1 Introduction
Let \(V=V(n,{\mathbb K})\) be an n-dimensional vector space over a field \({\mathbb K}\), we will denote by \({{\,\mathrm{{PG}}\,}}(V)\) as well as \({{\,\mathrm{{PG}}\,}}(n-1,{\mathbb K})\) the projective space induced by it.
We refer to [
5] for the definition of dimension, degree, smoothness and tangent space of an algebraic variety and for the methods and techniques used to study classical varieties.
The Veronese variety
\(\mathcal {V}_d\) of degree
d and dimension
\(n-1\) is a classical algebraic variety widely studied over fields of any characteristic [
5,
6] and it is the image of the Veronese map
$$\begin{aligned} \nu _d: (x_0,x_1,\ldots ,x_{n-1}) \in {{\,\mathrm{{PG}}\,}}(n-1,{\mathbb K}) \longrightarrow (\ldots ,X_I,\ldots ) \in {{\,\mathrm{{PG}}\,}}\left( \left( {\begin{array}{c}n+d-1\\ d\end{array}}\right) -1,{\mathbb K}\right) , \end{aligned}$$
where
\(X_I\) ranges over all the possible monomials of degree
d in
\(x_0,x_1,\ldots ,x_{n-1}\).
The Veronese map can be defined also by
$$\begin{aligned} \nu _{d} : \langle v \rangle \in {{\,\mathrm{{PG}}\,}}(n-1,{\mathbb K}) \longrightarrow \langle v \otimes v \otimes \cdots \otimes v \rangle \in {{\,\mathrm{{PG}}\,}}\left( \left( {\begin{array}{c}n+d-1\\ d\end{array}}\right) -1,{\mathbb K}\right) . \end{aligned}$$
Now, let
\(V_i\) be
\(n_i\)-dimensional vector spaces over the field
\({\mathbb K}\),
\(i=0,1,\ldots ,d-1\). A
Segre variety of type \((n_0,n_1,\ldots ,n_{d-1})\) in
\({{\,\mathrm{{PG}}\,}}\bigl (\bigotimes _{i=0}^{d-1} V_i \bigr )\) is the set
$$\begin{aligned} \Sigma _{n_0-1,n_1-1,\ldots ,n_{d-1}-1}= \bigl \{\langle v_0 \otimes v_1 \otimes \cdots \otimes v_{d-1} \rangle \,\,|\,\, v_i \in V_i \setminus \{0\}, \, i=0,1,\ldots ,d-1 \bigr \}.\nonumber \\ \end{aligned}$$
(1)
If
\(n_0=\cdots =n_{d-1}=n\), we write
\(\Sigma _{(n-1)^d}\) instead of
\(\Sigma _{n-1,n-1,\ldots ,n-1}\). Then it is clear that
\(\mathcal V_{d}\) turns out to be a linear section of the Segre variety product of
\({{\,\mathrm{{PG}}\,}}(n-1,{\mathbb K})\) for itself
d times.
If
\(\mathbf {\zeta }\) is a collineation of
\({{\,\mathrm{{PG}}\,}}\bigl (V^{\otimes d} \bigl )\) fixing
\(\Sigma _{(n-1)^d}\), then there exist
\(\zeta _i, i=0,1,\ldots ,d-1\) semilinear maps of
\({{\,\mathrm{{PG}}\,}}(V)\), with the same companion field automorphism, and a permutation
\(\tau \) on
\(\{0,1,\ldots ,d-1\}\) such that
$$\begin{aligned} \langle v_0 \otimes v_1 \otimes \cdots \otimes v_{d-1} \rangle ^{\mathbf {\zeta }}=\left\langle v_{\tau (0)}^{\zeta _0}\otimes v_{\tau (1)}^{\zeta _1}\otimes \cdots \otimes v_{\tau (d-1)}^{\zeta _{d-1}}\right\rangle , \end{aligned}$$
for a proof of this in positive characteristic see [
16].
