YFFTA663

This article appeared in a journal published by Elsevier.
The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Finite Fields and Their Applications 14 (2008) 785797
http://www.elsevier.com/locate/ffa
Two-weight cyclic codes constructed as the direct sum
of two one-weight cyclic codes
Gerardo Vega
Direccin General de Servicios de Cmputo Acadmico, Universidad Nacional Autnoma de Mxico,
04510 Mxico DF, Mexico
Received 25 April 2007; revised 7 January 2008
Available online 11 February 2008
Communicated by Gary L. Mullen
Abstract
We present a new family of two-weight cyclic codes that can be constructed as the direct sum of two
one-weight cyclic codes. For this new kind of two-weight cyclic codes, we give their weight enumerator
polynomials, and characterize those that are projective.
2008 Elsevier Inc. All rights reserved.
Keywords: One-weight cyclic codes; Two-weight cyclic codes; Gaussian sums
1. Introduction
The class of two-weight irreducible cyclic codes was extensively studied, see for example
[2,11]. As a remarkable work, in [7] the authors obtained the necessary and sufcient conditions
for an irreducible cyclic code to have at most two weights. Now we are going to study non-
irreducible two-weight cyclic codes that can be constructed as the direct sum of two one-weight
irreducible cyclic codes of the same dimension. In order to understand the motivation behind this,
we recall that in [10] it was proved that if C is a two-weight projective cyclic code of dimension
k over F
q
, then either:
E-mail address: gerardov@servidor.unam.mx.
1071-5797/$ see front matter 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.ffa.2008.01.002
786 G. Vega / Finite Fields and Their Applications 14 (2008) 785797
(1) C is irreducible, or
(2) if q = 2, C is the direct sum of two one-weight irreducible cyclic codes of length n =
(
q
k
1
q1
) where divides q 1 and = 1. Additionally, the two nonzero weights of C
are ( 1)q
k1
and q
k1
(direct sum here means direct sum as vector spaces).
An innite class of two-weight cyclic projective codes which are irreducible is known for
any q (deduced from semi-primitive codes) whereas, for the second case, a new class of two-
weight cyclic codes was recently presented in [9]. Additionally, in [1] some two-weight codes
(not necessarily projective ones) with composite parity-check polynomials were studied. Now,
by means of a new set of characterizations for the one-weight cyclic codes, that was introduced
in [8], we present a new family of two-weight cyclic codes that can be constructed as the direct
sum of two one-weight cyclic codes of the same length and dimension. For this new kind of
two-weight cyclic codes, we give their weight enumerator polynomials and characterize those
that are projective. As will be proved later, this new family gives a unied explanation for all
these two-weight cyclic codes that were presented in [1] and [9].
In order to achieve our goal, we will use several results related to character and Gaussian
sums, luckily all those results can be found in [3].
This work is organized as follows. In Section 2 we recall some facts about irreducible cyclic
codes and presents some of its consequences. We also recall, in Section 3, the set of characteri-
zations for the one-weight cyclic codes. Those characterizations will be the main tool in order to
achieve our goal. In Section 4 some preliminary results related to character and Gaussian sums,
are presented. In Section 5, we introduce a new family of two-weight cyclic codes that can be
constructed as the direct sum of two one-weight cyclic codes, and for this kind of two-weight
cyclic codes, we give their weight enumerator polynomials and characterize those that are pro-
jective. In Section 6 we prove that codes in [1] and [9] are subsets of this new family of 2-weight
cyclic codes, additionally we give some examples. Finally, Section 7 is devoted to conclusions.
2. Irreducible cyclic codes and some of its consequences
First of all, we set, for this section and for the rest of this work, the following:
Notation. By using p, q and k, we will denote three positive integers such that p is a prime
number and q is a positive power of p. For now on, will denote a xed primitive element
of F
q
k , and we will x =(q
k
1)/(q 1). Additionally, we will denote by Tr, the absolute
trace mapping from F
q
k to the prime eld F
p
, and by Tr
F
q
k
/F
q
the trace mapping from F
q
k
to F
q
.
Now, we began this section by recalling an important fact about irreducible cyclic codes (see
for example [4, p. 78]).
Proposition 1. Let C be an irreducible cyclic code of length n over F
q
with gcd(n, q) = 1 and
let F
q
k be the splitting eld of x
n
1 over F
q
. If the degree of the check polynomial of C is k and
if it is the minimal polynomial of F
q
k and if =
1
, then:
G. Vega / Finite Fields and Their Applications 14 (2008) 785797 787
(1) The polynomial representation of C is
I(C) =
_
c
(x) =
n1
i=0
Tr
F
q
k
/F
q
_
i
_
x
i
: F
q
k
_
.
(2) The map c
(x) of F
q
k to I(C) is a eld isomorphism and the inverse map is given by
=n
1
c
(),
where n
1
is calculated modulo p.
For now on, we are going to assume, without loss of generality, that any irreducible cyclic
code of length n has a monic generator polynomial g(x) given by
g(x) =
n1
m=0
g
m
x
nm1
. (1)
The following lemma gives us an explicit formula for the coefcients g
m
.
Lemma 2. By using the previous notation let = n
1
g(). Then the coefcients g
m
are given
by
g
m
=Tr
F
q
k
/F
q
_
m
_
, for m=0, 1, 2, . . . , n 1. (2)
Proof. Applying Proposition 1 to g(x) gives
g(x) =
n1
i=0
Tr
F
q
k
/F
q
_
n
1
g()
i
_
x
i
=
n1
m=0
Tr
F
q
k
/F
q
_
m
_
x
nm1
. 2
Since c
(x) is a eld isomorphism, then it is important to observe that , in previous

