The extended codes of a family of reversible MDS cyclic codesThanks: Z. Sun’s research was supported by The National Natural Science Foundation of China under Grant Number 62002093. C. Ding’s research was supported by the Hong Kong Research Grants Council, Proj. No. 16301123.

Zhonghua Sun Thanks: Z. Sun with the School of Mathematics, Hefei University of Technology, Hefei, 230601, Anhui, China (email: sunzhonghuas@163.com) Cunsheng Ding Thanks: C. Ding is with the Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China (email: cding@ust.hk)

Abstract

A linear code with parameters $[n,k,n-k+1]$ is called a maximum distance separable (MDS for short) code. A linear code with parameters $[n,k,n-k]$ is said to be almost maximum distance separable (AMDS for short). A linear code is said to be near maximum distance separable (NMDS for short) if both the code and its dual are AMDS. MDS codes are very important in both theory and practice. There is a classical construction of a $[q+1,2u-1,q-2u+3]$ MDS code for each $u$ with $1\leq u\leq\lfloor\frac{q+1}{2}\rfloor$ , which is a reversible and cyclic code. The objective of this paper is to study the extended codes of this family of MDS codes. Two families of MDS codes and several families of NMDS codes are obtained. The NMDS codes have applications in finite geometry, cryptography and distributed and cloud data storage systems. The weight distributions of some of the extended codes are determined.

Index Terms:

Cyclic code, extended code, linear code, locally recoverable code, MDS code, subfield code

I Introduction of motivations, objectives and methodology

By an $[n,k,d]$ code over ${\mathrm{GF}}(q)$ we mean a $k$ -dimensional linear subspace of ${\mathrm{GF}}(q)^{n}$ having minimum Hamming distance $d$ . The parameters of a linear code refer to its length, dimension and minimum distance. An $[n,k,d]$ code over ${\mathrm{GF}}(q)$ is called distance-optimal (respectively, dimension-optimal and length-optimal) if there is no $[n,k,d^{\prime}\geq d+1]$ (respectively, $[n,k^{\prime}\geq k+1,d]$ and $[n^{\prime}\leq n-1,k,d]$ ) linear code over ${\mathrm{GF}}(q)$ . An optimal code is a code that is length-optimal, or dimension-optimal, or distance-optimal, or meets a bound for linear codes.

Let $A_{i}({\mathcal{C}})$ denote the number of codewords of Hamming weight $i$ in a linear code ${\mathcal{C}}$ . The vector $(A_{0}({\mathcal{C}}),A_{1}({\mathcal{C}}),\ldots,A_{n}({\mathcal{C}}))$ is called the weight distribution of ${\mathcal{C}}$ , and the weight enumerator is defined by the polynomial $\sum_{i=0}^{n}A_{i}({\mathcal{C}})z^{i}.$ Throughout this paper, we use $\dim({\mathcal{C}})$ and $d({\mathcal{C}})$ to denote the dimension and minimum distance of a linear code ${\mathcal{C}}$ .

Denote $[n]=\{0,1,\cdots,n-1\}$ for each positive integer $n$ . We usually index the coordinates of the codewords in ${\mathcal{C}}$ with the elements in $[n]$ . An $[n,k,d]$ code ${\mathcal{C}}$ over ${\mathrm{GF}}(q)$ is called an $(n,k,d,q;r)$ -LRC (locally recoverable code) if for each $i\in[n]$ there is a subset $R_{i}\subseteq[n]\setminus\{i\}$ of size $r$ and a function $f_{i}(x_{1},x_{2},\ldots,x_{r})$ on ${\mathrm{GF}}(q)^{r}$ such that $c_{i}=f_{i}({\mathbf{c}}_{R_{i}})$ for each codeword ${\mathbf{c}}=(c_{0},c_{1},\ldots,c_{n-1})$ in ${\mathcal{C}}$ , where ${\mathbf{c}}_{R_{i}}$ is the projection of ${\mathbf{c}}$ at $R_{i}$ . The symbol $c_{i}$ is called the $i$ -th code symbol and the set $R_{i}$ is called the repair set or recovering set of the code symbol $c_{i}$ . In this definition of LRCs, the degrees of the functions $f_{i}$ are not restricted. If we require that each $f_{i}$ be a homogeneous function of degree $1$ in the definition above, then we say that ${\mathcal{C}}$ is $(n,k,d,q;r)$ -LLRC (linearly local recoverable code) and has linear locality $r$ . For general theory about minimum locality and minimum linear locality, the reader is referred to [35].

For any $(n,k,d,q;r)$ -LLRC, Gopalan et al. proved the following upper bound on the minimum distance $d$ [18]:

d\leq n-k-\left\lceil\frac{k}{r}\right\rceil+2.

(1)

The bound in (1) is similar to the Singleton bound, and is thus called the Singleton-like bound. If an $(n,k,d,q;r)$ -LLRC meets the Singleton-like bound with equality, then we say that the $(n,k,d,q;r)$ -LLRC is distance-optimal ( $d$ -optimal for short). If an $(n,k,d,q;r)$ -LLRC meets the Singleton-like bound minus one with equality, then we say that the $(n,k,d,q;r)$ -LLRC is almost distance-optimal (almost $d$ -optimal for short). Note that the Singleton-like bound is not tight for codes over small finite fields, as it is independent of the alphabet size $q$ .

For any $(n,k,d,q;r)$ -LLRC, Cadambe and Mazumdar developed the following bound on the dimension $k$ [5], [6]:

k\leq\min_{t\in\mathbb{{Z}}_{+}}[tr+k_{{\rm opt}}^{(q)}(n-t(r+1),d)],

where $\mathbb{{Z}}_{+}$ denotes the set of all positive integers, and $k_{{\rm opt}}^{(q)}(n,d)$ is the largest possible dimension of a code with length $n$ , minimum distance $d$ , and alphabet size $q$ . An $(n,k,d,q;r)$ -LLRC that attains the CM bound with equality is said to be dimension-optimal ( $k$ -optimal for short).

Any linear code with parameters $[n,k,n-k+1]$ for some positive integers $n$ and $k$ is called an MDS (maximum distance separable) code. Any linear code with parameters $[n,k,n-k]$ for some positive integers $n$ and $k$ is said to be almost maximum distance separable (AMDS for short). A linear code is said to be near maximum distance separable (NMDS for short) if both the code and its dual are almost maximum distance separable. By definition, an $[n,k]$ linear code ${\mathcal{C}}$ is NMDS if and only if $d({\mathcal{C}})+d({\mathcal{C}}^{\perp})=n$ , where $d({\mathcal{C}})$ and $d({\mathcal{C}}^{\perp})$ denote the minimum distance of ${\mathcal{C}}$ and ${\mathcal{C}}^{\perp}$ , respectively.

MDS codes are very important in theory and have important applications. For example, the Reed-Solomon codes are widely used in communication systems and data storage devices. It is easy to verify that each MDS code is a $d$ -optimal and $k$ -optimal LLRC. In theory, MDS codes over finite fields are the same as arcs in projective geometry, and some MDS codes are hrperovals and can be used to construct $t$ -designs [13, Chapter 12]. Motivated by the importance of MDS codes in both theory and practice, our first objective is to construct several families of MDS codes by extending a family of known MDS cyclic codes. Our methodology is the extending technique in coding theory.

The error-correcting capability of NMDS codes is slightly worse than that of MDS codes. However, NMDS codes have the following advantages over MDS codes:

•

Some infinite families of NMDS codes support infinite families of $t$ -designs for $t\in\{2,3,4\}$ [12, 22, 36, 40, 43], while MDS codes support only trivial designs.
•

The length $n$ of an NMDS code over ${\mathrm{GF}}(q)$ with parameters $[n,k,n-k]$ can be more than $q+1$ , while that of an MDS code with parameters $[n,k,n-k+1]$ over ${\mathrm{GF}}(q)$ is conjectured to be at most $q+1$ for $k\not\in\{3,n-3\}$ .

NMDS codes correspond also to certain objects in projective geometry [1, 2, 3, 4, 7, 8, 14, 15, 17, 32]. In addition, NMDS codes have applications in cryptography [33]. Recently, it was shown in [35] that an NMDS code is either a $d$ -optimal and $k$ -optimal LLRC or an almost $d$ -optimal and $k$ -optimal LLRC (see [28, 29] for further information). This application of NMDS codes in distributed and cloud data storage as LLRCs is the second motivation of this paper. Compared with MDS codes, there are a very small number of families of NMDS codes reported in the literature [1, 2, 3, 4, 7, 12, 14, 21, 28, 29, 36, 38, 40]. Our second objective is to construct several families of NMDS codes by extending a family of known MDS cyclic codes.

While a lot of infinite families of MDS codes over finite fields with variable parameters have been constructed in the literature, very limited results on the extended codes of MDS codes are known [23, Chapter 7], [31, Chapter 11]. Our third objective is to study the extended codes of a family of MDS cyclic codes.

The rest of this paper is organised as follows. Section II presents some auxiliary results about linear codes. Section III introduces a family of known MDS codes over ${\mathrm{GF}}(q)$ . Section IV investigates the extended codes of this family of known MDS codes over ${\mathrm{GF}}(q)$ . Section V summarizes this paper and makes concluding remarks.

II Preliminaries

II-A Some operations on linear codes

Let ${\mathcal{C}}$ be an $[n,k,d]$ code over ${\mathrm{GF}}(q)$ . The extended code $\overline{{\mathcal{C}}}$ of ${\mathcal{C}}$ is defined by

\overline{{\mathcal{C}}}=\left\{(c_{0},c_{1},\ldots,c_{n-1},c_{n})\in{\mathrm{GF}}(q)^{n+1}:(c_{0},c_{1},\ldots,c_{n-1})\in{\mathcal{C}}\mbox{ with }\sum_{i=0}^{n}c_{i}=0\right\}.

Let $\mathbf{H}$ and $\overline{\mathbf{H}}$ denote the check matrix of ${\mathcal{C}}$ and $\overline{{\mathcal{C}}}$ , respectively. Then we have the following known result whose proof is straightforward.

Lemma 1.

[23] Let ${\mathcal{C}}$ be an $[n,k,d]$ code over ${\mathrm{GF}}(q)$ . Then $\overline{{\mathcal{C}}}$ is an $[n+1,k,\overline{d}]$ code over ${\mathrm{GF}}(q)$ , where $\overline{d}=d$ or $d+1$ . In the binary case, $\overline{d}=d$ if $d$ is even, and $\overline{d}=d+1$ otherwise. In addition, the check matrix $\overline{\mathbf{H}}$ of $\overline{{\mathcal{C}}}$ can be deduced from that of ${\mathcal{C}}$ by

\displaystyle\overline{\mathbf{H}}=\begin{bmatrix}{\mathbf{1}}&1\\ \mathbf{H}&{\mathbf{0}}^{T}\end{bmatrix},

where ${\mathbf{1}}$ is the all-one vector of length $n$ and ${\mathbf{0}}$ is the all-zero vector of length $n$ .

The all-one vector ${\mathbf{1}}$ in ${\mathrm{GF}}(q)^{n}$ and the rows of a generator matrix of ${\mathcal{C}}$ span a linear subspace $\widetilde{{\mathcal{C}}}$ of ${\mathrm{GF}}(q)^{n}$ , which is called the augmented code of ${\mathcal{C}}$ . If ${\mathbf{1}}\in{\mathcal{C}}$ , then $\widetilde{{\mathcal{C}}}={\mathcal{C}}$ . Otherwise, ${\mathcal{C}}$ is a subcode of $\widetilde{{\mathcal{C}}}$ and $\dim(\widetilde{{\mathcal{C}}})=\dim({\mathcal{C}})+1$ . The following lemma will be useful.

Lemma 2.

Let $q$ be a power of a prime $p$ , $n$ be a positive integer with $n\equiv-1\pmod{p}$ . Let ${\mathcal{C}}$ be a linear code of length $n$ over ${\mathrm{GF}}(q)$ . Then $\overline{\widetilde{{\mathcal{C}}}}=\widetilde{\overline{{\mathcal{C}}}}$ .

Proof.

Let ${\mathbf{c}}=(c_{0},c_{1},\ldots,c_{n-1})\in{\mathcal{C}}$ . By definition, $\overline{{\mathcal{C}}}=\left\{\left({\mathbf{c}},-{\mathbf{c}}{\mathbf{1}}^{T}\right):{\mathbf{c}}\in{\mathcal{C}}\right\}$ , and

\displaystyle\widetilde{\overline{{\mathcal{C}}}}=\left\{\left(a{\mathbf{1}}+{\mathbf{c}},a-{\mathbf{c}}{\mathbf{1}}^{T}\right):a\in{\mathrm{GF}}(q),\,{\mathbf{c}}\in{\mathcal{C}}\right\},

(2)

where ${\mathbf{1}}$ is the all-one vector of length $n$ . Similarly, we have

\displaystyle\widetilde{{\mathcal{C}}}=\{a{\mathbf{1}}+{\mathbf{c}}:a\in{\mathrm{GF}}(q),\,{\mathbf{c}}\in{\mathcal{C}}\}

and

\displaystyle\overline{\widetilde{{\mathcal{C}}}}=\left\{\left(a{\mathbf{1}}+{\mathbf{c}},-an-{\mathbf{c}}{\mathbf{1}}^{T}\right):a\in{\mathrm{GF}}(q),\,{\mathbf{c}}\in{\mathcal{C}}\right\}.

Note that

\left(a-{\mathbf{c}}{\mathbf{1}}^{T}\right)-\left(-an-{\mathbf{c}}{\mathbf{1}}^{T}\right)=(n+1)a=0

for each $a\in{\mathrm{GF}}(q)$ . The desired result follows. ∎

The following lemma will also be needed later.

Lemma 3.

Let ${\mathcal{C}}$ be a linear code of length $n$ over ${\mathrm{GF}}(q)$ and the all-one vector ${\mathbf{1}}\in{\mathcal{C}}$ . Then the following hold.

1.

$(\overline{{\mathcal{C}}})^{\perp}=\widetilde{\overline{{\mathcal{C}}^{\perp}}}$ .
2.

If ${\mathbf{1}}\notin{\mathcal{C}}^{\perp}$ and ${\mathcal{C}}$ is an MDS code, then $d((\overline{{\mathcal{C}}})^{\perp})=d(\widetilde{{\mathcal{C}}^{\perp}})+1$ .

Proof.

1) Let $\mathbf{H}$ denote the check matrix of ${\mathcal{C}}$ . Since ${\mathbf{1}}\in{\mathcal{C}}$ , the sum of all the coordinates in each codewords of ${\mathcal{C}}^{\perp}$ is zero. As a result, we conclude that the code $\overline{{\mathcal{C}}^{\perp}}$ has generator matrix $\begin{bmatrix}\mathbf{H}&{\mathbf{0}}^{T}\end{bmatrix}$ , where ${\mathbf{0}}$ is the all-zero vector. The desired conclusion follows from Lemma 1.

2) By Result 1 and Equation (2), we deduce that

\displaystyle(\overline{{\mathcal{C}}})^{\perp}=\left\{\left(a{\mathbf{1}}+{\mathbf{c}},a\right):a\in{\mathrm{GF}}(q),\,{\mathbf{c}}\in{\mathcal{C}}^{\perp}\right\}.

(3)

Suppose ${\mathcal{C}}$ is an $[n,k,n-k+1]$ MDS code over ${\mathrm{GF}}(q)$ , then ${\mathcal{C}}^{\perp}$ is an $[n,n-k,k+1]$ MDS code over ${\mathrm{GF}}(q)$ . Since ${\mathbf{1}}\notin{\mathcal{C}}^{\perp}$ , we deduce that $\widetilde{{\mathcal{C}}^{\perp}}$ is an $[n,n-k+1]$ code over ${\mathrm{GF}}(q)$ . By the Singleton bound, $d(\widetilde{{\mathcal{C}}^{\perp}})\leq k=d({\mathcal{C}}^{\perp})-1$ . It follows that there exist $a\in{\mathrm{GF}}(q)\backslash\{0\}$ and ${\mathbf{c}}\in{\mathcal{C}}^{\perp}$ such that ${\mathtt{wt}}(a{\mathbf{1}}+{\mathbf{c}})=d(\widetilde{{\mathcal{C}}^{\perp}})$ . Note that $\left(a{\mathbf{1}}+{\mathbf{c}},a\right)\in(\overline{{\mathcal{C}}})^{\perp}$ , we get that

\displaystyle d((\overline{{\mathcal{C}}})^{\perp})\leq d(\widetilde{{\mathcal{C}}^{\perp}})+1.

(4)

On the other hand, let $\overline{{\mathbf{c}}}\in(\overline{{\mathcal{C}}})^{\perp}$ and $\overline{{\mathbf{c}}}\neq{\mathbf{0}}$ . By Equation (3), there exist ${\mathbf{c}}\in{\mathcal{C}}^{\perp}$ and $a\in{\mathrm{GF}}(q)$ such that $\overline{{\mathbf{c}}}=\left(a{\mathbf{1}}+{\mathbf{c}},a\right)$ . If $a=0$ , it follows from $\overline{{\mathbf{c}}}\neq{\mathbf{0}}$ that ${\mathbf{c}}\neq{\mathbf{0}}$ . Consequently,

\displaystyle{\mathtt{wt}}(\overline{{\mathbf{c}}})

\displaystyle={\mathtt{wt}}({\mathbf{c}})\geq d({\mathcal{C}}^{\perp})\geq d(\widetilde{{\mathcal{C}}^{\perp}})+1.