Let
\(\mathcal {L}_h\) be the set of all projective subspaces of dimension
h of
\({{\,\mathrm{{PG}}\,}}(n-1,{\mathbb K})\), and consider
$$\begin{aligned} g_{n,h}: \langle v_0, v_1,v_2,\ldots ,v_h \rangle \in \mathcal {L}_h \longrightarrow \langle v_0 \wedge v_1 \wedge v_2 \wedge \cdots \wedge v_h \rangle \in {{\,\mathrm{{PG}}\,}}\bigl ( \bigwedge \,^{h+1}V \bigr ), \end{aligned}$$
where
\(\wedge \) is the wedge product and
\(\bigwedge ^{h+1}V\) the
\((h+1)\)th exterior power of
V. This map is called
Grassmann embedding and its image
\(\mathcal {G}_{n,h}(V)\) is called
Grassmannian of subspaces of dimension h of
\({{\,\mathrm{{PG}}\,}}(V)\). It is well-known that
\(\mathcal {G}_{n,h}(V)\) is an algebraic variety which is the complete intersection of certain quadrics, see [
5].
Let
\({\mathbb K}\) be the Galois field
\({\mathbb F}_{q^t}\) of order
\(q^t\),
\(A={{\,\mathrm{{Aut}}\,}}({\mathbb K})\) be the automorphism group of
\({\mathbb K}\) and
\(\varvec{\sigma }=(\sigma _0,\ldots , \sigma _{d-1}) \in A^d\),
\(d \ge 1\). The aim of this paper is to study the following generalization of the Veronese map
$$\begin{aligned} \nu _{d,\varvec{\sigma }} : \langle v \rangle \in {{\,\mathrm{{PG}}\,}}(n-1,{\mathbb K}) \longrightarrow \langle v^{\sigma _0} \otimes v^{\sigma _1} \otimes \cdots \otimes v^{\sigma _{d-1}} \rangle \in {{\,\mathrm{{PG}}\,}}(n^d-1,{\mathbb K}) \end{aligned}$$
and some properties of its image that will be called here the
\((d,\varvec{\sigma })\)-
Veronese variety \(\mathcal V_{d,\varvec{\sigma }}\). For
\(d=t\),
\(\sigma \) a generator of
\(\textrm{Gal}({\mathbb F}_{q^t}|{\mathbb F}_{q})\) and
\(\varvec{\sigma }=(1,\sigma ,\sigma ^2,\ldots ,\sigma ^{t-1})\), the
\((t,\varvec{\sigma })\)-Veronese variety
\(\mathcal V_{t,\varvec{\sigma }}\) is the variety studied in [
9,
11,
13]. Such a variety is the Grassmann embedding of the Desarguesian spread of
\({{\,\mathrm{{PG}}\,}}(nt-1,{\mathbb F}_q)\) and it has been used to construct codes [
4] and (partial) ovoids of quadrics, see [
9,
12].
A
\([\nu ,\kappa ]\)-
linear code \({\mathcal C}\) is a subspace of the vector space
\({\mathbb F}_{q}^\nu \) of dimension
\(\kappa \). The
weight of a codeword is the number of its entries that are nonzero and the
Hamming distance between two codewords is the number of entries in which they differ. The distance
\(\delta \) of a linear code is the minimum distance between distinct codewords and it is equals to the minimum weight. A linear code of length
\(\nu \), dimension
\(\kappa \), and minimum distance
\(\delta \) is called a
\([\nu ,\kappa ,\delta ]\)-code. A matrix
H of order
\((\nu -\kappa ) \times \nu \) such that
$$\begin{aligned} {\textbf{x}}H^T={\textbf{0}}\quad \text {for all} \,\, {\textbf{x}} \in {\mathcal C}\end{aligned}$$
is called a
parity check matrix for
\({\mathcal C}\). The minimum weight, and hence the minimum distance, of
\({\mathcal C}\) is at least
w if and only if any
\(w-1\) columns of
H are linearly independent [
10, Theorem 10, p. 33]. Each linear
\([\nu ,\kappa ,\delta ]\)-code
\({\mathcal C}\) satisfies the following inequality
$$\begin{aligned} \delta \le \nu -\kappa +1, \end{aligned}$$
called
Singleton bound. If
\(\delta =\nu -\kappa +1\),
\({\mathcal C}\) is called
maximum distance separable or
MDS, while if
\(\delta =\nu -\kappa \) the code is called
almost MDS. These can be related to some subsets of points in the projective space. More precisely,
\({\mathcal C}\) is a
\([\nu ,\kappa ,\delta ]\)-linear code if and only if the columns of its parity check matrix
H can be seen as
\(\nu \) points in
\({{\,\mathrm{{PG}}\,}}(\nu -\kappa -1,q)\) each
\(\delta -1\) of which are in general position, [
2, Theorem 1]. Then, the existence of MDS or almost MDS linear codes is equivalent to the existence of arcs or tracks in projective spaces, respectively.
