCD corrections for top-quark pair production with an additional jet at a lepton collider is presented. The matrix element is decomposed into vector and axial-vector currents and the currents are further decomposed into Dirac spinor structures and form factors. The form factors are extracted with projectors. The Feynman integrals are reduced to a quasi-finite basis in 6 − 2ϵ dimensions using IBP identities and dimension-shift transformations. The majority of master integrals belong to a double-box integral family. The master integrals are computed by numerically solving systems of differential equations in the kinematic invariants. Asymptotic expansions of the master integrals in the top-quark mass variable are used to calculate initial conditions for the numerical differential equation solutions. The leading terms of the expansion are obtained with the expansion by regions and the higher orders are calculated by solving a system of equations obtained from applying the differential equation onto an ansatz of the expansion. The renormalized form factors and the leading-colour two-loop amplitude are evaluated numerically to high precision at a benchmark phase space point. The results are cross-checked with electroweak Ward identities and by numerically comparing the IR singularities with the expected singularity structure.

Dans cette thèse, nous nous intéressons à certains aspects concernant les phénomènes hadroniques à basse énergie sous 1 GeV, en dessous de laquelle la symétrie chirale de la Chromodynamique Quantique (QCD) est spontanément brisée. En dessous de cette échelle d'énergie, le spectre de QCD se réduit à un octet de mésons légers pseudo-scalaires (π, K and η). Mais à cause du confinement, QCD sous 1 GeV devient hautement non perturbative - il n'est donc plus possible de décrire à basse énergie la dynamique de ces mésons en termes de gluons et de quarks (ici seuls les quarks légers u, d et s sont concernés). Deux alternatives principales à cet obstacle majeur existent néanmoins: la QCD sur réseau ainsi que les Théories Effectives des Champs. La QCD sur réseau consiste à calculer de manière numériques les diverses observables hadroniques, alors que les théories effectives permettent de nouveau une approche analytique (et perturbative) adaptée à une échelle d'énergie donnée. Dans le cas de QCD à basse énergie, c'est la Théorie Chirale des Perturbations (ChiPT) qui joue le rôle de théorie effective. Cette théorie peut être construite à partir de deux saveurs de quarks légers (u et s) ou trois (u,d, et s). Il est alors possible d'utiliser certains résultats de calculs sur réseau (ainsi que certains résultats expérimentaux) afin d'extraire des valeurs numériques pour les divers paramètres libres que contient la théorie chirale. Il fut néanmoins observé que le développement en séries chirales de quelques observables hadroniques sont numériquement "malades" dans le cadre de la théorie à trois saveurs. En effet, des travaux antérieurs montrent qu'il pourrait exister une possible compétition numérique entre l'Ordre Dominant (LO) et l'Ordre Sous-Dominant (NLO): en place de la hiérarchie usuelle LO>>NLO, l'équivalence LO~NLO prévalerait. La partie principale de la thèse consiste ainsi à la description et l'utilisation d'une version alternative de ChiPT, nommée Théorie Chirale des Perturbations Ressommée (ReChiPT ). Quelques observables hadroniques de basse energie sont calculées puis étudiées dans ce cadre "ressommé", puis nous procédons à l'ajustement de certaines données de QCD sur réseau obtenues par des simulations à 2+1 quarks dynamiques sur ces observables exprimées en ReChiPT: les constantes de désintégrations et les masses de l'octet (π, K, η), ainsi que les facteurs de forme Kl3. Nous testons ensuite la validité de notre assertion concernant la possible compétition numérique observée dans les séries chirales. Enfin, dans la dernière partie, nous discutons plusieurs aspects analytiques et numériques concernant certaines quantités topologiques liées de manière intrinsèque à la très complexe structure du vide de QCD, dans le cadre de ChiPT (ressommé), et nous confrontons de nouveau cette étude à des données réseau 2+1.

