Tensor network algorithms (TNAs) represent one of the most recent developments in the field of numerical methods for the simulation of strongly-correlated manybody systems.Arising as a natural consequence of an improved understanding of the entanglement structures intrinsic to the ground-state manifolds of manybody Hilbert spaces, they correspond to algorithms exploiting entropic constraints expected to arise in various classes of many-body ground states.In this work we apply and develop TNAs for the simulation of various stronglycorrelated lattice models.Concretely, we apply tensor network renormalization (TNR) to the study of the classical Blume-Capel model (BCM).We propose to exploit the RG features specific to TNR to obtain an indicator for the vicinity of (multi-)critical points.We show that in the case of the BCM it leads to a location of its tricritical point matching the accuracy of state-of-the-art Monte Carlo approaches.This allows us to characterize the underlying c = 7/10 conformal field theory with an excellent accuracy.We then present a self-contained introduction to the most widely used techniques for the simulation of one-and two-dimensional quantum systems, where we cover matrix product states (MPS) and projected entangled-pair states (PEPS) in detail.We briefly discuss the multi-scale entanglement renormalization Ansatz (MERA).We apply infinite PEPS (iPEPS) to the simulation of the Kitaev-Heisenberg (KH) model, proposed as an effective low-energy theory for the so-called Iridate compounds of the form A 2 IrO 3 (A = Na, Li).We show the ability of iPEPS to accurately encode the complex ground-state physics of Kitaev's honeycomb model.When considering the KH model we confirm the existence of all previously found phases, locate all phase transitions in the phase diagram, finding good agreement with previous studies, and provide estimates for the survival regions of the spinliquid phases in the thermodynamic limit.We briefly discuss the nature of these transitions.We conclude this work with a study of various formulations of iPEPS on cylindrical geometries.We benchmark the proposed formulations by studying the transversefield Ising model and find good performance for a subset of the formulations studv ied.We then carry out a comparison between iPEPS and iMPS methods for the Heisenberg and Hubbard models and find a range of cylinder widths over which both methods exhibit comparable performance.We find evidence for the potential of iPEPS simulations on cylinders and argue that our findings provide support for future studies employing both MPS and PEPS methods in conjunction.vi Wir schlieSSen diese Arbeit mit einer Untersuchung verschiedener Formulierungen von iPEPS für zylindrische Geometrien ab.Wir bewerten die vorgeschlagenen Formulierungen durch die Untersuchung des Ising-Modells im transversalen Feld und finden gute Ergebnisse für eine Teilmenge der untersuchten Formulierungen.Wir führen dann einen Vergleich zwischen iPEPS und iMPS Methoden für die Heisenberg und Hubbard Modelle durch und finden einen Bereich von Zylinderbreiten, in dem beide Methoden vergleichbare Leistungen bringen.Wir finden Beweise für das Potenzial von iPEPS-Simulationen auf Zylinder und argumentieren, dass unsere Erkenntnisse Unterstützung für zukünftige Studien liefern, die sowohl MPSals auch PEPS-Methoden zusammen verwenden.viii Despite the considerable progress made towards the understanding of stronglycorrelated systems, much work remains to be done and, as is perhaps to be expected, the broad nature of the problems encountered means that no single approach is likely to be applicable in every circumstance.From a numerical perspective a wide variety of techniques have been developed over time.Of these, the most accurate is exact diagonalization (ED) which, as its name indicates, attempts to partially diagonalize the Hamiltonian operator.However, owing to the exponential growth of Hilbert space dimension as the system size is increased, this method rapidly becomes inapplicable.Indeed, even when targeting
Tensor network algorithms (TNAs) represent one of the most recent developments in the field of numerical methods for the simulation of strongly-correlated manybody systems.Arising as a natural consequence of an improved understanding of the entanglement structures intrinsic to the ground-state manifolds of manybody Hilbert spaces, they correspond to algorithms exploiting entropic constraints expected to arise in various classes of many-body ground states.In this work we apply and develop TNAs for the simulation of various stronglycorrelated lattice models.Concretely, we apply tensor network renormalization (TNR) to the study of the classical Blume-Capel model (BCM).We propose to exploit the RG features specific to TNR to obtain an indicator for the vicinity of (multi-)critical points.We show that in the case of the BCM it leads to a location of its tricritical point matching the accuracy of state-of-the-art Monte Carlo approaches.This