Here, we study the variety \(\mathcal V_{d,\varvec{\sigma }}\) and we will prove that it is the Grassmann embedding of a normal rational scroll and that any \(d+1\) points of it are in general position, i.e. any \(d+1\) points of \(\mathcal V_{d,\varvec{\sigma }}\) are linearly independent. Moreover, we give a characterization of \(d+2\) linearly dependent points of this variety and investigate how such a property is interesting for a linear code \({\mathcal C}_{d,\varvec{\sigma }}\) that can be associated to the variety.
2 The variety \(\mathcal V_{d,\varvec{\sigma }}\)
Let \(V=V(n,{\mathbb K})\) be an n-dimensional vector space over the field \({\mathbb K}\) and \({{\,\mathrm{{PG}}\,}}(V)={{\,\mathrm{{PG}}\,}}(n-1,{\mathbb K})\) be the induced projective space. In particular, if \({\mathbb K}\) is the Galois field of order \(q^t\), we will denote the projective space by \({{\,\mathrm{{PG}}\,}}(n-1,q^t)\)
Let
\(A={{\,\mathrm{{Aut}}\,}}({\mathbb K})\) be the automorphism group of
\({\mathbb K}\) and
\(\varvec{\sigma }=(\sigma _0,\ldots , \sigma _{d-1}) \in A^d\),
\(d \ge 1\), and define the map
$$\begin{aligned} \nu _{d,\varvec{\sigma }} : \langle v \rangle \in {{\,\mathrm{{PG}}\,}}(V) \longrightarrow \langle v^{\sigma _0} \otimes v^{\sigma _1} \otimes \cdots \otimes v^{\sigma _{d-1}} \rangle \in {{\,\mathrm{{PG}}\,}}\bigl (V^{\otimes d} \bigl ). \end{aligned}$$
(2)
Up to the action of the group P
\(\Gamma \)L(V), we may assume that
\(\sigma _0=1\). It is clear that the map
\(\nu _{d,\varvec{\sigma }}\) is an injection of
\({{\,\mathrm{{PG}}\,}}(V)\) into
\({{\,\mathrm{{PG}}\,}}(V^{\otimes d})\) by the injectivity of the Segre map.
We will call
\(\nu _{d,\varvec{\sigma }}\) the
\((d,\varvec{\sigma })\)-
Veronese embedding and, as defined before, its image
\(\mathcal V_{d,\varvec{\sigma }}\) the
\((d,\varvec{\sigma })\)-Veronese variety. Then
\(\mathcal V_{d,\varvec{\sigma }}\) is a rational variety of dimension
\(n-1\) in
\({{\,\mathrm{{PG}}\,}}(N-1,{\mathbb K})\),
\(N=n^d\) and it has as many points as
\({{\,\mathrm{{PG}}\,}}(n-1,q^t)\), see [
5,
6].
As a consequence of [
16, Theorems 3.5 and 3.8], one gets the following
Note that applying the map
$$\begin{aligned} \left\langle v\otimes v^{\sigma _1}\otimes \cdots \otimes v^{\sigma _{d-1}}\right\rangle ^{\mathbf {\zeta } }=\left\langle v^{\zeta _0}\otimes v^{\zeta _1\sigma _1}\otimes \cdots \otimes v^{\zeta _{d-1}\sigma _{d-1}} \right\rangle , \end{aligned}$$
where
\(\zeta _i\) is a bijective semilinear map, we get a subvariety of
\(\Sigma _{(n-1)^d}\) projectively equivalent to
\(\mathcal V_{d,\varvec{\sigma }}\).