With the constantly increasing precision of experimental data acquired at the current collider experiments Tevatron and LHC the theoretical uncertainty on the prediction of multiparticle final states has to decrease accordingly in order to have meaningful tests of the underlying theories such as the Standard Model. A pure leading order calculation, defined in the perturbative expansion of said theory in the interaction constant, represents the classical limit to such a quantum field theory and was already found to be insufficient at past collider experiments, e.g. LEP or Hera. Such a leading order calculation can be systematically improved in various limits. If the typical scales of a process are large and the respective coupling constants are small, the inclusion of fixed-order higher-order corrections then yields quickly converging predictions with much reduced uncertainties. In certain regions of the phase space, still well within the perturbative regime of the underlying theory, a clear hierarchy of the inherent scales, however, leads to large logarithms occurring at every order in perturbation theory. In many cases these logarithms are universal and can be resummed to all orders leading to precise predictions in these limits. Multiparticle final states now exhibit both small and large scales, necessitating a description using both resummed and fixed-order results. This thesis presents the consistent combination of two such resummation schemes with fixed-order results. The main objective therefor is to identify and properly treat terms that are present in both formulations in a process and observable independent manner. In the first part the resummation scheme introduced by Yennie, Frautschi and Suura (YFS), resumming large logarithms associated with the emission of soft photons in massive Qed, is combined with fixed-order next-to-leading matrix elements. The implementation of a universal algorithm is detailed and results are studied for various precision observables in e.g. Drell-Yan production or semileptonic B meson decays. The results obtained for radiative tau and muon decays are also compared to experimental data. In the second part the resummation scheme introduced by Dokshitzer, Gribov, Lipatov, Altarelli and Parisi (DGLAP), resumming large logarithms associated with the emission of collinear partons applicable to both Qcd and Qed, is combined with fixed-order next-to-leading matrix elements. While the focus rests on its application to Qcd corrections, this combination is discussed in detail and the implementation is presented. The resulting predictions are evaluated and compared to experimental data for a multitude of processes in four different collider environments. This formulation has been further extended to accommodate real emission corrections to beyond next-to-leading order radiation otherwise described only by the DGLAP resummation. Its results are also carefully evaluated and compared to a wide range of experimental data.

With the constantly increasing precision of experimental data acquired at the current collider experiments Tevatron and LHC the theoretical uncertainty on the prediction of multiparticle final states has to decrease accordingly in order to have meaningful tests of the underlying theories such as the Standard Model. A pure leading order calculation, defined in the perturbative expansion of said theory in the interaction constant, represents the classical limit to such a quantum field theory and was already found to be insufficient at past collider experiments, e.g. LEP or Hera. Such a leading order calculation can be systematically improved in various limits. If the typical scales of a process are large and the respective coupling constants are small, the inclusion of fixed-order higher-order corrections then yields quickly converging predictions with much reduced uncertainties. In certain regions of the phase space, still well within the perturbative regime of the underlying theory, a clear hierarchy of the inherent scales, however, leads to large logarithms occurring at every order in perturbation