Aliferis and Preskill (2008) proved a rigorous lower bound on the accuracy threshold for fault-tolerant quantum computation whose only entangling operation is a bias-preserving CPHASE gate, assuming a two-parameter local stochastic biased-noise model in which dephasing dominates, and excluding storage (idle) faults by measurement staggering. They explicitly left unanalyzed the extension to a more realistic Hamiltonian (coherent, non-Markovian) biased-noise model, tighter thresholds beyond their conservative combinatorial bounds, and architectures in which idle faults are present. Later literature has massively developed biased-noise error correction (asymmetric Bacon-Shor codes, tailored surface codes, XZZX codes, cat qubits) and shown numerically that bias can raise thresholds dramatically, but these results are restricted to stochastic Pauli noise (code-capacity or circuit-level Markovian models). The surviving open problem is a provable, physically realistic threshold theory exploiting noise bias: a rigorous accuracy threshold for Hamiltonian/coherent biased noise, provable thresholds with idle faults included, and tightened biased-noise threshold bounds for bias-preserving gate architectures.

Background

A quantum computer can be made scalable by quantum error correction and fault tolerance: quantum information is encoded so that errors can be detected and corrected faster than they accumulate, and every operation on encoded data is built from fault-tolerant gadgets (subcircuits designed so that a small number of physical faults inside a gadget cannot cause an uncorrectable logical error). The accuracy threshold is the largest physical error rate below which arbitrarily long computation becomes possible with polylogarithmic overhead; threshold theorems prove such a constant exists for specified noise models and gadget constructions.

Realistic noise is often highly asymmetric: in many platforms (superconducting, spin, and trapped-ion systems) dephasing, i.e. Pauli-ZZ phase errors, is far more frequent than bit-flip (XX/YY) errors. Aliferis and Preskill formulated a scheme in which the only entangling operation is a conditional-phase (CPHASE, equivalently CZCZ) gate together with single-qubit preparation of ∣+⟩=(∣0⟩+∣1⟩)/2|+\rangle=(|0\rangle+|1\rangle)/\sqrt{2} and measurement in the XX basis. These operations are bias-preserving: because CZCZ is diagonal in the computational basis, it commutes with phase errors, so frequent dephasing faults are never converted into rare bit-flip faults during syndrome extraction. Their scheme concatenates a repetition-style inner phase-flip code (a code that corrects only phase errors, by majority logic along one direction) with outer stabilizer codes (quantum error-correcting codes specified by commuting parity-check operators), using gate teleportation (realizing a gate by measurements on a prepared resource state) and postselection (repeating measurements and keeping only accepted outcomes) to build gadgets, and its threshold analysis uses level reduction: a fault analysis in which a level-(k+1)(k+1) gadget is fault-tolerant if faults at level kk propagate to at most one fault per block.

The noise model they analyzed is local stochastic biased noise, a two-parameter adversarial model: each faulty location has probability at most ε\varepsilon of a dephasing-type fault and at most εbias\varepsilon_\mathrm{bias} of any other fault, where the bias η≈ε/εbias\eta\approx\varepsilon/\varepsilon_\mathrm{bias} is large. In this model they obtained a rigorous lower bound on the accuracy threshold (about 2.09×10−32.09\times10^{-3} at bias 10310^3 and 3.51×10−33.51\times10^{-3} at bias 10410^4 with improved decoding, roughly five times the corresponding unbiased bound), by combinatorial counting of malignant fault paths (combinations of faults at gadget locations that propagate to an uncorrectable logical error) and optimization over small code parameters (n,r,t)(n,r,t). Two features of the analysis matter for the open problem. First, the local stochastic model abstracts noise as discrete Pauli faults; it is not a microscopic description of coherent (Hamiltonian) noise, in which the computer is weakly coupled to an environment by interaction Hamiltonians whose dephasing terms dominate, so that errors are unitary, non-commuting, non-Markovian, and can accumulate coherently. General threshold theorems for such Hamiltonian noise exist (Aharonov, Kitaev, and Preskill proved a threshold for non-Markovian noise with spatially decaying system-bath couplings, and Ng and Preskill treated Gaussian non-Markovian noise), but none of them exploits noise bias. Second, by staggering the repeated measurements inside each gadget, the construction eliminates storage locations where data qubits sit idle, so faults at idle locations are simply not counted; architectures in which idle faults or scheduling constraints are present are not analyzed.

