Efficient tomography of microwave photonic cluster states

Entanglement among a large number of qubits is a crucial resource for many quantum algorithms. Such many-body states have been efficiently generated by entangling a chain of itinerant photonic qubits in the optical or microwave domain. However, it has remained challenging to fully characterize the generated many-body states by experimentally reconstructing their exponentially large density matrices. Here, we develop an efficient tomography method based on the matrix-product-operator formalism and demonstrate it on a cluster state of up to 35 microwave photonic qubits by reconstructing its $2^{35} \times 2^{35}$ density matrix. The full characterization enables us to detect the performance degradation of our photon source which occurs only when generating a large cluster state. This tomography method is generally applicable to various physical realizations of entangled qubits and provides an efficient benchmarking method for guiding the development of high-fidelity sources of entangled photons.

💡 Research Summary

The paper addresses the long‑standing challenge of fully characterizing large‑scale photonic cluster states, whose density matrices grow exponentially with the number of qubits. By exploiting the inherent temporal locality of sequential photon emission, the authors show that the many‑body density matrix can be compactly represented as a matrix‑product operator (MPO). In this representation the number of free parameters scales linearly with the number of photons, because the MPO consists of a chain of small tensors whose bond dimension D is bounded by the square of the emitter’s internal Hilbert space dimension (D ≤ d²).

To reconstruct the MPO from experimental data, the authors develop a measurement protocol that requires only local correlations of a fixed small window L = 5 consecutive photons, regardless of the total chain length N. Each five‑photon block is fully characterized by the 4⁵ possible Pauli‑operator correlators C(a,b,c,d,e)ₛ, which can be obtained in parallel by measuring every fifth photon in the same basis. This clever choice reduces the number of distinct measurement settings to a constant independent of N, while still satisfying the “reconstructibility condition” D ≤ 4⌊(L‑1)/2⌋ for typical MPOs.

Experimentally, a superconducting qubit is used as a deterministic emitter to generate linear cluster states of up to 35 microwave photons. A quantum‑limited amplifier records the quadrature observables of each photon, enabling the extraction of the full set of five‑photon correlators. The authors first obtain an approximate MPO using a reformulated direct‑inversion formula, which also yields an estimate of the required bond dimension. They then refine the MPO by fitting the measured correlators with a Gauss–Newton weighted least‑squares algorithm. Because the algorithm starts from a good initial guess, it converges rapidly (linear or better) and is robust against the large statistical noise that plagues high‑order correlations. Moreover, the weighted fit naturally propagates measurement uncertainties to the MPO parameters, providing error bars on the reconstructed density matrix.

The reconstructed MPOs allow the authors to compute the full 2³⁵ × 2³⁵ density matrix, from which they evaluate the quantum state fidelity and the localizable entanglement (LE). For a 10‑qubit cluster, LE persists over seven consecutive photons; for larger clusters (20‑ and 35‑qubit) the entanglement length drops to five and three photons respectively. This degradation is not captured by extrapolations based on smaller clusters, demonstrating that high‑fidelity generation of a small cluster does not guarantee the same performance for larger ones. The fidelity analysis also reveals a subtle performance drop of the photon source that becomes apparent only when the chain length exceeds a certain threshold.

A systematic comparison (Tables I and II) shows that the proposed method outperforms conventional full tomography, process tomography with numerical extrapolation, direct inversion, and iterative maximum‑likelihood estimation. It requires only O(N) parameters, needs a constant number of measurement settings, tolerates statistical noise, and guarantees a physical (positive‑semidefinite) MPO without the need for post‑processing.

Finally, the authors argue that the MPO‑based tomography is platform‑agnostic and can be applied to optical photons, trapped‑ion chains, and other solid‑state emitters. By providing an efficient, scalable, and statistically sound tool for full state reconstruction, the method paves the way for reliable benchmarking of large entangled‑photon sources, a crucial step toward fault‑tolerant measurement‑based quantum computation and quantum networking.