arXiv:2610.32812
Exploiting the Weyl-Chamber Structure for Hardware-Aware Quantum Circuit Compression
Neel Misciasci, Isabel Nha Minh Le, Christian B. Mendl
Abstract
Quantum circuit compression based on tensor networks enables long-time simulation of quantum many-body dynamics on near-term hardware. While Riemannian optimization has been successfully employed for this purpose, it typically optimizes over the manifold of generic two-qubit unitaries. Consequently, the resulting gates must still be synthesized into the hardware's native gate set via KAK decomposition, an overhead that partially negates the original compression gain. We show that Riemannian-optimized gates already exhibit a predictable Weyl structure: near-$R_{ZZ}$ for the Ising model and near-fSim for the Heisenberg XXZ model. Building on this insight, we investigate whether parameterized optimization directly over the native gate set can offer a beneficial trade-off between expressivity and circuit efficiency by reducing the circuit's dimensionality. When comparing the two approaches, we find that the small residual Weyl parameters neglected by the latter are essential for high compression of the Ising propagator, whereas for the Heisenberg propagator, the native XXZ structure can be exploited directly to build a more efficient compressed circuit.
AI Summary
The paper asks whether optimizing compressed time-evolution circuits directly in a hardware-native gate family improves accuracy at fixed native entangling depth compared with optimizing generic two-qubit unitaries and then synthesizing them. It compares these approaches numerically for nonintegrable Ising and transverse-field XXZ chains of $N=50$ qubits with open boundaries. Both methods use matrix-product-operator contractions to evaluate gradients against a reference propagator. The baseline uses Riemannian optimization over generic two-qubit unitaries; the hardware-aware method optimizes native $R_{ZZ}$ or fSim parameters while retaining Riemannian updates for single-qubit gates. KAK analysis finds that optimized Ising gates lie near the $R_{ZZ}$ family, while XXZ gates lie near an fSim-like family. Comparisons account for synthesis into three $R_{ZZ}$ gates or an fSim plus an $R_{ZZ}$ gate, respectively. For Ising, native optimization provides little improvement over its Trotter initialization, and generic-unitary optimization generally wins beyond approximately 26 native layers. The paper attributes this to small residual Weyl coordinates omitted by the native ansatz. For XXZ, native optimization achieves roughly half an order of magnitude lower loss through approximately 35 layers, attributed to better initialization and sufficient fSim expressivity. At five Ising Trotter repetitions, parameter counts are 1044 versus 4050. The native method nevertheless incurs additional tensor contractions. Its deeper-circuit training becomes harder, and the mechanisms behind optimization plateaus remain unresolved. The comparison is limited to these models and tested regimes; conserved-quantity-sensitive simulations and fermionic systems are left for future work.
Central claim may not hold: The claim that residual Ising Weyl coordinates are essential is stronger than the evidence, which compares particular circuit layouts, initializations, and optimization procedures without isolating necessity. The Ising reference's stated cost of 603 R_ZZ gates is a layer count under the described synthesis, corresponding to 14,775 gates on the 50-qubit chain. The Weyl-chamber description excludes some SU(4) local-equivalence classes and incorrectly confines the entire fSim family to the ordered face alpha_x=alpha_y. The written open-chain Heisenberg RO initialization assigns only half the target transverse field to each endpoint, with no stated correction or evidence establishing whether this also affected the implementation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 15 pages, 7 figures