lemma, is uniquely determined. The following result gives an alternative formula for the eld
element .
Lemma 3. With the above conditions, is given by
1 =
_

q
__

q
2
_

_

q
k1
_
.
Proof. Since g(x) is the monic generator polynomial for an irreducible cyclic code of dimension
k and since g(x) is given by (1), then we have g
m
=0 for all 0 m < k 1 and g
k1
=1. Now,
by means of (2) we obtain the following matricial system
_
_
_
_
_
_
0
0
.
.
.
0
1
_
_
_
_
_
_
=
_
_
_
_
_
_
_
1 1 1 1

q

q
k2
q
k1
.
.
.
k2
(k2)q

(k2)q
k2
(k2)q
k1
k1
(k1)q

(k1)q
k2
(k1)q
k1
_
_
_
_
_
_
_
_
_
_
_
_
_
q
.
.
.
q
k2
q
k1
_
_
_
_
_
_
. (3)
The above square-matrix is clearly a Vandermonde matrix, and therefore its determinant (see
for example [5, p. 116]) is

k1
j=1
k
i=j+1
(
q
i1

q
j1
). Now, we replace the rst column of
this matrix with the column vector in the left-hand side of (3). The determinant of this new
square-matrix is equal to (1)
k+1
k1
j=2
k
i=j+1
(
q
i1
q
j1
). Then, the result follows by the
Kramers rule. 2
The following lemma will be useful in Section 4.
Lemma 4. Let p, q and k be as before and let h
(1)
(x) and h
(2)
(x) be two distinct irreducible
polynomials of degree k which are, respectively, the minimal polynomials of two eld elements
1
and
2
of F
q
k . Also let
1
and
2
in F
q
k such that 1 =
i
(
i

q
i
)(
i

q
2
i
) (
i

q
k1
i
), for
i =1, 2. For some positive integer dividing q 1, let G
be the multiplicative subgroup of order

