# A new construction of Algebraic Geometry code using Trace function

In this note, we give a construction of Algebraic-Geometry codes on algebraic function field F/ F_q using places of F (not necessarily of degree one) and trace functions from various extensions of F_q. We compute a bound on the dimension of this code. We also determine a bound on the minimum distance of this code in terms of B_r(F) ( the number of places of degree r in F), 1 ≤ r < ∞. This code is a generalization of the geometric Goppa code, with no restriction on the length of the code except the support condition on divisors defining the code. We obtained few quasi-cyclic codes over F_p as examples of these codes.

There are no comments yet.

## Authors

• 5 publications
• 6 publications
• ### On cyclic algebraic-geometry codes

In this paper we initiate the study of cyclic algebraic geometry codes. ...
06/01/2021 ∙ by Gustavo Cabaña, et al. ∙ 0

• ### Nearly linear time encodable codes beating the Gilbert-Varshamov bound

We construct explicit nearly linear time encodable error-correcting code...
12/28/2017 ∙ by Anand Kumar Narayanan, et al. ∙ 0

• ### Coding theory package for Macaulay2

In this Macaulay2 <cit.> package we define an object called linear code....
07/14/2020 ∙ by Taylor Ball, et al. ∙ 0

• ### Weierstrass Semigroups From a Tower of Function Fields Attaining the Drinfeld-Vladut Bound

For applications in algebraic geometric codes, an explicit description o...
11/07/2019 ∙ by Shudi Yang, et al. ∙ 0

• ### A new construction of nonlinear codes via rational function fields

It is well known that constructing codes with good parameters is one of ...
12/20/2019 ∙ by Lingfei Jin, et al. ∙ 0

• ### On the dimension and structure of the square of the dual of a Goppa code

The Goppa Code Distinguishing (GD) problem asks to distinguish efficient...
11/25/2021 ∙ by Rocco Mora, et al. ∙ 0

• ### Projective toric codes

Any integral convex polytope P in R^N provides a N-dimensional toric var...
03/23/2020 ∙ by Jade Nardi, et al. ∙ 0

##### This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

## 1. Introduction

A linear code is a -subspace of

, the n-dimensional standard vector space over a finite field

. Such codes are used for transmission of information. It was observed by Goppa in 1975 that we can use algebraic function fields over to construct a class of linear codes by choosing a divisor and some rational places of algebraic function field over .

A very useful method of constructing codes over is to use the trace mapping , if a code over is given. The codes so obtained are called trace codes. Many interesting codes over can be represented as trace codes. Trace code, its dimension and Hamming weigth have been studied in [1], [3], [4], [5], [6], [7], [8], [11], etc. Generalized hamming weights of trace codes have been studied in [9] and [10]. In [8]

, Conny Voss obtained an estimate for the weights of code words of trace codes by using the Hasse-Weil bound for the number of rational places over

. In this note, we give a construction of Algebraic-Geometry codes on algebraic function field using places of (not necessarily of degree one) and trace functions from various extensions of . This code is a generalization of the geometric Goppa code, with no restriction on the length of the code except the support condition on divisors defining the code. By using ideas from [8], we compute a bound on the dimension of this code. We also determine a bound on the minimum distance of this code in terms of  ( the number of places of degree in ), .

A linear code of length over is called a quasi-cyclic code of index if for every we have . Quasi-cyclic codes have been studied for many years. The algebraic structure has been studied in [18], [19], [20], [21], etc. In [17], the authors have studied geometric realisation of quasi-cyclic codes. In this note, we construct few quasi-cyclic codes over , where is prime, as examples of .

This note is organized as follows. In section , we recall some results about Goppa codes, Extensions of Algebraic function fields and Trace codes. In section , we give the definition of code . In section , we determine an upper bound on the dimension of code . In section , we determine lower bound on the minimum distance of code under various conditions. In section , we conclude the note listing examples of quasi-cyclic code over derived from the code .