Although many of the results also hold in the case of a general field, from now on it will be assumed, that
\({\mathbb K}\) is the Galois field
\({\mathbb F}_{q^t}\) of
\(q^t\) elements and
\(\varvec{\sigma }=(\sigma _0,\sigma _1,\ldots ,\sigma _{d-1}) \in G^d\) with
\(G={{\,\textrm{Gal}\,}}({\mathbb F}_{q^t} |\, {\mathbb F}_q)\). Moreover, since any element
\(\sigma _i \in G\) is a map of the type
\(\sigma _i : x \mapsto x^{q^{h_i}}\) with
\(0 \le h_i < t\) and
\(0 \le i \le d-1\), hereafter we will suppose that
$$\begin{aligned} \varvec{\sigma }=(\underbrace{\sigma _0,\ldots ,\sigma _0}_{d_0\, \text {times}},\underbrace{\sigma _1,\ldots ,\sigma _1}_{d_1\,\text {times}}, \ldots ,\underbrace{\sigma _{m},\ldots ,\sigma _{m}}_{d_m\, \text {times}}), \end{aligned}$$
where
\(0=h_0<h_1<\cdots<h_m<t\) and we will consider the vector
\(d_{\varvec{\sigma }}=(d_0,d_1,\ldots ,d_m)\) where
\(d_j\) is the occurrence of
\(\sigma _j\) in
\(\varvec{\sigma }\),
\(0 \le j \le m\). Clearly
\(d_0+d_1+\ldots +d_m=d\). If
\(\varvec{\sigma }\in G^d\), the integer
$$\begin{aligned} |\varvec{\sigma }|= \sum _{i=0}^{d-1}q^{h_i}=\sum _{i=0}^{m}d_iq^{h_i}. \end{aligned}$$
(3)
will be called
norm of
\(\varvec{\sigma }\).
Since we consider the ring of polynomials
\({\mathbb F}_{q^t}[x_0,x_1,\ldots ,x_{n-1}]\) actually as the quotient
\({\mathbb F}_{q^t}[x_0,x_1,\ldots ,x_{n-1}]/(x_0^{q^t}-x_0,x_1^{q^t}-x_1,\ldots ,x_{n-1}^{q^t}-x_{n-1})\),
from now on we assume \(|\varvec{\sigma }|< q^t\), so that distinct polynomials will be distinct functions over
\({\mathbb F}_{q^t}\). By injectivity of map in (
2), it is clear that
\((d,\varvec{\sigma })\)-Veronese variety
\(\mathcal V_{d,\varvec{\sigma }}\) has as many points as
\({{\,\mathrm{{PG}}\,}}(n-1,q^t)\).
Let
\(\{e_i \, \,|\,\,i=0,1,\ldots ,nd-1\}\) be the canonical basis of
\(V(nd,{\mathbb F}_{q^t})=V(nd,q^t)\) and let
\(\Pi \cong {{\,\mathrm{{PG}}\,}}(n-1,q^t)\) be the subspace of
\({{\,\mathrm{{PG}}\,}}(nd-1,q^t)\) spanned by
\(\{\langle e_i \rangle \,\,|\,\,0 \le i \le n-1\}\). Let
\(\phi \) be the collineation of
\({{\,\mathrm{{PG}}\,}}(nd-1,q^t)\) such that
$$\begin{aligned} \langle e_i \rangle \mapsto \langle e_{i+n} \rangle , \end{aligned}$$
where the subscripts are taken modulo
nd. As done in [
4, Sect. 4], for any
\(\langle v_i \rangle \in \Pi ^{\phi ^i}\), we can identify
\(v_0\otimes v_1\otimes v_2\otimes \cdots \otimes v_{d-1}\) with
\(v_0\wedge v_1\wedge v_{2}\wedge \cdots \wedge v_{d-1}\). Therefore,
\(\mathcal V_{d,\varvec{\sigma }}\) is the Grassmann embedding of the
d-fold normal rational scroll
$$\begin{aligned} S^{\varvec{\sigma }}_{n-1,n-1,\ldots ,n-1}=\bigl \{\left\langle P^{\sigma _0 },P^{\phi \sigma _1},P^{\phi ^2 \sigma _2},\ldots ,P^{\phi ^{d-1} \sigma _{d-1}}\bigr \rangle \, |\,P \in \Pi \right\} \end{aligned}$$
of
\({{\,\mathrm{{PG}}\,}}(nd-1,q^t)\), see [
5, Chap. 8] for a definition of normal rational scroll.
