On Toeplitz codes of index \(t\) and isometry codes (Q6098075): Difference between revisions

A square matrix is called Toeplitz, when it is in the following form: \[ A=\begin{pmatrix} a & a_1 & a_2 & \cdots & a_{n-2} & a_{n-1} \\ b_1 & a & a_1 & \ddots & \ddots & a_{n-2} \\ b_2 & b_1 & a & \ddots & \ddots & \vdots \\ \vdots & \ddots & \ddots& \ddots & \ddots & a_2 \\ b_{n-2} & \ddots & \ddots & \ddots & a & a_1 \\ b_{n-1} & b_{n-2} & \cdots & b_2 & b_1 & a \end{pmatrix}; \] that is, every diagonal of \(A\) parallel to the main diagonal has constant entries. A linear code of length \(tn\) is called a Toeplitz code of index \(t\) if it has a generator matrix of the form \(G=(I_n\ A_1\ A_2\ \cdots\ A_{t-1})\), where each \(A_i\) is a Toeplitz matrix and \(I_n\) is the \(n\)-square identity matrix. In this paper, the authors first prove that Toeplitz codes of index \(t\) meet the asymptotic Gilbert-Varshamov bound and hence form an asymptotic good family of codes. Then they give methods to construct LCD or self-orthogonal codes from Toeplitz codes which are optimal or almost optimal. Recall that, supposing \(C^\perp\) denotes the (Euclidean) dual of a linear code \(C\), the hull of a linear code \(C\) is \(\mathrm{Hull}(C)=C\cap C^\perp\). If the dimension of the hull of \(C\) is zero (resp. equals the dimension of \(C\)), then \(C\) is called an LCD code (resp. self-orthogonal code). Let \(n\geq m\geq 2\). The authors show that using certain linear transformations \(\mathbf{F}_{q^m} \to \mathbf{F}_q^n\), called isometries, one can map an \([N, K]\) linear code over \(\mathbf{F}_{q^m}\) with a \(l\)-dimensional hull to an \([nN, mK]\) linear code over \(\mathbf{F}_q\) with a \(lm\)-dimensional hull. In particular, the image of LCD or self-orthogonal codes under isometries is again LCD or self-orthogonal, respectively. Then, they consider two special families of Toeplitz codes and study when these two families are LCD or self-orthogonal and apply isometries on them to construct LCD or self-orthogonal families of codes, several of which are optimal or almost optimal. Finally, they consider (systematic) quasi-cyclic codes, that is, codes with a generator matrix of the form \(G=(I_n\ A_1\ A_2\ \cdots\ A_r)\) where each \(A_i\) is a circulant matrix. Noting that a circulant matrix is a Toeplitz matrix, it follows that every quasi-cyclic code is a Toeplitz code. They present conditions on a quasi-cyclic code equivalent to being LCD or sufficient to having a 1-dimensional hull. Then they apply isometries on such codes to find some optimal or almost optimal codes which are LCD or have small hulls.