Research

DLR, Cologne team reports state-loading method cutting two-qubit gate depth up to 100x

Researchers at the German Aerospace Center (DLR) Institute of Software Technology and the University of Cologne posted an arXiv paper, "Scalable Unitary State-Encoding for Generic Large Data," describing a deterministic state-preparation framework (the article does not say whether it has been peer reviewed, and arXiv papers are typically preprints). It combines tensor-network representations, Boolean logic minimization and Walsh-Hadamard transformations, and the authors report up to two orders of magnitude lower two-qubit gate depth than standard multiplexer compilers. In benchmarks on Lorentzian, polynomial and financial jump-diffusion distributions, depth plateaued at a fixed infidelity of about 5 x 10^-3, versus exponential scaling for Qiskit state-preparation tools. On the 20-physical-qubit IQM Emerald superconducting processor, a 2D Gaussian state was encoded at total circuit depth 56 (CZ depth 26) with state infidelity of 0.011 ± 0.002 over 256,000 shots.

Why it mattersLoading classical data into a quantum state is a bottleneck for algorithms such as the HHL linear solver, quantum PDE integration and quantum machine learning, because exact preparation for generic data needs gate depth that grows exponentially and can exceed hardware coherence limits. The reported approach trades a bounded, tunable infidelity for shallower circuits, and the hardware result is a single 20-qubit demonstration.

Read the full story at Quantum Computing Report →

Summarized by AI from the linked article.