Next to the emergence of nearly isolated quantum systems such as ultracold atoms with unprecedented experimental tunability, the conceptualization of the eigenstate thermalization hypothesis (ETH) by Deutsch and Srednicki in the late 20th century has sparked exceptional interest in the mechanism of quantum thermalization. The ETH conjectures that the expectation value of a local observable within the quantum state of an isolated, interacting quantum system converges to the thermal equilibrium value at large times caused by a loss of phase coherence, referred to as dephasing. The thermal behavior within the quantum expectation value is traced back to the level of individual eigenstates, who locally act as a thermal bath to subsystems of the full quantum system and are hence locally indistinguishable to thermal states. The ETH has important implications for the understanding of the foundations of statistical mechanics, the quantum-to-classical transition, and the nature of quantum entanglement. Irrespective of its theoretical success, a rigorous proof has remained elusive so far. $$ \ $$ An alternative approach to explain thermalization of quantum states is given by the concept of typicality. Typicality deals with typical states \(\Psi\) chosen from a subspace of Hilbert space with energy \(E\) and small fluctuations \(\delta\) around it. It assumes that the possible microstates of this subspace of Hilbert space are uniformly distributed random vectors. This is inspired by the microcanonical ensemble in classical statistical mechanics, which assumes equal weights for all accessible microstates with energy \(E\) within an energy allowance \(\delta\). It follows from the ergodic hypothesis, which states that the time spent in each part of phase space is proportional to its volume leading to large time averages being equated to ensemble averages. In typicality, the Hilbert space of quantum mechanics is hence treated as an analogue of classical phase space where statistical and thermodynamic properties can be defined. Since typicality merely shifts assumptions of statistical mechanics to the quantum realm, it does not provide a complete understanding of the emergence of thermalization on a fundamental microscopic level. $$ \ $$ To gain insights on quantum thermalization and derive it from a microscopic approach, we exclusively consider the fundamental laws of quantum mechanics. In the joint work with T. Hofmann, R. Thomale and M. Greiter, on which this thesis reports, we explore the ETH in generic local Hamiltonians in a two-dimensional spin-\(1/2\) lattice with random nearest neighbor spin-spin interactions and random on-site magnetic fields. This isolated quantum system is divided into a small subsystem weakly coupled to the remaining part, which is assumed to be large and which we refer to as bath. Eigenstates of the full quantum system as well as the action of local operators on those can then be decomposed in terms of a product basis of eigenstates of the small subsystem and the bath. Central to our analysis is the fact that the coupling between the subsystem and the bath, represented in terms of the uncoupled product eigenbasis, is given by an energy dependent random band matrix, which is obtained from both analytical and numerical considerations. $$ \ $$ Utilizing the methods of Dyson-Brownian random matrix theory for random band matrices, we analytically show that the overlaps of eigenstates of the full quantum system with the uncoupled product eigenbasis are described by Cauchy-Lorentz distributions close to their respective peaks. The result is supported by an extensive numerical study using exact diagonalization, where the numerical parameters for the overlap curve agree with the theoretical calculation. The information on the decomposition of the eigenstates of the full quantum system enables us to derive the reduced density matrix within the small subsystem given the pure density matrix of a single eigenstate. We show that in the large bath limit the reduced density matrix converges