Ultra Low Overhead Syndrome Extraction for the Steane code

TL;DR

The paper tackles the challenge of fault-tolerant syndrome extraction for the Steane code by introducing a dynamic protocol that uses two ZX-rewrite–derived circuits: a fault-tolerant primary circuit with 14 CNOTs and 4 ancillae, and a lean non-FT recovery circuit with 11 CNOTs and 3 ancillae. An adaptive flag mechanism discards flagged outputs and falls back to the recovery circuit only when needed, preserving fault tolerance with minimal overhead. Monte Carlo simulations under a circuit-level depolarizing noise model show the method achieves notable reductions in logical error rates—approximately $14.3\%$ relative to an optimized Steane method and $17.7\%$ relative to Reichardt's three-qubit approach—setting a new benchmark for distance-3 codes. The work also establishes CNOT-count optimality for the circuits, provides decoding strategies for flagged faults, and discusses extending the approach to automated optimization for larger codes.

Abstract

We establish a new performance benchmark for the fault-tolerant syndrome extraction of [[7, 1, 3]] Steane code with a dynamic protocol. Our method is built on two highly optimized circuits derived using fault-equivalent ZX-rewrites: a primary fault-tolerant circuit with 14 CNOTs and an efficient non-fault-tolerant recovery circuit with 11 CNOTs. The protocol uses an adaptive response to internal faults, discarding flagged measurements and falling back to the recovery circuit to correct potentially detrimental errors. Monte Carlo simulations confirm the efficiency of our protocol, reducing the logical error rate per cycle by an average of ~14.3% relative to the optimized Steane method [arXiv:2506.17181] and ~17.7% compared to the Reichardt's three-qubit method [arXiv:1804.06995], the leading prior techniques.

Figures (6)

• Figure 1: Qubit layout and stabilizer structure of the Steane code, as given in ryan-andersonRealizationRealTime2021.
• Figure 2: Optimized syndrome extraction circuits for of our dynamic protocol: (a) fault-tolerant primary circuit using 14 CNOTs and 4 ancillae, and (b) non-fault-tolerant recovery circuit using 11 CNOTs and 3 ancillae. The Steane code's self-duality yields structurally identical X-syndrome measurements.
• Figure 3: Circuit with (a) flagged errors and (b) an example of how an error spreads.
• Figure 4: Flowchart of the full syndrome extraction protocol for a single cycle.
• Figure 5: Logical error rate $p_{\mathrm{L}} / p_{\mathrm{phys}}^2$ as a function of physical error rate $p_{\mathrm{phys}}$, with Wilson confidence intervals.

Theorems & Definitions (8)

• Theorem 1
• proof
• Theorem 2
• proof
• Proposition 3
• proof
• Proposition 4
• proof