## 2. Preliminaries

### 2.1. Goppa code

[11], Chapter Goppa’s construction of linear code is described as follows:
Let be an algebraic function field of genus . Let be pairwise distinct places of of degree 1. Let and be a divisor of such that . The geometric Goppa code associated with and is defined by

 CL(D,G):={(x(P1),⋯,x(Pn))| x∈L(G)}⊆Fnq.

Then, is an code with parameters and .

### 2.2. Extensions of Algebraic Function Fields

Let denotes an algebraic function field of one variable with full constant field . The field is assumed to be perfect.

###### Definition 2.1.

[11], Chapter An algebraic function field is called an algebraic extension of if is an algebraic field extension and The algebraic extension of is called a finite extension if .

###### Definition 2.2.

[11], Chapter A place is said to be an extension of if .

###### Definition 2.3.

[11], Chapter Let be a subring of . An element is said to be integral over if for some monic polynomial , i.e. if there are such that

 zn+an−1zn−1+⋯+a1z+a0=0.

The next theorem describes a method which can often be used to determine all extensions of a place in . For convenience we introduce some notation.
is the residue class field of
is the residue class of
If is a polynomial with coefficients , we set

 ¯ψ(T):=∑¯ciTi∈¯F[T].
###### Theorem 2.4.

(Kummer) [11], Suppose that where is integral over , and consider the minimal polynomial of over . Let

 ¯ϕ(T)=r∏i=1γi(T)ϵi

be the decomposition of into irreducible factors over . Choose monic polynomials with

 ¯ϕi(T)=γi(T)  and  deg ϕi(T)=deg γi(T).

Then for , there are places satisfying

 Pi|P,  ϕi(y)∈Pi  and  f(Pi|P)≥degγi(T).

Moreover , for .
Under additional assumptions one can prove more: Suppose that atleast one of the following hypotheses is satisfied:

 ϵi=1  for  i=1,⋯,r,or
 {1,y,⋯,yn−1}  is an integral basis for   P.

Then there exists, for , exactly one place with and . The places are all places of lying over , and we have . The residue class field is isomorphic to , hence .

The next theorem we need gives an estimate for the number of places of a fixed degree . Given a function field of genus , we define

 Br:=Br(F):=|{P∈PF;deg P=r}|.
###### Theorem 2.5.

[11], The estimate

 |Br−qrr|≤(qq−1+2gq1/2q1/2−1)qr/2−1r<(2+7g)qr/2r

holds for all .

### 2.3. ( [11], VIII.1) Trace codes

Consider the field extension , let denote the trace mapping. For , we define .

###### Definition 2.6.

( [11], Def. VIII.1.1) Let be a code over . Then

 Tr(C):={Tr(c)∣c∈C}⊆Fnq

is called the trace code of .

A subcode of a code means an -subspace . By we denote the set

 Uq:={(aq1,⋯,aqn)∣(a1,⋯,an)∈U}.

A bound on the dimension of the trace code is given by the following proposition.

###### Proposition 2.7.

( [11], Theorem VIII.1.4) Let be a code over and be a subcode with the additional property . Then

 dim Tr(C)≤m.(dim C−dim U)+dim U∣∣Fq.

## 3. Definition of code

Let be an algebraic function field. Choose a positive integer . Choose arbitrary places of . Let . Let . Choose a divisor of with . Let denote the trace map from to .
Consider the map

Define . Then, is a linear code of length over .

###### Remark 3.1.

There is no restriction on the length of the code except its support doesn’t intersect with .

## 4. Dimension of code Ctr

We have is a vector space over . Hence,

 dimFq Ctr=l(G)−dimFq K

where is the kernel of . Note that . Let

 E:={f∈L(G) | f=hq−h for h∈F}.

For any we have . Therefore, the -vector space generated by denoted by is contained in and

Now, for a divisor we define the divisor , where denotes the greatest integer function. For , we have and . Consider the map

The kernel of the map is .Also .

From the above discussion, we have the following result regarding the dimension of ,

.