If $a\neq 0$ , ${\mathtt{wt}}(\overline{{\mathbf{c}}})=1+{\mathtt{wt}}(a{\mathbf{1}}+{\mathbf{c}})$ . It follows from ${\mathbf{1}}\notin{\mathcal{C}}^{\perp}$ that $a{\mathbf{1}}+{\mathbf{c}}\neq{\mathbf{0}}$ . Consequently,

{\mathtt{wt}}(\overline{{\mathbf{c}}})\geq d(\widetilde{{\mathcal{C}}^{\perp}})+1.

In summary,

\displaystyle d((\overline{{\mathcal{C}}})^{\perp})\geq d(\widetilde{{\mathcal{C}}^{\perp}})+1.

(5)

Combining Inequalities (4) and (5), we deduce that $d((\overline{{\mathcal{C}}})^{\perp})=d(\widetilde{{\mathcal{C}}^{\perp}})+1$ . This completes the proof. ∎

When ${\mathcal{C}}$ is a cyclic code of length $n$ over ${\mathrm{GF}}(q)$ , the augmented code of ${\mathcal{C}}$ has the following result.

Lemma 4.

Let ${\mathcal{C}}$ be the cyclic code of length $n$ over ${\mathrm{GF}}(q)$ with generator polynomial $g(x)$ . Then $\widetilde{{\mathcal{C}}}$ is the cyclic code of length $n$ over ${\mathrm{GF}}(q)$ with generator polynomial $\gcd(\frac{x^{n}-1}{x-1},g(x))$ .

Proof.

Let ${\mathcal{C}}^{\prime}$ be the cyclic code of length $n$ over ${\mathrm{GF}}(q)$ with generator polynomial $\frac{x^{n}-1}{x-1}$ , then

\widetilde{{\mathcal{C}}}={\mathcal{C}}^{\prime}+{\mathcal{C}}=\{{\mathbf{c}}_{1}+{\mathbf{c}}_{2}:{\mathbf{c}}_{1}\in{\mathcal{C}}^{\prime},{\mathbf{c}}_{2}\in{\mathcal{C}}\}.

The desired result follows from [23, Theorem 4.3.7]. ∎

II-B Cyclic codes and BCH codes over finite fields

An $[n,k,d]$ code ${\mathcal{C}}$ over ${\mathrm{GF}}(q)$ is called cyclic if $(c_{0},c_{1},\ldots,c_{n-1})\in{\mathcal{C}}$ implies

(c_{n-1},c_{0},c_{1},\ldots,c_{n-2})\in{\mathcal{C}}.

We identify any vector $(c_{0},c_{1},\ldots,c_{n-1})\in{\mathrm{GF}}(q)^{n}$ with the polynomial

c(x)=\sum_{i=0}^{n-1}c_{i}x^{i}\in{\mathrm{GF}}(q)[x]/(x^{n}-1).

Then a code ${\mathcal{C}}$ of length $n$ over ${\mathrm{GF}}(q)$ corresponds to a subset ${\mathcal{C}}(x)$ of the ring ${\mathrm{GF}}(q)[x]/(x^{n}-1)$ , where

{\mathcal{C}}(x):=\left\{\sum_{i=0}^{n-1}c_{i}x^{i}:{\mathbf{c}}=(c_{0},c_{1},\ldots,c_{n-1})\in{\mathcal{C}}\right\}.

A linear code ${\mathcal{C}}$ is cyclic if and only if ${\mathcal{C}}(x)$ is an ideal of the ring ${\mathrm{GF}}(q)[x]/(x^{n}-1)$ .

Note that every ideal of ${\mathrm{GF}}(q)[x]/(x^{n}-1)$ must be principal. Let ${\mathcal{C}}=(g(x))$ be a cyclic code of length $n$ over ${\mathrm{GF}}(q)$ , where $g(x)$ is monic and has the smallest degree among all the generators of ${\mathcal{C}}$ . Then $g(x)$ is unique and called the generator polynomial, and $h(x)=(x^{n}-1)/g(x)$ is referred to as the check polynomial of ${\mathcal{C}}$ .

Let $n$ be a positive integer with $\gcd(n,q)=1$ . Let $\mathbb{{Z}}_{n}$ denote the ring of integers modulo $n$ . For any integer $s$ with $0\leq s<n$ , the $q$ -cyclotomic coset of $s$ modulo $n$ is defined by

C_{s}^{(q,n)}=\{s,sq,sq^{2},\cdots,sq^{\ell_{s}-1}\}\bmod n\subseteq\mathbb{{Z}}_{n},

where $\ell_{s}$ is the smallest positive integer such that $s\equiv sq^{\ell_{s}}\pmod{n}$ , and is the size of the $q$ -cyclotomic coset. The smallest integer in $C_{s}^{(q,n)}$ is called the coset leader of $C_{s}^{(q,n)}$ . Let $\Gamma_{(n,q)}$ be the set of all the coset leaders. It is easily seen that $C_{s}^{(q,n)}\cap C_{t}^{(q,n)}=\emptyset$ for any two distinct elements $s$ and $t$ in $\Gamma_{(n,q)}$ , and

\displaystyle\bigcup_{s\in\Gamma_{(n,q)}}C_{s}^{(q,n)}=\mathbb{{Z}}_{n}.

Hence, the distinct $q$ -cyclotomic cosets modulo $n$ partition $\mathbb{{Z}}_{n}$ .

Let $m={\mathrm{ord}}_{n}(q)$ be the order of $q$ modulo $n$ , and let $\alpha$ be a generator of the group ${\mathrm{GF}}(q^{m})^{*}$ . Put $\beta=\alpha^{(q^{m}-1)/n}$ , then $\beta$ must be a primitive $n$ -th root of unity in ${\mathrm{GF}}(q^{m})$ . The minimal polynomial ${\mathsf{M}}_{\beta^{s}}(x)$ of $\beta^{s}$ over ${\mathrm{GF}}(q)$ is the monic polynomial of the smallest degree over ${\mathrm{GF}}(q)$ with $\beta^{s}$ as a root and is given by

\displaystyle{\mathsf{M}}_{\beta^{s}}(x)=\prod_{i\in C_{s}^{(q,n)}}(x-\beta^{i})\in{\mathrm{GF}}(q)[x],

which is irreducible over ${\mathrm{GF}}(q)$ .

Let $\delta$ be an integer with $2\leq\delta\leq n$ and let $h$ be an integer. A BCH code over ${\mathrm{GF}}(q)$ with length $n$ and designed distance $\delta$ , denoted by ${\mathcal{C}}_{(q,n,\delta,h)}$ , is a cyclic code with generator polynomial

\displaystyle{\mathrm{lcm}}({\mathsf{M}}_{\beta^{h}}(x),{\mathsf{M}}_{\beta^{h+1}}(x),\cdots,{\mathsf{M}}_{\beta^{h+\delta-2}}(x)).

(6)

If $h=1$ , the code ${\mathcal{C}}_{(q,n,\delta,h)}$ with the generator polynomial in (6) is called a narrow-sense BCH code. If $n=q^{m}-1$ , then ${\mathcal{C}}_{(q,n,\delta,h)}$ is referred to as a primitive BCH code.

BCH codes are a subclass of cyclic codes with interesting properties and applications. In many cases BCH codes are the best linear codes. For example, among all binary cyclic codes of odd lengths $n$ with $n\leq 125$ the best cyclic code is always a BCH code except for two special cases [10]. Reed-Solomon codes are also BCH codes and are widely used in communication devices and consumer electronics. In the past decade, a lot of progress on the study of BCH codes has been made (see, for example, [25, 27, 30, 34, 42, 44]).

A cyclic code ${\mathcal{C}}$ is said to be reversible if ${\mathcal{C}}\cap{\mathcal{C}}^{\perp}=\{{\mathbf{0}}\}$ . Such code is said to be linear complementary dual (LCD) code. The family of MDS codes presented in Section III are cyclic BCH codes and reversible.

II-C The weight distributions of MDS codes and NMDS codes

The weight distribution of MDS codes is known and given by the following lemma [31, p. 321]. We will use it later.

Lemma 5.

Let ${\mathcal{C}}$ be an $[n,k,d]$ code over ${\mathrm{GF}}(q)$ with $d=n-k+1$ , and let the weight enumerator of ${\mathcal{C}}$ be $1+\sum_{i=d}^{n}A_{i}z^{i}$ . Then

A_{i}=\binom{n}{i}(q-1)\sum_{j=0}^{i-d}(-1)^{j}\binom{i-1}{j}q^{i-j-d}

for all $d\leq i\leq n$ .

We have the following weight distribution formulas for NMDS codes.

Lemma 6.

[15] Let ${\mathcal{C}}$ be an $[n,k,n-k]$ NMDS code. Then the weight distributions of ${\mathcal{C}}$ and ${\mathcal{C}}^{\perp}$ are given by

\displaystyle A_{n-k+s}=\binom{n}{k-s}\sum_{j=0}^{s-1}(-1)^{j}\binom{n-k+s}{j}(q^{s-j}-1)+(-1)^{s}\binom{k}{s}A_{n-k}

(7)

for $s\in\{1,2,\cdots,k\}$ , and

\displaystyle A_{k+s}^{\perp}=\binom{n}{k+s}\sum_{j=0}^{s-1}(-1)^{j}\binom{k+s}{j}(q^{s-j}-1)+(-1)^{s}\binom{n-k}{s}A_{k}^{\perp}

(8)

for $s\in\{1,2,\cdots,n-k\}$ .

By definition, we have $\sum_{i=0}^{n}A_{i}=q^{k}$ and $\sum_{i=0}^{n}A_{i}^{\perp}=q^{n-k}$ . The weight distribution of an NMDS code ${\mathcal{C}}$ or its dual cannot be determined by Equations (7) and (8). It was shown in [12] that two $[n,k,n-k]$ NMDS codes over ${\mathrm{GF}}(q)$ could have different weight distributions. Thus, the weight distribution of an $[n,k,n-k]$ NMDS code over ${\mathrm{GF}}(q)$ depends on not only $n$ , $k$ and $q$ , but also some other parameters of the code. This is a major difference between MDS codes and NMDS codes. Actually, it could be very difficult to determine the weight distribution of an NMDS code. This difficulty will be justified in this paper later.

III The family of classical MDS cyclic codes

Starting from now on, let $q=p^{m}>2$ , where $p$ is a prime and $m$ is a positive integer. Let $\alpha$ be a primitive element of ${\mathrm{GF}}(q^{2})$ . Define $\beta=\alpha^{q-1}$ , then $\beta$ is a primitive $(q+1)$ -th root of unity in ${\mathrm{GF}}(q^{2})$ . Let $\mathbb{M}_{\beta^{i}}(x)$ denote the minimal polynomial of $\beta^{i}$ over ${\mathrm{GF}}(q)$ . It is easily verified that

x^{q+1}-1=\prod_{i=0}^{\lfloor\frac{q+1}{2}\rfloor}\mathbb{M}_{\beta^{i}}(x),

where $\mathbb{M}_{\beta^{0}}(x)=x-1$ ,

	$\displaystyle\mathbb{M}_{\beta^{i}}(x)$	$\displaystyle=(x-\beta^{i})(x-\beta^{q+1-i})$
		$\displaystyle=(x-\beta^{i})(x-\beta^{-i})$

for all $1\leq i\leq\frac{q}{2}$ , and

\displaystyle\mathbb{M}_{\beta^{\frac{q+1}{2}}}(x)

\displaystyle=x-\beta^{\frac{q+1}{2}}=x+1

for $q$ being odd.

For each $u$ with $1\leq u\leq\lfloor\frac{q+1}{2}\rfloor$ , define

\displaystyle g_{u}(x)=\mathbb{M}_{\beta^{u}}(x)\mathbb{M}_{\beta^{u+1}}(x)\cdots\mathbb{M}_{\beta^{\lfloor\frac{q+1}{2}\rfloor}}(x).

Let ${\mathcal{C}}_{u}$ be the cyclic code of length $q+1$ over ${\mathrm{GF}}(q)$ with generator polynomial $g_{u}(x)$ . We have the following result.

Theorem 7.

[31, Chapter 11] [9] For each $u$ with $1\leq u\leq\lfloor\frac{q+1}{2}\rfloor$ , ${\mathcal{C}}_{u}$ is a $[q+1,2u-1,q-2u+3]$ MDS cyclic code and $({\mathcal{C}}_{u})^{\perp}$ is a $[q+1,q+2-2u,2u]$ MDS cyclic code. Both codes are reversible.

The codes ${\mathcal{C}}_{u}$ and $({\mathcal{C}}_{u})^{\perp}$ are called the classical MDS cyclic codes. Reed-Solomon codes over ${\mathrm{GF}}(q)$ are MDS codes over ${\mathrm{GF}}(q)$ with length $q-1$ . Hence, the codes ${\mathcal{C}}_{u}$ and $({\mathcal{C}}_{u})^{\perp}$ in Theorem 7 are not Reed-Solomon codes. The objective of this paper is to study the extended codes of some of the MDS codes ${\mathcal{C}}_{u}$ .

IV The extended codes of the family of MDS cyclic codes

The objective of this section is to study the extended codes $\overline{{\mathcal{C}}_{u}}$ of the family of MDS codes ${\mathcal{C}}_{u}$ introduced in Section III and their duals. As will be seen later, we can settle the parameters of the extended code $\overline{{\mathcal{C}}_{u}}$ only for a few values of $u$ .

IV-A Some general results of the extended code $\overline{{\mathcal{C}}_{u}}$ and its dual

By Lemma 1, the extended code $\overline{{\mathcal{C}}_{u}}$ has length $q+2$ , dimension $2u-1$ and

d(\overline{{\mathcal{C}}_{u}})=d({\mathcal{C}}_{u})\mbox{ or }d(\overline{{\mathcal{C}}_{u}})=d({\mathcal{C}}_{u})+1.

As will be seen later, both cases can happen.

In the case that $d(\overline{{\mathcal{C}}_{u}})=d({\mathcal{C}}_{u})+1$ , $\overline{{\mathcal{C}}_{u}}$ is an MDS code and the parameters and weight distributions of $\overline{{\mathcal{C}}_{u}}$ are easily obtained from Lemma 5. A trivial verification shows that if $u=1$ , then $\overline{{\mathcal{C}}}_{u}$ is a $[q+2,1,q+2]$ MDS code over ${\mathrm{GF}}(q)$ . It will be proved later that $\overline{{\mathcal{C}}_{2}}$ and $\overline{{\mathcal{C}}_{q/2}}$ are MDS codes for even $q$ . The classical MDS code conjecture leads to the conjecture that $\overline{{\mathcal{C}}_{u}}$ is an AMDS code for the other cases.

In the case that $d(\overline{{\mathcal{C}}_{u}})=d({\mathcal{C}}_{u})$ , $\overline{{\mathcal{C}}_{u}}$ is a $[q+2,2u-1,q+3-2u]$ AMDS code over ${\mathrm{GF}}(q)$ , but its weight distribution cannot be derived from Lemma 6. As will be seen later, settling the weight distribution of $\overline{{\mathcal{C}}_{u}}$ is very difficult in general. In this case, we have

\displaystyle d((\overline{{\mathcal{C}}_{u}})^{\perp})\leq 2u-1.

If $d((\overline{{\mathcal{C}}_{u}})^{\perp})=2u-1$ , then $\overline{{\mathcal{C}}_{u}}$ is an NMDS code. We are much more interested in this case. As will be seen later, in some cases we indeed have $d((\overline{{\mathcal{C}}_{u}})^{\perp})<2u-1$ . The most difficult problem in this project is to determine $d((\overline{{\mathcal{C}}_{u}})^{\perp})$ . The following theorem may be useful in determining $d((\overline{{\mathcal{C}}_{u}})^{\perp})$ .

Theorem 8.

Let notation be the same as before. Let $2\leq u\leq\lfloor\frac{q+1}{2}\rfloor$ . Then

\displaystyle d((\overline{{\mathcal{C}}_{u}})^{\perp})=d({\mathcal{C}}_{(q,q+1,u,1)})+1.

Proof.

Recall that ${\mathcal{C}}_{u}$ has generator polynomial

g_{u}(x)=\mathbb{M}_{\beta^{u}}(x)\mathbb{M}_{\beta^{u+1}}(x)\cdots\mathbb{M}_{\beta^{\lfloor\frac{q+1}{2}\rfloor}}(x).

It is easily verified that $({\mathcal{C}}_{u})^{\perp}$ has generator polynomial

h_{u}(x)=(x-1)\mathbb{M}_{\beta}(x)\cdots\mathbb{M}_{\beta^{u-1}}(x).

It is easily seen that $g_{u}(x)\mid(\frac{x^{q+1}-1}{x-1})$ and $h_{u}(x)\nmid(\frac{x^{q+1}-1}{x-1})$ , then ${\mathbf{1}}\in{\mathcal{C}}_{u}$ and ${\mathbf{1}}\notin({\mathcal{C}}_{u})^{\perp}$ . By Result 2 of Lemma 3, we deduce that $d((\overline{{\mathcal{C}}_{u}})^{\perp})=d(\widetilde{{\mathcal{C}}_{u}^{\perp}})+1$ . It is easily seen that

\displaystyle\gcd\left(\frac{x^{n}-1}{x-1},h_{u}(x)\right)=\mathbb{M}_{\beta}(x)\mathbb{M}_{\beta^{2}}(x)\cdots\mathbb{M}_{\beta^{u-1}}(x).

