Quantum circuit compression using tensor networks allows longer simulations of quantum systems on today’s hardware; however, current methods require decomposing resulting circuits into a device’s native components, a step that reduces efficiency gains. Optimised circuits exhibit predictable structures mirroring underlying physical models, specifically near-RZZ for the Ising and near-fSim for the Heisenberg XXZ model. Tailoring compression techniques to a system’s specific characteristics improves efficiency when simulating complex systems on limited hardware.
Volkswagen AG and Technical University of Munich discovered that quantum circuits compressed using tensor networks exhibit predictable structures linked to underlying physical models; specifically, near-RZZ for magnetism known as the Ising model and near-fSim for the Heisenberg XXZ model. This finding addresses a key bottleneck in running these simulations on real hardware, converting optimised instructions into commands understood by specific processors, a process akin to translating between languages.
Even after optimisation, ‘translated’ gates retain subtle biases, like adding seasoning to a recipe, revealing they aren’t entirely generic but instead reflect the system being simulated. Tensor networks can be visualised like LEGO bricks: complicated data or calculations are represented as interconnected simpler parts.
Riemannian optimisation compresses quantum circuits exploiting Weyl symmetry and Hamiltonian dependence
It reduces two-qubit gate count by leveraging predictable “Weyl” structures within circuits, near RZZ for Ising propagators and near-fSim for Heisenberg XXZ models. Parameterised Optimisation (PO) struggles to match this performance with Ising propagators. However, it surpasses it when working with Heisenberg propagators due to its inherent compatibility with the XXZ structure; this reveals an important trade-off: retaining small residual Weyl parameters is essential for compressing the Ising propagator while native XXZ structure allows direct construction of more efficient compressed circuits for the Heisenberg model.
Optimised gates fall near RZZ rotations during propagation of Ising interactions, and close to fSim gates for Heisenberg XXZ models, demonstrating predictable structures in Riemannian optimisation yields. Analysis of the Ising propagator shows that an average Weyl coordinate consistently exceeds others by several orders of magnitude. This means resulting matrices closely resemble single RZZ rotations requiring only one entangling gate instead of three needed for a generic two-qubit operation; similarly, optimised gates with the Heisenberg XXZ model strongly favour local XXZ structure equivalent to a single fSim gate.
Direct construction of more efficient compressed circuits is possible through using native hardware capabilities, sharply reducing the number of required two-qubit operations in second-order Trotterization schemes. Quantum circuit compression extends simulations beyond current hardware limits, but simply reducing gate count isn’t enough. Translating these compressed circuits into processor’s native instructions can erode those gains, and direct optimisation over available quantum gates offers potential benefits while presenting a trade-off: retaining subtle “Weyl” structures within Ising propagators appears important for achieving high levels of compression.
Preserving these subtle mathematical structures within quantum calculations to achieve substantial compression is significant; it highlights a complex challenge that goes beyond merely minimising gate numbers.
This research revealed that quantum circuits optimized for compression exhibit mathematical patterns, specifically resembling RZZ or fSim gate structures, that align with the physics being simulated. This matters because it demonstrates a way to reduce computational overhead; rather than converting compressed circuits into standard instructions, optimization can directly utilize available quantum gates.
The researchers found retaining subtle Weyl structures is important for compressing simulations of Ising interactions, while those simulating Heisenberg interactions benefit from exploiting existing XXZ structure within the hardware. They tested this approach on systems up to fifty qubits and observed reduced numbers of necessary two-qubit operations.
👉 More information
🗞 Exploiting the Weyl-Chamber Structure for Hardware-Aware Quantum Circuit Compression
✍️ Neel Misciasci (Volkswagen AG); Isabel Nha Minh Le and Christian B. Mendl (Technical University of Munich)
🧠 ArXiv: https://arxiv.org/abs/2610.32812
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
