Add noise. The simulation gets easier.
Matrix-product-operator simulation of quantum circuits: represent the quantum state as a compressed tensor network and update it gate by gate. The truncation — discarding small singular values to keep the representation manageable — introduces error. In noiseless circuits, this error can grow exponentially with circuit depth, making deep circuits intractable.
Note (2026-11-07): the abstract of arXiv:2603.20400 studies matrix-product-operator truncation errors for 1D Haar-random circuits and non-integrable Ising Lindbladian dynamics under depolarizing or amplitude-damping noise, shows that the average purity relaxes to a steady value on a timescale inversely proportional to the noise rate, shows that truncation errors contract exponentially in system size and time, yields an empirical bound exponentially tighter in system size than the existing bound, and says the results are empirical evidence that MPO methods may efficiently sample from these systems. The inverse-noise-rate timescale in the abstract belongs to the purity relaxation, not to the contraction of the truncation error as the essay says, and the essay states as established what the authors call empirical evidence. The closing remark about noise as a source of convergence is my own. I compared this essay with the abstract only.
Add depolarizing or amplitude-damping noise (arXiv:2603.20400). The truncation errors now contract exponentially in both system size and evolution time. The reason: noise drives different density matrices toward the same steady state. Two states that were distinguishable before the noise become indistinguishable after it. The truncation error — which is the difference between the exact state and the truncated approximation — contracts because both are being driven toward the same attractor.
The contraction timescale is inversely proportional to noise strength. Stronger noise means faster contraction means smaller errors at shorter times. The system purity stabilizes at a steady value, and the MPO simulation tracks this steady purity accurately.
The implications: one-dimensional noisy random circuits at arbitrary depths and steady states of one-dimensional Lindbladian systems may be efficiently simulable with MPO methods. The noise that makes the quantum system “less quantum” (decoheres it, reduces entanglement, drives it toward a classical mixture) is exactly what makes the classical simulation tractable.
The structural insight: noise is not just a source of error in quantum computing. For simulation, it's a source of convergence. The same mechanism that destroys quantum advantage (decoherence) creates classical advantage (simulability). The difficulty of simulating a quantum system tracks its quantumness, and noise reduces both simultaneously.