## 5. Minimum distance of code Ctr

In this section we apply the techniques of [11], VIII. to compute bound on the minimum distance of . Suppose where . We first assume that for all , is coprime to and i.e degrees of are not times multiples of each other.

Given , there exists a unique effective divisor of smallest degree such that for all . One can describe as follows: choose a basis of then

 υP(A)=max{υP((fi)∞)|1≤i≤k}

for all . We associate with the divisor a second divisor defined by

 A0:=∑P∈supp AP.
###### Definition 5.1.

An element is called degenerate if where , and . Otherwise, is said to be non-degenerate. A -subspace of is said to non-degenerate if every element of is non-degenerate.

###### Remark 5.2.

Assume that is non-degenerate.

For , consider the field extension defined by . Consider the polynomial . Since, is non-degenerate by [2] Lemma , we have is irreducible over . Thus, is a Galois extension of degree . Also,

###### Lemma 5.3.

is constant field of .

###### Proof.

Let be the constant field of . By definition of constant field extension,

 [FL:L(x)]=[F:Fq(x)] ⇒ [FL:F]=[L(x):Fq(x)]=[L:Fq] ⇒ [L:Fq]=[FL:F]∣[Ef:F]=q ⇒ [L:Fq]=pd for some d

So, by Galois correspondence, there exists a field extension of such that . Thus, by Artin-Scherier extension, where . Since, . By Elementary Abelian -Extension, where . and are two generators of Artin extension. So, they are related as for and . This implies, . This is a contradiction as is assumed to be non-degenerate. ∎

### 5.1. Case I: All Pi have same degree β

The support of consists of two disjoint subsets where

 N:={Pi∈supp(D) | Tri(f(Pi))≠0}

and

 Z:={Pi∈supp(D) | Tri(f(Pi))=0}.

Hilbert’s Theorem 90 states that for ,

 Tr(γ)=0 ⇔ γ=αq−α for some α∈Fqβ.

We would like to determine how the places decompose in the extension .

Let and . The Artin-Scherier polynomial has no root in . Each irreducible factor of in has degree . So, by Kummer’s theorem there are no places of degree in over in .
Next we consider a place . Then can be written as , hence the polynomial factors over into distinct places of , all of degree .
The above considerations imply that

 Bβ(Ef)=q|Z|+|¯S|=q(n−|N|)+¯s.

Therefore, Applying Theorem , we get

 wf>n−(2+7g(Ef))qβ/2−1β−qβ−1β.
###### Theorem 5.4.

If and , then the minimum distance of is bounded from below by

 d>n−2qβ/2−1β−qβ−1β−7qβ/2−1β[qg+(q−1)2(−2+deg A+deg A0)].

### 5.2. Case II: Degrees of Pi are not same

Let be the set of distinct ’s ( where as defined before). For , let

 Zi:={Pj∈supp(D) | deg Pj=ti  and  Trj(f(Pj))=0  for some j,1≤j≤n}

and

 Ni:={Pj∈supp(D) | deg Pj=ti  and  Trj(f(Pj))≠0  for some j,1≤j≤n}.

Let

 Z:=r⋃i=1Zi   and   N:=r⋃i=1Ni.

For , let and . Then as above, for

 Bti(Ef)=q|Zi|+|¯Si|.

Therefore,

 wf>n−r∑i=1qti−1ti−(2+7g(Ef))r∑i=1qti/2ti.
###### Theorem 5.5.

If and , then the minimum distance of is bounded from below by

 d>n−r∑i=1qti−1ti−2r∑i=1qti/2ti−7r∑i=1qti/2ti[qg+(q−1)2(−2+deg A+deg A0)].

Now we consider the case when degrees of are times multiple of each other.

Suppose and be the set of distinct ’s such that for and , . For and , let

 Zi,j:={Pl∈supp(D) | deg Pl=ti,j  and  Trl(f(Pl))=0}

and

 Ni,j:={Pl∈supp(D) | deg Pl=t,ji  and  Trl(f(Pl))≠0}.

Let

 Z:=⋃Zi,j  and  N:=⋃Ni,j.

