Lean, informalized

Nielsen & Chuang · Chapter 10, Quantum error-correction · p. 447

Exercise 10.17 - The differences of neighbouring copies check the repetition code, with independent rows

A parity check matrix for the repetition code.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Exercise

The repetition code protects a message by copying it. Each of kk message symbols, taken from a field KK, is written down rr times, turning a message of length kk into a word of length rkrk. The exercise asks for a parity check matrix for this code: a matrix HH with linearly independent rows whose kernel is exactly the code.

The Repetition Code

Over a field KK, the generator matrix GG of the [rk,k][rk,k] repetition code, of length rkrk and encoding kk symbols, is the rk×krk \times k matrix whose rows are labelled by pairs (j,s)(j,s) of a message position j=1,…,kj = 1, \dots, k and a copy number s=1,…,rs = 1, \dots, r, whose columns are labelled by message positions ii, and whose entries are G(j,s),i=δij.G_{(j,s),\,i} = \delta_{ij}.(1)✦

The [rk,k][rk,k] repetition code CC over KK is the linear code of length rkrk, a subspace of KrkK^{rk} with coordinates labelled by the same pairs (j,s)(j,s), whose codewords are the images of the messages under GG: C={Gx:x∈Kk}.C = \{\, Gx : x \in K^k \,\}.(2)✦

With r≥1r \ge 1 copies, the parity check matrix HH of the repetition code has (r−1)k(r-1)k rows, labelled by pairs (j,t)(j,t) with j=1,…,kj = 1, \dots, k and t=1,…,r−1t = 1, \dots, r-1, and rkrk columns, labelled by the coordinates (i,s)(i,s) of the code; its entries are H(j,t),(i,s)=δij(δst−δs,t+1).H_{(j,t),\,(i,s)} = \delta_{ij}\,\bigl(\delta_{st} - \delta_{s,t+1}\bigr).(3)✦

The Claim

Theorem 1. Let KK be a field, kk a natural number and r≥1r \ge 1 a number of copies, let CC be the [rk,k][rk,k] repetition code over KK and HH its parity check matrix. The kernel of HH is the code, {y∈Krk:Hy=0}=C,\{\, y \in K^{rk} : Hy = 0 \,\} = C,(4)✦ and the rows h(j,t)h_{(j,t)} of HH are linearly independent over KK: ∑j=1k∑t=1r−1c(j,t)h(j,t)=0⇒c(j,t)=0for all (j,t).\sum_{j=1}^{k} \sum_{t=1}^{r-1} c_{(j,t)}\, h_{(j,t)} = 0 \quad\Longrightarrow\quad c_{(j,t)} = 0 \ \text{for all } (j,t).(5)✦

What Is Checked

The claim is stated over the repetition code (2), the image of the generator matrix (1), and over the matrix HH of (3).

What the library proves about these constructions is the theorem’s two statements, (4) and (5).