Tensor Network Simulation Quantum Circuits

Short answer: Tensor networks let classical computers simulate some quantum circuits by representing the quantum state as a network of smaller tensors, and recent papers push on how far this works. A 2023 Physical Review Letters paper proves that circuits of single-qubit and finite-ranged two-qubit gates can be simulated in subexponential time in the number of gates. These are classical algorithms with limits, and they do not mean every quantum circuit can be simulated cheaply.

What the method does

A 2023 paper in Quantum describes simulating quantum circuits on classical computers by representing quantum states as rooted tree tensor networks. The algorithm first determines a fixed tree structure suited to the entanglement the circuit is expected to generate. Gates are applied in sequence: single-qubit gates are absorbed into leaf nodes, and two-qubit gates are split by singular value decomposition, with the resulting virtual bond threaded through the tree. The authors analyze the applicability, computational cost and memory requirements, identify scenarios needing smaller bond dimensions than a matrix product state representation, and report numerical experiments for circuit layouts up to 37 qubits.

A bound on contraction time

A 2023 Physical Review Letters paper derives a rigorous upper bound on the classical computation time of finite-ranged tensor network contractions in two or more dimensions. It concludes that quantum circuits of single-qubit and finite-ranged two-qubit gates can be classically simulated in subexponential time in the number of gates. The authors also present and implement an algorithm that meets the bound and finds contraction orders with much lower computation times in practice. They report that in practically relevant cases it beats standard simulation schemes and, for certain circuits, a state-of-the-art method, with speedups of several orders of magnitude over naive contraction for two-dimensional circuits on as small as an 8 by 8 lattice. They say they obtain similarly efficient contraction schemes for Sycamore-type circuits, instantaneous quantum polynomial-time circuits and nonhomogeneous (2+1)-dimensional random circuits.

Bridging two simulation approaches

A 2026 arXiv paper says prominent approaches to classical simulation are based on stabilizer decompositions and tensor network contraction. It presents a unified framework placing the two under a common formalism, and two new algorithms for an n-qubit circuit C: one runs in time that scales as T to the power of the tree-width of C, and the other as T to the power of gamma times the rank-width, where T is the number of non-Clifford gates and gamma is about 3.42. The authors say the algorithms are simple, need only a linear amount of memory and are trivially parallelizable. They also introduce refined measures, focused tree-width and focused rank-width, that are never worse than their standard counterparts.

Why simulation matters

The same paper says classical simulation plays a critical role in developing, validating and understanding quantum algorithms, and that it is essential for validating quantum hardware, because without it the outputs of quantum processors cannot easily be verified. It notes the widely held belief that simulating generic quantum systems classically takes time exponential in system size, while many quantum circuits turn out to be simulable more efficiently.

What this does not tell you

The abstracts do not say which specific quantum hardware experiments these methods can or cannot reproduce. The results are about particular circuit families and measures such as tree-width, and a circuit with a large tree-width or many non-Clifford gates remains hard for these methods.

FAQ

What is a tensor network simulation?

A classical method that represents a quantum state as a network of smaller tensors and contracts it to compute results.

Does this mean quantum circuits are easy to simulate?

No. The papers cover specific circuit families, and costs grow with quantities like tree-width and the number of non-Clifford gates.

What did the 2023 PRL paper prove?

That circuits of single-qubit and finite-ranged two-qubit gates can be simulated in subexponential time in the number of gates.

Why do people simulate circuits classically?

To develop and validate algorithms and hardware, per the 2026 paper.

How big were the experiments?

The tree tensor network paper reports numerical experiments up to 37 qubits.