|q⟩ Bad Qubits

advanced · Physics · Stabilizer Formalism & CSS Codes

Stabilizer Groups

A stabilizer group is a special kind of subgroup of the Pauli group. It is the central object of this module: a stabilizer group can pin down a quantum state, define a quantum code, and be measured to detect errors. In this lesson we give the precise definition and the two conditions every stabilizer group must satisfy.

The definition

A subgroup SPnS \subseteq \mathcal{P}_n is a stabilizer group if it satisfies two conditions:

  1. It is abelian — every pair of elements commutes.
  2. It does not contain I-I (and therefore not ±iI\pm iI either).

The second condition rules out a trivial obstruction: if IS-I \in S, then for any state ψ|\psi\rangle we would need both gψ=ψg|\psi\rangle = |\psi\rangle and Iψ=ψ-I|\psi\rangle = |\psi\rangle, i.e. ψ=ψ-|\psi\rangle = |\psi\rangle, forcing ψ=0|\psi\rangle = 0. No nonzero state could be stabilized.

A state ψ|\psi\rangle is stabilized by SS when

gψ=ψfor every gS,g\,|\psi\rangle = |\psi\rangle \quad \text{for every } g \in S,

i.e. ψ|\psi\rangle is a +1+1 eigenvector of every element of SS. The set of all such states is the codespace of SS.

Why abelian and why no I-I

These two requirements are exactly what is needed for a nontrivial common +1+1 eigenspace to exist.

When both hold, the simultaneous +1+1 eigenspace is guaranteed to be nonempty, and its dimension is controlled by the number of independent generators.

Generators and the size of a stabilizer group

A stabilizer group is finitely generated. If SS has mm independent generators g1,,gmg_1, \dots, g_m (each of order 22, so gi2=Ig_i^2 = I), then

S=g1,,gm,S=2m,S = \langle g_1, \dots, g_m \rangle, \qquad |S| = 2^m,

because every element is a product g1a1gmamg_1^{a_1}\cdots g_m^{a_m} with each ai{0,1}a_i \in \{0,1\}. The generators are independent if no product of a nonempty subset of them equals II.

On nn physical qubits, mm independent stabilizer generators cut the 2n2^n-dimensional Hilbert space down to a codespace of dimension

dim(codespace)=2nm.\dim(\text{codespace}) = 2^{\,n-m}.

Each generator imposes one binary (±1\pm 1) eigenvalue constraint and halves the dimension. When m=nm = n the codespace is one-dimensional: a single state, the subject of the next lesson.

Commutation as binary algebra

Checking the abelian condition does not require multiplying matrices. As we saw with the Pauli group, two strings commute or anticommute, and this is determined by a parity. Writing each Pauli (ignoring phase) as a binary vector (xz)(x \mid z) — with xj=1x_j = 1 for an XX or YY on qubit jj and zj=1z_j = 1 for a ZZ or YY — two operators PP and QQ commute exactly when their symplectic inner product

P,Q  =  j=1n(xjPzjQ+zjPxjQ)mod2\langle P, Q \rangle \;=\; \sum_{j=1}^{n} \big(x^P_j z^Q_j + z^P_j x^Q_j\big) \bmod 2

is zero. A generating set is abelian iff this vanishes for every pair of generators. We will build the full binary-matrix picture (the check matrix) two lessons from now.

Try it

Use the symplectic inner product to verify that the proposed generators of the three-qubit bit-flip code, g1=Z0Z1g_1 = Z_0 Z_1 and g2=Z1Z2g_2 = Z_1 Z_2, commute — so they generate a valid stabilizer group.

Run your code to see the quantum state.

Sign in on the full site to ask questions and join the discussion.