The question of exploiting noise bias has since become a major research direction, but along different lines. Asymmetric (bias-tailored) code constructions include asymmetric Bacon-Shor codes (concatenations of repetition codes with very simple syndrome-extraction circuits; Brooks and Preskill), overhead analyses for biased noise (Webster, Bartlett, and Poulin), and bosonic cat qubits engineered to have exponentially suppressed bit-flips (Guillaud and Mirrahimi). For topological codes, Tuckett and coauthors tailored surface codes to biased noise and proved that a modified surface code has a 50% threshold for pure dephasing at code capacity, and the XZZX surface code of Bonilla Ataides and coauthors (a surface-code variant whose stabilizer generators form an XZZXXZZX pattern, adapting it to biased noise) numerically matches the hashing bound (the threshold achievable by random codes) for every single-qubit Pauli channel and surpasses previously known thresholds even with noisy syndrome measurements. Recent circuit-level studies (which do include idle and gate faults) find that the code-capacity advantages of bias-tailored codes are strongly reduced under realistic syndrome-extraction noise, and a 2026 review by Ruiz, Guillaud, Vuillot, and Mirrahimi concludes that in the CPHASE-restricted gate set used by Aliferis and Preskill the gadget complexity essentially cancels the benefit of the bias, with the benefit instead unlocked by bias-preserving CXCX gates or quantum non-demolition (repeatable, observable-preserving) multi-qubit ZZ measurements. All of these results, however, are formulated for stochastic (mostly Markovian) noise models, and their thresholds are numerical or semi-analytic rather than proven bounds; none extends the rigorous threshold analysis to coherent/Hamiltonian biased noise.

Problem Statement

Establish a provable, physically realistic accuracy-threshold theory for fault-tolerant quantum computation that exploits noise asymmetry, in the setting of bias-preserving architectures whose entangling operation is a biased CPHASE-type gate with single-qubit XX-basis preparation and measurement. Concretely, relative to the rigorous local-stochastic biased-noise lower bounds of Aliferis and Preskill (2008) for such schemes: (i) prove an accuracy threshold that exploits dephasing bias for a Hamiltonian (coherent, non-Markovian) biased-noise model, in which weak system-bath interaction Hamiltonians with dominant dephasing couplings replace the adversarial discrete-Pauli fault model, handling the facts that coherent errors are non-Pauli, non-commuting, and can accumulate rather than average out across gadgets; (ii) prove biased-noise thresholds for architectures in which storage (idle) faults and realistic scheduling constraints are present, rather than eliminated by measurement staggering, and characterize how idle faults degrade the bias advantage as an explicit function of the idle-to-gate fault ratio; and (iii) tighten the provable biased-noise threshold beyond the conservative combinatorial lower bounds. These three directions jointly support one objective, a rigorous threshold statement for biased noise that is simultaneously physically realistic (coherent, including idle faults) and quantitatively strong. A rigorous negative resolution counts equally: a proof that no bias-exploiting guaranteed threshold exists for a given bias-preserving architecture, or that no improvement over the stated conservative combinatorial bounds is attainable, resolves the corresponding direction. Device-level characterization of whether actual hardware satisfies the assumed bias and error rates at the level the threshold requires is out of scope for this theorem task.