(9)

It follows from Lemma 4 that $\widetilde{{\mathcal{C}}_{u}^{\perp}}$ is the narrow-sense BCH code of length $q+1$ over ${\mathrm{GF}}(q)$ with designed distance $u$ . The desired result follows. ∎

It follows from Equation (9) that $\left(\widetilde{{\mathcal{C}}_{u}^{\perp}}\right)^{\perp}$ is the cyclic code of length $q+1$ over ${\mathrm{GF}}(q)$ with generator polynomial $(x-1)g_{u}(x)$ , which is subcode of ${\mathcal{C}}_{u}$ . Specifically,

\displaystyle{\mathcal{C}}_{u}=\widetilde{\left(\widetilde{{\mathcal{C}}_{u}^{\perp}}\right)^{\perp}}.

(10)

To simplify notation, we will use ${\mathcal{C}}(u)$ to denote the code $\left(\widetilde{{\mathcal{C}}_{u}^{\perp}}\right)^{\perp}$ . With this new notation, Equation (10) becomes ${\mathcal{C}}_{u}=\widetilde{{\mathcal{C}}(u)}$ . Note that ${\mathcal{C}}(u)\subseteq{\mathcal{C}}_{u}$ , we get that $d({\mathcal{C}}_{u})\leq d({\mathcal{C}}(u))$ . Hence, ${\mathcal{C}}(u)$ is a $[q+1,2u-2,d]$ code over ${\mathrm{GF}}(q)$ , where $d\in\{q-2u+3,q-2u+4\}$ . Furthermore, we have the following result.

Theorem 9.

Let notation be the same as before. Let $2\leq u\leq\lfloor\frac{q+1}{2}\rfloor$ . Then $d(\overline{{\mathcal{C}}_{u}})=d({\mathcal{C}}(u))$ and $A_{q-2u+3}(\overline{{\mathcal{C}}_{u}})=A_{q-2u+3}({\mathcal{C}}(u))$ .

Proof.

Let ${\mathrm{Tr}}(x)$ denote the trace function from ${\mathrm{GF}}(q^{2})$ to ${\mathrm{GF}}(q)$ . By Delsarte’s theorem,

{\mathcal{C}}(u)=\left\{{\mathbf{c}}(a_{1},a_{2},\ldots,a_{u-1}):a_{i}\in{\mathrm{GF}}(q^{2})\right\},

and ${\mathcal{C}}_{u}=\left\{a{\mathbf{1}}+{\mathbf{c}}(a_{1},a_{2},\ldots,a_{u-1}):a\in{\mathrm{GF}}(q),a_{i}\in{\mathrm{GF}}(q^{2})\right\}$ , where

{\mathbf{c}}(a_{1},a_{2},\ldots,a_{u-1})=\left({\mathrm{Tr}}\left(\sum_{i=1}^{u-1}a_{i}\beta^{ij}\right)\right)_{j=0}^{q}.

For any $i$ with $1\leq i\leq u-1$ , $\beta^{i}\neq 1$ . Then $\sum_{j=0}^{q}\beta^{ij}=0$ . It is easily verified that

	$\displaystyle(a{\mathbf{1}}+{\mathbf{c}}(a_{1},a_{2},\ldots,a_{u-1})){\mathbf{1}}^{T}=$	$\displaystyle\penalty\ a+\sum_{j=0}^{q}{\mathrm{Tr}}\left(\sum_{i=1}^{u-1}a_{i}\beta^{ij}\right)$
	$\displaystyle=$	$\displaystyle\penalty\ a+{\mathrm{Tr}}\left(\sum_{i=1}^{u-1}a_{i}\sum_{j=0}^{q}\beta^{ij}\right)$
	$\displaystyle=$	$\displaystyle\penalty\ a.$

Therefore,

\overline{{\mathcal{C}}_{u}}=\{(a{\mathbf{1}}+{\mathbf{c}}(a_{1},a_{2},\ldots,a_{u-1}),-a):a\in{\mathrm{GF}}(q),\ a_{i}\in{\mathrm{GF}}(q^{2})\}.

Since $d({\mathcal{C}}_{u})=q-2u+3$ , we deduce that

{\mathtt{wt}}((a{\mathbf{1}}+{\mathbf{c}}(a_{1},a_{2},\ldots,a_{u-1}),-a))=q-2u+3

if and only if $a=0$ and ${\mathtt{wt}}({\mathbf{c}}(a_{1},a_{2},\ldots,a_{u-1}))=q-2u+3$ . As a result,

A_{q-2u+3}(\overline{{\mathcal{C}}_{u}})=A_{q-2u+3}({\mathcal{C}}(u)).

(11)

Notice that

d(\overline{{\mathcal{C}}_{u}}),d({\mathcal{C}}(u))\in\{q-2u+3,q-2u+4\},

by Equation (11), we deduce that $d(\overline{{\mathcal{C}}_{u}})=d({\mathcal{C}}(u))$ . This completes the proof. ∎

Theorems 8 and 9 show that ${\mathcal{C}}_{u}$ , ${\mathcal{C}}_{(q,q+1,u,1)}$ , $\overline{{\mathcal{C}}_{u}}$ and their duals have the following relationship:

\displaystyle\begin{matrix}{\mathcal{C}}_{u}:\penalty\ [q+1,2u-1,q-2u+3]&\stackrel{{\scriptstyle{\rm Dual}}}{{\longleftrightarrow}}&({\mathcal{C}}_{u})^{\perp}:\penalty\ [q+1,q-2u+2,2u]\\ \bigcup&&\bigcap\\ {\mathcal{C}}(u):\penalty\ [q+1,2u-2,d_{2}]&\stackrel{{\scriptstyle{\rm Dual}}}{{\longleftrightarrow}}&{\mathcal{C}}_{(q,q+1,u,1)}:\penalty\ [q+1,q-2u+3,d_{1}]\\ d_{2}\in\{q-2u+3,q-2u+4\}&&u\leq d_{1}\leq 2u-1\\ \overline{{\mathcal{C}}_{u}}:\penalty\ [q+2,2u-1,d_{2}]&\stackrel{{\scriptstyle{\rm Dual}}}{{\longleftrightarrow}}&(\overline{{\mathcal{C}}_{u}})^{\perp}:\penalty\ [q+2,q-2u+3,d_{1}+1].\\ \end{matrix}

Hence, the minimum distance of $(\overline{{\mathcal{C}}_{u}})^{\perp}$ (resp. $\overline{{\mathcal{C}}_{u}}$ ) can be determined if the minimum distance of the code ${\mathcal{C}}_{(q,q+1,u,1)}$ (resp. ${\mathcal{C}}(u)$ ) is known and vice versa. The following lemma comes from [37].

Lemma 10.

Let $q$ be a prime power and let $u$ with $2\leq u\leq\lfloor\frac{q+1}{2}\rfloor$ . Then the following hold.

1.

If $u\mid(q+1)$ , then $d({\mathcal{C}}_{(q,q+1,u,1)})=u$ .
2.

If $u\nmid(q+1)$ , then $d({\mathcal{C}}_{(q,q+1,u,1)})\geq u+1$ .

Theorem 11.

Let $q=p^{m}>2$ , where $p$ is a prime and $m$ is a positive integer. Let $u$ be an integer with $2\leq u\leq\lfloor\frac{q+1}{2}\rfloor$ . Then the following hold.

1.

If $u\mid(q+1)$ , then $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,2u-1,q-2u+3]$ and $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,q-2u+3,u+1]$ .
2.

If $u\nmid(q+1)$ and $(u+1)\mid(q+1)$ , then $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,2u-1,d]$ , where $d=q-2u+4$ if $u=p=2$ and $m$ is odd, $d=q-2u+3$ otherwise. The dual code $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,q-2u+3,u+2]$ .

Proof.

By Theorem 7, ${\mathcal{C}}_{u}$ has parameters $[q+1,2u-1,q-2u+3]$ . It follows from Lemma 1 that $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,2u-1,d]$ , where $d\in\{q-2u+3,q-2u+4\}$ . By Theorem 8, we obtain that $d((\overline{{\mathcal{C}}_{u}})^{\perp})=d({\mathcal{C}}_{(q,q+1,u,1)})+1$ .

1) Since $u\mid(q+1)$ , by Lemma 10, we deduce that $d({\mathcal{C}}_{(q,q+1,u,1)})=u$ . Therefore, $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,q+3-2u,u+1]$ . Since $u>2$ , $(\overline{{\mathcal{C}}_{u}})^{\perp}$ is not MDS. As a result, $\overline{{\mathcal{C}}_{u}}$ is not MDS. Therefore, $d=q-2u+3$ .

2) It is easily verified that ${\mathcal{C}}_{(q,q+1,u+1,1)}\subseteq{\mathcal{C}}_{(q,q+1,u,1)}$ . Consequently,

d({\mathcal{C}}_{(q,q+1,u,1)})\leq d({\mathcal{C}}_{(q,q+1,u+1,1)}).

Since $(u+1)\mid(q+1)$ , we deduce that $d({\mathcal{C}}_{(q,q+1,u+1,1)})=u+1$ . Since $u\nmid(q+1)$ , it follows from Lemma 10 that $d({\mathcal{C}}_{(q,q+1,u,1)})\geq u+1$ . Consequently, $d({\mathcal{C}}_{(q,q+1,u,1)})=u+1$ . It then following from Theorem 8 that $d((\overline{{\mathcal{C}}_{u}})^{\perp})=u+2$ . In this case, $(\overline{{\mathcal{C}}_{u}})^{\perp}$ is an MDS code if and only if $u=2$ . By assumption, $2\nmid(q+1)$ and $3\mid(q+1)$ , we deduce that $p=2$ and $m$ is odd. Therefore, we have the following hold.

•

If $u=p=2$ and $m$ is odd, $(\overline{{\mathcal{C}}_{u}})^{\perp}$ is a $[q+2,q-1,4]$ MDS code and $\overline{{\mathcal{C}}_{u}}$ is a $[q+2,3,q]$ MDS code.
•

Otherwise, $(\overline{{\mathcal{C}}_{u}})^{\perp}$ is not MDS. Consequently, $\overline{{\mathcal{C}}_{u}}$ is not MDS. Hence, $d=q-2u+3$ .

This completes the proof. ∎

Lemma 12.

[26] Let $u$ be a power of a prime $p$ and $q=u^{e}$ with $e\geq 2$ . Then $d({\mathcal{C}}_{(q,q+1,u,1)})=u+1$ .

Theorem 13.

Let $q=p^{m}>2$ , where $p$ is a prime and $m$ is a positive integer. If $u=p^{s}$ , where $s$ is a positive integer with $s\mid m$ and $s<m$ . Then $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,2u-1,d]$ , where $d=q-2u+4$ if $u=2$ , and $d=q-2u+3$ if $u>2$ . The dual code $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,q-2u+3,u+2]$ .

Proof.

By Theorem 7, ${\mathcal{C}}_{u}$ has parameters $[q+1,2u-1,q-2u+3]$ . By Lemma 1, $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,2u-1,d]$ , where $d\in\{q-2u+3,q-2u+4\}$ . It follows from Theorem 8 and Lemma 12 that $d((\overline{{\mathcal{C}}_{u}})^{\perp})=u+2$ . Hence, $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,q-2u+3,u+2]$ . Notice that $(\overline{{\mathcal{C}}_{u}})^{\perp}$ is an MDS code over ${\mathrm{GF}}(q)$ if and only if $u=2$ . Consequently, $d(\overline{{\mathcal{C}}_{u}})=q-2u+4$ for $u=2$ , and $d(\overline{{\mathcal{C}}_{u}})=q-2u+3$ for $u>2$ . This completes the proof. ∎

Example 14.

Let $q=3^{2}$ . By Theorems 11 and 13, we obtain the following results.

1.

If $u=2$ , the extended code $\overline{{\mathcal{C}}_{u}}$ has parameters $[11,3,8]$ and $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[11,8,3]$ . These two linear codes are optimal codes [16].
2.

If $u=3$ , the extended code $\overline{{\mathcal{C}}_{u}}$ has parameters $[11,5,6]$ and $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[11,6,5]$ . These two linear codes are optimal codes [16].
3.

If $u=4$ , the extended code $\overline{{\mathcal{C}}_{u}}$ has parameters $[11,7,4]$ and is an optimal code [16]. The code $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[11,4,6]$ and is a distance-almost-optimal code [16].

IV-B The case $u=2$

In this subsection, we study the parameters of $\overline{{\mathcal{C}}_{2}}$ and its dual. By Theorems 11 and 13, we have the following results.

Corollary 15.

Let $q$ be an odd prime power. Then $\overline{{\mathcal{C}}_{2}}$ is a $[q+2,3,q-1]$ NMDS code over ${\mathrm{GF}}(q)$ .

Proof.

The desired result follows directly from Result 1 of Theorem 11. ∎

Example 16.

Let $q=5$ . Then the extended code $\overline{{\mathcal{C}}_{2}}$ has parameters $[7,3,4]$ and $(\overline{{\mathcal{C}}_{2}})^{\perp}$ has parameters $[7,4,3]$ . These two linear codes are optimal codes [16].

Corollary 17.

Let $q=2^{m}$ with $m\geq 2$ . Then $\overline{{\mathcal{C}}_{2}}$ is a $[q+2,3,q]$ MDS code over ${\mathrm{GF}}(q)$ with weight enumerator

1+\frac{(q+2)(q^{2}-1)}{2}z^{q}+\frac{q(q-1)^{2}}{2}z^{q+2}.

Proof.

The desired conclusion on the parameters of $\overline{{\mathcal{C}}_{2}}$ follows from Theorem 13. The desired conclusion on the weight distribution follows from Lemma 5. This completes the proof. ∎

Any $[2^{m}+2,3,2^{m}]$ code over ${\mathrm{GF}}(2^{m})$ is called a hyperoval code, as it corresponds to a hyperoval in the Desarguesian projective plane ${\mathrm{PG}}(2,{\mathrm{GF}}(2^{m}))$ [13, Chapter 12].

Example 18.

Let $q=2^{3}$ . The extended code $\overline{{\mathcal{C}}_{2}}$ has parameters $[10,3,8]$ and weight enumerator $1+315z^{8}+196z^{10}$ . The dual code $(\overline{{\mathcal{C}}_{2}})^{\perp}$ has parameters $[10,7,4]$ .

IV-C The case $u=3$

In this subsection, we study the parameters of $\overline{{\mathcal{C}}_{3}}$ and its dual. Let ${\mathcal{C}}(3)$ be the dual of the BCH code ${\mathcal{C}}_{(q,q+1,3,1)}$ , then ${\mathcal{C}}(3)$ is the cyclic code of length $q+1$ over ${\mathrm{GF}}(q)$ with check polynomial $\mathbb{M}_{\beta}(x)\mathbb{M}_{\beta^{2}}(x)$ , where $\beta$ is a primitive $(q+1)$ -th root of unity in ${\mathrm{GF}}(q^{2})$ . The following lemmas will be needed later.

Lemma 19.

Let notation be the same as before. Let $U_{q+1}=\{1,\beta,\cdots,\beta^{q}\}$ . Then $d({\mathcal{C}}_{(q,q+1,3,1)})\leq 4$ if and only if there are four pairwise distinct elements $x,y,z,w$ in $U_{q+1}$ such that

xy+xz+xw+yz+yw+zw=0.

Proof.

For any $4$ pairwise distinct elements $x,y,z,w$ in $U_{q+1}$ , define a $4\times 4$ matrix $\mathbf{M}$ by

\begin{bmatrix}x^{-2}&y^{-2}&z^{-2}&w^{-2}\\ x^{-1}&y^{-1}&z^{-1}&w^{-1}\\ x&y&z&w\\ x^{2}&y^{2}&z^{2}&w^{2}\\ \end{bmatrix}.

It is clear that $d({\mathcal{C}}_{(q,q+1,3,1)})\leq 4$ if and only if there are four pairwise distinct elements $x,y,z,w$ in $U_{q+1}$ such that the system of homogeneous linear equations

\mathbf{M}{\mathbf{x}}^{T}={\mathbf{0}}

(12)

has a nonzero solution in ${\mathrm{GF}}(q)^{4}$ . By [26, Lemma 16], (12) has a nonzero solution in ${\mathrm{GF}}(q)^{4}$ if and only if ${\mathrm{rank}}(\mathbf{M})<4$ , where ${\mathrm{rank}}(\mathbf{M})$ denotes the rank of the matrix $\mathbf{M}$ . It is easily verified that

\det(\mathbf{M})=\sigma_{x,y,z,w}\times(xy+xz+xw+yz+yw+zw).

(13)

where $\sigma_{x,y,z,w}=\frac{(y-x)(z-x)(z-y)(w-x)(w-y)(w-z)}{(xyzw)^{2}}$ . Notice that $x,y,z,w$ are pairwise distinct, we get that $\sigma_{x,y,z,w}\neq 0$ . Therefore, ${\mathrm{rank}}(\mathbf{M})<4$ if and only if

xy+xz+xw+yz+yw+zw=0.

This completes the proof. ∎

The parameters of ${\mathcal{C}}_{(q,q+1,3,1)}$ and ${\mathcal{C}}(3)$ are documented in the following theorem.

Theorem 20.