theory. In many cases these logarithms are universal and can be resummed to all orders leading to precise predictions in these limits. Multiparticle final states now exhibit both small and large scales, necessitating a description using both resummed and fixed-order results. This thesis presents the consistent combination of two such resummation schemes with fixed-order results. The main objective therefor is to identify and properly treat terms that are present in both formulations in a process and observable independent manner. In the first part the resummation scheme introduced by Yennie, Frautschi and Suura (YFS), resumming large logarithms associated with the emission of soft photons in massive Qed, is combined with fixed-order next-to-leading matrix elements. The implementation of a universal algorithm is detailed and results are studied for various precision observables in e.g. Drell-Yan production or semileptonic B meson decays. The results obtained for radiative tau and muon decays are also compared to experimental data. In the second part the resummation scheme introduced by Dokshitzer, Gribov, Lipatov, Altarelli and Parisi (DGLAP), resumming large logarithms associated with the emission of collinear partons applicable to both Qcd and Qed, is combined with fixed-order next-to-leading matrix elements. While the focus rests on its application to Qcd corrections, this combination is discussed in detail and the implementation is presented. The resulting predictions are evaluated and compared to experimental data for a multitude of processes in four different collider environments. This formulation has been further extended to accommodate real emission corrections to beyond next-to-leading order radiation otherwise described only by the DGLAP resummation. Its results are also carefully evaluated and compared to a wide range of experimental data.:1. Introduction 1.1 Event generators 1.2 The event generator Sherpa 1.3 Outline of this thesis Part I YFS resummation & fixed order calculations 2 Yennie-Frautschi-Suura resummation 2.1 Resummation of virtual photon corrections 2.2 Resummation of real emission corrections 2.3 The Yennie-Frautschi-Suura form factor 3 A process independent implementation in Sherpa 3.1 The Algorithm 3.1.1 The master formula 3.1.2 Phase space transformation 3.1.3 Mapping of momenta 3.1.4 Event generation 3.2 Higher Order Corrections 3.2.1 Approximations for real emission matrix elements 3.2.2 Real emission corrections 3.2.3 Virtual emission corrections 4 The Z lineshape and radiative lepton decay corrections 4.1 The Z lineshape 4.1.1 Radiation pattern 4.1.2 Numerical stability 4.2 Radiative lepton decays 4.3 Summary and conclusions 5 Electroweak corrections to semileptonic B decays 5.1 Tree-level decay 5.2 Next-to-leading order corrections 5.2.1 Matching of different energy regimes 5.2.2 Short-distance next-to-leading order corrections 5.2.3 Long-distance next-to-leading order corrections 5.2.4 Structure dependent terms 5.2.5 Soft-resummation and inclusive exponentiation 5.3 Methods 5.3.1 BLOR 5.3.2 Sherpa/Photons 5.3.3 PHOTOS 5.4 Results 5.4.1 Next-to-leading order corrections to decay rates 5.4.2 Next-to-leading order corrections to differential rates 5.4.3 Influence of explicit short-distance terms 5.5 Summary and conclusions Part II DGLAP resummation & fixed order calculations 6 DGLAP resummation & approximate higher order corrections 6.1 Dokshitzer-Gribov-Lipatov-Altarelli-Parisi resummation 6.1.1 The naive parton model 6.1.2 QCD corrections to the parton model 6.1.3 Factorisation and the collinear counterterm 6.1.4 The DGLAP equations 6.2 Parton evolution 6.2.1 Approximate real emission cross sections 6.2.2 Parton evolution 6.2.3 Scale choices for the running coupling 6.3 Soft emission corrections 7 The reinterpretation and automisation of the POWHEG method 7.1 Decomposition of the real-emission cross sections 7.2 Construction of a parton shower 7.3 Matrix element corrections to parton showers 7.4 The reformulation of the POWHEG method 7.4.1 Approximate NLO cross sections 7.4.2 The POWHEG method and