# Clifford gates

**Clifford gates** are unitaries that map [[quantum-gate-pauli|Pauli operators]] to Pauli operators under conjugation: if $C$ is Clifford and $P$ is a Pauli, then $C P C^\dagger$ is also a Pauli (up to global phase). This closure property partitions quantum gates into two classes with profound consequences for simulation, error correction, and the boundary between classical and quantum advantage. Clifford-only circuits are efficiently simulatable classically; adding even one non-Clifford gate requires exponential resources.

## Single-qubit Clifford gates

- **[[quantum-gate-pauli|Pauli gates]]** (I, X, Y, Z): self-inverse; map qubits between orthogonal bases
- **[[quantum-gate-h|Hadamard (H)]]**: basis rotation; swaps $X \leftrightarrow Z$: $HXH = Z$, $HZH = X$
- **[[quantum-gate-s|S gate (phase gate)]]**: $\pi/2$ rotation around z-axis; $S = R_Z(\pi/2)$; $S^4 = I$
- **[[quantum-gate-s-dagger|S† (inverse phase gate)]]**: applies $-\pi/2$ phase; $S^\dagger S = I$
- Other single-qubit Cliffords: $(H S)^3 = I$ and its powers generate up to 24 distinct single-qubit Cliffords

## Two-qubit Clifford gates

- **[[quantum-gate-cx|CX (Controlled-NOT)]]**: controlled NOT; entangling gate; maps $X \otimes I \to X \otimes X$, $I \otimes Z \to I \otimes Z$
- **[[quantum-gate-cz|CZ]]**: controlled Z; maps $Z \otimes I \to Z \otimes Z$, $I \otimes X \to I \otimes X$
- **[[quantum-gate-swap|SWAP]]**: exchanges two qubits; $\mathrm{SWAP} = (\mathrm{CNOT})^3$ (up to circuit depth)
- **[[quantum-gate-iswap|iSWAP]]**: sometimes non-Clifford depending on normalization

## Matrix representations

Key single-qubit Clifford matrices:

$$H = \frac{1}{\sqrt{2}}\begin{pmatrix} 1 & 1 \\ 1 & -1 \end{pmatrix} \quad S = \begin{pmatrix} 1 & 0 \\ 0 & i \end{pmatrix}$$

CNOT in computational basis:

$$\mathrm{CNOT} = \begin{pmatrix} 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 1 & 0 \end{pmatrix}$$

## Conjugation and Pauli preservation

The defining property of Cliffords is how they conjugate Pauli operators. Examples:

- $H X H = Z$, $H Z H = X$ (H swaps X and Z)
- $S X S^\dagger = Y$, $S Z S^\dagger = Z$ (S maps X to Y)
- $\mathrm{CNOT} \cdot (X \otimes I) \cdot \mathrm{CNOT} = X \otimes X$
- $\mathrm{CNOT} \cdot (I \otimes Z) \cdot \mathrm{CNOT} = I \otimes Z$

In general, if $P$ is Pauli and $C$ is Clifford, then $C P C^\dagger = e^{i\phi} P'$ where $P'$ is also Pauli and $\phi$ is a global phase.

## Clifford vs non-Clifford distinction

**Non-Clifford gates** (like the [[quantum-gate-t|T gate]]) do NOT preserve the Pauli group under conjugation:

$$T X T^\dagger = \frac{1}{\sqrt{2}}(X + Y), \quad T Z T^\dagger = Z$$

The result $\frac{1}{\sqrt{2}}(X + Y)$ is not Pauli, breaking the closure property. This small addition enables universal quantum computation but requires expensive resource overhead in fault-tolerant systems.

## Bloch sphere action

Clifford gates permute the basis directions on the Bloch sphere:

- **H**: swaps x and z axes (basis rotation by 90° around y)
- **S**: leaves z unchanged, rotates x → y (phase rotation)
- **Paulis**: 180° rotations around x, y, z axes
- **CNOT**: on the second qubit, applies different rotations depending on first qubit state

Clifford circuits thus act as Bloch sphere symmetries, never creating superpositions in Pauli bases.

## Composition and group structure

The Clifford group $\mathcal{C}_n$ on $n$ qubits has size:

$$|\mathcal{C}_1| = 24 \quad (2^4 \cdot 3, \text{ single-qubit Cliffords})$$
$$|\mathcal{C}_2| = 11,520 \quad (\text{two-qubit Cliffords})$$
$$|\mathcal{C}_n| = 2^{n(n+1)} \prod_{k=1}^{n} (4^k - 1) / (4 - 1)$$

**Generation**: Any Clifford can be decomposed into H, S, and CNOT gates (generators). The group is finite but grows rapidly with qubit count.

## Uses

- **State preparation**: Clifford circuits prepare stabilizer states from computational basis; deterministic outcome
- **Quantum error correction**: Syndrome measurements and logical operations in stabilizer codes (surface codes, toric codes)
- **Stabilizer codes and error detection**: Measure stabilizer eigenvalues without collapsing encoded information
- **Circuit verification**: Clifford-only circuits are efficiently verifiable classically (useful for debugging)
- **Simulation**: Classical simulators (stabilizer tableau method) run ~1000-qubit Clifford circuits in seconds
- **Basis rotation**: Prepare for measurement in different bases (e.g., H before Z measurement gives X basis measurement)

## Clifford+T decomposition

Quantum algorithms decompose into a Clifford base layer plus non-Clifford gates (usually T). The **T-count** (number of T gates) is the primary cost metric in fault-tolerant quantum computing because:

- **Clifford gates**: nearly "free" (polynomial overhead, tableau simulation)
- **T gates**: expensive (magic state distillation required, ~100-1000 physical qubits per logical T)

Quantum compilers minimize T-count through circuit optimization. A single T gate nested in a Clifford circuit costs orders of magnitude more than the Clifford layer.

## Implementation

- **Superconducting qubits**: Hadamard ~10-50 ns; S ~<1 ns (virtual); CNOT ~20-100 ns (longer than single-qubit gates)
- **Trapped ions**: Single-qubit Cliffords ~1-5 μs via laser; CNOT slower (~100 μs)
- **Photonic**: Clifford gates via beam splitters, polarization rotators; fidelity ~99%
- **Fidelity**: Clifford gate fidelity typically 99-99.9%; errors primarily limit circuit depth before T gates

## Relations

- [[quantum-gate-pauli|Pauli gates]]: special case of Cliffords; map Paulis to Paulis trivially
- [[quantum-gate-h|Hadamard (H)]], [[quantum-gate-phase|Phase gates]] (S, S†): fundamental single-qubit Cliffords
- [[quantum-gate-cx|CX]], [[quantum-gate-cz|CZ]], [[quantum-gate-swap|SWAP]]: fundamental two-qubit Cliffords
- [[quantum-gate-t|T gate]]: non-Clifford; enables universal computation but expensive in fault tolerance
- [[quantum-gate-rotation|Rotation gates]]: S gate is $R_Z(\pi/2)$; most rotations are non-Clifford except at special angles
- [[quantum-gate-stabilizer-formalism|Stabilizer formalism]]: theoretical framework built on Clifford conjugation property
- [[quantum-gate-single-qubit|Single-qubit gates]]: Cliffords are subset of $\mathrm{SU}(2)$
- Error correction: stabilizer codes, surface codes, toric codes use Clifford stabilizer measurements