Let $q\geq 5$ be a prime power. Then the following hold.

1.

If $3\mid(q+1)$ , then ${\mathcal{C}}_{(q,q+1,3,1)}$ has parameters $[q+1,q-3,3]$ and ${\mathcal{C}}(3)$ has parameters $[q+1,4,q-3]$ .
2.

If $3\nmid(q+1)$ , then ${\mathcal{C}}_{(q,q+1,3,1)}$ is a $[q+1,q-3,4]$ NMDS code over ${\mathrm{GF}}(q)$ .

Proof.

It is clear that ${\mathcal{C}}_{(q,q+1,3,1)}$ has length $q+1$ and dimension $q-3$ , and ${\mathcal{C}}(3)$ has length $q+1$ and dimension $4$ , respectively. Notice that $d({\mathcal{C}}(3))\in\{q-3,q-2\}$ , and $d({\mathcal{C}}(3))=q-2$ if and only if ${\mathcal{C}}(3)$ is MDS. Hence, we only need to prove that $d({\mathcal{C}}_{(q,q+1,3,1)})=3$ for $3\mid(q+1)$ and $d({\mathcal{C}}_{(q,q+1,3,1)})=4$ for $3\nmid(q+1)$ . The rest of the proof is divided into the following two cases.

Case 1: $3\mid(q+1)$ . It follows from Lemma 10 that $d({\mathcal{C}}_{(q,q+1,3,1)})=3$ . The desired result follows.

Case 2: $3\nmid(q+1)$ . Suppose $q=p^{m}$ , where $p$ is a prime and $m$ is a positive integer. If $p\in\{2,3\}$ , the desired result was proved in [12]. Now suppose $p$ is an odd prime. Since $3\nmid(q+1)$ , by Lemma 10, $d({\mathcal{C}}_{(q,q+1,3,1)})\geq 4$ . Below we prove that $d({\mathcal{C}}_{(q,q+1,3,1)})\leq 4$ . Let $x\in U_{q+1}$ , $y=-x$ , $z\in U_{q+1}\backslash\{x,-x\}$ and $w=x^{2}z^{-1}$ . It is easily verified that $x,y,z,w\in U_{q+1}$ and

xy+xz+xw+yz+yw+zw=0.

Since $q$ is odd, we deduce that $y\neq x$ . We now prove that $x^{2}z^{-1}\in U_{q+1}\backslash\{x,-x,z\}$ .

•

If $x^{2}z^{-1}=x$ , we get that $z=x$ , a contradiction.
•

If $x^{2}z^{-1}=y$ , we get that $z=-x$ , a contradiction.
•

If $x^{2}z^{-1}=z$ , we deduce that $z=x$ or $-x$ , a contradiction.

By Lemma 19, $d({\mathcal{C}}_{(q,q+1,3,1)})\leq 4$ . In summary, $d({\mathcal{C}}_{(q,q+1,3,1)})=4$ . This completes the proof. ∎

The parameters of $\overline{{\mathcal{C}}_{3}}$ are documented in the following theorem.

Theorem 21.

Let $q\geq 5$ be a prime power. Then the following hold.

1.

If $3\mid(q+1)$ , then $\overline{{\mathcal{C}}_{3}}$ has parameters $[q+2,5,q-3]$ and $(\overline{{\mathcal{C}}_{3}})^{\perp}$ has parameters $[q+2,q-3,4]$ .
2.

If $3\nmid(q+1)$ , then $\overline{{\mathcal{C}}_{3}}$ is a $[q+2,5,q-3]$ NMDS code over ${\mathrm{GF}}(q)$ .

Proof.

By Theorem 7, ${\mathcal{C}}_{3}$ has parameters $[q+1,5,q-3]$ . Then $\overline{{\mathcal{C}}_{3}}$ has length $q+2$ and dimension $5$ . Consequently, $(\overline{{\mathcal{C}}_{3}})^{\perp}$ has length $q+2$ and dimension $q-3$ . By Theorems 9 and 20, we get that $d(\overline{{\mathcal{C}}_{3}})=d({\mathcal{C}}(3))=q-3$ . By Theorems 8 and 20, we obtain that

\displaystyle d((\overline{{\mathcal{C}}_{3}})^{\perp})=d({\mathcal{C}}_{(q,q+1,3,1)})+1=\begin{cases}4\penalty\ &{\rm if}\penalty\ 3\mid(q+1),\\ 5\penalty\ &{\rm if}\penalty\ 3\nmid(q+1).\end{cases}

This completes the proof. ∎

Example 22.

Let $q=5^{2}$ . Then the following hold.

1.

The code ${\mathcal{C}}_{3}$ has parameters $[26,5,22]$ and $({\mathcal{C}}_{3})^{\perp}$ has parameters $[26,21,6]$ .
2.

The BCH code ${\mathcal{C}}_{(q,q+1,3,1)}$ has parameters $[26,22,4]$ and ${\mathcal{C}}(3)$ has parameters $[26,4,22]$ .
3.

The extended code $\overline{{\mathcal{C}}_{3}}$ has parameters $[27,5,22]$ and $(\overline{{\mathcal{C}}_{3}})^{\perp}$ has parameters $[27,22,5]$ .

By Theorem 9, we have

A_{q-3}(\overline{{\mathcal{C}}_{3}})=A_{q-3}({\mathcal{C}}(3)).

(14)

When $q>5$ and $q\not\equiv 2\pmod{3}$ , by Theorem 21, $\overline{{\mathcal{C}}_{3}}$ is an NMDS code over ${\mathrm{GF}}(q)$ . From Lemma 6 and Equation (14), the weight distribution of $\overline{{\mathcal{C}}_{3}}$ is determined by $A_{q-3}({\mathcal{C}}(3))$ . It seems difficult to determine $A_{q-3}({\mathcal{C}}(3))$ in general. We will determine the weight distribution of $\overline{{\mathcal{C}}_{3}}$ for two special cases. The following two lemmas will be needed later.

Lemma 23.

[12] Let $q=2^{m}$ with $m\geq 4$ being even. Then ${\mathcal{C}}(3)$ is a $[q+1,4,q-3]$ NMDS code with the following weight distribution:

•

$A_{0}({\mathcal{C}}(3))=1$ .
•

$A_{q-3}({\mathcal{C}}(3))=\frac{(q-4)(q-1)q(q+1)}{24}$ .
•

$A_{q-2}({\mathcal{C}}(3))=\frac{(q-1)q(q+1)}{2}$ .
•

$A_{q-1}({\mathcal{C}}(3))=\frac{(q+1)q^{2}(q-1)}{4}$ .
•

$A_{q}({\mathcal{C}}(3))=\frac{(q-1)(q+1)(2q^{2}+q+6)}{6}$ .
•

$A_{q+1}({\mathcal{C}}(3))=\frac{3q^{4}-4q^{3}-3q^{2}+4q}{8}$ .
•

$A_{i}({\mathcal{C}}(3))=0$ for other $i$ .

Lemma 24.

[12] Let $q=3^{m}$ with $m\geq 2$ . Then ${\mathcal{C}}(3)$ is a $[q+1,4,q-3]$ NMDS code with the following weight distribution:

•

$A_{0}({\mathcal{C}}(3))=1$ .
•

$A_{q-3}({\mathcal{C}}(3))=\frac{(q-1)^{2}q(q+1)}{24}$ .
•

$A_{q-1}({\mathcal{C}}(3))=\frac{(q-1)q(q+1)(q+3)}{4}$ .
•

$A_{q}({\mathcal{C}}(3))=\frac{(q^{2}-1)(q^{2}-q+3)}{3}$ .
•

$A_{q+1}({\mathcal{C}}(3))=\frac{3(q-1)^{2}q(q+1)}{8}$ .
•

$A_{i}({\mathcal{C}}(3))=0$ for other $i$ .

The following family of NMDS codes were claimed but not proved in [35]. Here we prove they are indeed NMDS and settle the weight distribution of $\overline{{\mathcal{C}}_{3}}$ .

Theorem 25.

Let $q=2^{m}$ with $m\geq 4$ being even. Then $\overline{{\mathcal{C}}_{3}}$ is a $[q+2,5,q-3]$ NMDS code with the following weight distribution:

•

$A_{0}(\overline{{\mathcal{C}}_{3}})=1$ .
•

$A_{q-3}(\overline{{\mathcal{C}}_{3}})=\frac{q^{4}-4q^{3}-q^{2}+4q}{24}$ .
•

$A_{q-2}(\overline{{\mathcal{C}}_{3}})=\frac{q^{5}-4q^{4}+17q^{3}+4q^{2}-18q}{24}$ .
•

$A_{q-1}(\overline{{\mathcal{C}}_{3}})=\frac{3q^{4}-4q^{3}-3q^{2}+4q}{4}$ .
•

$A_{q}(\overline{{\mathcal{C}}_{3}})=\frac{3q^{5}-2q^{4}+17q^{3}+14q^{2}-20q-12}{12}$ .
•

$A_{q+1}(\overline{{\mathcal{C}}_{3}})=\frac{8q^{5}+13q^{4}-44q^{3}-13q^{2}+36q}{24}$ .
•

$A_{q+2}(\overline{{\mathcal{C}}_{3}})=\frac{3q^{5}-8q^{4}+7q^{3}-2q}{8}$ .
•

$A_{i}(\overline{{\mathcal{C}}_{3}})=0$ for other $i$ .

Proof.

The conclusions on the parameters of $\overline{{\mathcal{C}}_{3}}$ and $(\overline{{\mathcal{C}}_{3}})^{\perp}$ follow from Theorem 21. It remains to prove the weight distribution of the code $\overline{{\mathcal{C}}_{3}}$ . It follows from Lemma 23 and Equation (14) that

A_{q-3}(\overline{{\mathcal{C}}_{3}})=\frac{(q-4)(q-1)q(q+1)}{24}.

(15)

Recall that $\overline{{\mathcal{C}}_{3}}$ is an NMDS code over ${\mathrm{GF}}(q)$ . From Lemma 6 and Equation (15), the desired weight distribution follows. ∎

Example 26.

Let $q=2^{4}$ . Then $\overline{{\mathcal{C}}_{3}}$ has parameters $[18,5,13]$ and weight enumerator

\displaystyle 1+2040z^{13}+35700z^{14}+44880z^{15}+257295z^{16}+377400z^{17}+331260z^{18}.

The dual code $(\overline{{\mathcal{C}}_{3}})^{\perp}$ has parameters $[18,13,5]$ . Hence, $\overline{{\mathcal{C}}_{3}}$ is NMDS in this case.

Similar to the proof of Theorem 25, one can prove the following.

Theorem 27.

Let $q=3^{m}$ with $m\geq 2$ . Then $\overline{{\mathcal{C}}_{3}}$ is a $[q+2,5,q-3]$ NMDS code with the following weight distribution:

•

$A_{0}(\overline{{\mathcal{C}}_{3}})=1$ .
•

$A_{q-3}(\overline{{\mathcal{C}}_{3}})=\frac{q^{4}-q^{3}-q^{2}+q}{24}$ .
•

$A_{q-2}(\overline{{\mathcal{C}}_{3}})=\frac{q^{5}-4q^{4}+2q^{3}+4q^{2}-3q}{24}$ .
•

$A_{q-1}(\overline{{\mathcal{C}}_{3}})=\frac{3q^{4}+q^{3}-3q^{2}-q}{4}$ .
•

$A_{q}(\overline{{\mathcal{C}}_{3}})=\frac{3q^{5}-2q^{4}+2q^{3}+14q^{2}-5q-12}{12}$ .
•

$A_{q+1}(\overline{{\mathcal{C}}_{3}})=\frac{8q^{5}+13q^{4}-29q^{3}-13q^{2}+21q}{24}$ .
•

$A_{q+2}(\overline{{\mathcal{C}}_{3}})=\frac{3q^{5}-8q^{4}+6q^{3}-q}{8}$ .
•

$A_{i}(\overline{{\mathcal{C}}_{3}})=0$ for other $i$ .

Example 28.

Let $q=3^{2}$ . Then $\overline{{\mathcal{C}}_{3}}$ has parameters $[11,5,6]$ and weight enumerator

\displaystyle 1+240z^{6}+1440z^{7}+5040z^{8}+13880z^{9}+22320z^{10}+16128z^{11}.

The dual code $(\overline{{\mathcal{C}}_{3}})^{\perp}$ has parameters $[11,6,5]$ . Hence, $\overline{{\mathcal{C}}_{3}}$ is NMDS in this case.

IV-D The case $u=4$

In this subsection, we study the parameters of $\overline{{\mathcal{C}}_{4}}$ and its dual. Before studying $\overline{{\mathcal{C}}_{4}}$ , we need to do some preparations. Let ${\mathcal{C}}(4)$ denote the dual of the BCH code ${\mathcal{C}}_{(q,q+1,4,1)}$ , then ${\mathcal{C}}(4)$ is the cyclic code of length $q+1$ over ${\mathrm{GF}}(q)$ with check polynomial $\mathbb{M}_{\beta}(x)\mathbb{M}_{\beta^{2}}(x)\mathbb{M}_{\beta^{3}}(x)$ , where $\beta$ is a primitive $(q+1)$ -th root of unity in ${\mathrm{GF}}(q^{2})$ . When $q=2^{m}$ with $m\geq 4$ , the parameters of the BCH code ${\mathcal{C}}_{(q,q+1,4,1)}$ were studied in [36].

Lemma 29.

[36] Let $q=2^{m}$ with $m\geq 5$ being odd. Then ${\mathcal{C}}(4)$ is a $[q+1,6,q-5]$ NMDS code over ${\mathrm{GF}}(q)$ and

A_{q-5}({\mathcal{C}}(4))=\frac{(q-1)(q-8)}{30}\binom{q+1}{4}.

Theorem 30.

Let $q=2^{m}$ with $m\geq 4$ . Then the following hold.

1.

If $m$ is even, then $\overline{{\mathcal{C}}_{4}}$ has parameters $[q+2,7,q-5]$ and $(\overline{{\mathcal{C}}_{4}})^{\perp}$ has parameters $[q+2,q-5,6]$ .
2.
If $m$ is odd, then $\overline{{\mathcal{C}}_{4}}$ is a $[q+2,7,q-5]$ NMDS code with the following weight distribution:
- •
  
  $A_{0}(\overline{{\mathcal{C}}_{4}})=1$ .
- •
  
  $A_{q-5}(\overline{{\mathcal{C}}_{4}})=\frac{q^{6}-11q^{5}+25q^{4}-5q^{3}-26q^{2}+16q}{720}$ .
- •
  
  $A_{q-4}(\overline{{\mathcal{C}}_{4}})=\frac{q^{7}-11q^{6}+75q^{5}-155q^{4}+24q^{3}+166q^{2}-100q}{720}$ .
- •
  
  $A_{q-3}(\overline{{\mathcal{C}}_{4}})=\frac{3q^{6}-17q^{5}+27q^{4}+q^{3}-30q^{2}+16q}{48}$ .
- •
  
  $A_{q-2}(\overline{{\mathcal{C}}_{4}})=\frac{3q^{7}-7q^{6}+101q^{5}-133q^{4}-64q^{3}+140q^{2}-40q}{144}$ .
- •
  
  $A_{q-1}(\overline{{\mathcal{C}}_{4}})=\frac{8q^{7}+47q^{6}-69q^{5}+167q^{4}+141q^{3}-214q^{2}-80q}{144}$ .
- •
  
  $A_{q}(\overline{{\mathcal{C}}_{4}})=\frac{45q^{7}+93q^{6}+127q^{5}-255q^{4}-200q^{3}+402q^{2}+28q-240}{240}$ .
- •
  
  $A_{q+1}(\overline{{\mathcal{C}}_{4}})=\frac{264q^{7}+151q^{6}-845q^{5}+415q^{4}+181q^{3}-566q^{2}+400q}{720}$ .
- •
  
  $A_{q+2}(\overline{{\mathcal{C}}_{4}})=\frac{53q^{7}-133q^{6}+99q^{5}-19q^{4}+8q^{2}-8q}{144}$ .
- •
  
  $A_{i}(\overline{{\mathcal{C}}_{4}})=0$ for other $i$ .

Proof.

1) The first desired result follows directly from Theorem 13.

2) With Lemma 6, Theorem 9 and Lemma 29, similar to the proof of Theorem 25, we can prove the second desired result. Details are omitted here. ∎

Example 31.

Let $q=2^{4}$ . Then $\overline{{\mathcal{C}}_{4}}$ has parameters $[18,7,11]$ and is AMDS. The dual code $(\overline{{\mathcal{C}}_{4}})^{\perp}$ has parameters $[18,11,6]$ , but is not AMDS.

Example 32.

Let $q=2^{5}$ . Then $\overline{{\mathcal{C}}_{4}}$ has parameters $[34,7,27]$ and weight enumerator

	$\displaystyle 1+1014816z^{27}+34588312z^{28}+55814880z^{29}+686184752z^{30}+$
	$\displaystyle 2244500192z^{31}+6875142087z^{32}+12784990240z^{33}+11677503088z^{34}.$

The dual code $(\overline{{\mathcal{C}}_{4}})^{\perp}$ has parameters $[34,27,7]$ . Hence, $\overline{{\mathcal{C}}_{4}}$ is NMDS in this case.

Theorem 30 documents another family of NMDS codes in this paper.

