| Tensor network | |
|---|---|
| Name | Tensor network |
| Caption | Graphical depiction of a tensor network |
| Field | Quantum physics; Condensed matter physics |
| Introduced | 1990s |
| Notable use | Density matrix renormalization group, quantum simulation |
Tensor network
A tensor network is a structured representation of high-order tensors as a network of low-rank tensors connected by contracted indices, used to describe the correlations and entanglement structure of quantum many-body states. In Quantum physics tensor networks provide scalable parametrizations of wavefunctions and operators, enabling understanding and efficient computation in strongly correlated Condensed matter physics, quantum information, and quantum computing research. They matter because they combine tools from linear algebra, statistical mechanics, and computer science to make otherwise intractable problems accessible while highlighting how entanglement organizes physical systems.
Tensor networks arose from efforts to compress and analyze the exponentially large state spaces of quantum systems. Early practical connections include the density matrix renormalization group (DMRG) and later formalizations relating network geometry to entanglement scaling laws such as the area law for entanglement entropy. They serve both as variational ansätze for ground and excited states and as analytical frameworks linking microscopic models (e.g., Heisenberg model, Hubbard model) to universal behaviour governed by conformal field theory and renormalization group ideas. Major institutions active in tensor network research include Perimeter Institute for Theoretical Physics, Max Planck Institute for Physics, Microsoft Research and university groups at MIT, Harvard University, and University of Cambridge.
Tensor networks are graphs whose vertices represent tensors and edges represent index contractions. The most used architectures include matrix product states (MPS), projected entangled pair states (PEPS), multiscale entanglement renormalization ansatz (MERA), and tree tensor networks (TTN). MPS underpin the success of DMRG and efficiently represent 1D gapped ground states consistent with the area law. PEPS generalize MPS to higher dimensions for lattice models such as the square lattice or triangular lattice, but carry higher computational cost. MERA encodes scale invariance and connects to real-space renormalization group transformations and holographic duality conjectures in theoretical work related to AdS/CFT correspondence. TTNs provide hierarchical decompositions useful in quantum chemistry and data compression. Formal tools include tensor decompositions like the singular value decomposition (SVD), canonical forms, and graphical calculus used in publications by researchers such as Steven R. White, Guifre Vidal, and Norbert Schuch.
Tensor networks are widely applied to compute ground states, excited spectra, and dynamics in models like the quantum Ising model and Bose–Hubbard model. They enable study of topological phases (e.g., fractional quantum Hall effect, topological order), symmetry-protected phases, and quantum phase transitions. In quantum information theory they are used to quantify entanglement, design error-correcting codes, and study measurement-induced phase transitions. Tensor networks also inform algorithm design for near-term quantum devices, hybrid variational algorithms such as variational quantum eigensolver (VQE) comparisons, and benchmarks for noisy intermediate-scale quantum (NISQ) hardware produced by companies like IBM, Google Quantum AI, and Rigetti Computing.
Algorithms for tensor networks include variational optimization, time-evolution schemes (TEBD, time-dependent DMRG), and contraction strategies for PEPS and MERA. Complexity considerations relate to tensor contraction order (an NP-hard problem in general) and approximations that trade accuracy for polynomial scaling in bond dimension. Software ecosystems supporting tensor network research include ITensor, TeNPy, and libraries from academic groups and industry. Benchmarks and competitions (e.g., at conferences like the Quantum Information Processing or NeurIPS workshops) compare scaling, precision, and resource use. Theoretical complexity results connect to classical simulation limits of quantum circuits via reductions to tensor network contraction and to hardness results stemming from computational complexity theory.
Tensor networks guide and interpret quantum simulation experiments performed on platforms such as ultracold atoms in optical lattices (groups at Harvard, MIT), trapped ions (e.g., Monroe group), and superconducting qubits (e.g., IBM Quantum). They help design initial states with low entanglement for adiabatic preparation and are used to post-process experimental tomography data via matrix product operator (MPO) reconstructions. Emulation of tensor network states has practical overlap with analog simulators of spin chains and digital quantum simulation protocols run on devices from Google Sycamore and IonQ. Collaborations between national labs like Lawrence Berkeley National Laboratory and universities often aim to validate tensor network predictions against measured correlations and entanglement witnesses.
Limitations of tensor networks include inefficient scaling for highly entangled states (volume-law entanglement), challenges contracting two-dimensional PEPS at large bond dimension, and rigorous error bounds for dynamics. Open problems include better algorithms for contraction order optimization, connections to quantum complexity classes, and extensions to finite-temperature and open quantum systems. From a social-justice perspective, access to high-performance computing, specialized software (e.g., proprietary toolchains), and experimental facilities concentrates advantages in wealthy institutions and industry, perpetuating inequities in who can contribute to quantum research. Equitable access initiatives—open-source projects (e.g., ITensor community editions), international collaboration programs, and training at minority-serving institutions—are essential to democratize benefits from advances in quantum simulation and to ensure that outcomes (from quantum materials to computing resources) serve broad public interest rather than narrow commercial control.
Category:Quantum many-body theory Category:Computational physics