(q 1)/ of F
q
. Suppose that for some integer , with 1 p
< q
k
, we have
p
1

2
(mod G
),
then
p
1

2
(mod G
).
Proof. We have
1 =
_
1
_
1

q
1
__
1

q
2
1
_

_
1

q
k1
1
__
p
=
p
1
_
1

_
1
_
q
__
1

_
1
_
q
2 _

_
1

_
1
_
q
k1 _
.
Now,
p
1
=
2
, for some G
q
, thus
1 =
p
1

(k1)
_
2

q
2
__
2

q
2
2
_

_
2

q
k1
2
_
.
Therefore,
p
1
=
2
(k1)
. 2
3. Characterizations for one-weight irreducible cyclic codes
Let C be an [n, k] linear code over F
q
, whose dual weight is at least 2. It is well known (see
for example [10]) that if, additionally, C is a one-weight code, then its length n must be given by
n =
q
k
1
q 1
, (4)
for some positive integer . Since the minimal distance of the dual of any nonzero cyclic code
over F
q
is greater than 1, it follows that the length n of all one-weight cyclic codes of dimen-
sion k, is given by (4). Here we are going to assume n q
k
1. Thus, since all the one-weight
cyclic codes are irreducible, then for this particular case the zeros of x
n
1, which form a cyclic
group, lie in the extension eld F
q
k and therefore n divides q
k
1. That is, for some integer ,
we have (
q
k
1
q1
) =(q 1)(
q
k
1
q1
), which implies that divides q 1. Since we are interested
in the construction of codes as the direct sum of two one-weight cyclic codes, then the length n,
for such constructed codes, will always be assumed to be given by (4), where | (q 1).
The following set of characterizations for the one-weight cyclic codes, that was introduced
in [8], will be the main tool of this work.
Theorem 5. Let q and k be integers, where q is a prime power and k > 0. Let be a primitive
element of F
q
k . For a positive integer a, let h
a
(x) F
q
[x] be the minimal polynomial of
a
.
Let be the last positive integer for which x
is congruent, modulo h
a
(x), to some element
in F
q
. For each positive integer , such that | gcd(a, q 1), set
=
(q1)
gcd(a,q1)
and n
(q
k
1)/(q 1). Then, the following ve statements are equivalent:
(A) gcd(a, (q
k
1)/(q 1)) =1.
(B) h
a
(x) | x
n
1, deg(h
a
(x)) =k and =(q
k
1)/(q 1).
(C) h
a
(x) | x
n
1, deg(h
a
(x)) =k and the Hamming weight of the polynomial x
n
1/h
a
(x)
is
q
k1
.
(D) h
a
(x) | x
n
1 and deg(h
a
(x)) = k. Additionally, if g(x) = x
n
1/h
a
(x), then for any
nonzero codeword c(x) in g(x) there exist a uniquely determined integer s, 0 s <
(q
k
1)/(q 1), and a uniquely determined eld element d F
q
, such that c(x) =dx
s
g(x).
(E) h
a
(x) is the parity-check polynomial for a one-weight cyclic code of dimension k.
4. Some preliminary results on character and Gaussian sums
Before our next result, we recall some already known results on character and Gaussian sums.
In order to do this, let p, q, k and be as before. We began with the denition for the canonical
additive character of F
q
k :
(c) =e
2i Tr(c)/p
, for all c F
q
k .
On the other hand, any multiplicative character of F
q
k is dened by
j
_
l
_
=e
2ijl/(q
k
1)
, for j, l =0, 1, . . . , q
k
2.
Commonly
0
is referenced as the trivial multiplicative character. For the canonical additive
character , of F
q
, and for any multiplicative character =
0
, of F
q
, the following two or-
thogonal properties will be useful for us
cF
q
(c) =0 and
cF
q
(c) =0. (5)
Additionally, for any multiplicative character of F
q
k and for the canonical additive character
of F
q
k , the Gaussian sum G(, ) is dened by
G(, ) =
cF
q
k
(c)(c).
Two useful properties of G(, ) are (see for example [3, pp. 193194]):
G
_
p
,
_
=G(, ) and if =
0
then G(, )G(, ) =q
k
. (6)
Another important property of Gaussian sums is the so-called expansion of the restriction of to
F
q
k
in terms of the multiplicative characters of F
q
k , with Gaussian sums as Fourier coefcients
(see for example [3, p. 195]):
(c) =
1
(q
k
1)
(c)G(, ) for c F
q
k
. (7)
Lemma 6. Let p, q, k, and be as before and let h
(1)
(x), h
(2)
(x),
1
,
2
,
1
and
2
be as in
Lemma 4. Let a
1
and a
2
be two integers in such way that
1
=
a
1
and
2
=
a
2
. Assume that
a
2
is a unit in the ring Z
, where a
2
is its inverse in such ring. Let = 1 + a
2
(a
1
a
2
), where
the arithmetic operations in the denition of are taking in Z. Let be the canonical additive
character and set = gcd(a
1
a
2
, q 1) and = (q 1)/. Suppose that at least one of the
following two conditions holds:
(1) p =2, k =2 and =1, or
(2) for some integer , with 1 p
< q
k
, we have p
1 (mod ),
then, for all t =1, 2, . . . , 1, the following two equalities are true
G(
t
, )G
_
t
,
_
=q
k
and
t
(
1
)
t
(
2
) =1.
Proof. Suppose condition (1) holds. If c F
q
2
, then Tr
F
q
2
/F
q
(c) =0 c F
q
={
s
| 0 s <
(q 1)}. Thus, since =q
2
1, we have
G(
t
, ) =
q
2
2
i=0
t
_
i
_
i
_
=q 1
1
i=1
t
_
i
_
=q.
In a similar way G(
t
, ) = q. From Lemma 3, and since p = 2 and k = 2, we have
i
=
Tr
F
q
2
/F
q
(
a
i
)
1
F
q
, for i =1, 2. Thus,
t
(
1
) =
t
(
2
) =1.
Now, suppose condition (2) is satised. Since
has order , then