Starting from now on, let $q\geq 9$ be an odd prime power. Let ${\mathrm{Tr}}(x)$ denote the trace function from ${\mathrm{GF}}(q^{2})$ to ${\mathrm{GF}}(q)$ . To determine the parameters of ${\mathcal{C}}_{(q,q+1,4,1)}$ and ${\mathcal{C}}(4)$ , we need the following preparations.

Suppose $x_{1},x_{2},x_{3},x_{4},x_{5}$ are five pairwise distinct elements in $U_{q+1}=\{1,\beta,\cdots,\beta^{q}\}$ . Define a $6\times 5$ matrix $\mathbf{M}$ by

\displaystyle\begin{bmatrix}x_{1}^{-3}&x_{2}^{-3}&x_{3}^{-3}&x_{4}^{-3}&x_{5}^{-3}\\ x_{1}^{-2}&x_{2}^{-2}&x_{3}^{-2}&x_{4}^{-2}&x_{5}^{-2}\\ x_{1}^{-1}&x_{2}^{-1}&x_{3}^{-1}&x_{4}^{-1}&x_{5}^{-1}\\ x_{1}&x_{2}&x_{3}&x_{4}&x_{5}\\ x_{1}^{2}&x_{2}^{2}&x_{3}^{2}&x_{4}^{2}&x_{5}^{2}\\ x_{1}^{3}&x_{2}^{3}&x_{3}^{3}&x_{4}^{3}&x_{5}^{3}\\ \end{bmatrix}.

For any $i\in\{\pm 1,\pm 2,\pm 3\}$ , let $\mathbf{M}[i]$ denote the submatrix of $\mathbf{M}$ obtained by deleting the row $(x_{1}^{i},x_{2}^{i},x_{3}^{i},x_{4}^{i},x_{5}^{i})$ of the matrix $\mathbf{M}$ . Let $I=\{1,2,3,4,5\}$ , the elementary symmetric polynomial (ESP) of degree $i$ in $5$ variables $x_{1},x_{2},\cdots,x_{5}$ , written $\sigma_{5,i}$ , is defined by

\sigma_{5,i}=\sum_{J\subseteq I,|J|=i}\prod_{j\in J}x_{j}.

Lemma 33.

Let $q\geq 9$ be an odd prime power and $4\mid(q-1)$ . Then $d({\mathcal{C}}_{(q,q+1,4,1)})=5$ if and only if there are five pairwise distinct elements $x_{1},x_{2},x_{3},x_{4},x_{5}$ in $U_{q+1}$ such that $\sigma_{5,2}=0$ .

Proof.

Since $4\mid(q-1)$ , we get that $4\nmid{(q+1)}$ . By Lemma 10, we deduce that $d({\mathcal{C}}_{(q,q+1,4,1)})\geq 5$ . Therefore, $d({\mathcal{C}}_{(q,q+1,3,1)})=5$ if and only if there are five pairwise distinct elements $x_{1},x_{2},x_{3},x_{4},x_{5}$ in $U_{q+1}$ such that the system of homogeneous linear equations

\mathbf{M}{\mathbf{x}}^{T}={\mathbf{0}}

(16)

has a nonzero solution in ${\mathrm{GF}}(q)^{5}$ . By [26, Lemma 16], (16) has a nonzero solution in ${\mathrm{GF}}(q)^{5}$ if and only if ${\mathrm{rank}}(\mathbf{M})\leq 4$ .

It is easily verified that

	$\displaystyle\det(\mathbf{M}[-3])$	$\displaystyle=\frac{\prod_{1\leq j<i\leq 5}(x_{i}-x_{j})}{\sigma_{5,5}}\times\sigma_{5,2}^{q},$
	$\displaystyle\det(\mathbf{M}[-2])$	$\displaystyle=\frac{\prod_{1\leq j<i\leq 5}(x_{i}-x_{j})}{\sigma_{5,5}}\times(\sigma_{5,1}^{q}\sigma_{5,2}^{q}-\sigma_{5,5}^{q}\sigma_{5,2}),$
	$\displaystyle\det(\mathbf{M}[-1])$	$\displaystyle=\frac{\prod_{1\leq j<i\leq 5}(x_{i}-x_{j})}{\sigma_{5,5}}\times(\sigma_{5,2}^{2q}-\sigma_{5,5}^{q}\sigma_{5,1}^{q}\sigma_{5,2}),$
	$\displaystyle\det(\mathbf{M}[1])$	$\displaystyle=\frac{\prod_{1\leq j<i\leq 5}(x_{i}-x_{j})}{\sigma_{5,5}^{3}}\times(\sigma_{5,2}^{2}-\sigma_{5,5}\sigma_{5,1}\sigma_{5,2}^{q}),$
	$\displaystyle\det(\mathbf{M}[2])$	$\displaystyle=\frac{\prod_{1\leq j<i\leq 5}(x_{i}-x_{j})}{\sigma_{5,5}^{3}}\times(\sigma_{5,1}\sigma_{5,2}-\sigma_{5,5}\sigma_{5,2}^{q}),$
	$\displaystyle\det(\mathbf{M}[3])$	$\displaystyle=\frac{\prod_{1\leq j<i\leq 5}(x_{i}-x_{j})}{\sigma_{5,5}^{3}}\times\sigma_{5,2}.$

It then follows that ${\mathrm{rank}}(\mathbf{M})\leq 4$ if and only if $\sigma_{5,2}=0$ . This completes the proof. ∎

Lemma 34.

Let $q\geq 9$ be an odd prime power and $4\mid(q-1)$ . Then the following hold.

1.

${\mathrm{Tr}}(\beta^{i})\neq{\mathrm{Tr}}(\beta^{j})$ for any $i\neq j$ with $1\leq i,j\leq\frac{q-1}{2}$ .
2.

${\mathrm{Tr}}(\beta^{i})\notin\{0,2,-2\}$ for any $1\leq i\leq\frac{q-1}{2}$ .
3.

${\mathrm{Tr}}(\beta^{i})=-1$ with $1\leq i\leq\frac{q-1}{2}$ if and only if $3\mid(q+1)$ and $i=\frac{q+1}{3}$ .

Proof.

1) It is easily seen that

x^{q+1}-1=(x-1)(x+1)\prod_{i=1}^{\frac{q-1}{2}}(x^{2}-{\mathrm{Tr}}(\beta^{i})x+1).

Since $\gcd(q+1,q)=1$ , we deduce that $x^{q+1}-1$ has no multiple roots. Hence, ${\mathrm{Tr}}(\beta^{i})\neq{\mathrm{Tr}}(\beta^{j})$ for any $i\neq j$ with $1\leq i,j\leq\frac{q-1}{2}$ .

2) Since $4\mid(q-1)$ , we deduce that

x^{q+1}-1=x^{2}(x^{q-1}-1)+x^{2}-1\equiv x^{2}-1\pmod{x^{2}+1}.

It follows that $\gcd(x^{q+1}-1,x^{2}+1)=\gcd(x^{2}-1,x^{2}+1)=1$ . It then follows that ${\mathrm{Tr}}(\beta^{i})\neq 0$ for any $1\leq i\leq\frac{q-1}{2}$ . We now prove that ${\mathrm{Tr}}(\beta^{i})\notin\{2,-2\}$ for any $1\leq i\leq\frac{q-1}{2}$ . Since

{\mathrm{Tr}}(\beta^{\frac{q+1}{2}-i})={\mathrm{Tr}}(-\beta^{-i})=-{\mathrm{Tr}}(\beta^{i}),

to prove the desired result we only need to prove ${\mathrm{Tr}}(\beta^{i})\neq 2$ for any $1\leq i\leq\frac{q-1}{2}$ . It is easily verified that ${\mathrm{Tr}}(\beta^{i})=2$ with $1\leq i\leq\frac{q-1}{2}$ if and only if $\beta^{2i}-2\beta^{i}+1=0$ , i.e., $\beta^{i}=1$ with $1\leq i\leq\frac{q-1}{2}$ , a contradiction. The desired result follows.

3) It is clear that ${\mathrm{Tr}}(\beta^{i})=-1$ if and only if $\beta^{2i}+\beta^{i}+1=0$ . Notice that

\beta^{3i}-1=(\beta^{i}-1)(\beta^{2i}+\beta^{i}+1),

and $\beta^{i}\neq 1$ for any $1\leq i\leq\frac{q-1}{2}$ . Hence, $\beta^{2i}+\beta^{i}+1=0$ with $1\leq i\leq\frac{q-1}{2}$ if and only if $\beta^{3i}=1$ . It follows that ${\mathrm{Tr}}(\beta^{i})=-1$ with $1\leq i\leq\frac{q-1}{2}$ if and only if $3\mid(q+1)$ and $i=\frac{q+1}{3}$ . This completes the proof. ∎

Lemma 35.

Let $q=p^{m}$ and $4\mid(q-1)$ , where $p$ is an odd prime and $m$ is a positive integer. Then the following hold.

1.

For any $j$ with $1\leq j\leq q$ ,

$\sum_{i=1}^{\frac{q-1}{2}}{\mathrm{Tr}}(\beta^{ji})=-1-(-1)^{j}.$
2.

If $3\mid(q-1)$ ,

$-\sum_{i=1}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)^{-1}=\frac{q-1}{6}\bmod{p},$

where $i\bmod{p}$ denotes the unique integer $j$ such that $0\leq j\leq p-1$ and $p\mid(i-j)$ .
3.

If $3\mid(q+1)$ ,

$-\sum_{i=1,\atop i\neq\frac{q+1}{3}}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)^{-1}=\frac{q-1}{2}\bmod{p}.$

Proof.

1) It is easily seen that

	$\displaystyle\sum_{i=1}^{\frac{q-1}{2}}{\mathrm{Tr}}(\beta^{ji})$	$\displaystyle=\sum_{i=1}^{\frac{q-1}{2}}[(\beta^{j})^{i}+(\beta^{j})^{q+1-i}]$
		$\displaystyle=\sum_{i=0}^{q}(\beta^{j})^{i}-1-(\beta^{j})^{\frac{q+1}{2}}$
		$\displaystyle=-1-(-1)^{j}$

for any $1\leq j\leq q$ . The desired result follows.

2) If $3\mid(q-1)$ , then

x\frac{x^{q+1}-1}{x-1}=(x^{2}+x+1)\sum_{j=0}^{\frac{q-1}{3}}x^{3j}-1.

Hence,

(\beta^{2i}+\beta^{i}+1)\sum_{j=0}^{\frac{q-1}{3}}\beta^{3ij}=1

for each $1\leq i\leq q$ . It follows that

	$\displaystyle({\mathrm{Tr}}(\beta^{i})+1)^{-1}$	$\displaystyle=\beta^{i}(\beta^{2i}+\beta^{i}+1)^{-1}$
		$\displaystyle=\sum_{j=0}^{\frac{q-1}{3}}\beta^{(3j+1)i}$
		$\displaystyle=\sum_{j=0}^{\frac{q-7}{6}}(\beta^{(3j+1)i}+\beta^{-(3j+1)i})+\beta^{(\frac{q+1}{2})i}$
		$\displaystyle=(-1)^{i}+\sum_{j=0}^{\frac{q-7}{6}}{\mathrm{Tr}}(\beta^{(3j+1)i}),$

where the third equation follows from the fact that $\beta^{[3(\frac{q-1}{3}-j)+1]i}=\beta^{(q-3j)i}=\beta^{-(3j+1)i}$ . Consequently,

	$\displaystyle-\sum_{i=1}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)^{-1}$	$\displaystyle=-\sum_{i=1}^{\frac{q-1}{2}}[(-1)^{i}+\sum_{j=0}^{\frac{q-7}{6}}{\mathrm{Tr}}(\beta^{(3j+1)i})]$
		$\displaystyle=-\sum_{i=1}^{\frac{q-1}{2}}\sum_{j=0}^{\frac{q-7}{6}}{\mathrm{Tr}}(\beta^{(3j+1)i})$
		$\displaystyle=-\sum_{j=0}^{\frac{q-7}{6}}\sum_{i=1}^{\frac{q-1}{2}}{\mathrm{Tr}}(\beta^{(3j+1)i})$
		$\displaystyle=\sum_{j=0}^{\frac{q-7}{6}}[1+(-1)^{3j+1}]$
		$\displaystyle=\frac{q-1}{6}\bmod{p},$

where the fourth equation follows from Result 1. The desired result follows.

3) If $3\mid(q+1)$ , then

(x^{2}+x+1)\sum_{j=0}^{\frac{q-5}{3}}(2+3j)x^{3j}(x-1)+3\sum_{j=0}^{\frac{q-2}{3}}x^{3j}=1.

Let $i$ be an integer with $1\leq i\leq\frac{q-1}{2}$ and $i\neq\frac{q+1}{3}$ , then $\sum_{j=0}^{\frac{q-2}{3}}\beta^{3ij}=0$ . It follows that

(\beta^{2i}+\beta^{i}+1)\sum_{j=0}^{\frac{q-5}{3}}(2+3j)\beta^{3ij}(\beta^{i}-1)=1.

Consequently,

	$\displaystyle({\mathrm{Tr}}(\beta^{i})+1)^{-1}$	$\displaystyle=\beta^{i}(\beta^{2i}+\beta^{i}+1)^{-1}$
		$\displaystyle=\sum_{j=0}^{\frac{q-5}{3}}(2+3j)(\beta^{(2+3j)i}-\beta^{(1+3j)i})$
		$\displaystyle=\sum_{j=0}^{\frac{q-5}{3}}(2+3j)\beta^{(2+3j)i}-\sum_{j=1}^{\frac{q-2}{3}}[2+3(\frac{q-2}{3}-j)]\beta^{[1+3(\frac{q-2}{3}-j)]i}$
		$\displaystyle=\sum_{j=0}^{\frac{q-5}{3}}(2+3j)\beta^{(2+3j)i}+\sum_{j=1}^{\frac{q-2}{3}}3j\beta^{-(2+3j)i}$
		$\displaystyle=2\sum_{j=0}^{\frac{q-5}{3}}\beta^{(2+3j)i}+3\sum_{j=1}^{\frac{q-5}{3}}j{\mathrm{Tr}}(\beta^{(2+3j)i})-2\beta^{i}$
		$\displaystyle=-2{\mathrm{Tr}}(\beta^{i})+3\sum_{j=1}^{\frac{q-5}{3}}j{\mathrm{Tr}}(\beta^{(2+3j)i}),$

where the sixth equation follows from the fact that

	$\displaystyle\sum_{j=0}^{\frac{q-5}{3}}\beta^{(2+3j)i}-\beta^{i}$	$\displaystyle=\beta^{2i}\frac{(\beta^{3i})^{\frac{q-2}{3}}-1}{\beta^{3i}-1}-\beta^{i}$
		$\displaystyle=\frac{\beta^{qi}-\beta^{2i}}{\beta^{3i}-1}-\beta^{i}$
		$\displaystyle=-{\mathrm{Tr}}(\beta^{i}).$

Therefore,

	$\displaystyle-\sum_{i=1,\atop i\neq\frac{q+1}{3}}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)^{-1}$	$\displaystyle=\sum_{i=1,\atop i\neq\frac{q+1}{3}}^{\frac{q-1}{2}}[2{\mathrm{Tr}}(\beta^{i})-3\sum_{j=1}^{\frac{q-5}{3}}j{\mathrm{Tr}}(\beta^{(2+3j)i})]$
		$\displaystyle=\sum_{i=1}^{\frac{q-1}{2}}[2{\mathrm{Tr}}(\beta^{i})-3\sum_{j=1}^{\frac{q-5}{3}}j{\mathrm{Tr}}(\beta^{(2+3j)i})]-2{\mathrm{Tr}}(\beta^{\frac{q+1}{3}})+3\sum_{j=1}^{\frac{q-5}{3}}j{\mathrm{Tr}}(\beta^{\frac{2(q+1)}{3}})$
		$\displaystyle=3\sum_{j=1}^{\frac{q-5}{3}}j[1+(-1)^{2+3j}]+2-3\sum_{j=1}^{\frac{q-5}{3}}j$
		$\displaystyle=\frac{q-1}{2}\bmod{p}.$

This completes the proof. ∎

Lemma 36.

Let $q\geq 9$ be an odd prime power, $4\mid(q-1)$ and $3\nmid(q+1)$ . Then there exist two distinct integers $i$ and $j$ with $1\leq i,j\leq\frac{q-1}{2}$ such that $({\mathrm{Tr}}(\beta^{i})+1)({\mathrm{Tr}}(\beta^{j})+1)+1=0$ .

Proof.

If $5\mid(q+1)$ , then

\displaystyle({\mathrm{Tr}}(\beta^{\frac{q+1}{5}})+1)({\mathrm{Tr}}(\beta^{\frac{2(q+1)}{5}})+1)+1=2\sum_{i=0}^{4}\beta^{\frac{i(q+1)}{5}}=0.

The desired result follows. Below we suppose $5\nmid(q+1)$ .

Let

\Delta=\left\{i:\ 1\leq i\leq\frac{q-1}{2}\right\},

then $|\Delta|=\frac{q-1}{2}$ . By Result 3 of Lemma 34, ${\mathrm{Tr}}(\beta^{i})\neq-1$ for any $i\in\Delta$ . Let

S_{1}=\left\{{\mathrm{Tr}}(\beta^{i})+1:\ i\in\Delta\right\}

and $S_{2}=\left\{-({\mathrm{Tr}}(\beta^{i})+1)^{-1}:\ i\in\Delta\right\}$ . By Result 1 of Lemma 34, we deduce that $|S_{1}|=|S_{2}|=|\Delta|$ . It is clear that

|S_{1}\cap S_{2}|=|S_{1}|+|S_{2}|-|S_{1}\cup S_{2}|=2|\Delta|-|S_{1}\cup S_{2}|.