to a thermal density matrix with canonical Boltzmann probabilities determined by renormalized energies of the small subsystem which are shifted from their bare values due the influence of the coupling to the bath. The behavior of the reduced density matrix is confirmed through a finite size scaling analysis of the numerical data. Within our calculation, we make use of the pivotal result, that the density of states of a local random Hamiltonian is given by a Gaussian distribution under very general circumstances. As a consequence of our analysis, the quantum expectation value of any local observable in the subsystem agrees with its thermal expectation value, which proves the validity of the ETH in the equilibrium phase for the considered class of random local Hamiltonians and elevates it from hypothesis to theory. $$ \ $$ Our analysis of quantum thermalization solely relies on the application of quantum mechanics to large systems, locality and the absence of integrability. With the self-averaging property of large random matrices, random matrix theory does not entail a statistical assumption, but is rather applied as a mathematical tool to extract information about the behavior of large quantum systems. The canonical distribution of statistical mechanics is derived without resorting to statistical assumptions such as the concepts of ergodicity or maximal entropy, nor assuming any characteristics of quantum states such as in typicality. In future research, with this microscopic approach it may become possible to exactly pinpoint the origin of failure of quantum thermalization, e.g. in systems that exhibit many body localization or many body quantum scars. The theory further enables the systematic investigation of equilibration, i.e. to study the time scales on which thermalization takes place. / Neben der Entwicklung experimentell zugänglicher nahezu isolierter Quantensysteme wie ultrakalter Gase hat die Formulierung der Eigenstate Thermalization Hypothesis (ETH) durch Deutsch und Srednicki im späten 20. Jahrhundert ein gesteigertes Interesse am Mechanismus der Quantenthermalisierung geweckt. Die ETH postuliert, dass der Erwartungswert einer lokalen Observablen innerhalb des Quantenzustands eines isolierten, wechselwirkenden Quantensystems bei großen Zeiten zum thermischen Gleichgewichtswert konvergiert. Dies vollzieht sich durch den Verlust der Phasenkohärenz im Erwartungswert der lokalen Observable, was als Dephasing bekannt ist. Das thermische Verhalten innerhalb des Quantenerwartungswerts wird auf die Ebene einzelner Eigenzustände zurückgeführt, die lokal als thermisches Bad für Untersysteme des gesamten Quantensystems wirken und daher lokal nicht von thermischen Zuständen unterscheidbar sind. Die ETH hat wichtige Auswirkungen auf das Verständnis der Grundlagen der statistischen Mechanik, des Übergangs von der Quanten- zur klassischen Physik und der Natur der Quantenverschränkung. Ungeachtet ihres theoretischen Erfolges ist ein rigoroser Beweis der Hypothese bisher nicht erfolgt. $$ \ $$ Ein alternativer Ansatz zur Erklärung der Thermalisierung von Quantenzuständen ist das Konzept der typicality. Typicality befasst sich mit typischen Zuständen \(\Psi\), die aus einem Unterraum des Hilbertraums mit Energie \(E\) und kleinen Fluktuationen \(\delta\) ausgewählt werden. Dabei wird angenommen, dass die möglichen Mikrozustände dieses Unterraums des Hilbertraums gleichmäßig verteilte Zufallsvektoren sind. Dies ist ein aus dem klassischen mikrokanonischen Ensemble übertragener Ansatz, der von einer Gleichgewichtung aller Mikrozustände mit der Energie \(E\) in einem Energiebereich \(\delta\) ausgeht. Das geht auf die ergodische Hypothese zurück, die besagt, dass die verbrachte Zeit in jedem Teil des klassischen Phasenraums proportional zu dessen Volumen ist. Dies führt schlussendlich zu einer Gleichsetzung der Mittelwerte bei großen Zeiten mit Ensemblemittelwerten. Der Hilbertraum in der Quantenmechanik wird mit typicality daher als Analogon des klassischen Phasenraums behandelt, in dem