Surface Codes For Generic Single-Qubit Coherent Error
Quantum Surface Codes
Quantifying the Hidden Threat: Surface Code Resists Generic Single-Qubit Coherent Errors
Large-scale quantum processing is hindered by noise, especially coherent mistakes. The surface code, the industry-leading quantum error correction architecture, was tested for noise endurance by Cambridge researchers due to these widespread defects. The maximum-likelihood error thresholds for general single-qubit coherent defects are “surprisingly high”.
Surface codes encode one or more logical qubits using a 2D lattice of physical qubits, dispersing quantum information across several qubits to prevent errors. Measurement qubits (stabilizers) periodically detect data qubit faults without deleting encoded data. Classical computation is used to determine and correct physical qubits after detecting the incorrect “syndrome”. Surface codes are a top choice for fault-tolerant quantum computers due to their high error threshold and near-neighbor interactions for most operations.
Coherent Errors' Character
In quantum computing, error-correcting codes safeguard encoded quantum information to reduce noise and gain quantum advantage. Common QEC models assume that errors behave probabilistically and that noise is stochastic or incoherent (decoherence).
However, quantum hardware exhibits unitary time evolution. Any unwanted temporal evolution, such as systematic small rotations or incorrect gate operations, causes coherent errors. Generic single-qubit coherent errors are local unitary rotations about any axis.
The unique problem for QEC is that quantum interference effects might cause coherent errors to accumulate constructively and possibly faster than incoherent errors. Previous investigations had only succeeded in analyzing the surface code's resistance against extremely particular coherent scenarios, such as axes-only rotations. Error correction against generic axis rotations has received little research.
An Important Statistical Mechanics Development
The researchers solved this problem with a smart theoretical framework, statistical mechanical mapping. This strategy creates a new classical physics model for quantum error dynamics.
The error amplitudes for local unitary rotations are complex-coupling partition functions. The classical model is a random-bond Ising model (RBIM) with complicated couplings and four-spin interactions. This version of the Ashkin-Teller model has disordered complex coupling.
The complicated couplings prevent Monte Carlo simulations and other classical numerical methods from examining this model. Researchers solved this via the transfer matrix method. The complex partition function is represented by this non-unitary (1+1)-dimensional quantum circuit.
Phases of Quantum Error Correction
The behavior of the related 1D quantum circuit can be utilized to precisely identify the two regimes that determine QEC success or failure, separated by the maximum-likelihood threshold:
The Error-Correcting Phase: Coding distance exponentially reduces logical error rate. This phase corresponds to a gapped one-dimensional quantum Hamiltonian with spontaneous transfer matrix space symmetry breaking. Interestingly, the 1D quantum states follow an entanglement area law. Low entanglement allows numerical simulation approaches like Matrix Product States (MPSs) to accurately calculate thresholds and sampling syndromes.
The non-correcting phase: The logical error rate reduces with code distance above the threshold, but slowly, following a power-law decay to a non-zero constant. This behavior is very different from incoherent errors, where the logical error rate generally exceeds the threshold. Logarithmic entanglement increase characterizes this transfer matrix space non-correcting phase. The authors interpret this phase as having quasi-long-range order, like an interacting variation of the “metallic” phase seen for simpler rotations.
Unexpected Decoder Limits and Tolerance
Mapped the error-correcting phase and determined the entanglement transition point using the standard deviation of the entanglement entropy to estimate the surface code's maximum tolerance.
Generic coherent errors have higher probability thresholds than equivalent incoherent faults estimated using the conventional Pauli twirl approximation. When comparing rotations in the direction, the coherent maximum probability threshold was much greater than the incoherent Pauli twirl threshold. This suggests that, unlike past assumptions based on incoherent models, the surface code is resilient to systematic unitary errors.
But the analysis also found vulnerabilities in current decoding standards. The minimum weight perfect matching (MWPM) decoder had far lower thresholds than the incoherent Pauli twirl thresholds and maximum-likelihood bound. This shows that current decoders cannot handle generic coherent noise.
This research provided accurate theoretical thresholds and an effective computational algorithm that could (approximately) sample complex error strings via the MPS simulation to develop new decoding algorithms that better incorporate and take advantage of the surface code's high inherent tolerance against generic single-qubit coherent errors. The flexible framework can be used to evaluate more complex non-Pauli mistakes or a wider family of quantum codes, such as the colour code.