(17)

It follows from Result 2 of Lemma 34 that $S_{1}\cup S_{2}\subseteq{\mathrm{GF}}(q)\backslash\{-1,0,1\}$ . It follows from Equation (17) that $|S_{1}\cap S_{2}|\geq q-1-(q-3)=2$ . Define

N:=\left|\left\{i\in\Delta:({\mathrm{Tr}}(\beta^{i})+1)^{2}+1=0\right\}\right|,

then $N\in\{0,1,2\}$ . Clearly, to prove the desired conclusion, it suffices to prove $|S_{1}\cap S_{2}|\geq 3$ or $N\leq 1$ .

Suppose on the contrary that $|S_{1}\cap S_{2}|=2$ and $N=2$ , then $S_{1}\cup S_{2}={\mathrm{GF}}(q)\backslash\{-1,0,1\}$ and $S_{1}\cap S_{2}=\{a,-a\}$ , where $a\in{\mathrm{GF}}(q)^{*}$ and $a^{2}=-1$ . We seek a contradiction by considering two cases.

Case 1: $p=3$ . It is easily verified that

	$\displaystyle({\mathrm{Tr}}(\beta^{i})+1)^{2}+1$	$\displaystyle=\beta^{2i}+\beta^{-2i}-\beta^{i}-\beta^{-i}+1$
		$\displaystyle=\beta^{-2i}(\beta^{4i}-\beta^{3i}+\beta^{2i}-\beta^{i}+1)$
		$\displaystyle=\beta^{-2i}\left(\frac{\beta^{5i}+1}{\beta^{i}+1}\right).$

It follows that $({\mathrm{Tr}}(\beta^{i})+1)^{2}+1=0$ with $i\in\Delta$ if and only if $10\mid(q+1)$ and $i\in\{\frac{q+1}{10},\frac{3(q+1)}{10}\}$ . By assumption, $5\nmid(q+1)$ . Hence, $N=0$ , i.e., we get a contradiction.

Case 2: $p>3$ . Since $3\nmid(q+1)$ , we deduce that $3\mid(q-1)$ . Since $S_{1}\cup S_{2}={\mathrm{GF}}(q)\backslash\{-1,0,1\}$ and $S_{1}\cap S_{2}=\{a,-a\}$ , we have

$\displaystyle 0=\sum_{s\in S_{1}\cup S_{2}}s$	$\displaystyle=\sum_{s\in S_{1}}s+\sum_{s\in S_{1}}s-\sum_{s\in S_{1}\cap S_{2}}s$
	$\displaystyle=\sum_{i=1}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)-\sum_{i=1}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)^{-1}$
	$\displaystyle=(\frac{q-1}{2}+\frac{q-1}{6})\bmod{p},$	(18)

where the last equation follows from Lemma 35. Clearly,

\frac{q-1}{2}+\frac{q-1}{6}=\frac{2(q-1)}{3}\not\equiv 0\pmod{p}.

Hence, Equation (18) is impossible, i.e., we get a contradiction.

Collecting the conclusions in Cases 1 and 2 yields $|S_{1}\cap S_{2}|\geq 3$ or $N\leq 1$ . This completes the proof. ∎

Lemma 37.

Let $q>5$ be an odd prime, $4\mid(q-1)$ and $3\mid(q+1)$ . Then there exist two distinct integers $i$ and $j$ with $1\leq i,j\leq\frac{q-1}{2}$ such that $({\mathrm{Tr}}(\beta^{i})+1)({\mathrm{Tr}}(\beta^{j})+1)+1=0$ .

Proof.

Similar to Lemma 36, if $5\mid(q+1)$ , we have

\displaystyle({\mathrm{Tr}}(\beta^{\frac{q+1}{5}})+1)({\mathrm{Tr}}(\beta^{\frac{2(q+1)}{5}})+1)+1=0.

The desired result follows. Below we suppose $5\nmid(q+1)$ .

Let

\Gamma=\left\{i:\ 1\leq i\leq\frac{q-1}{2},\ i\neq\frac{q+1}{3}\right\},

then $|\Gamma|=\frac{q-3}{2}$ . By Result 3 of Lemma 34, ${\mathrm{Tr}}(\beta^{i})\neq-1$ for any $i\in\Gamma$ . Let

T_{1}=\left\{{\mathrm{Tr}}(\beta^{i})+1:\ i\in\Gamma\right\}

and $T_{2}=\left\{-({\mathrm{Tr}}(\beta^{i})+1)^{-1}:\ i\in\Gamma\right\}$ . By Result 1 of Lemma 34, we deduce that $|T_{1}|=|T_{2}|=|\Gamma|$ . Similar to Lemma 36, we have

|T_{1}\cap T_{2}|=2|\Gamma|-|T_{1}\cup T_{2}|,

(19)

and $T_{1}\cup T_{2}\subseteq{\mathrm{GF}}(q)\backslash\{-1,0,1\}$ . Similar to Lemma 36, to prove the desired result, it suffices to prove that $|T_{1}\cap T_{2}|\geq 3$ or

T:=|\{i\in\Gamma:\ ({\mathrm{Tr}}(\beta^{i})+1)^{2}+1=0\}|\leq 1.

We first prove $|T_{1}\cap T_{2}|\geq 2$ by distinguishing the following two cases.

Case 1: Suppose $|T_{1}\cap T_{2}|=0$ , then $T_{1}\cup T_{2}={\mathrm{GF}}(q)\backslash\{-1,0,1\}$ . By Lemma 35, we obtain that

	$\displaystyle 0$	$\displaystyle=\sum_{t\in T_{1}\cup T_{2}}t=\sum_{t\in T_{1}}t+\sum_{t\in T_{2}}t$
		$\displaystyle=\sum_{i=1,\atop i\neq\frac{q+1}{3}}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)-\sum_{i=1,\atop i\neq\frac{q+1}{3}}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)^{-1}$
		$\displaystyle=\sum_{i=1}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)-\sum_{i=1,\atop i\neq\frac{q+1}{3}}^{\frac{q-1}{2}}({\mathrm{Tr}}(\beta^{i})+1)^{-1}=-1,$

a contradiction.

Case 2: Suppose $|T_{1}\cap T_{2}|=1$ , then there is an element $a\in{\mathrm{GF}}(q)\backslash\{-1,0,1\}$ such that $T_{1}\cap T_{2}=\{a\}$ . It is easily verified that if $a\in T_{1}\cap T_{2}$ , then $-a^{-1}\in T_{1}\cap T_{2}$ . Hence, $a^{2}=-1$ . It follows from Equation (19) that $|T_{1}\cup T_{2}|=q-4$ . Consequently, there is an element $b\in{\mathrm{GF}}(q)\backslash\{-1,0,1\}$ such that $T_{1}\cup T_{2}={\mathrm{GF}}(q)\backslash\{-1,0,1,b\}$ . By Lemma 35, we obtain that

-b=\sum_{t\in T_{1}\cup T_{2}}t=\sum_{t\in T_{1}}t+\sum_{t\in T_{2}}t-\sum_{t\in T_{1}\cap T_{2}}t=-1-a,

i.e., $b=1+a$ . We now prove that $-a\in T_{1}\cup T_{2}$ . Clearly, to prove the desired result, we only need to prove $1+a\neq-a$ . Suppose $1+a=-a$ , then $a=\frac{q-1}{2}$ . Consequently,

a^{2}=(\frac{q-1}{2})^{2}\equiv-1\pmod{q}.

(20)

Suppose $q=4k+1$ , by Equation (20), we get that $k\equiv 1\pmod{q}$ . It follows that $5\equiv 0\pmod{q}$ . Consequently, $q=5$ , a contradiction. That is to say, $-a\in T_{1}\cup T_{2}$ . Notice that $-(-a)^{-1}=-a$ , we deduce that $-a\in T_{1}\cap T_{2}$ , which contradicts the fact that $T_{1}\cap T_{2}=\{a\}$ .

Collecting the conclusions in Cases 1 and 2 yields $|T_{1}\cap T_{2}|\geq 2$ .

Finally, we prove that $|T_{1}\cap T_{2}|\geq 3$ or $T\leq 1$ . Suppose on the contrary that $|T_{1}\cap T_{2}|=2$ and $T=2$ , then $T_{1}\cap T_{2}=\{a,-a\}$ , where $a\in{\mathrm{GF}}(q)^{*}$ and $a^{2}=-1$ . It follows from Equation (19) that $|T_{1}\cup T_{2}|=q-5$ . Consequently, there exist two distinct elements $t_{1}$ and $t_{2}$ in ${\mathrm{GF}}(q)\backslash\{-1,0,1\}$ such that $T_{1}\cup T_{2}={\mathrm{GF}}(q)\backslash\{-1,0,1,t_{1},t_{2}\}$ . By Lemma 35, we obtain that

-(t_{1}+t_{2})=\sum_{t\in T_{1}\cup T_{2}}t=\sum_{t\in T_{1}}t+\sum_{t\in T_{2}}t-\sum_{t\in T_{1}\cap T_{2}}t=-1.

Therefore,

t_{1}+t_{2}=1.

(21)

It is easily verified that if $t\in T_{1}\cup T_{2}$ , then $-t^{-1}\in T_{1}\cup T_{2}$ . It follows that $t_{2}=-(t_{1})^{-1}$ . By Equation (21), we deduce that

t_{1}^{2}-t_{1}-1=0.

(22)

It is easily verified that $x^{2}-x-1$ is reducible over ${\mathrm{GF}}(q)$ if and only if $5$ is a quadratic residue modulo $q$ . By [24], $5$ is a quadratic residue modulo $q$ if and only if $q\equiv\pm 1\pmod{5}$ . We seek a contradiction by considering the following two cases.

Case 1: $5\nmid(q-1)$ . By assumption, $5\nmid(q+1)$ . Consequently, $x^{2}-x-1$ is irreducible over ${\mathrm{GF}}(q)$ . It follows that Equation (22) is impossible.

Case 2: $5\mid(q-1)$ . It is easily verified that $3\notin\{-1,0,1,t_{1},t_{2}\}$ . Consequently, $3\in T_{1}\cup T_{2}$ . By Result 2 of Lemma 34, $3\notin T_{1}$ . Hence, $3\in T_{2}$ . It follows that there is an integer $i$ with $1\leq i\leq\frac{q-1}{2}$ such that ${\mathrm{Tr}}(\beta^{i})+1=-3^{-1}$ . It then follows that $3\beta^{2i}+4\beta^{i}+3=0$ . Notice that $20\mid(q-1)$ , we deduce that $3x^{2}+4x+3$ is reducible over ${\mathrm{GF}}(q)$ . Hence, $\beta^{i}\in{\mathrm{GF}}(q)$ , which contradicts the fact that $1\leq i\leq\frac{q-1}{2}$ .

Collecting the conclusions in Cases 1 and 2 yields $|T_{1}\cap T_{2}|\geq 3$ or $T\leq 1$ . This completes the proof. ∎

The parameters of ${\mathcal{C}}_{(q,q+1,4,1)}$ and ${\mathcal{C}}(4)$ are documented in the following theorem.

Theorem 38.

Let $q\geq 9$ be an odd prime power. Then ${\mathcal{C}}_{(q,q+1,4,1)}$ has parameters $[q+1,q-5,d]$ and ${\mathcal{C}}(4)$ has parameters $[q+1,6,q-5]$ , where

1.

if $4\mid(q+1)$ , then $d=4$ ;
2.

if $4\mid(q-1)$ and $3\nmid(q+1)$ , then $d=5$ ;
3.

if $4\mid(q-1)$ and $3\mid(q+1)$ , then $5\leq d\leq 6$ , and $d=5$ provided that $5\nmid q$ .

Proof.

It is clear that ${\mathcal{C}}_{(q,q+1,4,1)}$ has length $q+1$ and dimension $q-5$ , and ${\mathcal{C}}(4)$ has length $q+1$ and dimension $6$ , respectively. Notice that $d({\mathcal{C}}(4))\in\{q-5,q-4\}$ , and $d({\mathcal{C}}(4))=q-4$ if and only if ${\mathcal{C}}(4)$ is MDS. Therefore, we only need to study the minimum distance of ${\mathcal{C}}_{(q,q+1,4,1)}$ .

Case I: $4\mid(q+1)$ . By Lemma 10, we obtain that $d({\mathcal{C}}_{(q,q+1,4,1)})=4$ . The desired result follows.

Case II: $4\mid(q-1)$ and $3\nmid(q+1)$ . By Lemma 36, there exist two distinct integers $i$ and $j$ with $1\leq i,j\leq\frac{q-1}{2}$ such that $({\mathrm{Tr}}(\beta^{i})+1)({\mathrm{Tr}}(\beta^{j})+1)+1=0$ . Let $x_{1}=\beta^{i}$ , $x_{2}=\beta^{-i}$ , $x_{3}=\beta^{j}$ , $x_{4}=\beta^{-j}$ and $x_{5}=1$ , then $|\{x_{1},x_{2},x_{3},x_{4},x_{5}\}|=5$ . It is easily verified that

	$\displaystyle\sigma_{5,2}$	$\displaystyle=x_{1}x_{2}+(x_{1}+x_{2})(x_{3}+x_{4})+x_{3}x_{4}+(x_{1}+x_{2}+x_{3}+x_{4})x_{5}$
		$\displaystyle=2+{\mathrm{Tr}}(\beta^{i}){\mathrm{Tr}}(\beta^{j})+{\mathrm{Tr}}(\beta^{i})+{\mathrm{Tr}}(\beta^{j})$
		$\displaystyle=({\mathrm{Tr}}(\beta^{i})+1)({\mathrm{Tr}}(\beta^{j})+1)+1=0.$

By Lemma 33, we deduce that $d({\mathcal{C}}_{(q,q+1,4,1)})=5$ .

Case III: $4\mid(q-1)$ and $3\mid(q+1)$ . In this case, $6\mid(q+1)$ . It is clear that

{\mathcal{C}}_{(q,q+1,6,1)}\subseteq{\mathcal{C}}_{(q,q+1,4,1)}.

Consequently, $d({\mathcal{C}}_{(q,q+1,4,1)})\leq d({\mathcal{C}}_{(q,q+1,6,1)})$ . By Lemma 10, we deduce that

d({\mathcal{C}}_{(q,q+1,4,1)})\in\{5,6\}.

Below we prove that $d({\mathcal{C}}_{(q,q+1,4,1)})=5$ for $5\nmid q$ .

Suppose $q=p^{m}$ , where $p>5$ is an odd prime and $m$ is a positive integer. Since $3\mid(q+1)$ , we deduce that $m$ is odd. Let $\gamma=\beta^{\frac{q+1}{p+1}}$ , then $\gamma$ is a primitive $(p+1)$ -th root of unity in ${\mathrm{GF}}(p^{2})$ . Let ${\mathcal{C}}$ be the cyclic code of length $p+1$ over ${\mathrm{GF}}(p)$ with generator polynomial $\mathbb{M}_{\gamma}(x)\mathbb{M}_{\gamma^{2}}(x)\mathbb{M}_{\gamma^{3}}(x)$ , then ${\mathcal{C}}$ is the narrow-sense BCH code of length $p+1$ over ${\mathrm{GF}}(p)$ with designed distance $4$ . Since $4\mid(q-1)$ and $m$ is odd, we get that $4\mid(p-1)$ . Similar to Case II, we can prove that $d({\mathcal{C}})=5$ . Consequently, there are five pairwise distinct integers $i_{1},i_{2},i_{3},i_{4},i_{5}$ with $0\leq i_{j}\leq p$ and $c_{i}\in{\mathrm{GF}}(p)^{*}$ such that

c_{1}x^{i_{1}}+c_{2}x^{i_{2}}+c_{3}x^{i_{3}}+c_{4}x^{i_{4}}+c_{5}x^{i_{5}}\in{\mathcal{C}}.

It is easily verified that

c_{1}x^{{\frac{q+1}{p+1}}i_{1}}+c_{2}x^{\frac{q+1}{p+1}i_{2}}+c_{3}x^{\frac{q+1}{p+1}i_{3}}+c_{4}x^{\frac{q+1}{p+1}i_{4}}+c_{5}x^{\frac{q+1}{p+1}i_{5}}\in{\mathcal{C}}_{(q,q+1,4,1)}.

It then follows that $d({\mathcal{C}}_{(q,q+1,4,1)})=5$ .

Collecting all the conclusions in Cases I, II and III, we complete the proof of this theorem. ∎

The parameters of $\overline{{\mathcal{C}}_{4}}$ and $(\overline{{\mathcal{C}}_{4}})^{\perp}$ are documented in the following theorem.

Theorem 39.

Let $q\geq 9$ be an odd prime power. Then $\overline{{\mathcal{C}}_{4}}$ has parameters $[q+2,7,q-5]$ and $(\overline{{\mathcal{C}}}_{4})^{\perp}$ has parameters $[q+2,q-5,d]$ , where

1.

if $4\mid(q+1)$ , then $d=5$ ;
2.

if $4\mid(q-1)$ and $3\nmid(q+1)$ , then $d=6$ ;
3.

if $4\mid(q-1)$ and $3\mid(q+1)$ , then $6\leq d\leq 7$ , and $d=6$ provided that $5\nmid q$ .