statistische und thermodynamische Eigenschaften definiert werden können. Da typicality lediglich Annahmen der statistischen Mechanik auf den Quantenbereich überträgt, kann sie kein vollständiges mikroskopisches Bild der Entstehung von Thermalisierung liefern. $$ \ $$ Um Erkenntnisse über die Quantenthermalisierung zu gewinnen und sie aus einem mikroskopischen Ansatz abzuleiten, stützen wir uns ausschließlich auf die grundlegenden Gesetze der Quantenmechanik. In der gemeinsamen Arbeit mit T. Hofmann, R. Thomale und M. Greiter, von der diese Arbeit berichtet, untersuchen wir die ETH in generischen lokalen Hamiltonians in einem zweidimensionalen Spin-\(1/2\)-Gitter mit zufälligen Spin-Spin-Wechselwirkungen zwischen nächsten Nachbarn und zufälligen lokalen Magnetfeldern. Dieses isolierte Quantensystem wird in ein kleines Untersystem aufgeteilt, das schwach an den verbleibenden Teil gekoppelt ist, der als groß angenommen und als Bad bezeichnet wird. Die Eigenzustände des gesamten Quantensystems sowie die Wirkung lokaler Operatoren auf diese können dann in Form einer Produktbasis von Eigenzuständen des kleinen Untersystems und des Bades zerlegt werden. Von zentraler Bedeutung für unsere Analyse ist die Tatsache, dass die Kopplung zwischen dem Untersystem und dem Bad, die in Form der ungekoppelten Produkteigenbasis dargestellt wird, durch eine energieabhängige Zufallsbandmatrix gegeben ist, welche sowohl aus analytischen als auch numerischen Überlegungen gewonnen wird. $$ \ $$ Unter Verwendung der Methoden der mathematischen Theorie für zufällige Bandmatrizen finden wir analytisch heraus, dass der Überlapp von Quanteneigenzuständen mit der ungekoppelten Produkteigenbasis durch Cauchy-Lorentzverteilungen in den Badenergien in der Nähe ihrer jeweiligen Peaks beschrieben werden. Das Ergebnis wird durch eine umfangreiche numerische Studie mit exakter Diagonalisierung bestätigt, bei der die numerischen Parameter für die Überlapps mit der theoretischen Berechnung übereinstimmen. Die Information über die Form der Quanteneigenzustände ermöglicht es uns, die reduzierte Dichtematrix in dem kleinen Untersystem aus der reinen Dichtematrix eines einzelnen Eigenzustandes des isolierten Quantensystems abzuleiten. Wir zeigen, dass sie im Limes großer Bäder zu einer thermischen Dichtematrix mit kanonischen Boltzmann-Gewichten auf der Diagonalen konvergiert. Dies wird mithilfe einer numerischen Skalierungsanalyse für endliche Systeme bestätigt. In unseren Berechnungen verwenden wir das zentrale Ergebnis, dass die Zustandsdichte eines lokalen zufälligen Hamiltonians unter allgemeinen Bedingungen durch eine Gauß-Verteilung gegeben ist. Aus unserer Analyse folgt, dass der Quantenerwartungswert jeder lokalen Observablen in dem Untersystem mit ihrem thermischen Erwartungswert übereinstimmt, was die Gültigkeit der ETH in der Gleichgewichtsphase für die betrachtete Klasse von Hamiltonians beweist. $$ \ $$ Unsere Analyse der Quantenthermalisierung beruht ausschließlich auf der Anwendung der Quantenmechanik auf große Systeme, der Lokalität und der fehlenden Integrabilität. Stützend auf der mathematischen Eigenschaft des Self-averaging von großen Zufallsmatrizen impliziert die Zufallsmatrixtheorie keine statistische Annahme, sondern wird vielmehr als mathematisches Instrument eingesetzt, um Informationen über das Verhalten großer Quantensysteme zu extrahieren. Die kanonische Verteilung der statistischen Mechanik wird abgeleitet, ohne auf die Konzepte der Ergodizität oder der maximalen Entropie zurückzugreifen und ohne irgendwelche Eigenschaften von Quantenzuständen anzunehmen wie es etwa bei typicality der Fall ist. Mit diesem mikroskopischen Ansatz könnte es zudem in zukünftiger Forschung möglich werden, den Ursprung des Nichterfüllens der Quantenthermalisierung, z.B. in Systemen mit Vielteilchenlokalisierung oder Quanten-Scar-Zuständen, exakt zu bestimmen. Die Theorie könnte außerdem eine systematische Untersuchung der Equilibrierung ermöglichen, d.h. die Bestimmung der Zeitskalen, auf denen Thermalisierung stattfindet.