allows us to characterize the underlying c = 7/10 conformal field theory with an excellent accuracy.We then present a self-contained introduction to the most widely used techniques for the simulation of one-and two-dimensional quantum systems, where we cover matrix product states (MPS) and projected entangled-pair states (PEPS) in detail.We briefly discuss the multi-scale entanglement renormalization Ansatz (MERA).We apply infinite PEPS (iPEPS) to the simulation of the Kitaev-Heisenberg (KH) model, proposed as an effective low-energy theory for the so-called Iridate compounds of the form A 2 IrO 3 (A = Na, Li).We show the ability of iPEPS to accurately encode the complex ground-state physics of Kitaev's honeycomb model.When considering the KH model we confirm the existence of all previously found phases, locate all phase transitions in the phase diagram, finding good agreement with previous studies, and provide estimates for the survival regions of the spinliquid phases in the thermodynamic limit.We briefly discuss the nature of these transitions.We conclude this work with a study of various formulations of iPEPS on cylindrical geometries.We benchmark the proposed formulations by studying the transversefield Ising model and find good performance for a subset of the formulations studv ied.We then carry out a comparison between iPEPS and iMPS methods for the Heisenberg and Hubbard models and find a range of cylinder widths over which both methods exhibit comparable performance.We find evidence for the potential of iPEPS simulations on cylinders and argue that our findings provide support for future studies employing both MPS and PEPS methods in conjunction.vi Wir schlieSSen diese Arbeit mit einer Untersuchung verschiedener Formulierungen von iPEPS für zylindrische Geometrien ab.Wir bewerten die vorgeschlagenen Formulierungen durch die Untersuchung des Ising-Modells im transversalen Feld und finden gute Ergebnisse für eine Teilmenge der untersuchten Formulierungen.Wir führen dann einen Vergleich zwischen iPEPS und iMPS Methoden für die Heisenberg und Hubbard Modelle durch und finden einen Bereich von Zylinderbreiten, in dem beide Methoden vergleichbare Leistungen bringen.Wir finden Beweise für das Potenzial von iPEPS-Simulationen auf Zylinder und argumentieren, dass unsere Erkenntnisse Unterstützung für zukünftige Studien liefern, die sowohl MPSals auch PEPS-Methoden zusammen verwenden.viii Despite the considerable progress made towards the understanding of stronglycorrelated systems, much work remains to be done and, as is perhaps to be expected, the broad nature of the problems encountered means that no single approach is likely to be applicable in every circumstance.From a numerical perspective a wide variety of techniques have been developed over time.Of these, the most accurate is exact diagonalization (ED) which, as its name indicates, attempts to partially diagonalize the Hamiltonian operator.However, owing to the exponential growth of Hilbert space dimension as the system size is increased, this method rapidly becomes inapplicable.Indeed, even when targeting
Strongly correlated quantum systems are one of the most interesting and challenging phenomena in modern condensed matter physics. Unfortunately, the theoretical access even to their low energy properties is limited by the exponential growth of the Hilbert space in a many-body setting. To mitigate this computational barrier, tensor network methods are one of the most competitive approaches to classically simulate complex quantum systems. In this thesis we will discuss recent developments in the field of tensor networks to efficiently and accurately simulate two-dimensional quantum many-body states in the thermodynamic limit. Hereby, we introduce the variational update globally optimizing the wave-function, encoded as tensor network in the form of an infinite projected entangled pair state (iPEPS), in order to obtain ground states of many-body Hamiltonian models. We focus in the discussion on the well-established method to approximately contract this two-dimensional structure, the corner transfer matrix renormalization group (CTMRG), and discuss its native application on square lattice and triangular lattice based models. Furthermore, we report the development of a refined numerical algorithm, the split-CTMRG, to significantly reduce its computational complexity, pushing the limits of feasible simulations within modern hardware capabilities. Finally, the method advancements are applied to a currently widely studied model of frustrated magnetism, the Heisenberg maple-leaf antiferromagnet. We are able to find a large phase of vanishing magnetization in contradiction to previous works and, to the best of our knowledge, the lowest reported variational ground state energy at the isotropic point of the model. As an outlook, we describe a potential adaption of the previously developed methods to variationally simulate two-dimensional many-body thermal states.
