Misguich and Colleagues Introduce Matrix-Product States for Quantum Many-Body Physics

Grégoire Misguich and colleagues at Université Paris-Saclay are advancing quantum many-body physics through the investigation of tensor-network methods, with a focus on matrix-product states. Their work details a foundational understanding of this language, encompassing graphical notation and entanglement control for efficient representation. It sharply details key algorithms such as DMRG and time-evolution methods, extending to higher-dimensional systems with projected entangled-pair states and offering insights into mixed states, quantum channels, and open quantum systems. Practical Julia code examples enable application of this thorough set of tools.

Representing many-body quantum systems with compressed tensor networks

Matrix-product states (MPS) form the cornerstone of this approach, acting as a simplified way to describe the behaviour of many interacting quantum particles, much like approximating a complex wave with a combination of simpler waves. A tensor network represents a quantum state not as a single, enormous set of numbers, but as a network of smaller mathematical objects called tensors, linked together in a specific arrangement; it is visualised as a highly organised spreadsheet where relationships between different parts of the system are clearly shown. The power of tensor networks lies in their ability to efficiently represent states that are only weakly entangled, effectively compressing the information needed to describe the system. Traditional methods for simulating quantum systems suffer from the ‘curse of dimensionality’, where the computational resources required grow exponentially with the number of particles. Tensor networks circumvent this by exploiting the limited entanglement present in many physical systems, reducing the computational cost significantly. Université Paris-Saclay utilises this method to efficiently represent the behaviour of many interacting quantum particles, focusing on algorithms like DMRG and time-evolution, chosen for their ability to handle entanglement and to overcome limitations of traditional computational methods. These notes, prepared for the Les Houches Summer School in June 2026, are accompanied by Julia code examples utilising ITensor libraries, facilitating both research and pedagogical applications. The Julia implementation leverages the ITensor framework, a widely used library for performing tensor network calculations, providing optimised routines for common operations like tensor contraction and singular value decomposition. Understanding the underlying mathematical principles, including linear algebra and concepts from quantum information theory, is crucial for effectively utilising these techniques.

Extending quantum simulation timescales via unified tensor network development

Advanced tensor network techniques have achieved a six-fold increase in the maximum simulated time-step for quantum many-body systems, extending from approximately 0.1 to 0.6. This threshold crossing enables the study of dynamics over timescales inaccessible to earlier methods, important for modelling non-equilibrium phenomena and thermalisation processes. Prior computational constraints restricted investigations to short-timescale behaviour, limiting the ability to observe long-term dynamics or the emergence of steady states. This new framework overcomes those limitations by employing optimised algorithms and efficient data structures. The improvement in timescale accessibility is particularly significant for studying systems driven far from equilibrium, such as those subjected to strong external fields or rapid changes in parameters. The team at Université Paris-Saclay detail a unified framework encompassing matrix-product states and projected entangled-pair states, alongside representations of mixed states, offering a thorough set of tools for quantum simulations. This unification allows researchers to seamlessly switch between different tensor network representations depending on the specific problem and dimensionality of the system. The representation of mixed states is crucial for modelling systems at finite temperature or those subject to decoherence, while the inclusion of quantum channels allows for the simulation of open quantum systems interacting with their environment. The framework also incorporates techniques for handling bond dimensions, which control the accuracy of the MPS representation and the computational cost of the simulation.

Université Paris-Saclay develops tensor-network methods in quantum many-body physics, with a focus on matrix-product states (MPS). The notes establish the fundamental language of tensor networks, including graphical notation, bond dimensions, and the role of entanglement in efficient representation. Core MPS algorithms, such as contractions, correlation functions, and the density-matrix renormalization group (DMRG), are introduced alongside time-evolution methods. DMRG, in particular, is a variational algorithm that iteratively optimises the MPS representation to minimise the energy of the system, providing highly accurate ground state estimates. The material briefly discusses projected entangled-pair states (PEPS) as a higher-dimensional extension of MPS, with ideas for approximate contraction. PEPS offer a way to represent quantum states in two or more spatial dimensions, but their computational cost is significantly higher than MPS, requiring sophisticated contraction algorithms. Representations of mixed states, quantum channels, and Lindblad dynamics are presented, with applications to thermal states and open quantum systems; accompanying Julia code examples, based on the ITensor and ITensorMPS libraries, support both research and educational use. Lindblad dynamics provides a framework for describing the evolution of open quantum systems, accounting for the effects of dissipation and decoherence.

Tensor networks and matrix-product states for modelling entanglement and open quantum systems

Université Paris-Saclay have created a detailed resource for tackling quantum many-body problems, employing tensor networks and matrix-product states to represent complex systems. The framework elegantly handles entanglement, a uniquely quantum connection between particles, and extends to modelling systems interacting with their environment. Entanglement is a key resource in quantum information processing and plays a crucial role in the behaviour of many condensed matter systems. The ability to accurately represent and manipulate entanglement is therefore essential for simulating these systems. However, the notes acknowledge a limited exploration of projected entangled-pair states, which raises a tension because effectively simulating higher-dimensional quantum systems often demands these more complex techniques, potentially restricting the immediate applicability of the framework to certain physical scenarios. While MPS is well-suited for one-dimensional systems, extending to higher dimensions requires more sophisticated tensor network structures like PEPS or multiscale entanglement renormalization ansatz (MERA). Further research is needed to develop efficient algorithms for contracting these higher-dimensional tensor networks.

These methods are supported for wider adoption when investigating thermal states, open quantum systems, and dynamics beyond timescales previously accessible by providing accompanying Julia code examples. This work moves beyond previous fragmented approaches by detailing the underlying language of tensor networks, including concepts like bond dimensions and entanglement, and the algorithms needed for practical computation. The team’s unified approach offers a thorough set of tools for quantum simulations, and the accompanying resources enable wider adoption of these methods within the research community. The availability of well-documented and optimised code examples is crucial for accelerating the development of new applications and fostering collaboration among researchers. The emphasis on pedagogical clarity makes these notes a valuable resource for students and researchers entering the field of tensor network methods, providing a solid foundation for further exploration and innovation.

These notes successfully detail tensor-network methods, particularly matrix-product states, offering a unified language and algorithms for quantum simulations. This is important because accurately representing entanglement is essential for modelling complex quantum systems and their interactions. The authors provide accompanying Julia code examples to support wider adoption of these techniques for investigating thermal states and open quantum systems. Further research is needed to improve algorithms for higher-dimensional systems, as current methods are best suited for one-dimensional scenarios.