Proof.

By Theorem 7, ${\mathcal{C}}_{4}$ has parameters $[q+1,7,q-5]$ . Then $\overline{{\mathcal{C}}_{4}}$ has length $q+2$ and dimension $7$ . Consequently, $(\overline{{\mathcal{C}}_{4}})^{\perp}$ has length $q+2$ and dimension $q-5$ . By Theorems 9 and 38, we obtain that $d(\overline{{\mathcal{C}}_{4}})=d({\mathcal{C}}(4))=q-5$ . By Theorems 8 and 38, we deduce that

d((\overline{{\mathcal{C}}_{4}})^{\perp})=d({\mathcal{C}}_{(q,q+1,4,1)})+1.

The desired conclusions on the minimum distance of $(\overline{{\mathcal{C}}_{4}})^{\perp}$ follow from Theorem 38. This completes the proof. ∎

Example 40.

Let $q=5^{2}$ . Then the following hold.

1.

The code ${\mathcal{C}}_{4}$ has parameters $[26,7,20]$ and $({\mathcal{C}}_{4})^{\perp}$ has parameters $[26,19,8]$ .
2.

The BCH code ${\mathcal{C}}_{(q,q+1,4,1)}$ has parameters $[26,20,5]$ and ${\mathcal{C}}(4)$ has parameters $[26,6,20]$ .
3.

The code $\overline{{\mathcal{C}}_{4}}$ has parameters $[27,7,20]$ and $(\overline{{\mathcal{C}}_{4}})^{\perp}$ has parameters $[27,20,6]$ .

For the case $u=4$ , there is only one case where the minimum distance of $(\overline{{\mathcal{C}}_{u}})^{\perp}$ remains open, but we have the following conjecture.

Conjecture 41.

Let $q=5^{m}$ with $m\geq 3$ being odd, then $d({\mathcal{C}}_{(q,q+1,4,1)})=5$ and $d((\overline{{\mathcal{C}}_{4}})^{\perp})=6$ .

IV-E The case $u=\lfloor\frac{q-1}{2}\rfloor$

Let $q>5$ be a prime power and $u=\lfloor\frac{q-1}{2}\rfloor$ . Our task in this subsection is to settle the parameters of the code $\overline{{\mathcal{C}}_{u}}$ and its dual $(\overline{{\mathcal{C}}_{u}})^{\perp}$ . First of all, we have the following result.

Corollary 42.

Let $q>5$ be an odd prime power and $u=\frac{q-1}{2}$ . Then $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,q-2,4]$ and $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,4,\frac{q+3}{2}]$ .

Proof.

Since $q>5$ , $u>2$ . Note that $q+1\equiv 2\pmod{u}$ , we deduce that $u\nmid(q+1)$ . Clearly, $(u+1)\mid(q+1)$ . The desired results follow from Theorem 11. ∎

Starting from now on, let $q=2^{m}$ with $m\geq 3$ and $u=2^{m-1}-1$ , then ${\mathcal{C}}_{(q,q+1,u,1)}$ has checked polynomial $(x-1)\mathbb{M}_{\beta^{u}}(x)\mathbb{M}_{\beta^{u+1}}(x)$ , where $\beta$ is a primitive $(q+1)$ -th root of unity in ${\mathrm{GF}}(q^{2})$ . By Delsarte’s theorem, we have

{\mathcal{C}}_{(q,q+1,u,1)}=\left\{{\mathbf{c}}_{(a,b,c)}:a\in{\mathrm{GF}}(q),\ b,\ c\in{\mathrm{GF}}(q^{2})\right\},

(23)

where ${\mathbf{c}}(a,b,c)=(a+{\mathrm{Tr}}(b\beta^{ui}+c\beta^{(u+1)i}))_{i=0}^{q}$ , and ${\mathrm{Tr}}(x)$ denote the trace function from ${\mathrm{GF}}(q^{2})$ to ${\mathrm{GF}}(q)$ . Let ${\mathcal{C}}(u)$ denote the dual of the BCH code ${\mathcal{C}}_{(q,q+1,u,1)}$ , then ${\mathcal{C}}(u)$ is the cyclic code of length $q+1$ over ${\mathrm{GF}}(q)$ with generator polynomial $(x-1)\mathbb{M}_{\beta^{u}}(x)\mathbb{M}_{\beta^{u+1}}(x)$ . To determine the parameters of ${\mathcal{C}}_{(q,q+1,u,1)}$ and ${\mathcal{C}}(u)$ , we first prove the following lemma.

Lemma 43.

Let $q=2^{m}$ with $m\geq 3$ . Then there are three pairwise distinct integers $i,j,k$ with $1\leq i,j,k\leq 2^{m-1}$ such that ${\mathrm{Tr}}(\beta^{i}+\beta^{j}+\beta^{k})=0$ .

Proof.

It is easily verified that

x^{q+1}+1=(x+1)\prod_{i=1}^{2^{m-1}}(x^{2}+{\mathrm{Tr}}(\beta^{i})x+1),

and $\gcd(q+1,q)=1$ . Consequently,

{\mathrm{Tr}}(\beta^{i})\neq{\mathrm{Tr}}(\beta^{j})

(24)

for any two distinct integers $i$ and $j$ with $1\leq i,j\leq 2^{m-1}$ . We first prove that ${\mathrm{Tr}}(\beta^{i})\neq 0$ for each $1\leq i\leq 2^{m-1}$ . Suppose ${\mathrm{Tr}}(\beta^{i})=0$ , then $x^{2}+1$ divides $x^{q+1}+1$ , which contradicts the fact that $\gcd(2,q+1)=1$ . We now prove that ${\mathrm{Tr}}(\beta)\neq 1$ . Suppose ${\mathrm{Tr}}(\beta)=1$ , then $\beta^{2}+\beta+1=0$ . It follows that $\beta^{3}=1$ , which contradicts the fact that $\beta$ is a primitive $(q+1)$ -th root of unity. Let

S_{1}=\{{\mathrm{Tr}}(\beta^{i}):1\leq i\leq 2^{m-1}\}

and $S_{2}=\{{\mathrm{Tr}}(\beta){\mathrm{Tr}}(\beta^{i}):1\leq i\leq 2^{m-1}\}$ , then $|S_{1}|=|S_{2}|=2^{m-1}$ . It is easily verified that

$\displaystyle{\mathrm{Tr}}(\beta){\mathrm{Tr}}(\beta^{i})$	$\displaystyle=(\beta+\beta^{-1})(\beta^{i}+\beta^{-i})$
	$\displaystyle=\beta^{i+1}+\beta^{-(i+1)}+\beta^{i-1}+\beta^{-(i-1)}$
	$\displaystyle={\mathrm{Tr}}(\beta^{i+1}+\beta^{i-1}).$	(25)

Notice that ${\mathrm{Tr}}(\beta^{2^{m-1}+1})={\mathrm{Tr}}(\beta^{2^{m-1}})$ , we have

{\mathrm{Tr}}(\beta){\mathrm{Tr}}(\beta^{2^{m-1}})={\mathrm{Tr}}(\beta^{2^{m-1}}+\beta^{2^{m-1}-1}).

(26)

It is clear that ${\mathrm{Tr}}(\beta){\mathrm{Tr}}(\beta)={\mathrm{Tr}}(\beta^{2})$ , then $|S_{1}\cap S_{2}|\geq 1$ . If $|S_{1}\cap S_{2}|=1$ , then

|S_{1}\cup S_{2}|=|S_{1}|+|S_{2}|-|S_{1}\cap S_{2}|=q-1.

Notice that $S_{1}\cup S_{2}\subseteq{\mathrm{GF}}(q)^{*}$ , we get that

\displaystyle S_{1}\cup S_{2}=\{{\mathrm{Tr}}(\beta^{i}):1\leq i\leq 2^{m-1}\}\cup\{{\mathrm{Tr}}(\beta){\mathrm{Tr}}(\beta^{i}):2\leq i\leq 2^{m-1}\}={\mathrm{GF}}(q)^{*}.

It then follows that

\displaystyle 0=\sum_{x\in{\mathrm{GF}}(q)^{*}}x=\sum_{i=1}^{2^{m-1}}{\mathrm{Tr}}(\beta^{i})+\sum_{i=2}^{2^{m-1}}{\mathrm{Tr}}(\beta){\mathrm{Tr}}(\beta^{i})={\mathrm{Tr}}(\beta)^{2},

a contradiction. Therefore, $|S_{1}\cap S_{2}|\geq 2$ . Then there is an integer $i$ with $2\leq i\leq 2^{m-1}$ such that

{\mathrm{Tr}}(\beta){\mathrm{Tr}}(\beta^{i})={\mathrm{Tr}}(\beta^{j}),

where $1\leq j\leq 2^{m-1}$ . If $2\leq i\leq 2^{m-1}-1$ , by Equation (IV-E), we deduce that

{\mathrm{Tr}}(\beta^{i+1}+\beta^{i-1})={\mathrm{Tr}}(\beta^{j}).

Since ${\mathrm{Tr}}(\beta^{l})\neq 0$ for each $1\leq l\leq 2^{m-1}$ , we deduce that $j\notin\{i-1,i+1\}$ . The desired result follows. If $i=2^{m-1}$ , by Equation (26), we obtain that

{\mathrm{Tr}}(\beta^{2^{m-1}}+\beta^{2^{m-1}-1})={\mathrm{Tr}}(\beta^{j}).

Similarly, we have $j\notin\{2^{m-1},2^{m-1}-1\}$ . This completes the proof. ∎

Theorem 44.

Let $q=2^{m}$ with $m\geq 3$ and $u=2^{m-1}-1$ . Then ${\mathcal{C}}_{(q,q+1,u,1)}$ has parameters $[q+1,5,q-5]$ and ${\mathcal{C}}(u)$ has parameters $[q+1,q-4,5]$ .

Proof.

It is clear that ${\mathcal{C}}_{(q,q+1,u,1)}$ has length $q+1$ and dimension $5$ . We now study the minimum distance of ${\mathcal{C}}_{(q,q+1,u,1)}$ . Let ${\mathbf{c}}$ be a nonzero codeword of ${\mathcal{C}}_{(q,q+1,u,1)}$ . By Equation (23), there is $(a,b,c)\in{\mathrm{GF}}(q)\times{\mathrm{GF}}(q^{2})^{2}$ and $(a,b,c)\neq(0,0,0)$ such that

{\mathbf{c}}=(a+{\mathrm{Tr}}(b\beta^{ui}+c\beta^{(u+1)i}))_{i=0}^{q}.

It is easily verified that $a+{\mathrm{Tr}}(b\beta^{ui}+c\beta^{(u+1)i})=f_{a,b,c}(\beta^{i})$ , where

f_{a,b,c}(x)=a+bx^{2^{m-1}-1}+cx^{2^{m-1}}+c^{q}x^{2^{m-1}+1}+b^{q}x^{2^{m-1}+2}.

Note that

f_{(a,b,c)}(x)^{2}=a^{2}+b^{2}x^{q-2}+c^{2}x^{q}+c^{2q}x^{q+2}+b^{2q}x^{q+4}.

Consequently,

	$\displaystyle f_{(a,b,c)}(\beta^{i})^{2}$	$\displaystyle=a^{2}+b^{2}\beta^{-3i}+c^{2}\beta^{-i}+c^{2q}\beta^{i}+b^{2q}\beta^{3i}$
		$\displaystyle=(b^{2}+c^{2}\beta^{2i}+a^{2}\beta^{3i}+c^{2q}\beta^{4i}+b^{2q}\beta^{6i})\beta^{-3i}.$

Let $g_{a,b,c}(x)=b^{2}+c^{2}x^{2}+a^{2}x^{3}+c^{2q}x^{4}+b^{2q}x^{6}$ , then $f_{(a,b,c)}(\beta^{i})=0$ if and only if $g_{a,b,c}(\beta^{i})=0$ . Since $(a,b,c)\neq(0,0,0)$ , $g_{a,b,c}(x)\neq 0$ . Consequently, $g_{a,b,c}(x)$ has at most $6$ solutions. Therefore,

{\mathtt{wt}}({\mathbf{c}})\geq q+1-6=q-5.

It follows that $d({\mathcal{C}}_{(q,q+1,u,1)})\geq q-5$ .

We now prove that $g_{a,b,c}(x)$ has $6$ solutions $x\in\{\beta^{i}:1\leq i\leq q\}$ for some $(a,b,c)\neq(0,0,0)$ . By Lemma 43, there are three pairwise distinct integers $i,j,k$ with $1\leq i,j,k\leq 2^{m-1}$ such that

{\mathrm{Tr}}(\beta^{i}+\beta^{j}+\beta^{k})=0.

Clearly, $|\{\beta^{i},\beta^{-i},\beta^{j},\beta^{-j},\beta^{k},\beta^{-k}\}|=6$ . Put $u={\mathrm{Tr}}(\beta^{i})$ , $v={\mathrm{Tr}}(\beta^{j})$ , $w={\mathrm{Tr}}(\beta^{k})$ , then

	$\displaystyle\mathbb{M}_{\beta^{i}}(x)\mathbb{M}_{\beta^{j}}(x)\mathbb{M}_{\beta^{k}}(x)$	$\displaystyle=(x^{2}+ux+1)(x^{2}+vx+1)(x^{2}+wx+1)$
		$\displaystyle=x^{6}+(uv+uw+vw+1)x^{4}+uvwx^{3}+(uv+uw+vw+1)x^{2}+1.$

Set

a=(uvw)^{q/2},\ b=1,\ c=(uv+uw+vw+1)^{q/2}.

Then $\{\beta^{i},\beta^{-i},\beta^{j},\beta^{-j},\beta^{k},\beta^{-k}\}$ is a set of $6$ solutions of $g_{a,b,c}(x)$ . Consequently, the corresponding codeword ${\mathbf{c}}$ has weight $q-5$ . Therefore, $d({\mathcal{C}}_{(q,q+1,u,1)})=q-5$ .

It is clear that ${\mathcal{C}}(u)$ has length $q+1$ and dimension $q-4$ . Note that ${\mathcal{C}}_{(q,q+1,u,1)}$ is not MDS, then ${\mathcal{C}}(u)$ is not MDS. As a result, $d({\mathcal{C}}(u))\leq 5$ . Notice that $\beta^{i}$ is a zero of ${\mathcal{C}}(u)$ for each

i\in\{2^{m-1}-1,2^{m-1},2^{m-1}+1,2^{m-1}+2\}.

By the BCH bound, we deduce that $d({\mathcal{C}}(u))\geq 5$ . Therefore, $d({\mathcal{C}}(u))=5$ . This completes the proof. ∎

The parameters of the extended code $\overline{{\mathcal{C}}_{u}}$ are documented in the following theorem.

Theorem 45.

Let $q=2^{m}$ with $m\geq 3$ , and $u=2^{m-1}-1$ . Then $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,q-3,5]$ and $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,5,q-4]$ .

Proof.

By Theorem 7, ${\mathcal{C}}_{u}$ has parameters $[q+1,q-3,5]$ . Then $\overline{{\mathcal{C}}_{u}}$ has length $q+2$ and dimension $q-3$ . Consequently, $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has length $q+2$ and dimension $5$ . By Theorems 9 and 44, we obtain that $d(\overline{{\mathcal{C}}_{u}})=d({\mathcal{C}}(u))=5$ . By Theorems 8 and 44, we deduce that

d((\overline{{\mathcal{C}}_{u}})^{\perp})=d({\mathcal{C}}_{(q,q+1,u,1)})+1=q-4.

This completes the proof. ∎

Example 46.

Let $q=2^{4}$ and $u=2^{3}-1=7$ . Then the following hold.

1.

The code ${\mathcal{C}}_{u}$ has parameters $[17,13,5]$ and $({\mathcal{C}}_{u})^{\perp}$ has parameters $[17,4,14]$ .
2.

The BCH code ${\mathcal{C}}_{(q,q+1,u,1)}$ has parameters $[17,5,11]$ and ${\mathcal{C}}(u)$ has parameters $[17,12,5]$ .
3.

The extended code $\overline{{\mathcal{C}}_{u}}$ has parameters $[18,13,5]$ and $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[18,5,12]$ .

IV-F The case $u=\lfloor\frac{q+1}{2}\rfloor$

Let $q>2$ be a prime power and $u=\lfloor\frac{q+1}{2}\rfloor$ . Our task in this subsection is to settle the parameters of the code $\overline{{\mathcal{C}}_{u}}$ and its dual $(\overline{{\mathcal{C}}_{u}})^{\perp}$ . By Theorem 11, we have the following result.

Corollary 47.

Let $q$ be an odd prime power and $u=\frac{q+1}{2}$ . Then $\overline{{\mathcal{C}}_{u}}$ has parameters $[q+2,q,2]$ and $(\overline{{\mathcal{C}}_{u}})^{\perp}$ has parameters $[q+2,2,\frac{q+3}{2}]$ .