its accuracy 7.5 The single-singularity projectors 7.6 Theoretical ambiguities 7.7 MC@NLO 7.8 Realisation of the POWHEG method in the Sherpa Monte Carlo 7.8.1 Matrix elements and subtraction terms 7.8.2 The parton shower 7.8.3 Implementation & techniques 7.8.4 Automatic identification of Born zeros 7.9 Results for processes with trivial colour structures 7.9.1 Process listing 7.9.2 Tests of internal consistency 7.9.3 Comparison with tree-level matrix-element parton-shower merging 7.9.4 Comparison with experimental data 7.9.5 Comparison with existing POWHEG 7.10 Results for processes with non-trivial colour structures 7.10.1 Comparison with experimental data 7.11 Summary and conclusions 8 MENLOPS 8.1 Improving parton showers with higher-order matrix elements 8.1.1 The POWHEG approach 8.1.2 The ME+PS approach 8.2 Merging POWHEG and ME+PS - The MENLOPS 8.3 Results 8.3.1 Merging Systematics 8.3.2 ee -> jets 8.3.3 Deep-inelastic lepton-nucleon scattering 8.3.4 Drell-Yan lepton-pair production 8.3.5 W+jets Production 8.3.6 Higgs boson production 8.3.7 W-pair+jets production 8.4 Summary and conclusions Summary Appendix A Details on the YFS resummation implementation A.1 The YFS-Form-Factor A.1.1 Special cases A.2 A.2.1 Avarage photon multiplicity A.2.2 Photon energy A.2.3 Photon angles A.2.4 Photons from multipoles A.3 Massive dipole splitting functions A.3.1 Final State Emitter, Final State Spectator A.3.2 Final State Emitter, Initial State Spectator A.3.3 Initial State Emitter, Final State Spectator B Formfactors and higher order matrix elements for semileptonic B decays B.1 Form factor models of exclusive semileptonic B meson decays B.1.1 Form factors for B -> D l nu B.1.2 Form factors for B -> pi l nu B.1.3 Form factors for B -> D0* l nu B.2 NLO matrix elements B.2.1 Real emission matrix elements B.2.2 Virtual emission matrix elements B.3 Scalar Integrals B.3.1 General definitions B.3.2 Tadpole integrals B.3.3 Bubble integrals B.3.4 Triangle integrals C Explicit form of the leading order Altarelli-Parisi splitting functions C.1 Collinear limit of real emission matrix elements C.1.1 q -> gq splittings C.1.2 q -> qg splittings C.1.3 g -> qq splittings C.1.4 g -> gg splittings Bibliography

In dieser Arbeit adressieren wir die Schwierigkeit, Präzisionsvorhersagen mit dem kompletten Satz theoretischer Unsicherheiten in der perturbativen Quantenchromodynamik im Rahmen von Monte-Carlo-Simulationen zu treffen, angesichts der zunehmenden Komplexität der dazu nötigen Berechnungen. Die Anforderungen an die Rechenleistung können so groß sein, dass nicht in jeder Anwendung die bestmögliche Präzision erzielt wird. Wir präsentieren eine Reweighting-Methode für den Monte-Carlo-Ereignisgenerator SHERPA. Diese erstellt Variationen der nominellen Vorhersage mit vergleichsweise geringem zusätzlichen Zeitaufwand. Die Methode ist kompatibel mit aktuellen Multijet-Berechnungen nächsthöherer Ordnung, die mit Korrekturen von allen Ordnungen durch einen Partonschauer versehen sind. Zusätzlich diskutieren wir neue Entwicklungen für einen weiteren Reweighting-Ansatz, der auf QCD-Interpolationsgittern beruht. Diese ermöglichen noch schnellere Variationen für Berechnungen fester Ordnung. Solche Gitter können für Monte-Carlo-Simulationen automatisiert erstellt werden mithilfe von Interfaces wie MCgrid. Unsere Verbesserungen für MCgrid ermöglichen die Erstellung vielseitigerer Gitter, die eine größere Klasse von Berechnungen, Gitter-Implementierungen und Skalenvariationen unterstützen. Darüber hinaus diskutieren wir, auf welche Weise solche Gitter für die Unterstützung von Resummationseffekten erweitert werden müssten. Neben dem Reweighting studieren wir noch die Verwendung von Extrapolationsmethoden für die Vorhersage von Jet-Raten hoher Multiplizitäten, welche an zukünftigen Hochenergiebeschleunigern allgegenwärtig sein werden. Diese Methoden basieren auf dem Skalierungsverhalten der Jet-Raten. Eingebettet ist diese Studie in eine allgemeinere Diskussion der zu erwartenden Jet-Aktivität an einem Proton-Proton-Beschleuniger mit einer Schwerpunktsenergie von 100 TeV.

