Decoherence and Error Channels

Quantum states are fragile: interaction with the environment causes decoherence, collapsing superpositions and entanglement. The principal error channels are:

Bit-flip channel: Applies $X$ with probability $p$:

\[\mathcal{E}_{\text{bf}}(\rho) = (1-p)\rho + p\, X\rho X\]

Phase-flip channel: Applies $Z$ with probability $p$:

\[\mathcal{E}_{\text{pf}}(\rho) = (1-p)\rho + p\, Z\rho Z\]

Depolarising channel: Applies each Pauli with probability $p/3$:

\[\mathcal{E}_{\text{dep}}(\rho) = \left(1 - p\right)\rho + \frac{p}{3}(X\rho X + Y\rho Y + Z\rho Z)\]

Amplitude damping: Models energy relaxation (spontaneous emission) from $

1\rangle$ to $

0\rangle$ with rate $\gamma$:

\[K_0 = \begin{pmatrix}1&0\\0&\sqrt{1-\gamma}\end{pmatrix}, \quad K_1 = \begin{pmatrix}0&\sqrt{\gamma}\\0&0\end{pmatrix}\] \[\mathcal{E}(\rho) = K_0\rho K_0^\dagger + K_1\rho K_1^\dagger\]

Any single-qubit error channel can be decomposed in the Pauli basis ${I, X, Y, Z}$, so correcting arbitrary Pauli errors suffices for correcting arbitrary single-qubit errors — a key linearity argument.

The No-Cloning Theorem and QEC Strategy

No-cloning theorem: There is no unitary $U$ such that $U

\psi\rangle

0\rangle =

\psi\rangle

\psi\rangle$ for all states $

\psi\rangle$.

Proof: Suppose $U

\psi\rangle

0\rangle =

\psi\rangle

\psi\rangle$ and $U

\phi\rangle

0\rangle =

\phi\rangle

\phi\rangle$. Taking inner products and using unitarity:

\[\langle\psi|\phi\rangle = \langle\psi|\phi\rangle^2\]

This forces $\langle\psi

\phi\rangle \in {0, 1}$, so only orthogonal or identical states can be cloned. General superpositions cannot.

Despite this, QEC works by encoding information in entangled states (not copies) and performing syndrome measurements — projective measurements that identify which error occurred without measuring the encoded state.

QEC conditions (Knill-Laflamme): A code with code space $\mathcal{C}$ corrects error set ${E_k}$ if and only if:

\[\langle \bar{\psi}|E_k^\dagger E_l|\bar{\phi}\rangle = c_{kl}\,\delta_{\bar{\psi}\bar{\phi}}\]

for all code words $

\bar{\psi}\rangle,

\bar{\phi}\rangle$ and some Hermitian matrix $C = (c_{kl})$.

The Three-Qubit and Shor Codes

Three-qubit bit-flip code: Encode $

\psi\rangle = \alpha

0\rangle + \beta

1\rangle$ as:

\[|\bar{0}\rangle = |000\rangle, \quad |\bar{1}\rangle = |111\rangle\]

Syndrome measurements $Z_1Z_2$ and $Z_2Z_3$ identify which qubit flipped without measuring $\alpha, \beta$:

Syndrome $(Z_1Z_2, Z_2Z_3)$	Error
$(+1, +1)$	None
$(-1, +1)$	$X_1$
$(-1, -1)$	$X_2$
$(+1, -1)$	$X_3$

This corrects single bit-flips but is blind to phase errors.

Shor code (9 physical qubits → 1 logical qubit): Concatenates bit-flip and phase-flip codes:

\[|\bar{0}\rangle = \frac{(|000\rangle + |111\rangle)^{\otimes 3}}{2\sqrt{2}}, \quad |\bar{1}\rangle = \frac{(|000\rangle - |111\rangle)^{\otimes 3}}{2\sqrt{2}}\]

The outer code corrects phase errors between the three blocks; the inner codes correct bit-flips within each block. The Shor code corrects any single-qubit error using 8 additional qubits — the first demonstration that QEC is possible in principle.

Distance: A code with distance $d$ can detect $d-1$ errors and correct $\lfloor(d-1)/2\rfloor$ errors. The Shor code has $d = 3$, correcting any 1-qubit error.