p
. From (6),
we have G(
t
, ) =G(
p
t
, ) =G(
t
, ) and therefore we obtain G(
t
, )G(
t
, ) =
G(
t
, )G(
t
, ) = q
k
. For the second equality we rst take a
2
p
a
2
(mod ). But
| (a
2
a
1
) and a
2
a
2
= 1 + for some integer , thus a
1
p
a
2
(mod ) and therefore,
if G
is the multiplicative subgroup of order (q 1)/ of F

q
, we have
p
1

2
(mod G
).
Now, by virtue of Lemma 4, we have
2
=
p
1
for some G
, and hence
t
(
1
)
t
(
2
) =
t
(
1
)
p
t
(
1
)
t
() =1, because
t
() =1. 2
We end this section with the following:
Lemma 7. Assume the same notation and hypothesis as in the previous lemma. Also, for some
positive integer n, assume
n
1
=
n
2
= 1. Let B be a set of pairs of multiplicative characters
of F
q
k . For x F
q
k
we set
S(x) =
(,)B
(
1
)G(, )(
2
)G(, )(x)
n1
m=0
(
1
)
m
(
2
)
m
.
If B ={(
u
1
,
u
2
(q1)u
1
) | 0 u
1
< q
k
1, 0 u
2
< }, then
S(x) =n +nq
k
1
t =1
t
(x).
Proof. By using B we have
S(x) =
q
k
2
u
1
=0
u
1
(
1
)G(
u
1
, )
1
u
2
=0
u
2
(q1)u
1
(
2
)G(
u
2
(q1)u
1
, )
u
1
(x)
n1
m=0
u
1
(
1
)
m
u
2
(q1)u
1
(
2
)
m
.
The inner sum in the last expression is a nite geometric series that vanishes if
u
1
(
1
)
u
2
(q1)u
1
(
2
) =1, because of
u
1
(
1
)
n
u
2
(q1)u
1
(
2
)
n
=
u
1
_
n
1
_
u
2
(q1)u
1
_
n
2
_
=
u
1
(1)
u
2
(q1)u
1
(1) =1.
On the other hand,
u
1
(
1
)
u
2
(q1)u
1
(
2
) =1
1
_
a
1
u
1
+a
2
(u
2
(q1)u
1
)
_
=1
a
2
u
2
(q 1) (a
2
a
1
)u
1
_
mod q
k
1
_
,
which implies that (q 1) | (a
2
a
1
)u
1
, but gcd(a
1
a
2
, q 1) = and therefore u
1
= t
for t = 0, 1, . . . , 1. Now, from the previous congruence we have a
2
a
2
u
2
(q 1)
a
2
(a
2
a
1
)u
1
(mod q
k
1), but a
2
a
2
= 1 + for some integer , thus u
2
(q 1)
a
2
(a
2
a
1
)u
1
(mod q
k
1) and hence, for each t , there exist a uniquely 0 u
2
< such
that u
2
a
2
yt (mod ), where y = (a
2
a
1
)/. With this we conclude that
u
1
=
t
and
u
2
(q1)u
1
=
a
2
(a
2
a
1
) t t
=
t
. Thus
S(x) =n +n
1
t =1
t
(
1
)G(
t
, )
t
(
2
)G
_
t
,
_
t
(x),
and by means of previous lemma we obtain the desired result. 2
5. A new family of two-weight cyclic codes
We begin this section recalling some already know results.
Let C be an [n, k] linear code over F
q
. Let w
1
, w
2
, . . . , w
N
be the nonzero weights in C and
suppose that the weight of the dual of C is at least 2. For 1 i N, let A
i
be the number of
words of weight w
i
in C. The two rst identities of Pless (see [6] for the general result), for C,
are:
(1)

N
i=1
A
i
=q
k
1,
(2)

N
i=1
A
i
w
i
=n(q 1)q
k1
.
The following two results were proved in [9].
Proposition 8. Let C be a two-weight code of length n and dimension k over F
q
with nonzero
weights w
1
and w
2
. Assume that the minimum weight of the dual of C is at least 2. Then C is a
projective code if and only if
n
2
(q 1)
_
q(w
1
+w
2
) 1
_
n +
(q
k
1)w
1
w
2
(q 1)q
k2
=0.
Corollary 9. If C is a two-weight cyclic code of length n = (q
k
1)/(q 1) and dimension
2k over F
q
, with = 1 and nonzero weights w
1
= ( 1)q
k1
and w
2
= q
k1
, then C is a
projective code.
Suppose that C
1
and C
2
are two different one-weight cyclic codes over F
q
of the same
length n and dimension k, where their generator polynomials are, respectively, g
1
(x) =
n1
m=0
g
(1)
m
x
nm1
and g
2
(x) =

