Nondegenerate quantum error-correcting code

For a specified correctable set of distinct Pauli operators modulo phase, a nondegenerate code has mutually orthogonal error images $E\mathcal{X}$. Equivalently, the matrix in the Knill--Laflamme condition is diagonal for that Pauli error set. Each error therefore requires a separate syndrome subspace of dimension equal to the logical code dimension. Degenerate codes can have distinct errors with the same action on the code and do not satisfy this orthogonality counting argument.

For unknown-location Pauli operators of weight at most $t$, there are $N(t)={\sum }_{j=0}^t{3}^j\left(\genfrac{}{}{0ex}{}{n}{j}\right)$ physically distinct errors modulo phase, including the identity. A nondegenerate quantum code assigns them mutually orthogonal error-image subspaces, each of dimension $2^k$, within the $2^n$-dimensional physical Hilbert space. Hence $N(t)2^k\le 2^n$. The hypothesis of nondegeneracy is essential to this dimension count.

A nondegenerate quantum code is perfect at correction radius $t$ if its distinct correctable Pauli error images exhaust the physical Hilbert space. Each image has dimension $2^k$, and their orthogonal direct sum is the whole $2^n$-dimensional space precisely when the quantum Hamming bound is an equality. This perfectness concerns unknown-location errors of weight at most $t$; it does not imply correction of every higher-weight error.