The question whether and how closed quantum systems equilibrate is still debated today. In this thesis a generic spin system is analysed and criteria to classify unique equilibration dynamics are developed. Furthermore, the eigenstate thermalization hypothesis is investigated as a possible cause for the unique equilibrium. For both problems novel numerical methods for solving the time-dependent Schroedinger equation based on series expansions and typicality are developed. Furthermore, the problem of markovian dynamics on the level of single measurements is discussed.

Work is one of the cornerstones of classical thermodynamics. However, a direct transfer of this concept to quantum systems has proved problematic, especially for non-equilibrium processes. Unlike in the classical case, quantum work cannot be defined unambiguously. Depending on the specific setting and the imposed assumptions, different definitions are well motivated. In particular, in quantum thermodynamics, a clear distinction must be made between the measurement, storage, and use of work, since these three facets of the concept are not necessarily compatible with each other. The present thesis is mainly concerned with the measurement aspect. With the help of illustrative scenarios several approaches to quantum work measurements, their advantages and drawbacks are discussed. The focus will be on the question to what extent quantumness plays a decisive role in such scenarios, both in a qualitative and quantitative sense. Based on the gedankenexperiment of a Szilárd machine a criterion is proposed which can be used to verify genuine quantum correlations between the work medium in a heat engine and its thermal environment. In a Szilárd scenario a Maxwell's demon determines the state of the work medium and uses this information to extract work. We split this model into a bipartite setting. The demon only has access to the environment and, thus, can only indirectly measure the state of the work medium. By sharing the acquired information with another agent, the latter can extract work. The question of the quantumness of the experiment can then be reduced to the question of the maximum attainable work in the context of a suitable quantum steering scenario. For the constructed setting a bound for the work output achievable for classical correlations between the engine and the environment is derived. Work extraction beyond this classical limit thus proves the quantum nature of the machine. The verification of non-classical correlations by means of quantum steering is motivated by the fact that such a scenario reflects the typical asymmetry of a thermodynamic setup. While the machine itself is considered to be controllable and characterized in detail, no requirements are imposed on the correlated environment and the measurements performed on it. Consequently, this verification of a truly quantum heat engine is semi-device-independent. In a second scenario, the compatibility of average work and work fluctuations in a driven system is discussed. Fluctuation theorems play an important role in classical non-equilibrium thermodynamics. The best-known example is the Jarzynski equality. This equation establishes a connection between the free energy difference of two equilibrium states and the fluctuating work measured in a non-equilibrium process. A transfer of the Jarzynski equality to quantum systems succeeds most simply if the work definition is based on a so-called two-point measurement scheme. This approach determines the work as the difference of two projective energy measurements. The disadvantage of this definition is the unavoidable disturbance of the quantum state by the measurement, which makes a determination of the correct average work impossible. By means of a generalized two-point measurement scheme, it is shown how this contradiction between fluctuating and average work can be overcome. The approach is based on the concept of joint measurability. Unsharp measurements with a smaller disturbance of the quantum state can be measured jointly and allow for the determination of the correct average work. Nevertheless, the connection between measured fluctuations and the change of free energy can be preserved by means of a modified Jarzynski equality, as elucidated in this thesis. Even though the two-point measurement scheme - both in its projective form and in the generalized variant presented in this thesis - satisfies a Jarzynski equality, the operationality and the associated experimental significance are to be assessed differently than in the classical case. In classical thermodynamics, the Jarzynski relation can be used practically to determine, for example, the change of free energy in RNA molecules. However, it is crucial for such an experiment that the non-equilibrium work can actually be measured without requiring detailed knowledge of the system under consideration. In contrast, the two-point measurement scheme defines work as the energy difference of the system between the beginning and the end of the process. Crucially, for the measurement of these energies the Hamiltonians have to be