n1
m=0
g
(2)
m
x
nm1
, whereas h
(1)
(x) and h
(2)
(x) are, respec-
tively, their parity check polynomials. We construct the code C as the direct sum of the codes C
1
and C
2
, and hence, a typical word c(x) C is of the form c(x) =c
1
(x) c
2
(x), for some words
c
1
(x) C
1
and c
2
(x) C
2
. The codeword c(x) will be nontrivial if the two codewords c
1
(x)
and c
2
(x) are different from zero. Clearly, nontrivial codewords are of special interest in order to
determine the weights in C. Under these conditions on c
1
(x) and c
2
(x) we know, by virtue of the
equivalence between (D) and (E) in Theorem 5, that there must exist uniquely determined inte-
gers s
1
and s
2
, 0 s
1
, s
2
< (q
k
1)/(q 1), and uniquely determined eld elements d
1
, d
2
F
q
,
such that c(x) = d
1
x
s
1
g
1
(x) d
2
x
s
2
g
2
(x). Now, let
1
and
2
be two xed roots of h
(1)
(x) and
h
(2)
(x), respectively. Additionally, let
1
and
2
be the two eld elements, obtained by means of
Lemma 3, such that 1 =
i
(
i

q
i
)(
i

q
2
i
) (
i

q
k1
i
) and g
(i)
m
= Tr
F
q
k
/F
q
(
i
m
i
), for
i = 1, 2, and m = 0, 1, 2, . . . , n 1. Then by using Lemma 2 we known that the weight of c(x)
is n Z(d
1
, s
1
, d
2
, s
2
), where
Z(d
1
, s
1
, d
2
, s
2
) =
_
m, 0 m < n: Tr
F
q
k
/F
q
_
d
1
m+s
1
1
d
2
m+s
2
2
_
=0
_
.
Because of the F
q
-linearity and the cyclicity of C, this number is equal to
Z(d, s, 1, 0) =
_
m, 0 m < n: Tr
F
q
k
/F
q
_
d
1
m+s
1

2
m
2
_
=0
_
,
where d = d
1
2
d
1
and s = s
1
s
2
. Consequently, the Hamming weight of any nontrivial code-
word in C can be determined via computing Z(d, s, 1, 0), for some integer s and some eld
element d F
q
. Keeping in mind this, we present the following:
Theorem 10. Let p, q, k, and be as before. For a positive integer a, let h
a
(x) F
q
[x] be the
minimal polynomial of
a
. Also, let a
1
, a
2
and be integers such that a
1
q
i
a
2
(mod q
k
1),
for all i 0, =gcd(a
1
a
2
, q 1) and a
2
Z
, where a
2
is the inverse of a
2
in Z
. For some
integer , such that | gcd(a
1
, a
2
, q 1), we set =
(q1)
gcd(a
1
,a
2
,q1)
, n = , = (q 1)/ and
=(q 1)/. Suppose that at least one of the following two conditions holds:
(1) p =2, k =2, =1 and a
1
is a unit in the ring Z
, or
(2) for some integer , with 1 p
< q
k
, we have
_
1 + a
2
(a
1
a
2
)
_
p
1 (mod ), (8)
then the following four assertions are true:
(a) h
a
1
(x) and h
a
2
(x) are the parity-check polynomials for two different one-weight cyclic codes
of length n and dimension k.
(b) | and > /.
(c) If C is the cyclic code with parity-check polynomial h
a
1
(x)h
a
2
(x), then C is an [n, 2k] two-
weight cyclic code with weight enumerator polynomial
A(z) =1 +(/)n(q 1)z
((/))q
k1
+
_
q
2k
1 (/)n(q 1)
_
z
q
k1
.
(d) C is a projective code if and only if =.
Proof. Part (a): If condition (2) holds, then, from (8), we have a
2
a
1
p
1 (mod ) and
since gcd(p
, ) = 1, then gcd(a
1
, ) = 1. Thus, under (1) or (2) we conclude gcd(a
1
, ) =
gcd(a
2
, ) = 1. Therefore, by virtue of the equivalence between (A) and (E) in Theorem 5, and
since a
1
q
i
a
2
(mod q
k
1) for all i 0, we have that h
a
1
(x) and h
a
2
(x) are the parity-check
polynomials for two different one-weight cyclic codes of length n and dimension k.
Part (b): By the equivalence between statements (A) and (B) in Theorem 5, the orders of
h
a
1
(x) and h
a
2
(x) divide the length n. Then there must exist integers l
i
, i = 1, 2, such that
n =(q
k
1)/ =l
i
(q
k
1)/gcd(a
i
, q
k
1), thus | a
i
, i =1, 2, and therefore | . Suppose
< / then > q 1, which is impossible. Now, suppose = / then = q 1 and =
q
k
1. Therefore, from (8), a
2
(a
2
a
1
)p
1 (mod q
k
1). But ( =q 1) | (a
2
a
1
),
then q 1 | p
1, and hence p
=q
for some positive integer . Multiplying both sides of (8)

