An independent eighteen-unit textbook on exact QEC conditions, quantum stabilizers, CSS and topological codes, noisy syndrome extraction, likelihood-based decoders, thresholds, LDPC, bosonic protection, leakage, logical gates, cellular-automaton recovery and research reproduction. Substantial original math, worked calculations, embedded original lectures and primary-source documentation are taught directly on QuantumSpark.
Published syllabus
Quantum error correction protects a subspace of quantum information against a specified span of physical error operators, not against an unconstrained list of informal 'bit flips'. Start with an encoding isometry from a logical Hilbert space into n physical qubits and a projector onto the resulting codespace. The distinction between detecting an error, identifying its operator representative and recovering its logical action is central: a syndrome may be sufficient for a correctable coset without revealing an individual physical fault.
The repetition construction corrects X errors only, while Hadamard conjugation exchanges X and Z. Shor's nine-qubit code combines nested repetition structures to protect arbitrary single-qubit Pauli errors. Rather than memorizing its codewords, reconstruct the stabilizer generators and logical operators, then verify the distance and each syndrome response. Explicit recovery channels reveal why measuring a syndrome need not measure a logical amplitude.
Modulo irrelevant phases, an n-qubit Pauli operator is represented by two binary vectors. The first records X support; the second records Z support; simultaneous bits describe Y up to a phase. Pauli commutation becomes a symplectic bilinear form over the binary field. This representation turns code construction into finite-dimensional linear algebra and makes machine-checkable commutativity constraints possible.
CSS quantum codes are built from two classical binary codes with compatible parity checks. The central constraint is orthogonality of X and Z check spaces, which guarantees quantum stabilizer commutation. The two syndrome types detect the opposite Pauli component, so keeping the x/z roles straight is the first essential implementation check.
A decoder receives syndrome information and must choose a recovery that restores the correct logical equivalence class. The most likely individual Pauli error need not belong to the most likely coset, because a coset may contain many moderately probable representatives. This is quantum degeneracy: stabilizer-equivalent physical errors have the same encoded effect despite different microscopic histories.
The toric code places qubits on the edges of a periodic square cellulation. Star and plaquette checks define the stabilizer group: their overlap has even parity, ensuring commutation. Z chains end on violated X stars, while X dual chains end on violated Z plaquettes. A contractible closed loop is a stabilizer product; a loop winding around the torus represents a nontrivial logical operator.
A planar surface-code patch uses boundaries to encode one or more logical degrees of freedom without periodic lattice connectivity. Rough and smooth boundary types correspond to different anyon condensation rules, so the orientation of X and Z strings and the definition of distance must be derived from a precise patch layout. A drawing with a square grid but no check schedule or boundary labels is not a complete code specification.
Under perfect measurements a stabilizer syndrome is a snapshot of Pauli errors. Under realistic repeated extraction, a single measurement fault flips one outcome and generates a pair of detection events in time, while a data error can create persistent changes in measured checks. The decoding input is therefore a space-time event history with carefully chosen initial and final temporal boundaries.
Minimum-weight perfect matching associates detector events in a graph with paths of faults; edge weights are derived from an assumed physical likelihood model. For independent graphlike error mechanisms, shortest total matching weight approximates maximum-likelihood fault inference. Under degeneracy and correlated faults, it can disagree with optimal logical-coset decoding.
A threshold is a property of a family of codes, a specified physical noise model, measurement schedule, and decoder, not an intrinsic number attached to the term 'surface code'. Below threshold a family may show decreasing logical error as code distance increases. Above threshold larger codes may fail more frequently. Real finite-distance curves have corrections and need not intersect at a single precise value.
Quantum low-density parity-check codes use stabilizer generators of bounded weight and bounded per-qubit participation as code size grows. Achieving asymptotically favorable rate and distance is an algebraic achievement, but converting a sparse abstract check graph into a two-dimensional nearest-neighbor device can introduce long connections, swap depth or modular interconnects that degrade practical overhead.
Color codes place stabilizers on suitably colored cellulations, enabling useful transversal or gauge-fixed gate structures in specific dimensions. In a subsystem code, physical degrees of freedom factor into protected logical and disposable gauge subsystems. Noncommuting gauge generators can be measured while their center supplies commuting stabilizers; this is not equivalent to pretending every gauge generator is a stabilizer.
Concatenation encodes each physical qubit of an outer code into a lower-level inner code. In idealized independent noise, one level transforms a physical fault rate into a smaller effective logical fault rate when the noise lies beneath a code-dependent threshold. Repeating levels can suppress logical failure, but overhead grows and the actual threshold depends on the exact fault-tolerant gadgets, including syndrome extraction and interactions between encoded blocks.
Bosonic QEC encodes logical states in one or more oscillators rather than a single physical two-level system per code block. Photon loss, dephasing and phase-space displacement are physically distinct errors; the appropriate code depends on the dominant jump operators and the available control and readout. Truncating an infinite Fock space is a numerical approximation requiring a tail-weight audit.
A physical quantum device rarely produces exactly symmetric Pauli noise. Dephasing-dominant channels, coherent phase drift, measurement resets, erasure flags and population leakage change decoder design. An XZZX-type surface code rotates stabilizer structure to exploit biased Pauli channels under suitable gates and measurements. The achievable improvement is conditional on bias surviving the entire syndrome cycle.
Error correction alone stores information; useful computation must perform logical gates while controlling fault propagation. A transversal gate acts across corresponding positions of code blocks and can limit spread of a single physical fault. However, no finite-dimensional exact code admits a universal set of purely transversal logical gates under standard assumptions. Gate synthesis, code deformation and resource-state injection fill the gap.
A passive self-correcting memory stores logical information without active high-bandwidth syndrome processing for a lifetime that improves as system size grows at fixed nonzero temperature. This is stronger than topological ground-state degeneracy or an active surface-code logical-error threshold. In the ordinary two-dimensional toric code, anyon pairs can be created and separated at bounded energy cost, so ordinary local thermal dynamics does not yield an arbitrarily growing passive lifetime.
A research-quality QEC result must be a falsifiable comparison in one precisely specified model. The code family, physical error channel, syndrome schedule, decoder, logical observable, baseline, hardware limits and random-sampling scheme jointly define the claim. A decoder advertised as improving surface-code accuracy under phenomenological Z noise may fail under full circuit-level correlated leakage, so both results should be reported rather than collapsing them into an attractive threshold headline.