By (
2), a point of
\({{\,\mathrm{{PG}}\,}}(n-1,q^t)\) with homogeneous coordinates
\((x_0,x_1,\ldots ,x_{n-1})\) is mapped by
\(\nu _{d,\varvec{\sigma }}\) into a point of coordinates
$$\begin{aligned} \left( \ldots ,\displaystyle \prod _{j=0}^{m}X_{I_j}^{\sigma _j},\ldots \right) , \end{aligned}$$
where
\(X_{I_j}\) is a monomial of degree
\(d_j\) in the variables
\(x_0,x_1,\ldots ,x_{n-1}\). Hence, the
\((d,\varvec{\sigma })\)-Veronese variety
\(\mathcal V_{d,\varvec{\sigma }}\) is contained in a projective space of vector space dimension
$$\begin{aligned} N=N_0N_1\cdots N_m, \quad N_j=\left( {\begin{array}{c}n+d_j-1\\ d_j\end{array}}\right) ,\quad j=0,1,\ldots ,m. \end{aligned}$$
(4)
Let
\(\sigma _0,\sigma _1,\ldots \sigma _m\) distinct automorphisms in
\(\varvec{\sigma }\) and
\(d_{\varvec{\sigma }}\) the vector of their occurrences and suppose that
\(d_i\sigma _i\ne d_j\sigma _j\) for all
\(i,j=0,1,\ldots ,m\) distinct, then we get exactly
\(N=n^d\) distinct monomials of type
\(\displaystyle \prod \nolimits _{j=0}^{m}X_{I_j}^{\sigma _j}\). This is not the case anymore if
\(d_i\sigma _i=d_j\sigma _j\) for some
\(i\ne j\). For example, if
\(q=2\),
\(\varvec{\sigma }=(1,1,2)\), then
\(d_0=2,d_1=1\) and hence
\(d_0=d_1\sigma _1\). Then
$$\begin{aligned} (x_0,x_1)\otimes (x_0,x_1)\otimes \left( x_0^2,x_1^2\right) =\left( x_0^4,x_0^2x_1^2,x_0^3x_1,x_0x_1^3,x_0^3x_1,x_0x_1^3,x_0^2x_1^2,x_1^4\right) , \end{aligned}$$
and we get 5 distinct monomials and
\(\mathcal V_{3,\varvec{\sigma }}\) is in fact contained in a projective space of vector space dimension less than
\(N=6\).
Recall that an
r-
hypersurface of
\({{\,\mathrm{{PG}}\,}}(n-1,q^t)\) is a variety such that its points have coordinates vanish an
r-form of
\({\mathbb F}_{q^t}[X_0,\ldots ,X_{n-1}].\) If
\(r=2\), an
r-hypersurface is called
quadric. In [
14], it is shown a lower bound on the degree of an
r-hypersurface
\(\mathcal {D}\) of
\({{\,\mathrm{{PG}}\,}}(n-1,q^t)\) after which
\(\mathcal {D}\) could contain all points of the projective space. More precisely,
Let
I be a multi-index of the form
\(I=I_0I_1\cdots I_m\), where
\(I_j\) is a multi-index corresponding to a monomial in
\(x_0,x_1,\ldots ,x_{n-1}\) of degree
\(d_j\). Once we have labelled the coordinates of
\({{\,\mathrm{{PG}}\,}}(N-1,q^t)\) according to the multi-index
I, we can define a natural linear map
\(\psi \) that sends the hyperplane of
\({{\,\mathrm{{PG}}\,}}(N-1,q^t)\) of equation
\(\displaystyle \sum \nolimits _{I}a_Iz_I=0\) to the
\(\varvec{\sigma }\)-
hypersurface of equation
$$\begin{aligned} \sum _{I}a_I \displaystyle \prod _{j=0}^{m}X_{I_j}^{\sigma _j}=0. \end{aligned}$$
Then, by Theorem
2.4, we get the following result.