The two-dimensional (2D) Hubbard model, widely believed to capture the essential physics of high-T_c cuprate compounds, has attracted immense research interest ever since the discovery of superconductivity in these materials. Despite its superficial simplicity, the Hubbard model incubates abundant fascinating phenomena owing to the strong interactions between charge carriers, encompassing antiferromagnetism, the pseudogap, and unconventional superconductivity. Yet these very interactions induce strong correlations, rendering canonical perturbative approaches unreliable. With advances in modern computational hardware, numerical techniques — notably tensor network methods and Artificial Intelligence (AI) algorithms — have become indispensable to our expedition. Tensor networks encode the amplitude information of many-body quantum states in a network of interconnected tensors. Leveraging the entanglement area law, tensor networks provide efficient, systematic, and controllable representations of quantum states that would otherwise require an exponentially large number of parameters. Past decades have witnessed the development of various tensor network ansatzes, such as Matrix Product States/Operators (MPS/MPO) for one-dimensional systems and Projected Entangled-Pair States (PEPS) for two-dimensional systems, together with diverse algorithms for optimizing these ansatzes for both zero- and finite temperature scenarios. Over the past years, AI technologies have revolutionized the way we analyze scientific data. The transformer architecture, in particular, has demonstrated exceptional capabilities in the domain of natural language processing. Its attention mechanism excels at capturing long-range correlations in sequential data, making it a promising tool for studying strongly correlated many-body systems. The interpretable nature of the attention mechanism further offers valuable insights into the AI's perception of the underlying physics. In this thesis, we assemble an all-inclusive numerical toolchain for analyzing strongly correlated lattice systems. As a demonstration, we investigate the zero- and finite-temperature properties of the 2D Hubbard model on a square lattice via the infinite PEPS (iPEPS) and the eXponential Tensor Renormalization Group (XTRG), respectively. The native two-dimensional iPEPS tensor network faithfully embodies the entanglement area law and helps settle long-standing debates regarding superconducting order in the ground state of the Hubbard model. The XTRG algorithm efficiently constructs thermal density matrices across a broad temperature range via a highly effective exponential-cooling protocol, thereby enabling detailed probes of pairing and pseudogap phenomena. Afterwards, we generate a comprehensive snapshot dataset which furnishes subsequent training of an optimized encoder-only transformer model. Our novel AI architecture features a Markovian interpretation of the attention design and affords improved parallelism. The collective technological stack opens a brand new avenue for exploring the rich physics of strongly correlated many-body systems.
Quantum many-body systems far out of equilibrium can lead to rich physics and are challenging to understand due to the exponentially large Hilbert space.We employ large scale tensor network techniques in numerical simulations and investigate the dynamics of different model systems, where we focus on the dimensional crossover in two-dimensional systems and on many-body localization.Moreover, we introduce a new method to efficiently simulate the dynamics of mixed states using minimally entangled purified states.The developed codes are made available as an open-source library.v Kurzfassung Quantenvielteilchensysteme im Nichtgleichgewicht bieten reichhaltige Physik, der exponentiell große Hilbertraum stellt jedoch eine Herausforderung für deren Verständnis dar.Wir verwenden umfangreiche numerische Simulationen mit Tensor-Netzwerk-Methoden und erforschen die Dynamik in verschiedenen Modellsystemen, wobei wir uns auf auf den dimensionalen Übergang in zwei-dimensionalen Systemen und auf Vielteilchenlokalisierung konzentrieren.Außerdem präsentieren wir eine neue Methode, um effizient die Dynamik von gemischten Zuständen mit minimal verschränkten reduzierten Zuständen zu simulieren.