by a
2
, we get a
1
q
a
2
(mod q
k
1), which is a contradiction.
Part (c): Clearly C is a cyclic code of length n and dimension 2k. Let g
i
(x) =x
n
1/h
a
i
(x) =
n1
m=0
g
(i)
m
x
nm1
, i =1, 2. Let s be an integer and d F
q
. In accordance with previous discus-
sion, we are going to compute Z(d, s, 1, 0). For this, let h
(1)
(x) = h
a
1
(x), h
(2)
(x) = h
a
2
(x) and
1
,
2
,
1
and
2
be as in such discussion. If
is the canonical additive character of F

q
, then,
by the rst orthogonal property in (5), we have
1
q
cF
q
_
c
_
dg
(1)
m+s
g
(2)
m
__
=
_
1 if dg
(1)
m+s
=g
(2)
m
,
0 otherwise,
where the subscript m+s is taken modulo n. Thus,
Z(d, s, 1, 0) =
1
q
n1
m=0
cF
q
_
Tr
F
q
k
/F
q
_
dc
1
m+s
1
__
_
Tr
F
q
k
/F
q
_
c
2
m
2
__
.
If denotes the canonical additive character of F
q
k , then
and are related by
(Tr
F
q
k
/F
q
()) =() for all F
q
k . Therefore,
Z(d, s, 1, 0) =
1
q
cF
q
n1
m=0
_
dc
1
m+s
1
_
_
c
2
m
2
_
=
n
q
+
1
q
cF
q
n1
m=0
_
dc
1
m+s
1
_
_
c
2
m
2
_
. (9)
Now, from (7), we know that (dc
1
m+s
1
)(c
2
m
2
) is equal to
1
(q
k
1)
2
_
dc
1
m+s
1
_
G(, )
_
c
2
m
2
_
G(, ),
where the sums are extended over all multiplicative characters and of F
q
k . Substituting
in (9), we obtain
Z(d, s, 1, 0) =
n
q
+
1
q(q
k
1)
2
(
1
)G(, )
(
2
)G(, )
_
d
s
1
_
n1
m=0
(
1
)
m
(
2
)
m
cF
q
()(c).
Remark. The previous part of this proof is very similar to the proof of Theorem 8.84 in [3,
p. 450].
If the restriction of to F
q
is nontrivial, then, by the second orthogonal property in (5), we
have
cF
q
()(c) = 0. Consequently, it sufces to extend the previous sum over the set B of
pairs of characters and for which is trivial in F
q
, so that Z(d, s, 1, 0) is equal to
n
q
+
q 1
q(q
k
1)
2
(,)B
(
1
)G(, )(
2
)G(, )
_
d
s
1
_
n1
m=0
(
1
)
m
(
2
)
m
.
Since F
q
= {
s
| 0 s < (q 1)}, then B = {(
u
1
,
u
2
(q1)u
1
) | 0 u
1
< q
k
1,
0 u
2
< }. Therefore by Lemma 7 we have
Z(d, s, 1, 0) =
n
q
+
n(q 1)
q(q
k
1)
2
+
n(q 1)q
k
q(q
k
1)
2
1
t =1
t
_
d
s
1
_
.
Since
has order , we have