La physique actuelle cherche, à côté des expériences, à vérifier et déduire les lois de la nature en simulant les modèles physiques sur d'énormes ordinateurs. Cette thèse explore comment accélérer ces simulations en améliorant les programmes qui les font tourner. L'application de référence est la chromodynamique quantique sur réseaux (LQCD pour "Lattice Quantum Chromodynamics"), une branche de la théorie quantique des champs, tournant sur le plus récent des supercalculateurs d'IBM, le Blue Gene/Q.Dans un premier temps, on améliore le code source de tmLQCD, un programme de LQCD, dont l'opération clef pour la performance est un stencil à 8 points en dimension 4. On étudie deux stratégies d'optimisation différentes: la première se donne comme priorité d'améliorer la localité spatiale et temporelle; la seconde utilise le préchargement matériel de flux de données. Sur le Blue Gene/Q, la première stratégie permet d'atteindre 20% de la performance crête théorique. La seconde, avec jusqu'à 54% de la performance crête est bien meilleure mais utilise 4 fois plus de mémoire car elle stocke les résultats dans l'ordre où les utilise le stencil suivant, ce qui requiert de dupliquer des données. Les autres techniques exploitées sont la programmation directe du système de communication (appelé MUSPI chez IBM), un mécanisme allégé de gestion des threads, le préchargement explicite de certaines données (à l'aide de l'instruction dcbt) et la vectorisation manuelle (en utilisant les instructions SIMD de largeur 4; appelé QPX par IBM). Le préchargement de liste et la mémoire transactionnelle - deux nouveaux mécanismes du Blue Gene/Q - n'améliorent pas les performances.Dans un second temps, on présente la réalisation d'une extension appelé Molly au compilateur LLVM, pour optimiser automatiquement le programme, et plus précisément la distribution des données et des calculs entre les nœuds d'un cluster tel que le Blue Gene/Q. Molly représente les tableaux par des polyèdres entiers et utilise l'extension existante Polly qui représente les boucles et les instructions par des polyèdres. Partant de la spécification de la distribution des données et de l'emplacement des calculs, Molly ajoute le code qui gère les flots de données entre les nœuds de calcul. Molly peut aussi permuter l'ordre des données en mémoire. La tâche principale de Molly est d'agréger les données dans des ensembles qui sont envoyés dans le même tampon au même destinataire, pour éviter l'overhead des transferts trop petits. Nous présentons un algorithme qui minimise le nombre de transferts pour des boucles non-paramétrées, basé sur les antichaînes du flot des données. De plus, nous implémentons une heuristique qui tient compte de la manière dont le programmeur a écrit son code. Les primitives de communication asynchrone sont insérées juste après que les données soient disponibles - respectivement juste avant qu'elles soient utilisées. Une bibliothèque runtime implémente ces primitives en utilisant MPI. Molly gère la distribution pour tout code représentable dans le modèle polyédrique, mais fonctionne mieux pour du code à stencil tel LQCD. Compilé avec Molly, le code LQCD atteint 2,5% de la performance crête. L'écart de performance est surtout dû au fait que les autres optimisations ne sont pas faites, par exemple la vectorisation. Les versions futures de Molly pourraient aussi gérer efficacement les codes non à stencil et exploiter les autres optimisations qui ont rendu le code LQCD optimisé à la main si rapide. / Motivated by modern day physics which in addition to experiments also tries to verify and deduce laws of nature by simulating the state-of-the-art physical models using oversized computers, this thesis explores means of accelerating such simulations by improving the simulation programs they run. The primary focus is Lattice Quantum Chromodynamics (QCD), a branch of quantum field theory, running on IBM newest supercomputer, the Blue Gene/Q.In a first approach, the source code of tmLQCD, a Lattice QCD program, is improved to run faster on the Blue Gene machine. Its most performance-relevant operation is a 8-point stencil in 4 dimensional space. Two different optimization strategies are perused: One with the priority of improving spatial and temporal locality, and a second making use of the hardware's data stream prefetcher. On Blue Gene/Q the first strategy reaches up to 20% of the peak theoretical floating point operation performance of that machine. The second strategy with up to 54% of peak is much faster at the cost of using 4 times more memory by storing the data in the order they will be used in the next stencil operation, duplicating data where necessary.Other techniques exploited are direct programming of the messaging hardware (called MUSPI by IBM), a low-overhead work distribution mechanism for threads, explicit data