The Five-Qubit Code

The Shor code uses nine physical qubits to protect one logical qubit against an arbitrary single-qubit error. Can we do better? The five-qubit code, discovered by Laflamme, Miquel, Paz, and Zurek in 1996, is the answer: it is the smallest possible code that corrects any single-qubit error. No four-qubit code can.

The quantum Hamming bound

To correct any single-qubit Pauli error on $n$ qubits encoding $k$ logical qubits, the distinct error syndromes must fit into the available syndrome space. There are $3n$ single-qubit Pauli errors ( $X$ , $Y$ , $Z$ on each of $n$ qubits) plus the identity, and the code uses $n - k$ stabilizer generators giving $2^{\,n-k}$ syndromes. The quantum Hamming bound requires

(3n + 1)\,2^{k} \le 2^{n}.

For $k = 1$ this reads $3n + 1 \le 2^{\,n-1}$ . The smallest $n$ satisfying it is $n = 5$ , where $16 = 16$ holds with equality. A code meeting the bound with equality is called perfect: every syndrome corresponds to exactly one correctable error, with none wasted. The five-qubit code is the unique smallest perfect code — denoted $[[5,1,3]]$ : five physical qubits, one logical qubit, distance three.

Its stabilizers

The code is defined by four stabilizer generators. A common choice is the cyclic set

\begin{aligned} S_1 &= X\,Z\,Z\,X\,I, \\ S_2 &= I\,X\,Z\,Z\,X, \\ S_3 &= X\,I\,X\,Z\,Z, \\ S_4 &= Z\,X\,I\,X\,Z, \end{aligned}

each obtained from the previous one by a cyclic shift of the five qubit positions. The logical codewords are the simultaneous $+1$ eigenstates of all four. Because the generators mix $X$ and $Z$ on the same qubits, the code is not a CSS code (unlike Steane's, in the next lesson) — its $X$ -type and $Z$ -type checks are entangled together.

Why it matters

Optimality. It proves a hard floor: you cannot protect a qubit against arbitrary single-qubit noise with fewer than five qubits.
Density of syndromes. All $4\cdot 4 = 16$ syndromes are used — one for "no error" and one each for the $3\times 5 = 15$ single-qubit Pauli errors. The decoder is a clean bijection.
A benchmark. Its compactness makes it a favorite test bed for demonstrating full QEC cycles on early hardware with few qubits.

The trade-off

Perfection in qubit count comes at a price. The five-qubit code lacks the transversal-gate structure that makes CSS codes (Steane, surface) convenient for fault-tolerant computation: most of its logical gates require more elaborate constructions. In practice, codes are chosen for their fault-tolerant gate set and high threshold, not minimal qubit count — which is why the surface code, despite using far more qubits, dominates real devices. Still, as the provably smallest code that does the job, the five-qubit code marks the exact boundary of what quantum error correction can achieve.