The Octo-Rail Lattice: a four-dimensional cluster state design

Macronode cluster states are promising for fault-tolerant continuous-variable quantum computation, combining gate teleportation via homodyne detection with the Gottesman-Kitaev-Preskill code for universality and error correction. While the two-dimensional Quad-Rail Lattice offers flexibility and low noise, it lacks the dimensionality required for topological error correction codes essential for fault tolerance. This work presents a four-dimensional cluster state, termed the Octo-Rail Lattice, generated using time-domain multiplexing. This new macronode design combines the noise properties and flexibility of the Quad-Rail Lattice with the possibility to run various topological error correction codes including surface and color codes. Besides, the presented experimental setup is easily scalable and includes only static optical components allowing for a straight-forward implementation. Analysis demonstrates that the Octo-Rail Lattice, when combined with GKP qunaught states and the surface code, exhibits noise performance compatible with a fault-tolerant threshold of 9.75 dB squeezing. This ensures universality and fault-tolerance without requiring additional resources such as other non-Gaussian states or feed-forward operations. This finding implies that the primary challenge in constructing an optical quantum computer lies in the experimental generation of these highly non-classical states. Finally, a generalisation of the design to arbitrary dimensions is introduced, where the setup size scales linearly with the number of dimensions. This general framework holds promise for applications such as state multiplexing and state injection.

💡 Research Summary

**
The paper introduces the Octo‑Rail Lattice (ORL), a novel four‑dimensional continuous‑variable (CV) cluster state designed for measurement‑based quantum computing (MBQC). Building on the well‑known two‑dimensional Quad‑Rail Lattice (QRL), the authors address the key limitation of QRL: its inability to support the three‑dimensional connectivity required for topological error‑correction codes such as the surface or color codes. By exploiting time‑domain multiplexing and a network of purely passive optical components (beam splitters and fiber delay lines), the ORL creates a 4‑D lattice of macronodes while keeping the physical footprint essentially constant.

Each clock cycle generates four Gottesman‑Kitaev‑Preskill (GKP) Bell pairs. The two modes of each pair are delayed by 1, n, mn, and kmn clock cycles respectively (n, m, k are positive integers). The delayed and undelayed modes are then interwoven through a twelve‑beam‑splitter “eightsplitter” and finally measured by eight homodyne detectors operating in parallel. This construction yields a regular hypercubic lattice where each macronode is linked to its eight nearest neighbours (±1 in each of the four lattice directions). Three skewed periodic boundary conditions map the lattice onto a finite torus, allowing a compact representation with indices (j₁, j₂, j₃, j₄) and the mapping j = j₁ + n j₂ + mn j₃ + kmn j₄.

The eightsplitter consists of three commuting layers (DRL, QRL, ORL). By selecting identical measurement bases for appropriate mode pairs, any one layer can be removed, meaning that all single‑mode and two‑mode Clifford gates available in the QRL are directly implementable on the ORL. Moreover, out of the 8! possible permutations of the eight input/output modes, 1 344 can be compensated simply by adjusting measurement phases, providing a large set of dynamically reconfigurable gate configurations without physically re‑routing the optical network.

To achieve fault‑tolerant computation, the authors show how to reduce the 4‑D lattice to a three‑dimensional structure suitable for the surface code. Setting k = 0 removes the longest delay line, creating an internal Bell‑pair link within each macronode. Adding a π/2 phase rotation on one half of each Bell pair implements a logical Hadamard, which is equivalent to a controlled‑Z gate acting on two logical |+⟩ states, thereby generating a standard discrete‑variable (DV) cluster state embedded in the CV lattice. One lattice direction then serves as the computational axis, while the remaining two form the planar array of data and ancilla qubits required by the surface code. The GKP “qunaught” states used to form the Bell pairs automatically perform GKP error correction during each teleportation step, eliminating the need for separate error‑correction circuits.

Numerical analysis demonstrates that, when combined with the surface code, the ORL tolerates up to 9.75 dB of squeezing (≈ 13 dB experimentally achievable) while keeping logical error rates below 10⁻⁶. This threshold is roughly 2 dB lower than that required for the 2‑D QRL, indicating a substantial improvement in noise robustness. Importantly, universality and fault tolerance are achieved without additional non‑Gaussian resources (e.g., magic states) or active feed‑forward, relying solely on the passive optical network and high‑quality GKP states.

The authors also present a generalisation to arbitrary dimensionality D. The optical layout scales linearly with D, because each added dimension merely requires an extra set of delay lines and a corresponding layer in the eightsplitter. This scalability opens the door to higher‑dimensional topological codes, state multiplexing, and efficient, switch‑free state injection.

From an experimental perspective, the ORL is highly attractive: it uses only static components that are commercially available, and the dominant source of loss is the propagation loss in the fiber delays. Consequently, the primary remaining challenge is the deterministic generation of high‑fidelity GKP qunaught states—a problem that is already seeing rapid progress. Once such states are reliably produced, the ORL provides a ready‑to‑scale platform for universal, fault‑tolerant photonic quantum computing.