Stabiliser Formalism and CSS Codes

The stabiliser formalism provides a compact description of many important codes. A stabiliser group $\mathcal{S}$ is an Abelian subgroup of the $n$-qubit Pauli group $\mathcal{P}_n = {\pm 1, \pm i} \times {I,X,Y,Z}^{\otimes n}$ satisfying $-I \notin \mathcal{S}$.

The code space is the simultaneous $+1$ eigenspace of all stabilisers:

\[\mathcal{C}(\mathcal{S}) = \{|\psi\rangle : S|\psi\rangle = |\psi\rangle \;\forall\, S \in \mathcal{S}\}\]

For $n$ qubits and $n-k$ independent generators, the code encodes $k$ logical qubits. Syndrome measurement of generator $S_j$ yields $+1$ (no error of that type) or $-1$ (error anti-commuting with $S_j$), collapsing the system into an error subspace without disturbing the encoded state.

CSS codes (Calderbank-Shor-Steane) are constructed from two classical linear codes $C_1 \supseteq C_2$. For parity-check matrices $H_X$ and $H_Z$ with $H_X H_Z^T = 0$:

\[X\text{-stabilisers: } X^{H_X \text{ rows}}, \quad Z\text{-stabilisers: } Z^{H_Z \text{ rows}}\]

This separates bit and phase error correction, enabling transversal gates and more efficient decoding.

Surface Code

The surface code (Kitaev toric code variant) is the leading practical QEC code. On an $L \times L$ grid of $2L^2 - 1$ qubits with data qubits on vertices and syndrome qubits interspersed:

$X$-stabilisers (plaquettes): $A_p = \prod_{i \in \text{plaquette } p} X_i$

$Z$-stabilisers (vertices): $B_v = \prod_{i \in \text{star of }v} Z_i$

The logical operators are $\bar{X} = X^{\otimes L}$ (horizontal path) and $\bar{Z} = Z^{\otimes L}$ (vertical path), each with weight $L$.

Distance: $d = L$ — errors must span the grid to cause a logical failure.

Error threshold: Under independent depolarising noise, the logical error rate satisfies:

\[p_L \approx A \left(\frac{p}{p_{\text{th}}}\right)^{\lfloor d/2 \rfloor + 1}\]

The surface code threshold is $p_{\text{th}} \approx 1\%$ (via Monte Carlo simulation), meaning if physical error rates are below $1\%$, adding more qubits exponentially suppresses logical errors.

Overhead: To achieve logical error rate $\varepsilon$ with physical error rate $p < p_{\text{th}}$:

\[d \approx \frac{\log(1/\varepsilon)}{\log(p_{\text{th}}/p)}, \quad n_{\text{physical}} \approx 2d^2\]

For $\varepsilon = 10^{-12}$ and $p = 10^{-3}$: $d \approx 20$, $n_{\text{physical}} \approx 800$ qubits per logical qubit.

Fault-Tolerant Gates and Magic State Distillation

Not all gates are naturally fault-tolerant on the surface code. Transversal gates (applied independently to corresponding qubits of each code block) are automatically fault-tolerant but cannot form a universal gate set (Eastin-Knill theorem).

The surface code supports transversal Clifford gates ${H, S, \text{CNOT}}$, which generate the Clifford group. The Clifford group is not universal; adding the $T$ gate achieves universality.

Magic state distillation provides fault-tolerant $T$ gates:

1. Prepare many noisy copies of the magic state $

   T\rangle = \frac{1}{\sqrt{2}}(

   0\rangle + e^{i\pi/4}

   1\rangle)$ using noisy physical $T$ gates.

2. Use stabiliser operations (fault-tolerant) to distill a few high-fidelity magic states from many noisy ones.
3. Apply a perfect $T$ gate via gate teleportation using the magic state.

15-to-1 distillation: 15 noisy $

T\rangle$ states with error rate $p$ produce 1 state with error rate $\sim 35p^3$:

Input error $p$	Output error	Distillation rounds
$10^{-3}$	$\sim 3.5 \times 10^{-8}$	1
$10^{-3}$	$\sim 1.3 \times 10^{-22}$	2 (cascade)

The overhead for a single logical $T$ gate is 15–1000 physical qubits per logical qubit, making $T$ gates the dominant resource cost in fault-tolerant quantum computation. Reducing magic state distillation overhead is an active research area.