prefetching of data (using dcbt instruction) and manual vectorization (using QPX; width-4 SIMD instructions). Hardware-based list prefetching and transactional memory - both distinct and novel features of the Blue Gene/Q system -- did not improve the program's performance.The second approach is the newly-written LLVM compiler extension called Molly which optimizes the program itself, specifically the distribution of data and work between the nodes of a cluster machine such as Blue Gene/Q. Molly represents arrays using integer polyhedra and uses another already existing compiler extension Polly which represents statements and loops using polyhedra. When Molly knows how data is distributed among the nodes and where statements are executed, it adds code that manages the data flow between the nodes. Molly can also permute the order of data in memory. Molly's main task is to cluster data into sets that are sent to the same target into the same buffer because single transfers involve a massive overhead. We present an algorithm that minimizes the number of transfers for unparametrized loops using anti-chains of data flows. In addition, we implement a heuristic that takes into account how the programmer wrote the code. Asynchronous communication primitives are inserted right after the data is available respectively just before it is used. A runtime library implements these primitives using MPI.Molly manages to distribute any code that is representable by the polyhedral model, but does so best for stencils codes such as Lattice QCD. Compiled using Molly, the Lattice QCD stencil reaches 2.5% of the theoretical peak performance. The performance gap is mostly because all the other optimizations are missing, such as vectorization. Future versions of Molly may also effectively handle non-stencil codes and use make use of all the optimizations that make the manually optimized Lattice QCD stencil so fast.

In this thesis we aim to test if Landau\'s two-component superfluid model is compatible with the quark-gluon-plasma description. We follow the test proposed by Chernodub et. al. [Two-component liquid model for the quark-gluon plasma. Theor. Math. Phys., v. 170, p. 211–216, 2012]. We start by reviewing the building process of a field theory with gauge symmetries and discussing the conservation laws associated to the theory’s symmetries. We explore the thermodynamic approach to quantum theory and the interesting fact that, when combined with a field theory, the path-integral formulation for quantum field theories emerges naturally. We also present the necessity of introducing a momentum cutoff into the theory and show that embedding space-time on a lattice is a way to introduce this cutoff and renormalize the theory. As a bonus, this also allows the numerical and non-perturbative evaluation of observables. We overview the phenomenological aspects of relativistic heavy-ion collisions and Landau’s two-component model for superfluids, along with a quantum-field-theory motivation for it, and explain details of the test proposed by Chernodub et. al.. Lastly, we show the implementation details of our simulation along with results. We do not see evidence that the proposed superfluid model is able to describe the plasma. We speculate that this might be caused by the absence of fermions in our simulations. / Nesta tese nosso objetivo é testar se o modelo de Landau de duas componentes para superfluidos é compatível com a descrição do plasma de quarks e glúons. Seguimos o teste proposto por Chernodub et. al. [Two-component liquid model for the quark-gluon plasma. Theor. Math. Phys., v. 170, p. 211–216, 2012]. Começamos revisando o processo de construção de uma teoria de campo com simetria de gauge e discutindo as leis de conservação associadas às simetrias da teoria. Exploramos a abordagem termodinâmica para teoria quântica e o interessante fato de que, quando combinada com uma teoria de campo, a formulação de integrais de trajetória para teorias quânticas de campo emerge naturalmente. Também apresentamos a necessidade de se introduzir um corte de momento na teoria, e mostramos que embutir o espaço-tempo em uma rede é um meio de introduzir o corte na teoria e renormalizá-la. Como um bônus, isso também permite o cálculo numérico e não-perturbativo de observáveis. Apresentamos um panorama dos aspectos fenomenológicos da colisão de íons pesados relativísticos e o modelo de duas componentes de Landau para superfluidos, bem como uma motivação de teoria quântica de campo para ele, e explicamos detalhes do teste proposto por Chernodub et. al.. Por fim, mostramos os detalhes de nossa implementação juntamente com nossos resultados. Não vemos evidência de que o modelo de superfluidod proposto seja capaz de descrever o plasma. Nós especulamos que isto possa ser causado pela ausência de férmions em nossas simulações.

