A computational method used to simulate quantum many-body systems by representing wavefunctions as networks of tensors.Tensor networks are a family of computational techniques that represent the quantum state of a many-body system as a network of interconnected tensors — multidimensional arrays of numbers. Instead of storing the full wavefunction, which grows exponentially with the number of particles, a tensor network compresses the information by exploiting the local structure of interactions. This makes it possible to simulate quantum systems that would otherwise be intractable on classical computers, such as chains of interacting spins, superconducting circuits, or the behavior of particles in a quantum fluid.
Tensor networks were introduced in the 1990s, with the Density Matrix Renormalization Group developed by Steven White in 1992 as a precursor. The formal framework of tensor networks, including Matrix Product States, was established in the early 2000s by Guifré Vidal and others.
The key idea is to decompose the huge wavefunction into smaller tensors connected by contracted indices, forming a graph. The geometry of this graph (e.g., a chain, a tree, or a more complex lattice) reflects the entanglement structure of the system. Algorithms like the Density Matrix Renormalization Group (DMRG) iteratively optimize the tensors to find the ground state or time evolution of the system. By adjusting the 'bond dimension' — the size of the connections between tensors — one can trade accuracy for computational cost, making the method flexible and powerful.
Tensor networks are used to simulate quantum many-body systems in condensed matter physics, to design quantum circuits, and in machine learning for compressing neural networks.
Tensor networks can also be used to simulate classical statistical mechanics problems, such as the Ising model, by mapping them to quantum systems.