Let then for some such that and . The polynomial has no root in . It decomposes into places of degree . Now for some with , by Kummer’s theorem has extensions all of degree .

Let and and and .

The above considerations imply that

 Bti,j−1(Ef)=∣¯Si,j−1∣+q|Zi,j−1|+m|Ni,j|.

Applying Theorem for we get,

 ∣Zi,j∣<(2+7g(Ef)qti,j/2−1ti,j+qti,j−1ti,j.

Now, . Proceeding as above, we get

 wf>n−r∑i=1si−1∑j=1[2q(ti,j/2)−1ti,j+qti,j−1ti,j−7q(ti,j/2)−1ti,j[qg+(q−1)2(−2+deg A+deg A0))]].
###### Theorem 5.6.

If and , then the minimum distance of is bounded from below by

 d>n−r∑i=1si−1∑j=1[2q(ti,j/2)−1ti,j+qti,j−1ti,j−7q(ti,j/2)−1ti,j[qg+(q−1)2(−2+deg A+deg A0))]].

## 6. Examples

### 6.1. Trace code

For , if , then we get the code is the trace code of geometric goppa code.

### 6.2. Some examples of quasi-cyclic codes over Fp

We let where is prime. For a positive integer it is well known that the number of monic irreducible polynomials of degree over , denoted by , is given by

 (6.1) N(d,q)=1d∑a∣dμ(a)qda.

The trace of a polynomial of degree over is the coefficient of in .

For , let denotes the number of monic irreducible polynomials over of degree and trace . Carlitz’s formula [12] for when is

 Nγ(d,q)=1qd∑a∣d, p∤aμ(a)qda.
###### Lemma 6.1.

[13], Lemma If and are non-zero elements of then .

###### Proposition 6.2.

[13], Corollary Write with . Then

 N0(d,q)=1dq∑a∣bμ(a)qda−ϵd∑a∣bμ(a)qdap

where if and if .

###### Proposition 6.3.

[14],Theorem Let be a subfield of two isomorphic finite fields and . If is a field isomorphism from onto that fixes the elements of , then for ,

 TrK1(a1)=TrK2(σ(a1)).

Fix a prime and positive integers and such that . Consider the finite field where is a prime. Then from Lemma and Proposition , we have for any we have .

Let denotes the trace function

 Tr:Fpd→Fp,a↦a+ap+⋯apd−1.

Choose such that mod or in other words (such a always exist since ). Let denotes the set of all monic irreducible polynomials of degree over . We know that automorphism group of rational function field is the projective linear group over i.e any is of the form

 ϕ(x)=ax+bcx+d where a,b,c,d∈Fp such % that ad−bc≠0.

Let given by .

Since for any i.e for monic irreducible polynomial of degree , we have is monic and irreducible of degree i.e and conversely. Order the elements of as

 p0,1,p0,2,⋯,p0,mp1,1,p1,2,⋯,p1,m⋮⋮⋮⋮pp−1,1pp−1,2⋯,pp−1,m,

where for , the -th row consists of elements of with trace and for ,   . Then the corresponding set of places of is

 P0,1,P0,2,⋯,P0,mP1,1,P1,2,⋯,P1,m⋮⋮⋮⋮Pp−1,1Pp−1,2⋯,Pp−1,m.

Choose a positive integer . Then the code with and is

 (6.2) Ctr={(Tri,j(z(Pi,j)))0≤i≤p−1,1≤j≤m∣z∈L(rP∞)}
###### Proposition 6.4.

The code as in equation is quasi-cyclic code of length and index .

###### Proof.

We have for as defined before and . Let be defined by , for and . We extend the action of to as

 (6.3) θ((Tri,j(z(Pi,j)))0≤i≤p−1,1≤j≤m):=(Trπ(i),j(z(θ(Pi,j))))0≤i≤p−1,1≤j≤m.

We claim that .

We have the following diagram,

where

 ψ1(f(x) mod pi,j(x))=f(x)(Pi,j),
 ψ3(f(x) mod pi+1,j(x))=f(x)(Pi+1