In the following, we generalize some results proved in [
3, Sect. 2] for the SLP-variety.
In order to prove the next Corollary, we need the following
Since we have assumed
\(|\varvec{\sigma }|<q^t\), Lemma
2.9 always applies to
\(\mathcal {V}_{d,\varvec{\sigma }}\).
Finally, since the algebraic variety \(\Sigma _{(n-1)^d}\) has dimension \(d(n-1)\) and degree \(\left( {\begin{array}{c}d(n-1)\\ n-1,n-1,\cdots ,n-1\end{array}}\right) =\frac{(d(n-1))!}{d(n-1)!}\), a general subspace of \({{\,\mathrm{{PG}}\,}}(N-1,q^t)\) of codimension \(d(n-1)\) contains at most \(\frac{(d(n-1))!}{d(n-1)!}\) points of \(\mathcal {V}_{d,\varvec{\sigma }}\).
Moreover, the Segre variety is smooth and hence the tangent space
\(T_P(\Sigma _{(n-1)^d})\) to
\(\Sigma _{(n-1)^d}\) at a point
\(P= \langle v_0\otimes v_1\otimes \cdots \otimes v_{d-1} \rangle \) has dimension
\(d(n-1)\) and it spanned by the
d subspaces
$$\begin{aligned} \langle \langle v_0 \otimes v_1 \otimes \cdots v_{i-1} \otimes u_{i} \otimes v_{i+1} \otimes \cdots \otimes v_{d-1} \rangle \,|\, \langle u_i\rangle \in {{\,\mathrm{{PG}}\,}}(n-1,q^t)\rangle \cong {{\,\mathrm{{PG}}\,}}(n-1,q^t). \end{aligned}$$
These subspaces pairwise intersect only in
P and they are the maximal subspaces contained in
\(\Sigma _{(n-1)^d}\) through the point
P, and
\(\Sigma _{(n-1)^d}\) does not share with
\(\mathcal V_{d,\varvec{\sigma }}\) the property proved in Corollary
2.7. We have, in fact,
\(T_P(\Sigma _{(n-1)^d})\cap \mathcal V_{d,\varvec{\sigma }}=P\) for each
\(P \in \mathcal V_{d,\varvec{\sigma }}\).
3 The code \(\mathcal {C}_{d,\varvec{\sigma }}\)
As we have seen in Example
2.3, the SLP-variety turns out to be a variety of a subgeometry of order
q, even though the array
\(\varvec{\sigma }\) is defined on a finite field of order
\(q^t\), hence among all the possible choice of
\(\varvec{\sigma }\) and
n, for
q ‘big enough’
\(\mathcal V_{t,\sigma }\) is the variety with the most ‘dense’ set of points of a projective space with the property that any
\(d+1\) points are independent. In this case, since
\(d=t\) and, as proved in [
3],
\(t+2\) linearly dependent points are contained in a normal rational curve of degree
t of
\({{\,\mathrm{{PG}}\,}}(t,q)\),
\(q>t\).
For the classical Veronese variety of degree
d, hence for
\(\varvec{\sigma }={\textbf {1}}\), Theorem
2.11 implies that
\(d+2\) linearly dependent points are contained in the Veronese embedding of degree
d of a line, hence in a normal rational curve of degree
d of
\({{\,\mathrm{{PG}}\,}}(d,q^t)\).
Finally, for a general
\((d,\varvec{\sigma })\)-Veronese variety, if
\(d+2>q'+1\), with
\(q'\) defined as in Corollary
2.11, every
\(d+2\) points of
\(\mathcal {V}_{d,\varvec{\sigma }}\) are linearly independent, hence, for ‘small’
\(q'\), it provides a dense set of points with that property. More precisely, we get
\(\frac{q^{nt}-1}{q^t-1}\) points in
\({{\,\mathrm{{PG}}\,}}(N-1,q^t)\) such that any
\(d+2\) of them are in general position. Sets of points with properties of this sort are studied for their connections with linear codes.
If
H is the matrix whose columns are the coordinates vectors of the points of the variety
\(\mathcal V_{d,\varvec{\sigma }}\), we get a code
\({\mathcal C}_{d,\varvec{\sigma }}\) and we may study the minimum distance of it and characterize the codewords of minimum weight (for an overview on this topic, see, e.g., [
1]).
