Classical Error Correction Basics

If noise flips a single one of the three copies, a straightforward majority vote recovers the original value: two of the three copies still agree, and as long as errors are rare and independent, that majority is almost certainly correct.

It is tempting to assume quantum error correction could work by the same recipe — copy the qubit's state three times, then take a majority vote at the end — but two independent facts rule this out completely.

First, the no-cloning theorem forbids copying an unknown quantum state in the first place: there is no gate implementing |ψ⟩ → |ψ⟩|ψ⟩|ψ⟩ for an arbitrary |ψ⟩, so the repetition code's very first step is already unavailable.

Second, even setting that aside, a majority vote would require directly measuring the qubits in order to compare them — and measurement collapses superposition, destroying precisely the quantum information the whole procedure was meant to protect.

Quantum error correction therefore has to solve a strictly harder problem than its classical counterpart: detect and correct errors on a qubit's state without ever directly measuring, and therefore collapsing, the information stored in it, and without ever making a literal copy of it.

That sounds close to paradoxical, but Preskill's lecture notes show it is entirely solvable: encode one logical qubit across several physical qubits in an entangled way, and measure only carefully chosen joint properties of the block, called syndromes, which reveal whether and where an error occurred without ever revealing the logical qubit's actual state — the construction taken up in the next topic.