References

  1. Panos Aliferis and John Preskill, "Fault-tolerant quantum computation against biased noise", Phys. Rev. A 78, 052331 (2008). DOI: 10.1103/PhysRevA.78.052331; arXiv:0710.1301. https://arxiv.org/abs/0710.1301
  2. Panos Aliferis, Frederico Brito, David P. DiVincenzo, John Preskill, Matthias Steffen, Barbara M. Terhal, "Fault-tolerant computing with biased-noise superconducting qubits: a case study", New J. Phys. 11, 013061 (2009). DOI: 10.1088/1367-2630/11/1/013061; arXiv:0806.0383. https://arxiv.org/abs/0806.0383
  3. Dorit Aharonov, Alexei Kitaev, John Preskill, "Fault-tolerant quantum computation with long-range correlated noise", Phys. Rev. Lett. 96, 050504 (2006). DOI: 10.1103/PhysRevLett.96.050504; arXiv:quant-ph/0510231. https://arxiv.org/abs/quant-ph/0510231
  4. Barbara M. Terhal and Guido Burkard, "Fault-tolerant quantum computation for local non-Markovian noise", Phys. Rev. A 71, 012336 (2005). DOI: 10.1103/PhysRevA.71.012336; arXiv:quant-ph/0402104. https://arxiv.org/abs/quant-ph/0402104
  5. Hui Khoon Ng and John Preskill, "Fault-tolerant quantum computation versus Gaussian noise", arXiv:0810.4953 (2008). https://arxiv.org/abs/0810.4953
  6. Peter Brooks and John Preskill, "Fault-tolerant quantum computation with asymmetric Bacon-Shor codes", Phys. Rev. A 87, 032310 (2013). DOI: 10.1103/PhysRevA.87.032310; arXiv:1211.1400. https://arxiv.org/abs/1211.1400
  7. Paul Webster, Stephen D. Bartlett, David Poulin, "Reducing the overhead for quantum computation when noise is biased", arXiv:1509.05032 (2015). https://arxiv.org/abs/1509.05032
  8. David K. Tuckett, Andrew S. Darmawan, Christopher T. Chubb, Sergey Bravyi, Stephen D. Bartlett, Steven T. Flammia, "Tailoring surface codes for highly biased noise", Phys. Rev. X 9, 041031 (2019). DOI: 10.1103/PhysRevX.9.041031; arXiv:1812.08186. https://arxiv.org/abs/1812.08186
  9. David K. Tuckett, Stephen D. Bartlett, Steven T. Flammia, Benjamin J. Brown, "Fault-tolerant thresholds for the surface code in excess of 5% under biased noise", Phys. Rev. Lett. 124, 130501 (2020). DOI: 10.1103/PhysRevLett.124.130501; arXiv:1907.02554. https://arxiv.org/abs/1907.02554
  10. J. Pablo Bonilla Ataides, David K. Tuckett, Stephen D. Bartlett, Steven T. Flammia, Benjamin J. Brown, "The XZZX surface code", Nature Communications 12, 2172 (2021). DOI: 10.1038/s41467-021-22274-1; arXiv:2009.07851. https://arxiv.org/abs/2009.07851
  11. Jérémie Guillaud and Mazyar Mirrahimi, "Repetition Cat Qubits for Fault-Tolerant Quantum Computation", Phys. Rev. X 9, 041053 (2019). DOI: 10.1103/PhysRevX.9.041053; arXiv:1904.09474. https://arxiv.org/abs/1904.09474
  12. Yingkai Ouyang, "Avoiding coherent errors with rotated concatenated stabilizer codes", npj Quantum Information 7, 87 (2021). DOI: 10.1038/s41534-021-00429-8; arXiv:2010.00538. https://arxiv.org/abs/2010.00538
  13. Josu Etxezarreta Martinez, Paul Schnabl, Javier Oliva del Moral, Reza Dastbasteh, Pedro M. Crespo, Ruben M. Otxoa, "Leveraging biased noise for more efficient quantum error correction at the circuit-level with two-level qubits", arXiv:2505.17718 (2025). https://arxiv.org/abs/2505.17718
  14. César Benito, I. Jesán Velázquez-Reséndiz, Alejandro Bermúdez, "Optimizing bias-tailored quantum error correction beyond code-capacity noise", arXiv:2606.17709 (2026). https://arxiv.org/abs/2606.17709
  15. Christophe Vuillot, Diego Ruiz, Jérémie Guillaud, Mazyar Mirrahimi, "Quantum non-demolition measurements as a practical primitive for fault-tolerant computation against biased noise", arXiv:2605.24262 (2026). https://arxiv.org/abs/2605.24262
  16. Diego Ruiz, Jérémie Guillaud, Christophe Vuillot, Mazyar Mirrahimi, "Biased-noise qubits: a guide to efficient fault-tolerance using the hierarchy of errors", arXiv:2607.20143 (2026). https://arxiv.org/abs/2607.20143
  17. Jagannath Das, Sayandip Dhara, Pedro Medina, Arthur Pesah, Arpit Dua, "Clifford-deformed zero-rate LDPC codes with 50% biased noise thresholds", arXiv:2605.15348 (2026). https://arxiv.org/abs/2605.15348