Starting from now on, let $q=2^{m}$ with $m\geq 2$ and $u=2^{m-1}$ . Recall that ${\mathcal{C}}_{u}$ is a $[q+1,q-1,3]$ MDS code over ${\mathrm{GF}}(q)$ and has generator polynomial $g_{u}(x)=\mathbb{M}_{\beta^{u}}(x)$ , where $\beta$ is a primitive $(q+1)$ -th root of unity in ${\mathrm{GF}}(q^{2})$ . Let ${\mathcal{C}}(u)$ be the dual of ${\mathcal{C}}_{(q,q+1,u,1)}$ , then ${\mathcal{C}}(u)$ is the cyclic code of length $q+1$ over ${\mathrm{GF}}(q)$ with generator polynomial $(x-1)\mathbb{M}_{\beta^{u}}(x)$ .

Lemma 48.

Let $q=2^{m}$ with $m\geq 2$ and $u=2^{m-1}$ . Then ${\mathcal{C}}_{(q,q+1,u,1)}$ is a $[q+1,3,q-1]$ MDS code over ${\mathrm{GF}}(q)$ .

Proof.

To prove the desired result, we only need to prove that ${\mathcal{C}}(u)$ is an MDS code over ${\mathrm{GF}}(q)$ . By definition, ${\mathcal{C}}(u)$ has length $q+1$ , dimension $q-2$ and $d({\mathcal{C}}(u))\leq 4$ . Since $\gcd(u,q+1)=1$ , we deduce that $\gamma=\beta^{u}$ is also a primitive $(q+1)$ -th root of unity. Notice that $\gamma^{-1}=\beta^{-u}$ , then $\gamma^{i}$ is a root of $(x-1)\mathbb{M}_{\beta^{u}}(x)$ for each $i\in\{-1,0,1\}$ . By the BCH bound, $d({\mathcal{C}}(u))\geq 4$ . Therefore, ${\mathcal{C}}(u)$ is a $[q+1,q-2,4]$ MDS code over ${\mathrm{GF}}(q)$ . This completes the proof. ∎

Theorem 49.

Let $q=2^{m}$ with $m\geq 2$ and $u=2^{m-1}$ . Then the following hold:

1.

${\mathcal{C}}_{u}$ is a $[q+1,q-1,3]$ MDS code over ${\mathrm{GF}}(q)$ .
2.

$({\mathcal{C}}_{u})^{\perp}$ is a $[q+1,2,q]$ MDS code over ${\mathrm{GF}}(q)$ .
3.

$\overline{{\mathcal{C}}_{u}}$ is a $[q+2,q-1,4]$ MDS code over ${\mathrm{GF}}(q)$ .
4.

$(\overline{{\mathcal{C}}_{u}})^{\perp}$ is a $[q+2,3,q]$ MDS code over ${\mathrm{GF}}(q)$ .

Proof.

The first two desired conclusions follow from Theorem 7. To prove that the latter two desired conclusions, we only need to prove $d((\overline{{\mathcal{C}}_{u}})^{\perp})=q$ . From Theorem 8 and Lemma 48, we deduce that $d((\overline{{\mathcal{C}}_{u}})^{\perp})=d({\mathcal{C}}_{(q,q+1,u,1)})+1=q$ . This completes the proof. ∎

The codes $(\overline{{\mathcal{C}}_{u}})^{\perp}$ are the second family of hyperoval codes presented in this paper.

V Summary and concluding remarks

It is observed that only a small number of families of NMDS codes with known weight distributions have been reported in [12, 22, 28, 29, 36, 38, 40, 41], as two NMDS codes over ${\mathrm{GF}}(q)$ with the same parameters $[n,k,n-k]$ may have different weight distributions. The first contribution of this paper is the several families of NMDS codes documented in Corollary 15, Theorems 20, 21 and 30, and the settlement of the weight distributions of three families of NMDS codes (see Theorems 25, 27 and 30). One of the infinite families of NMDS codes is both distance-optimal and dimension-optimal locally recoverable codes [35]. The second contribution of this paper is the MDS codes $\overline{{\mathcal{C}}_{2}}$ documented in Corollary 17 and the MDS codes ${\mathcal{C}}_{q/2}$ and $\overline{{\mathcal{C}}_{q/2}}$ documented in Theorem 49 for even $q$ . They are also both distance-optimal and dimension-optimal locally recoverable codes. The third contribution of this paper is the settlement of the parameters of $\overline{{\mathcal{C}}_{u}}$ and its dual for $u\in\left\{2,3,4,\lfloor\frac{q-1}{2}\rfloor,\lfloor\frac{q+1}{2}\rfloor\right\}$ .

In this paper, we settled the parameters of the extended code $\overline{{\mathcal{C}}_{u}}$ for only a few values of $u$ . It would be very challenging to study the parameters of the code $\overline{{\mathcal{C}}_{u}}$ and its dual $(\overline{{\mathcal{C}}_{u}})^{\perp}$ for $5\leq u\leq\lfloor\frac{q-3}{2}\rfloor$ . The reader is cordially invited to attack this problem. This shows that studying the extended code of a known linear code could be very challenging, even if the parameters of the linear code are known. This may explain why not much work on extended codes has been reported in the literature.

Although the NMDS codes $\overline{{\mathcal{C}}_{2}}$ , $\overline{{\mathcal{C}}_{3}}$ and $\overline{{\mathcal{C}}_{4}}$ have small dimensions (i.e., $3$ , $5$ and $7$ , respectively), they are interesting and valuable due to the following applications. These NMDS codes have applications in cryptography [33], finite geometry [8, 15, 17] and distributed data storage systems (see the explanations in Section I). Another application of these NMDS codes with small dimensions is the construction of subfield codes (i.e., the trace codes) with larger dimensions. We inform the reader that the binary subfield codes of $\overline{{\mathcal{C}}_{2}}$ , $\overline{{\mathcal{C}}_{3}}$ and $\overline{{\mathcal{C}}_{4}}$ are distance-optimal or have the best parameters for $q=2^{m}$ with $m\in\{3,4,5,6\}$ in every case except one case (see Table I). The reader is referred to [11, 19, 20, 39] for information about subfield codes, which are different from the subfield subcodes of linear codes.

Table I: The binary subfield codes of

\overline{{\mathcal{C}}_{u}}

$u$	$m$	subfield code	best distance	best dimension
$2,3$	$3$	$[10,7,2]$	optimal	$9$
$2,3$	$4$	$[18,9,6]$	optimal	optimal
$2,3$	$5$	$[34,11,12]$	optimal	$12$
$2,3$	$6$	$[66,13,26]$	best known	best known
$4$	$3$	$[10,9,2]$	optimal	optimal
$4$	$4$	$[18,17,2]$	optimal	optimal
$4$	$5$	$[34,21,4]$	$6$	$27$
$4$	$6$	$[66,25,16]$	best known	$29$ is best known

According to our experimental data, the two families of NMDS codes documented in Theorems 25, 27 and 30 do not support $2$ -designs, only special NMDS codes support $2$ -designs [12, 36, 40].

References

[1] V. Abatangelo, B. Larato, “Near-MDS codes arising from algebraic curves,” Disc. Math., vol. 301, no. 1, pp. 5–19, Sept. 2005.
[2] V. Abatangelo, B. Larato, “Elliptic near-MDS codes over $\mathbb{F}_{5}$ ,” Des. Codes Cryptogr., vol. 46, pp. 167–174, Feb. 2008.
[3] A. Aguglia, L. Giuzzi, A. Sonnino, “Near-MDS codes from elliptic curves,” Des. Codes Cryptogr., vol. 89, pp. 965–972, May 2021.
[4] D. Bartoli, M. Giulietti, G. Zini, “Complete $(k,3)$ -arcs from quartic curves,” Des. Codes Cryptogr., vol. 79, pp. 487–505, Jun. 2016.
[5] V. R. Cadambe, A. Mazumdar, “An upper bound on the size of locally recoverable codes,” In International Symposium on Network Coding, NetCod 2013, Calgary, AB, Canada, June 7-9, 2013, pp. 1–5, 2013.
[6] V. R. Cadambe, A. Mazumdar, “Bounds on the size of locally recoverable codes,” IEEE Trans. Inf. Theory, vol. 61, no. 11, pp. 5787–5794, Nov. 2015.
[7] M. Ceria, A. Cossidente, G. Marino, F. Pavese, “On near-MDS codes and caps,” Des. Codes Cryptogr., vol. 91, pp. 1095–1110, Mar. 2023.
[8] M. A. de Boer, “Almost MDS codes,” Des. Codes Cryptogr., vol. 9, pp. 143–155, Oct. 1996.
[9] C. Dahl, J. P. Pedersen, “Cyclic and pesudo-cyclic MDS codes of length $q+1$ ,” J. Combinat. Theory, Series A, vol. 59, no. 1, pp. 130–133, Jan. 1992.
[10] C. Ding, Codes from Difference Sets, World Scientific, Singapore, 2015.
[11] C. Ding, Z. Heng, “The subfield codes of ovoid codes,” IEEE Trans. Inf. Theory, vol. 65, no. 8, pp. 4715–4729, Aug. 2019.
[12] C. Ding, C. Tang, “Infinite families of near MDS codes holding $t$ -designs,” IEEE Trans. Inf. Theory, vol. 66, no. 9, pp. 5419–5428, Sept. 2020.
[13] C. Ding, C. Tang, Designs from Linear Codes, Second Edition, World Scientific, Singapore, 2022.
[14] S. Dodunekov, I. Landgev, “On near-MDS codes,” J. Geometry, vol. 54, pp. 30–43, Nov. 1995.
[15] A. Faldum, W. Willems, “Codes of small defect,” Des. Codes Cryptogr., vol. 10, pp. 341–350, Mar. 1997.
[16] M. Grassl, Bounds on the minimum distance of linear codes and quantum codes, Online available at http://www.codetables.de.
[17] M. Giulietti, F. Pasticci, “On the completeness of certain $n$ -tracks arising from elliptic curves,” Finite Fields Appl., vol. 13, no. 4, pp. 988–1000, Nov. 2007.
[18] P. Gopalan, C. Huang, H. Simitci, S. Yekhanin, “On the locality of codeword symbols,” IEEE Trans. Inf. Theory, vol. 58, no. 11, pp. 6925–6934, Nov. 2012.
[19] Z. Heng, C. Ding, “The subfield codes of hyperoval and conic codes,” Finite Fields Appl., vol. 56, pp. 308–331, Mar. 2019.
[20] Z. Heng, C. Ding, W. Wang, “Optimal binary linear codes from maximal arcs,” IEEE Trans. Inf. Theory, vol. 66, no. 9, pp. 5387–5394, Sept. 2020.
[21] Z. Heng, C. Li, W. Wang, “Constructions of MDS, near MDS and almost MDS codes from cyclic subgroups of $\mathbb{F}^{*}_{q^{2}}$ ,” IEEE Trans. Inf. Theory, vol. 68, no. 12, pp. 7817–7831, Dec. 2022.
[22] Z. Heng, X. Wang, “New infinite families of near MDS codes holding $t$ -designs,” Disc. Math., vol. 346, no. 10, 113538, Oct. 2023.
[23] W. C. Huffman, V. Pless, Fundamentals of Error-Correcting Codes, Cambridge University Press, Cambridge, 2003.
[24] K. Ireland, M. Rosen, A classical introduction to modern number theory, second ed., Grad. Texts in Math., vol. 84, Springer-Verlag, 1990.
[25] C. Li, P. Wu, F. Liu, “On two classes of primitive BCH codes and some related codes,” IEEE Trans. Inf. Theory, vol. 65, no. 6, pp. 3830–3840, Jun. 2009.
[26] Q. Liu, C. Ding, S. Mesnager, C. Tang, V. D. Tonchev, “On infinite families of narrow-sense antiprimitive BCH codes admitting $3$ -transitive automorphism groups and their consequences,” IEEE Trans. Inf. Theory, vol. 68, no. 5, pp. 3096–3107, May 2021.
[27] S. Li, “The minimum distance of some narrow-sense primitive BCH codes,” SIAM J. Disc. Math., vol. 31, no. 4, pp. 2530–2569, Apr. 2017.
[28] X. Li, Z. Heng, Constructions of near MDS codes which are optimal locally recoverable codes, arXiv:2204.11208v1
[29] X. Li, Z. Heng, “A construction of optimal locally recoverable codes,” Cryptogr. Commun., vol. 15, pp. 553–563, May 2023.
[30] Y. Liu, R. Li, Q. Fu, L. Lu, Y. Rao, “Some binary BCH codes with length $n=2^{m}+1$ ,” Finite Fields Appl., vol. 55, pp. 109–133, Jan. 2019.
[31] F. J. MacWilliams, N. J. A. Sloane, The Theory of Error-Correcting Codes, North-Holland, Amsterdam, 1977.
[32] S. Marcugini, A. Milani, F. Pambianco, “NMDS codes of maximum length over $F_{q}$ , $8\leq q\leq 11$ ,” IEEE Trans. Inf. Theory, vol. 48, no. 4, pp. 963–966, Apr. 2002.
[33] S. Mehta, S. Sen, “Secret sharing using near-MDS codes,” in Codes, Cryptology and Information Security, pp. 195–214, 2019.
[34] X. Shi, Q. Yue, Y. Wu, “The dual-containing primitive BCH codes with the maximum designed distance and their applications to quantum codes,” Des. Codes Cryptogr., vol. 87, no. 9, pp. 2165–2183, Sept. 2019.
[35] P. Tan, C. Fan, C. Ding, C. Tang, Z. Zhou, “The minimum locality of linear codes,” Des. Codes Cryptgr., vol. 91, pp. 83–114, Jan. 2023.
[36] C. Tang, C. Ding, “An infinite family of linear codes supporting $4$ -designs,” IEEE Trans. Inf. Theory, vol. 67, no. 1, pp. 244–254, Jan. 2021.
[37] K. Tzeng, C. Hartmann, “On the minimum distance of certain reversible cyclic codes (corresp.),” IEEE Trans. Inf. Theory, vol. 16, no. 5, pp. 644–646, Sept. 1970.
[38] Q. Wang, Z. Heng, “Near MDS codes from overal polynomials,” Disc. Math., vol. 344, no. 4, 112277, Apr. 2021.
[39] X. Wang, D. Zheng, “The subfield codes of several classes of linear codes,” Cryptogr. Commun., vol. 12, pp. 1111–1131, Nov. 2020.
[40] G. Xu, X. Cao, L. Qu, “Infinite families of $3$ -designs and $2$ -designs from almost MDS codes,” IEEE Trans. Inf. Theory, vol. 68, no. 7, pp. 4344–4353, Jul. 2022
[41] L. Xu, C. Fan, D. Han, “Near-MDS codes from maximal arcs in PG $(2,q)$ ,” arXiv:2208.08578 [cs.IT].
[42] H. Yan, H. Liu, C. Li, S. Yang, “Parameters of LCD BCH codes with two lengths,” Adv. in Math. of Comm., vol. 12, no. 3, pp. 579–594, Jul. 2018.
[43] Q. Yan, J. Zhou, “Infinite families of linear codes supporting more $t$ -designs,” IEEE Trans. Inf. Theory, vol. 68, no. 7, pp. 4365–4377, Jul. 2022.
[44] H. Zhu, M. Shi, X. Wang, Tor Helleseth, “The $q$ -ary antiprimitive BCH codes,” IEEE Trans. Inf. Theory, vol. 68, no. 3, pp. 1683–1695, Mar. 2022.

The extended codes of a family of reversible MDS cyclic codesThanks: Z. Sun’s research was supported by The National Natural Science Foundation of China under Grant Number 62002093. C. Ding’s research was supported by the Hong Kong Research Grants Council, Proj. No. 16301123.

Abstract

Index Terms:

I Introduction of motivations, objectives and methodology

II Preliminaries

II-A Some operations on linear codes

Lemma 1.

Lemma 2.

Proof.

Lemma 3.

Proof.

Lemma 4.

Proof.

II-B Cyclic codes and BCH codes over finite fields

II-C The weight distributions of MDS codes and NMDS codes

Lemma 5.

Lemma 6.

III The family of classical MDS cyclic codes

Theorem 7.

IV The extended codes of the family of MDS cyclic codes

IV-A Some general results of the extended code 𝒞u¯\overline{{\mathcal{C}}_{u}} and its dual

Theorem 8.

Proof.

Theorem 9.

Proof.

Lemma 10.

Theorem 11.

Proof.

Lemma 12.

Theorem 13.

Proof.

Example 14.

IV-B The case u=2u=2

Corollary 15.

Proof.

Example 16.

Corollary 17.

Proof.

Example 18.

IV-C The case u=3u=3

Lemma 19.

Proof.

Theorem 20.

Proof.

Theorem 21.

Proof.

Example 22.

Lemma 23.

Lemma 24.

Theorem 25.

Proof.

Example 26.

Theorem 27.

Example 28.

IV-D The case u=4u=4

Lemma 29.

Theorem 30.

Proof.

Example 31.

Example 32.

Lemma 33.

Proof.

Lemma 34.

Proof.

Lemma 35.

Proof.

Lemma 36.

Proof.

Lemma 37.

Proof.

Theorem 38.

Proof.

Theorem 39.

Proof.

Example 40.

Conjecture 41.

IV-E The case u=⌊q−12⌋u=\lfloor\frac{q-1}{2}\rfloor

Corollary 42.

Proof.

Lemma 43.

IV-A Some general results of the extended code $\overline{{\mathcal{C}}_{u}}$ and its dual

IV-B The case $u=2$

IV-C The case $u=3$

IV-D The case $u=4$

IV-E The case $u=\lfloor\frac{q-1}{2}\rfloor$

IV-F The case $u=\lfloor\frac{q+1}{2}\rfloor$