The Tensor Network (TN) algorithm for strongly related electronic quantum lattice systems has become the uppermost algorithm for studying strongly related electronic quantum lattice systems in numerical computation. Focusing on the TN algorithm based on 2-D spin systems, this paper discusses the theoretical principles of the Projected Entangled Pair States (PEPS) algorithm and its implementation process. It also explores the infinite PEPS (iPEPS) algorithm for simulating the spin-1/2 infinite 2-D square lattice quantum XYX model. The critical phenomena of the two-dimensional XYX quantum system are analyzed from the perspectives of ground-state fidelity and order parameters, and the quantum critical point is identified. It is anticipated that with the further development of TN algorithms, the single-site fidelity theory of quantum phase transitions (QPTs) can be applied to various quantum many-body systems in condensed matter physics, thereby deepening our understanding of the extremely rich states of matter in the quantum world. Research indicates that quantum information theory provides a universal method for the study of QPTs, which can identify several ground-state fidelities and order parameters through the ground-state wave function of the system, several ground state fidelity and local order parameters can be identified to fully describe some quantum phase transition points, greatly enriching the traditional Landau theory of symmetry breaking QPTs.
The efficient simulation of a quantum many-body system is often incredibly difficult due to the exponential scaling of the Hilbert space with increasing number of interacting particles. In practice, it is only possible to exactly simulate small clusters. In many cases, tensor-network-based methods are capable of overcoming this challenge, providing reliable simulations of much larger or even infinite-sized systems through a low-rank factorization of the state coefficients. Tensor operations are typically highly parallelizable, and therefore capable of utilizing the high performance of modern general-purpose graphics processing units (GPU). A particular tensor-network ansatz called infinite projected-entangled pair states (iPEPS) may be used to simulate infinite two-dimensional lattice systems, and are the primary subject of this thesis. As part of this work, a GPU-based open-source software library called Ace-TN was developed to accelerate iPEPS calculations. Benchmarks are presented showing efficient single- and muli-gpu execution of the core iPEPS simulation techniques. iPEPS simulations are performed for quantum spin systems with bond directional interactions, referred to as quantum compass models (QCM) or Kitaev models depending on pre-established naming conventions. The QCM defined on square and kagome lattices are considered with the addition of an external magnetic field. Each system possesses one-dimensional gauge-like symmetries in zero field, relating subextensively-many ground states that are capable of being represented by iPEPS. These states are shown to be metastable in non-zero field through adiabatic evolution from the zero-field ground state, each separated by a small energy gap. Several field-dependent phases and phase transitions are found in each system. Prominent nematic phases are found in both the square- and kagome-lattice systems. The square-lattice system contains a field value that admits an exact solution of the ground state which is characteristic of the surrounding phase. The inclusion of isotropic Heisenberg coupling to the kagome-lattice system is also explored, which produces a frustrated parameter regime between two dual quantum spin-liquid phases.
We present a general graph-based projected entangled-pair state (gPEPS) algorithm to approximate ground states of nearest-neighbor local Hamiltonians on any lattice or graph of infinite size. By introducing the structural matrix, which codifies the details of tensor networks on any graphs in any dimension d, we are able to produce a code that can be essentially launched to simulate any lattice. We further introduce an optimized algorithm to compute simple tensor updates as well as expectation values and correlators with a mean-field-like effective environments. Though not being variational, this strategy allows to cope with PEPS of very large bond dimension (e.g., D=100) and produces remarkably accurate results in the thermodynamic limit in many situations, and specially when the correlation length is small and the connectivity of the lattice is large. We prove the validity of our approach by benchmarking the algorithm against known results for several models, i.e., the antiferromagnetic Heisenberg model on a chain, star and cubic lattices, the hardcore Bose-Hubbard model on square lattice, the ferromagnetic Heisenberg model in a field on the pyrochlore lattice, as well as the three-state quantum Potts model in field on the kagome lattice and the spin-1 bilinear-biquadratic Heisenberg model on the triangular lattice. We further demonstrate the performance of gPEPS by studying the quantum phase transition of the 2d quantum Ising model in transverse magnetic field on the square lattice, and the phase diagram of the Kitaev-Heisenberg model on the hyperhoneycomb lattice. Our results are in excellent agreement with previous studies.