Quantum error correction (QEC) protects quantum information from decoherence and gate errors by encoding a single logical qubit into multiple physical qubits. Unlike classical error correction, QEC must handle continuous errors (arbitrary rotations, not just bit flips), cannot clone the state to create redundancy, and must detect errors without measuring (and thus collapsing) the encoded data. The key insight is that syndrome measurements can project continuous errors onto a discrete set (bit-flip, phase-flip, or both) without revealing the encoded information, enabling correction by applying the appropriate Pauli operator.
Quantum computation is inherently fragile. Qubits interact with their environment (decoherence), and gate operations have finite precision. Without error correction, errors accumulate and the computation becomes useless after a small number of steps. Quantum error correction is the set of techniques that make fault-tolerant quantum computation possible — it is the bridge between the theoretical power of quantum algorithms and practical quantum hardware.
The fundamental challenge is that quantum errors are continuous: a qubit can rotate by any small angle, not just flip discretely. Classical error correction handles discrete bit flips using redundancy (copying bits), but the no-cloning theorem forbids copying an unknown quantum state. QEC solves both problems with an elegant trick: encode the logical qubit into an entangled state of multiple physical qubits, and use syndrome measurement to project continuous errors onto the discrete Pauli group {I, X, Y, Z} without learning anything about the encoded state.
The simplest example is the 3-qubit bit-flip code, which encodes |0_L> = |000> and |1_L> = |111>. A general state alpha|0_L> + beta|1_L> becomes alpha|000> + beta|111>. If a bit flip occurs on the second qubit, the state becomes alpha|010> + beta|101>. To detect this error, measure the parity of qubits 1 and 2 (are they the same or different?) and the parity of qubits 2 and 3. These parity measurements reveal which qubit flipped without measuring the encoded value — the syndrome {different, different} uniquely identifies a flip on qubit 2. Apply X to qubit 2 to correct the error. Crucially, the measurements are multi-qubit parity checks, not single-qubit measurements: they extract error information while preserving the superposition of the logical qubit.
The bit-flip code does not handle phase errors (Z maps |0> to |0> and |1> to -|1>). The 3-qubit phase-flip code handles phase errors by encoding in the Hadamard basis: |0_L> = |+++> and |1_L> = |--->. Shor's 9-qubit code concatenates the two, protecting against both bit and phase errors simultaneously. The key theoretical result is that any single-qubit error (an arbitrary 2x2 matrix, a continuous rotation) can be decomposed into Pauli components I, X, Y, Z. If a code can correct each Pauli error separately, it can correct any single-qubit error — including continuous rotations, depolarizing noise, or amplitude damping. This discretization of errors by syndrome measurement is what makes quantum error correction tractable despite the continuous nature of quantum noise.