Nesta tese obtemos, de um ponto de vista matemáticamente rigoroso, a parte inferior do espectro de energia-momento de dois modelos de teorias quânticas de campo com tempo imaginário em redes de dimensão $d+1$ (resultados explícitos para o caso $d = 3$ e matrizes de Dirac) que contém férmions: um modelo puramente fermiônico com interação quártica nos campos fermiônicos de $N$ componentes (modelo de Quatro-Férmions) e um modelo de cromodinâmica quântica. Para o modelo de Quatro-Férmions, $\\kappa$ é o parâmetro de hopping, $M_0$ é a massa bare dos férmions e $\\lambda$ é o parâmetro de interação. Uma expansão de polímeros garante a existência das funções de correlação no limite termodinâmico, na região onde $|\\frac{\\kappa}|$ é pequeno. A análise do espectro é baseada em representações espectrais para funções de correlação de dois e quatro férmions. A análise das funções de correlação adequadas é simplificada pelo uso de simetrias, em particular, de uma {\\em nova} simetria de Reflexão Temporal que aparece no nível das funções de correlação. A determinação do espectro é executada através de um estudo detalhado das taxas de decaimento das funções de correlação. Até próximo ao limiar de três partículas, o espectro de energia e momento exibe curvas de dispersão isoladas que são identificadas com partículas e estados ligados de duas partículas. No subespaço de uma partícula, o espectro consiste em uma curva de dispersão isolada. A massa da partícula é de ordem $-\\ln \\kappa$. O espectro de duas partículas aparece como soluções de uma equação de Bethe-Salpeter, resolvida primeiro em uma aproximação em escada. O espectro de duas partículas contém uma banda de duas partículas livres de largura finita. A existência de estados ligados acima ou abaixo da banda de duas partículas depende do fato do modelo apresentar ou não dominação gaussiana. Um parâmetro $\\aleph$ é dado para medir a dominação gaussiana. Para $\\aleph=0$, nenhum estado ligado ocorre. Para $\\aleph>0$, o estado ligado ocorre abaixo da banda de duas partículas. Para $\\aleph<0$, o estado ligado aparecem acima desta banda. Os resultados obtidos nesta aproximação em escada podem ser estendidos para o modelo completo através de um controle rigoroso das contribuições que diferenciam essas duas situações. Em uma segunda parte, idéias análogas são aplicadas para analisar o espectro do modelo de cromodinâmica quântica. Em particular, nós mostramos a existência dos pentaquarks no regime de acoplamento forte (acoplamento entre as plaquetas $0 <\\beta= \\frac{g^2_0} \\ll \\kappa $). O modelo possui simetria de calibre $SU(3)_c$ e de sabor $SU(2)_f$. Os pentaquark revelados são superposições de estados ligados de mésons e bárions. Apenas estados com um número ímpar de férmions e abaixo do limiar de energia meson-bárion são considerados. O pentaquark é determinado usando uma aproximação em escada para uma equação Bethe-Salpeter. Na ordem dominante em $\\beta$, a massa deste estado é aproximadamente $-5 \\ln\\kappa$ e sua energia de ligação é de ordem $\\textrm(\\kappa^2)$. O estado mais fortemente ligado tem isospin $I=\\frac$. Para $I=\\frac$ não há estados ligados. Estes resultados mostram uma dependência nos spins dos méson e bárion. Esta análise mostra que um potencial de troca de quark-anti-quark de $\\textrm(\\kappa^2)$ é a interação dominante, mas não há uma interpretação de troca de mésons. / In this thesis, we obtain, from a mathematically rigorous point of view, the low-lying energy-momentum spectrum of two $3+1$ dimensional imaginary time lattice quantum filed theory with fermion fields (we give explicit results for the case $d = 3$ and Dirac matrices): a pure fermionic model with quartic interaction in the $N$-component fermion field and a quantum chromodynamics model. For the Four-Fermion model, $\\kappa$ denotes the hopping parameter, $M_0$ the fermion bare mass and $\\lambda$ the interaction