1
t =1
t
_
d
s
1
_
=
_
1 if d
s
1
=1,
1 otherwise
thus,
Z(d, s, 1, 0) =
_
n ( (/))q
k1
if d
s
1
=1,
n q
k1
otherwise
and hence the weights of C are the desired ones. By using the two rst Pless identities, we can
see that the weight enumerator polynomial for C is the required one.
Part (d): This follows as a direct application of Proposition 8 and Corollary 9. 2
6. In perspective with the already known codes
Here we are going to show that codes in [1] and [9], are just very easy instances of Theo-
rem 10. Additionally, in the second part of this section we show that one can use Theorem 10 in
order performs a systematic search for the two-weight cyclic codes constructed as the direct sum
of two one-weight cyclic codes. In order to do this, we will keep using the same notation as in
Theorem 10.
In Theorem 7 of [9], it was showed that if q and k are odd numbers, and if a
1
=(q
k
+1)/2 and
a
2
=1, then the cyclic code C of length n =q
k
1 over F
q
with check polynomial h
a
1
(x)h
a
2
(x)
is a 2-weight cyclic code of dimension 2k and the two nonzero weights are (q 1)q
k1
and
((q 1)/2)q
k1
. Additionally, if q =3, then C is a projective code.
Since q and k are odd numbers, then = gcd(a
1
a
2
, q 1) = gcd((q
k
1)/2, q 1) =
(q 1)/2. Therefore (1 a
2
(a
2
a
1
)) = (q
k
+ 1)/2 1 (mod ), since a
2
= 1 and =
(q
k
1)/2. Now n = q
k
1, thus = q 1 and = 1, therefore, from Theorem 10, the two
nonzero weights are the desired ones, whereas the projective condition is now very obvious. By
using similar arguments we can do the same for Theorem 8 and Corollary 9 in [9].
In [1], it was showed that if a
1
and a
2
are in such way that gcd(k, N
1
) = gcd(k, N
2
) =
gcd(t, N
2
) = 1, where N
1
and N
2
divide q 1, a
1
= (q
k
1)/N
1
+1, a
2
= t (q
k
1)/N
2
+1,
N = lcm(N
1
, N
2
), u = gcd(N,
N
N
1
t
N
N
2
) and gcd(a
1
, a
2
, N
1
, N
2
) = 1, then the cyclic code C
of length n = q
k
1 over F
q
with check polynomial h
a
1
(x)h
a
2
(x) is a 2-weight cyclic code of
dimension 2k and the weight enumerator polynomial of C is
A(z) =1 +
_
q
k
1
_
N
u
z
q
k1
(q1)(1
u
N
)
+
_
q
2k
1
_
q
k
1
_
N
u
_
z
q
k1
(q1)
.
Since gcd(N
1
N
2
, N
2
t N
1
) =gcd(N
1
, N
2
) then u =1 (it seems that the author of [1] did not
realize that, in general, u = 1). Now, in the proof of the main theorem in [1], it was showed that
h
N
(x) is the parity-check polynomial for a one-weight cyclic code of length (q
k
1)/N and
nonzero weight q
k1
(q 1)/N. Let s be the smallest positive integer such that q
k
1 divides
N(q
s
1), but N | (q 1) then (q
k
1) | (q 1)(q
s
1). The last condition is impossible if
s < k, hence deg(h
N
(x)) = k. Therefore, from the equivalence of (E) and (A) in Theorem 5,
we can now conclude that gcd(, N) = 1. Thus, since a
1
a
2
= (
N
N
1
t
N
N
2
)(q 1)/N, then
=gcd(a
1
a
2
, q 1) =(q 1)/N. Now divides (q
k
1)/N
2
and divides (q
k
1)/N
i
,
for i =1, 2, hence a
2
=1 and a
1
a
2
1 (mod ) and, as a consequence, (1 a
2
(a
2
a
1
))
1 (mod ). Finally, since n = q
k
1, thus = q 1 and = 1, therefore, from Theorem 10,
the weight enumerator polynomial is the desired one.
Some examples. Theorem 10 gives a method for the systematic search for the two-weight cyclic
codes constructed as the direct sum of two one-weight cyclic codes. The following ones are
examples of such method.
(1) Let q =2. Thus = = = 1, which is in conict with part (b) of Theorem 10. In conclu-
sion, it does not exist a two-weight binary cyclic code, constructed as the direct sum of two
one-weight binary cyclic codes, which satises Theorem 10. This is in concordance with
[10], in the sense that all the two-weight projective cyclic codes, over F
2
, are irreducible.
(2) Let q = 16, k = 2, = 1 and a
2
= 93, then q 1 = 15 and = 17. Without taking into
consideration conjugates, there are 64 possible values for a
1
satisfying gcd(a
1
, 17) = 1
and gcd(a
1
93, 15) = 1. For example, if a
1
= 61 then h
61
(x)h
93
(x) is the parity-check
polynomial for a two-weight projective cyclic code of length 255, dimension 4 and weight
enumerator polynomial A(z) = 1 + 3825z
224
+ 61710z
240
. It is interesting to see how the
pair a
1
= 61 and a
2
= 93, satisfy condition (1) but not condition (2) in Theorem 10 (there
a
2
= 15). For the same values of q, k and a
2
, we take now = 5 and p
= 1. Thus,
a
1
=8 is the only possible value satisfying gcd(a
1
93, 15) =5 and (1 +15(a
1
93)) 1
(mod 85). Therefore h
8
(x)h
93