Clearly, the order of the columns of
H is arbitrary, so that Definition
3.1 makes sense only up to code equivalence, as a permutation of the columns that is not usually an automorphism of the code, see [
3, Remark 3.3].
As showed in [
3, Theorem 3.5], the following result holds
Now, as in [
3, Theorem 3.7], by the characterizations of sets of
\(d+2\) points of
\(\mathcal V_{d,\varvec{\sigma }}\) which are linearly dependent yields a characterization of the minimum weight codewords of the associated code. More precisely,
Suppose
\(d \ge q'\) where
\({\mathbb {F}}_{q'}\) is the largest subfield of
\({\mathbb {F}}_{q^t}\) fixed by
\(\sigma _i\), for all
\(\sigma _i\) in
\(\varvec{\sigma }\). By Theorem
2.11, the code
\({\mathcal C}_{d,\varvec{\sigma }}\) is a linear code with minimum distance
\( d+3 \le \delta \le N+1\). If the Singleton bound is reached, then it is an MDS code. Let
N be as in (
4) with
\(\displaystyle \sum \nolimits _{i=0}^{m}d_i=d\). If
\(n=2\), then
$$\begin{aligned} N=\displaystyle \prod _{i=0}^{m}(d_i+1) \end{aligned}$$
and the minimum is reached for
\(m=1,d_0=d-1,d_1=1\), so
\(N=2d\). If
\(\sigma \) is such that
\(\textrm{Fix}(\sigma )\cap {\mathbb {F}}_{q^t}={\mathbb {F}}_p\), where
p is the characteristic of the field, since we should have
\(d\ge p\), the smallest possible
\(d=p\) and in this case
$$\begin{aligned} \varvec{\sigma }=(\underbrace{1,1,\ldots ,1}_{p-1\, \text {times}},\sigma ) \end{aligned}$$
(5)
getting that
\(\mathcal {V}_{d,\varvec{\sigma }}\) is a set of
\(q^t+1\) points in
\({{\,\mathrm{{PG}}\,}}(2p-1,q^t)\) such that any
\(p+2\) of them are in general position. So the code
\({\mathcal C}_{d,\varvec{\sigma }}\) is a
\([q^t+1,q^t-2p+1]\)-linear code with minimum distance at least
\(p+3\) and the Singleton bound
\(2p+1\). Now, if
\(\sigma : x \mapsto x^p\), then
\(\mathcal V_{p,\varvec{\sigma }}\) is the normal rational curve of
\({{\,\mathrm{{PG}}\,}}(2p-1,q^t)\); hence
\({\mathcal C}_{p,\varvec{\sigma }}\) is an MDS code.
Furthermore for
\(p\in \{2,3\}\), the following cases can also occur
-
for \(p=2\), \(\sigma : x \mapsto x^{2^h}\), \(1< h < et\), \(\mathcal V_{2,\varvec{\sigma }}\) is either the Segre arc or the normal rational curve (for \(h=et-1\)), hence \({\mathcal C}_{2,\varvec{\sigma }}\) is an MDS code,
-
for
\(p=3\),
\(\sigma : x \mapsto x^{3^h}\),
\(1< h < et\),
\(\mathcal V_{3,\varvec{\sigma }}\) is a
\((3^{et}+1)\)-track of
\({{\,\mathrm{{PG}}\,}}(5,3^{et})\); hence
\({\mathcal C}_{3,\varvec{\sigma }}\) is a so called
almost MDS code, [
2], see next Theorem
3.6.
Clearly, as
p gets larger, the minimum distance gets smaller than the Singleton bound. Before showing the announced result, we recall the following theorem due to Thas [
15] and of which Kaneta and Maruta gives an elementary proof,
Actually, the result above holds for \(q^{t}=27,81\) as well, this is verified by the software MAGMA, obtaining an infinite family of almost MDS codes or, equivalently, an infinite family of \((3^{et}+1)\)-tracks of \({{\,\mathrm{{PG}}\,}}(5,3^{et})\) with \(et>2\).
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.