Two families of few-weight codes over a finite chain ring (Q6041875)

The study of codes over finite rings started in the early 1970s, with a great impulse given by the work of \textit{A. R. Hammons jun.} et al. [IEEE Trans. Inf. Theory 40, No. 2, 301--319 (1994; Zbl 0811.94039)] showing that non-linear binary codes such as Kerdock and Preparata can be seen as the image under the Gray map of linear codes over \(\mathbb{Z}_4\). Codes with few weights have found applications in secret sharing schemes, authentication codes, association schemes, among others. In this paper, the authors construct two families of few-weight codes over finite chain rings, meeting the Griesmer bound, thus giving an affirmative answer to the two related open questions raised in the 2018's article [\textit{J. Li} et al., Des. Codes Cryptography 86, No. 12, 2837--2855 (2018; Zbl 1442.94057)]: do there exist \((a)\) new linear codes over the finite chain ring \(R = \mathbb{F}_q + u \mathbb{F}_q\) \((u^2 = 0)\) meeting the Griesmer bound by a construction different from the one in [\textit{J. Li} et al., Des. Codes Cryptography 86, No. 12, 2837--2855 (2018; Zbl 1442.94057)], and \((b)\) Gray images producing linear codes with new parameters over \(\mathbb{F}_q\)? Briefly, after the Introduction and Preliminaries, in Section 3 the authors define two families of codes over the ring \(R=\mathbb{F}_q+u\mathbb{F}_q\) \((u^2=0)\), together with their corresponding codes over \(\mathbb{F}_q\) obtained under Gray images, and give the parameters (Theorem 3.1 and 3.4) and weight distributions of all of considered codes (Tables 1--4). Section 4 is devoted to the proof of these facts, which is done in great detail. Finally, in Section 5, they consider optimality of the codes considered under the Griesmer bound and showed that half of the codes over \(R\) and all the codes over \(\mathbb{F}_q\) meets the Griesmer bound (Theorems 5.3 and 5.4). In particular, this answers affirmatively the two questions (\(a\)) and (\(b\)) above, raised in [Li et al., loc. cit.]. Next we explain these facts in more detail. Let \(q=p^m\) with \(p\) prime. Consider the local ring \[ R=\mathbb{F}_q+u\mathbb{F}_q = \{\alpha+u\beta : \alpha, \beta \in \mathbb{F}_q\} \qquad (u^2=0), \] with unique maximal ideal \(M=\langle u \rangle\). Hence, \(R \simeq \mathbb{F}_q^2\) as an \(\mathbb{F}_q\)-vector space. The invertible elements of \(R\) are \(R^{*} = R \setminus M = \mathbb{F}_q^* + u \mathbb{F}_q\). Let \(Q = q^s = p^sm\) and \(R^{(s)} = \mathbb{F}_Q + u \mathbb{F}_Q\) where \((u^2= 0)\) and \(s \ge 2\). Then \(R^{(s)} / R\) is a Galois extension of \(R\) and the Galois group is \(\mathrm{Gal}(R^{(s)}/R) = \langle \sigma_q \rangle\), where \(\sigma_q\) is the \(R\)-automorphism of \(R^{(s)}\) defined by \[ \sigma_q(\alpha + u \beta) = \alpha^q + u \beta^q \] with \(\alpha,\beta \in \mathbb{F}_Q\). Then, we have the trace mapping \(\operatorname{Tr}_R^{R^{(s)}} : R^{(s)} \rightarrow R\) given by \[ \operatorname{Tr}_R^{R^{(s)}} (\alpha + u \beta) = \operatorname{Tr}_q^Q (\alpha) + u \operatorname{Tr}_q^Q (\beta) = \sum_{i=0}^{s-1} \sigma_q^i(\alpha+u\beta) \] where \(\alpha, \beta \in \mathbb{F}_Q\). If \(\mathbb{F}_q = \{ x_1 , x_2 , \ldots, x_q \}\), the Gray map over \(R\) is the map \(\psi : R \rightarrow \mathbb{F}_q^q\) defined by \[ \psi(a_0 + ua_1 ) = ( x_1 a_0 + a_1 , x_2 a_0 + a_1 , \ldots, x_q a_0 + a_1 ) \] for \(a = a_0 + ua_1 \in R\). Then \(\psi\) is an \(\mathbb{F}_q\)-linear map. For any \(n \ge 1\), one has the extended Gray map over \(R^n\), \(\Phi:R^n \rightarrow \mathbb{F}_q^{qn}\) as follows \(\Phi( v ) = (\psi( v_1 ), \ldots, \psi( v_n ))\) where \(v = ( v_1 , \ldots, v_n ) \in R^n\). Then \(\Phi\) is an \(\mathbb{F}_q\)-linear map. Under the previous notations, given the set \[ D_\varepsilon = \{ t \in R^{(s)*} : \operatorname{Tr}_R^{R^{(s)}} ( t ) = \varepsilon \} = \{ d_1 , d_2 ,\ldots, d_n \}, \] where \(\varepsilon \in \{0,1\}\), the authors define the linear code associated to \(D\) over \(R\) as \[ \mathcal{C}_{D_\varepsilon} = \big \{ c_a := ( \operatorname{Tr}_R^{R^{(s)}}( ad_1 ), \operatorname{Tr}_R^{R^{(s)}}( ad_2 ), \ldots, \operatorname{Tr}_R^{R^{(s)}}( ad_n )) \in R^n : a \in R^{( s )} \big \} \] and also the \(q\)-ary codes \(\Phi(\mathcal{C}_{D_\varepsilon})\) over \(\mathbb{F}_q\) via the Gray map. In Theorems 3.1 and 3.4, the authors show that: \begin{itemize} \item \(C_{D_0}\) is a \([\frac{Q(Q-q)}{q^2}, s-1,\frac{Q^2(q-1)}{q^3}]_R\) linear code over \(R\) with weight distribution given by \[ A_0=1, \qquad A_{\frac{Q^2(q-1)}{q^3}} = \tfrac{Q-q}q, \qquad A_{\frac{Q(q-1)(Qq+Q-q^2)}{q^4}}= \tfrac{Q(Q-q)}{q^2}. \] \item \(C_{D_1}\) is a \([\frac{Q^2}{q^2}, s,\frac{Q^2(q-1)}{q^3}]_R\) linear code over \(R\) with weight distribution given by \[ A_0=1, \qquad A_{\frac{Q^2(q-1)}{q^3}} = Q-q, \qquad A_{\frac{Q^2(q^2-1)}{q^4}} = Q(Q-q), \qquad A_{\frac{Q^2}{q^2}}= (q-1)(Q+1). \] \item \(\Phi(C_{D_0})\) is a \([\frac{Q(Q_q)}{q^2}, s-1,\frac{Q^2(q-1)}{q^3}]\) linear code over \(\mathbb{F}_q\) with weight distribution given by \[ A_0=1, \qquad A_{\frac{Q(q-1)(Q-q)}{q^2}} = \tfrac{Q(Q-q)}{q^2}, \qquad A_{\frac{Q^2(q-1)}{q^2}}= \tfrac{Q-q}{q}. \] \item \(\Phi(C_{D_1})\) is a \([\frac{Q^2}{q}, 2s,\frac{Q^2(q-1)}{q^2}]\) linear code over \(\mathbb{F}_q\) with weight distribution given by \[ A_0=1, \qquad A_{\frac{Q^2(q-1)}{q^2}} = Q^2-q, \qquad A_{\frac{Q^2}{q}}= q-1. \] \end{itemize} The authors note that compared with the known linear codes with two weights in the literature (see, e.g., [\textit{K. Ding} and \textit{C. Ding}, IEEE Trans. Inf. Theory 61, No. 11, 5835--5842 (2015; Zbl 1359.94687); \textit{Z. Heng} and \textit{Q. Yue}, Finite Fields Appl. 38, 72--92 (2016; Zbl 1338.94100); \textit{Z. Heng} et al., Des. Codes Cryptography 89, No. 8, 1993--2007 (2021; Zbl 1480.94035); Li et al., loc. cit.; \textit{G. Luo} and \textit{X. Cao}, Cryptogr. Commun. 10, No. 6, 1119--1135 (2018; Zbl 1419.94073); \textit{G. Luo} et al., Cryptogr. Commun. 10, No. 2, 301--318 (2018; Zbl 1412.94234); \textit{M. Shi} et al., IEEE Trans. Inf. Theory 63, No. 10, 6240--6246 (2017; Zbl 1390.94910); \textit{M. Shi} et al., Discrete Appl. Math. 219, 176--181 (2017; Zbl 1393.94935)], the parameters of the linear codes \(\mathcal{C}_{D_0}\), \(\Psi(\mathcal{C}_{D_0})\) and \(\Psi(\mathcal{C}_{D_1})\) are different. In Section 5, the authors recall the classical Griesmer bound \(n\ge \sum_{i=0}^{k-1} \lceil \frac{d}{q^i}\rceil\) for any \([n,k,d]\) code over \(\mathbb{F}_q\) and the Griesmer bound, due to [\textit{K. Shiromoto} and \textit{L. Storme}, Discrete Appl. Math. 128, No. 1, 263--274 (2003; Zbl 1026.94025)], for linear codes over finite commutative local rings. Namely, let \(R\) be a finite commutative local ring with maximal ideal \(M\). If \(\mathcal{C}\) is a linear code of length \(n\) over \(R\), then \[ n\ge \sum_{i=0}^{k(\mathcal{C})-1} \left\lceil\tfrac{d(\mathcal{C})}{q^i} \right\rceil \tag{1} \] where \(k(\mathcal{C})\) is the rank of the minimal free \(R\)-submodules of \(R^n\) which contain \(\mathcal{C}\) and \(R/M=\mathbb{F}_q\). If \(\mathcal{C}\) is a linear code with parameters \([n , k , d]\), and no \([ n , k , d + 1 ]\) code exists, then the code \(\mathcal{C}\) is called optimal. Thus, an \([ n , k , d ]\)-code is called Griesmer-optimal if \(\sum_{i=0}^{k-1} \lceil \frac{d+1}{q^i} \rceil >n\) In Theorems 5.3 and 5.4, the authors show that the code \(\mathcal{C}_{D_0}\) meets the Griesmer bound (1) and that the codes \(\Psi(\mathcal{C}_{D_\varepsilon})\) are Griesmer-optimal for \(\varepsilon=0,1\) (with \(s\ge q\) for \(\varepsilon=0)\).