parameter. A polymer expansion show the existence of the model correlation functions in the thermodynamic limit, in the region where $|\\frac{\\kappa}|$ is small enough. The analysis of the spectrum is based on spectral representations of two- and four- point correlation functions. The analysis of such adequate correlation functions is simplified by the help of symmetries, in particular, by a {\\em new} Time Reflection symmetry, which appear in the level of correlation functions. The exact determination of the spectrum is done using a detailed study of the decay rates of the correlations. Up to near the 3 particle threshold, the energy-momentum spectrum exhibits isolated dispersion curves that are identified as particles and bound states. In the one-particle subspace, the spectrum consist in just a isolated dispersion curve. The mass of the associated particle is of order $-\\ln \\kappa$. The two-particle spectrum shows up as solutions of a Bethe-Salpeter equation, which is solved first in a ladder approximation. The two-particle spectrum contains a two free particles band of finite width. The existence of bound states above or below the band depends on wherever the model Gaussian domination holds. A parameter $\\aleph$ is given to measure the Gaussian domination. For $\\aleph=0$, no bound state occurs. For $\\aleph>0$, a bound state appears bellow the two-particles band. For $\\aleph<0$, the bound state appears above this band. The result obtained in this ladder approximation can be extended to the full model by a rigorous control of the contributions that differ these two cases. In a second part, analog ideas are applied to analyze the spectrum of a quantum chromodynamics model. In particular, we show the existence of pentaquarks in the strong coupling regime (plaquette coupling $0 <\\beta= \\frac{g^2_0} \\ll \\kappa $). The model has a $SU(3)_c$ gauge symmetry and a $SU(2)_f$ flavor symmetry. The reveled pentaquarks are superpositions of meson-baryon bound states. Only states with an odd number of fermions and bellow the meson-baryon threshold are considered. The pentaquark are determined using a ladder approximation to the Bethe-Salpeter equation. In the dominant order in $\\beta$, the bound state mass is $\\approx -5 \\ln\\kappa$ and the binding energy is of order $\\textrm(\\kappa^2)$. The most strongly bounded bound state has isospin $I=\\frac$. For $I=\\frac$, there is no bound state. These results shows a dependence in the spins of the meson and baryon. This analysis show that a $\\textrm(\\kappa^2)$ quark-anti-quark exchange potential is the dominant interaction, although there is not a meson exchange interpretation.

Em nossas simulações, efetuamos uma cuidadosa análise numérica dos algoritmos de Monte Carlos empregados na termalização de teorias de gauge e sistemas de spins contínuos. Dentre eles, apresentamos uma nova proposta que permite reduzir em cerca de 25% os tempos computacionais. Aplicamos este novo algoritmo ao estudo numérico da transição de desconfinamento da teoria de Yang-Mills (YM) SU(2) tridimensional, a temperatura finita, em redes com volumes de 502 X 4 sítios. Por fim, também utilizamos técnicas de Escala de tamanho finito (FSS), Dinâmica de Tempos Curtos e Métodos Variacionais para extrair os expoentes críticos e espectro de massas de blindagem desta teoria. / In our numerical simulations, we have done a careful analysis of Monte-Carlo algorithms usually applied in the thermalization of gauge theories and continuous-spin systems. Among them we present a new proposal that enables a reduction of roughly 25% in the computational time. We apply this new algorithm to a numerical study of the at finite temperature, using lattices with 502 x 4 sites. We also use Finite-Size-Scaling techniques, Short-Time Dynamics and Variational Methods to extract critical exponents and the spectrum of screening masses of the theory.