Quantum hardware developer Nord Quantique has detailed an error-mitigation methodology that curtails State Preparation and Measurement (SPAM) error rates within single-mode bosonic grid-state qubits. Published as a preprint on arXiv (“Quantum error correction of a grid-state qubit with state preparation and measurement errors below 10−3“), the research team leveraged high-performance autonomous quantum error correction (QEC) to implement a repeat-until-success initialization sequence alongside a multi-round repeated measurement framework.

The combined optimization suppresses total SPAM errors to below 0.1% (7(7)×10−4 average across all cardinal states), delivering a 100-fold reduction in operational noise compared to traditional Gottesman-Kitaev-Preskill (GKP) architectures. This optimization brings bosonic initialization and readout parameters into parity with the baseline performance metrics of established superconducting transmon qubit platforms.

                         [ Nord Quantique Hardware Stack ]
  Cavity Topology     ──► Double-post microwave cavity machined from high-purity aluminum.
  Qubit Interface     ──► Transmon auxiliary control qubit dispersively coupled to the storage mode.
  Initialization Core ──► Post-selected stabilization utilizing an sBs tracking loop.
  Readout Optimization──► Multi-round repeated finite-energy measurements (RFE) with an all-agree rule.

The Mechanics of sBs Post-Selected Stabilization

Traditional single-mode grid state preparation methods rely on pre-calculated, open-loop sequences of Echoed Conditional Displacements (ECD) and auxiliary qubit rotations. While effective in low-noise simulations, this approach is vulnerable to physical transmon decoherence during extended gate operations, which injects residual noise into the final state envelope.

Nord Quantique addresses this limitation by repurposing the intrinsic sBs stabilization protocol—which acts as an autonomous QEC layer—to actively drive the system’s storage mode from a vacuum or squeezer state directly into the GKP code-space manifold.

The state preparation pipeline functions as a repeat-until-success system:

  1. Presqueezing Injection: An optimized ECD sequence prepares a 9 dB squeezed state in the storage mode. This step reduces the total number of subsequent correction cycles needed because the input is already an approximate eigenstate of a code stabilizer.
  2. Iterative sBs Steering: The system executes continuous sBs rounds consisting of four auxiliary qubit rotations interleaved with echoed conditional displacements. This process sets a bounded finite-energy envelope parameter (Δ=0.38, averaging ∼2.96 photons per mode).
  3. Conditional Post-Selection: Every stabilization block is paired with an inline mid-circuit auxiliary readout. By applying an all-agree post-selection policy, the classical controller discards iterations that register an auxiliary projection outside of the ground state (∣g⟩), filtering out gate faults before computation begins.

Repeated Finite-Energy Readout and Magic State Synthesis

To complement the initialization pipeline, the team upgraded the final readout phase by converting the conventional one-bit phase-estimation loop into a Repeated Finite-Energy (RFE) measurement framework. Instead of extracting a Pauli expectation value from a single measurement pass, the system executes up to eight consecutive finite-energy measurements back-to-back within a single-shot execution window.

This repetition suppresses errors caused by finite-energy envelopes and auxiliary qubit readout noise. The all-agree voting filter ensures that final logical operator assessments are insulated against single-photon loss during readout, driving the measurement survival probability to 39% while dropping the read error below the system’s structural transmon readout limit (V=0.95(1)).

[ sBs Stabilization Block ] ──► [ Mid-Circuit Check ] ──► [ Pass: Keep | Fail: Discard ] ──► [ Multi-Round RFE Readout ]

Crucially, the team demonstrated that these high-fidelity preparation subroutines seamlessly translate to H-type magic states (∣ψH​⟩∝∣±Z⟩±∣±X⟩), which serve as the primary resource needed to inject non-Clifford gates into fault-tolerant circuits.

Starting from a vacuum state, the system generated high-fidelity magic states after eight stabilization rounds, yielding a total SPAM error of 8(5)×10−3. When integrated into a continuous, non-postselected execution thread operating in the restless regime (zero idle time between consecutive 2.5μs sBs cycles), the processor maintained a steady logical error rate of 8.1(2)×10−3 per QEC round. This baseline stability rate confirms that suppressing SPAM errors provides a clear performance boost during initialization and readout without causing any downstream degradation to the platform’s autonomous error-correction logic.

Review the official corporate brief via the Nord Quantique Article Hub. The peer-reviewed market deployment timelines can be audited on the Business Wire Infrastructure Wire, and the underlying mathematical proofs detailing the auxiliary readout parameters, bootstrap resampling intervals, and GKP lattice spacing metrics can be reviewed on the arXiv Quantum Physics.

July 13, 2026

Print Friendly, PDF & Email