(x) is the parity-check polynomial for a two-weight cyclic
code of length 255, dimension 4 and weight enumerator polynomial A(z) = 1 + 765z
160
+
64770z
240
.
(3) Let q = 9, k = 3, = 2, p
= 3
5
and a
2
= 456. Then q 1 = 8, = 91 and a
2
= 1. Thus,
a
1
=94 and a
1
=458 are the only two possible values for a
1
satisfying gcd(a
1
456, 8) =2
and (1 + (a
1
456))3
5
1 (mod 182). For each one of these values, we have two pos-
sible scenarios: If a
1
= 94, then h
94
(x)h
456
(x) is the parity-check polynomial for a two-
weight projective cyclic code of length 364, dimension 6 and weight enumerator polynomial
A(z) = 1 + 2912z
243
+ 528528z
324
. Additionally, h
94
(x)h
456
(x) is the parity-check poly-
nomial for a two-weight no projective cyclic code of length 728, dimension 6 and weight
enumerator polynomial A(z) = 1 + 2912z
486
+ 528528z
648
. For a
1
= 458, we have the
same scenarios. For the same values of q, k and , we take now p
= 1 and a
2
= 1.
Thus, a
1
= 183 and a
1
= 547 are the only possible values satisfying gcd(a
1
1, 8) = 2 and
a
1
1 (mod 182). Therefore h
183
(x)h
1
(x) and h
547
(x)h
1
(x) are the parity-check polyno-
mials for two two-weight cyclic codes with same length, dimension and weight enumerator
polynomial as before. Observe that h
547
(x)h
1
(x) is the same polynomial presented as exam-
ple in [1].
7. Conclusion
By means of some new characterizations for one-weight cyclic codes, we have introduced a
new family of two-weight cyclic codes constructed as the direct sum of two one-weight cyclic
codes of the same length and dimension. For this new kind of two one-weight cyclic codes, we
gave their weight enumerator polynomials and characterize those that are projective. By means
of some examples we also outlined a method for the systematic search of this new kind of codes.
How was already proved, this new family of two-weight cyclic codes gives a unied explanation
for all these two-weight cyclic codes that were presented in [1] and [9]. In fact, we believe
that any two-weight cyclic code constructed as the direct sum of two one-weight cyclic codes,
satises either condition (1) or condition (2) in Theorem 10. Finally, condition (1) in Theorem
10 recalls us, once again, that nite elds of characteristic 2 impose a special treatment.
Acknowledgments
The author would like to thank the anonymous referees for valuable remarks and suggestions.
References
[1] T. Helleseth, Some two-weight codes with composite parity-check polynomials, IEEE Trans. Inform. Theory 22
(1976) 631632.
[2] P. Langevin, A new class of two weight codes, in: Finite Fields and Their Applications, Glasgow, 1995, in: London
Math. Soc. Lecture Note Ser., vol. 233, Cambridge Univ. Press, Cambridge, UK, 1996, pp. 181187.
[3] R. Lidl, H. Niederreiter, Finite Fields, Cambridge Univ. Press, 1983.
[4] J.H. van Lint, Introduction to Coding Theory, Grad. Texts in Math., vol. 86, Springer-Verlag, 1982.
[5] F.J. MacWilliams, N.J.A. Sloane, The Theory of Error-Correcting Codes, North-Holland, Amsterdam, The Nether-
lands, 1977.
[6] V. Pless, Power moment identities on weight distributions in error-correcting codes, Inf. Contr. 6 (1962) 147152.
[7] B. Schmidt, C. White, All two-weight irreducible cyclic codes?, Finite Fields Appl. 8 (2002) 117.
[8] G. Vega, Determining the number of one-weight cyclic codes when length and dimension are given, in: C. Carlet,
B. Sunar (Eds.), Proceedings of International Workshop on the Arithmetic of Finite Fields, WAIFI 2007, in: Lecture
Notes in Comput. Sci., vol. 4547, Springer, 2007, pp. 284293.
[9] G. Vega, J. Wolfmann, New classes of 2-weight cyclic codes, Des. Codes Cryptogr. 42 (3) (2007) 327334.
[10] J. Wolfmann, Are 2-weight projective cyclic codes irreducible?, IEEE Trans. Inform. Theory 51 (2005) 733737.
[11] J. Wolfmann, Projective two-weight irreducible cyclic and constacyclic codes, Finite Fields Appl. 14 (2) (2008)
351360.

YFFTA663

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

YFFTA663

Загружено:

Авторское право:

Доступные форматы

This article appeared in a journal published by Elsevier.

(x) is a eld isomorphism, then it is important to observe that , in previous

be the multiplicative subgroup of order

has order , then

is the multiplicative subgroup of order (q 1)/ of F

for some positive integer . Multiplying both sides of (8)

is the canonical additive character of F

and are related by

has order , we have

Вам также может понравиться