known and the free energy difference could therefore be calculated directly from this knowledge without reference to the Jarzynski equality. Thus, the operationality of the quantum Jarzynski relation differs fundamentally from its classical counterpart. In this thesis we develop a measurement scheme which, in principle, allows us to employ a quantum version of the Jarzynski equation without knowledge of the Hamiltonians. The crucial point is to include the apparatus that drives the system out of equilibrium in the quantum picture and to define the work measurement on that very apparatus. Such a work measurement can only be meaningfully defined as a quantum expectation value and work fluctuations cannot directly be measured, in contrast to the classical case. The work along a classical microstate trajectory can be determined in a single run. The trajectory itself does not need to be known for this purpose; its existence is sufficient. Quantum trajectories do not exist unless they are objectified by a measurement. It is shown how measurements on the environment of the system can provide information about the trajectories. A conditioning of the measured work on these trajectories then allows for the determination of work fluctuations in the quantum system. For these fluctuations an inequality is conjectured whose limit is given by the classical Jarzynski equation. Numerical results support the conjecture. A proof is still missing. By means of the presented framework, the free energy difference of a quantum system can, in principle, be determined without knowledge of the underlying Hamiltonian. However, as is shown, this requires an optimization over several external parameters, since the inequality in general provides only an upper bound. Thus, the operationality of the model enforces a quantum disadvantage. The methods presented in this thesis can be applied to various scenarios in quantum thermodynamics. Especially the framework for work measurements on an external apparatus offers an alternative to common approaches when the system under investigation and especially its Hamiltonian is not known in advance. The focus on operationality will help to better understand to what extend the work quantities defined and measured in quantum thermodynamic systems differ from the classical concept of work.

Wir untersuchen die Dynamik von offenen Quantensystemen sowohl im Gleichgewicht als auch im Nichtgleichgewicht. Unser Fokus liegt dabei auf der quantenoptischen Dispersionswechselwirkung zwischen einem mikroskopischen Teilchen und einer komplexen elektromagnetischen Umgebung. Wir sind der Meinung, dass Langzeitkorrelationen in dem System essenziell zum Verständnis der Dynamik des Teilchens beitragen können. Unter Langzeitkorrelationen verstehen wir die Beiträge zur Autokorrelationsfunktion von Quantenoperatoren, die als ein inverses Potenzgesetz in der Verzögerungszeit skalieren. Das Einbeziehen von Langzeitkorrelationen in unser theoretisches Modell sichert die Selbstkonsistenz unserer Beschreibung und ermöglicht es uns, die Rückkopplung der Umgebung auf das Teilchen vollständig zu berücksichtigen. Darüber hinaus erlaubt es uns die Vorhersage von bisher übersehenen Effekten und Mechanismen, die das Verhalten von Dispersionskräften im Gleichgewicht und Nichtgleichgewicht bestimmen. Unsere Beispiele reichen von der Wechselwirkungsentropie des magnetischen Casimir-Polder-Effekts, über den Einfluss von Materialeigenschaften und geometrischen Überlegungen auf experimentelle Aufbauten, bis hin zur Thermodynamik von Quantenreibung. Wir geben den Leser_innen außerdem eine Orientierungshilfe, wann und wie Langzeitkorrelationen in theoretische Modellbildungen einbezogen werden müssten und welche Auswirkungen im Zusammenhang mit quantenoptischen Dispersionskräften zu erwarten sind. / We explore the dynamics of open quantum systems in both equilibrium and nonequilibrium situations. Our focus lies on the quantum-optical dispersion interaction between a microscopic particle and a complex electromagnetic environment. We argue that long-time correlations in the system can be essential for understanding the dynamics of the particle. We define long-time correlations as those contributions to the autocorrelation function of quantum operators which scale as an inverse power law in the time delay. Incorporating long-time correlations into our theoretical model safeguards the self-consistency of our description and allows us to consider the full back-action of the environment on the particle. Moreover, it leads us to the prediction of previously overlooked effects and mechanisms determining dispersion forces in equilibrium and nonequilibrium. Our examples range from the interaction entropy of the magnetic Casimir-Polder effect, over the impact of material properties and geometric considerations for experimental setups, all the way down to the thermodynamics of quantum friction. We further provide the reader with a guideline when and how to include long-time correlations into theoretical models and what effects can be expected to emerge in the context of quantum-optical dispersion forces.