On Euclidean self-dual codes and isometry codes (Q2668221)

This work introduces new methods and algorithms to construct Euclidean self-dual codes over large finite fields. A new algorithm to construct orthogonal matrices is presented and this algorithm is applied to construct self-dual codes over finite fields. The orthogonal group of index \(n\) over a \(\mathbb{F}_q\) is defined by \(\mathcal{O}_n(q)=\{A \in \mathrm{GL}(n, q)\mid AA^T = I_n\}.\) In order to efficiently apply this algorithm, the author searches for matrices such that each row is having as many non-zero entries as possible. The construction is very efficient for small lengths. However sometimes these short codes can be used to construct longer ones in smaller field. Under the condition of \(q\equiv1\pmod{4}\) and \(\alpha=\omega^\frac{q-1}{4}\) for some primitive element \(\omega\in\mathbb{F}_{q}\) there exist some elements \(\alpha,\beta,\gamma,\delta,\lambda,\theta\) such that \(G_n=\left( \begin{array}{c|c|c|c} \theta & \beta \ldots \beta & \alpha & \lambda(L_1+\cdots+L_n) \\ \hline \delta & & \gamma & \\ \vdots & B & \vdots & \alpha L \\ \delta & & \gamma & \\ \end{array} \right),\) where \(L\in\mathcal{O}_n(q)\) and \(L_i\) -- its \(i\)-th row, generates a self-orthogonal \([2n+2,\geq n]\) code and in case \(\delta=0\) and \(\theta\neq 0\) the code is self-dual with parameters \([2n+2,n+1].\) Denoting by \(J_n\) the \(n\times n\) all-ones matrix, here there are two cases for the matrix \(B=J_n-I_n\) and \(B=J_n+I_n.\) The inner code as a linear map with some specific properties called an isometry is studied. Dual preserving linear maps are used in order to construct self-orthogonal codes over small finite prime fields using the method of concatenation. Self-dual codes over \(\mathbb{F}_{2^3}\) and \(\mathbb{F}_{3^2}\) up to length 40 as well as more than 30 new binary optimal self-orthogonal codes, one new binary self-dual code with parameters \([88, 44, 16]\) are constructed as an application of these new methods.