Is the Steane code the only $ [\![7,1,3]\!] $ CSS code?

This paper claims there are 10 non-equivalent $ [\![7,1,3]\!] $ stabilizer codes. How many of these are CSS codes? Is it just the Steane code?


1 Answer 1


The CSS construction takes a classical linear $[n,k_1]$ code $C_1$ and a classical linear $[n,k_2]$ code $C_2\subset C_1$ such that $C_1$ and $C_2^\perp$ both have code distance at least $d$, and yields a quantum stabilizer code $[\![n, k_1-k_2, d]\!]$. Thus, the question can be answered by counting suitable pairs of classical codes. By the classical Hamming bound $k_1,k_2\leqslant 4$, so there are only three cases to consider.

Case one: $[7,2]+[7,1]$

A classical $[7,1]$ code consists of two code words $a$ and $b$. By a simple counting argument, we can find two distinct $i,j\in\{0,1,\dots,6\}$ such that $a_i=a_j$ and $b_i=b_j$. But then the $7$-bit string $e_{i,j}$ with $1$ on positions $i$ and $j$ and zeroes everywhere else belongs to $C_2^\perp$. However, its Hamming weight is $2$, so $C_2^\perp$ cannot have distance $3$. Therefore, there are no suitable classical codes with $k_1=2$ and $k_2=1$.

Case two: $[7,3]+[7,2]$

The above argument applies in this case, too, albeit we now have to use the fact that the code is linear. In more detail, a classical linear $[7,2]$ code $C_2$ is a subspace of $\mathbb{Z}_2^7$ spanned by two linearly independent $7$-bit strings $a$ and $b$. Once again, we can find two distinct positions $i,j\in\{0,1,\dots,6\}$ such that $a_i=a_j$ and $b_i=b_j$. But then $e_{i,j}\in C_2^\perp$, so $C_2^\perp$ has distance at most $2$ and once again we find no suitable classical codes with $k_1=3$ and $k_2=2$.

Case three: $[7,4]+[7,3]$

Let $G\in\mathbb{Z}_2^{4\times 7}$ be the generator matrix of a $[7,4]$ code $C_1$. By adding and rearranging rows we can transform the matrix into the standard form $$ G=[I_4|P]\tag1 $$ where $I_4$ is the $4\times 4$ identity matrix and $P\in\mathbb{Z}_2^{4\times 3}$. But $C_1$ has distance at least $3$, so every row of $P$ must have Hamming weight at least two. For the same reason, the rows of $P$ must be distinct. But there are only four possible $3$-bit strings that satisfy these requirements: $111$, $110$, $101$, and $011$. Therefore, up to permutations of rows and columns, the generator matrix is $$ G=\begin{bmatrix} 1 & 0 & 0 & 0 & 1 & 1 & 1\\ 0 & 1 & 0 & 0 & 1 & 1 & 0\\ 0 & 0 & 1 & 0 & 1 & 0 & 1\\ 0 & 0 & 0 & 1 & 0 & 1 & 1\\ \end{bmatrix}.\tag2 $$ This fixes the $C_1$ code. Our task is then reduced to the task of enumerating all $3$-dimensional subspaces of $C_1$ whose dual is a linear code of distance at least three. However, $C_1$ is a set of mere sixteen elements, so there are fewer than ${15 \choose 3}=455$ eligible subspaces. The exhaustive search is left as a simple coding exercise for the reader.

  • 3
    $\begingroup$ I see, this is very nice and not only do these sorts of arguments prove that the Steane code is the only $ [\![7,1,3]\!] $ CSS code but they also show that no $ [\![6,1,3]\!] $ CSS code exists (or generally no $ [\![n,1,3]\!] $ CSS code exists for $ n \leq 6 $). $\endgroup$ Mar 4 at 13:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.