AAECC (2008) 19:509–525

DOI 10.1007/s00200-008-0086-9

On cyclic self-dual codes

Xiaoshan Kai · Shixin Zhu

Received: 12 June 2008 / Revised: 13 November 2008 / Published online: 29 November 2008
© Springer-Verlag 2008

Abstract We investigate cyclic self-dual codes over F2r . We give a decomposition

of a repeated-root cyclic codes over F pr . The decomposition is used to analyze cyclic
self-dual codes over F2r . We obtain a necessary and sufficient condition for the exis-
tence of nontrivial cyclic self-dual codes over F2r , and prove that all cyclic self-dual
codes over F2r are Type I. Finally we classify cyclic self-dual codes of some lengths
over F4 , F8 , and F16 .

Keywords Cyclic codes · Self-dual codes · Generator polynomial

1 Introduction

Cyclic codes over finite fields have been studied in the late 1950s. A lot of research
has been given to cyclic codes in that cyclic codes are an important class of codes
from either a theoretical or a practical viewpoint. Most of the research focused on the
situation when the length and the characteristic of the field are coprime. Cyclic codes
of length n over a field of characteristic p where p divides n are called repeated-
root cyclic codes and have been studied by Castagoli et al. [4] and Van Lint [12].
Even though it was shown that repeated-root cyclic codes are asymptotically bad,

The research is supported by the National Natural Science Foundation of China under Grant No.
60673074 and HFUT Research Grant 081003F.

X. Kai (B) · S. Zhu

Department of Mathematics, Hefei University of Technology,
230009 Hefei, Anhui, People’s Republic of China
e-mail: kxs6@sina.com
S. Zhu
e-mail: zhushixin@hfut.edu.cn

123
510 X. Kai, S. Zhu

they still remain interesting objects [6,11]. Recently, there has been much interest in
self-dual codes over finite fields and rings. However, not much work has been done
for cyclic self-dual codes. Binary cyclic self-dual codes were discussed by Sloane and
Thompson [10], who proved that binary cyclic self-dual codes are Type I. Later, Ned-
eloaia gave the weight distributions of all binary cyclic self-dual codes of length up to
110 [8]. It is well known that there are no cyclic self-dual codes over Fq when q is odd.
The purpose of this paper is to investigate self-dual repeated-root cyclic codes over
the finite field F2r . We first give a decomposition of a repeated-root cyclic code over
F pr using the Discrete Fourier Transform. Then the decomposition is used to explore
cyclic self-dual codes over F2r . We give a necessary and sufficient condition for the
existence of nontrivial cyclic self-dual codes over F2r . This enables us to construct
all cyclic self-dual codes over F2r . We prove that all cyclic self-dual codes over F2r
are Type I, which extends the result of [10, Corollary 2]. Finally we classify cyclic
self-dual codes of some lengths over F4 , F8 , and F16 .

2 Decomposing repeated-root cyclic codes over Fpr

Let C be a linear code of length N over F pr , then the code C is a cyclic code if C is
invariant under the permutation of F Npr :

(c0 , c1 , . . . , c N −1 ) → (c N −1 , c0 , . . . , c N −2 ).

A cyclic code of length N over F pr can be identified as ideals in the quotient ring
F pr [x]/x N −1. Given N -tuples u = (u 0 , u 1 , . . . , u N −1 ) , v = (v0 , v1 , . . . , v N −1 ) ∈
F Npr , their Euclidean inner product or dot product is defined as usual u · v = u 0 v0 +
u 1 v1 + · · · + u N −1 v N −1 . Two N -tuples u, v are called orthogonal if u · v = 0. For a
linear code C of length N over F pr , its dual code C ⊥ is the set of N -tuples over F pr
that are orthogonal to all codewords of C:

C ⊥ = {u |u · v = 0 for all v ∈ C } .

A code C of length N over F pr is called self-orthogonal if C ⊆ C ⊥ , and it is called

self-dual if C = C ⊥ . We say that two codes are equivalent if one can be obtained
from the other by permuting the coordinates. The Frobenius automorphism of F pr is
defined by a f = a p . It generates the group of automorphisms of F pr , which is cyclic
of order r . For the remainder of this paper, let N = p s n, where s is a positive integer
and n is prime to p.
s
We denote Rs ( pr , u) = F pr [u]/u p − 1. There exists a module isomorphism
ϕ : Rs ( p , u) → F pr defined by
r n N

s s

ϕ a0,0 + a0,1 u + · · · + a0, ps −1 u p −1 , . . . , an−1,0 + an−1,1 u + · · · + an−1, ps −1 u p −1
 
= a0,0 , a1,0 , . . . , an−1,0 , a0,1 , a1,1 , . . . , an−1,1 , . . . , a0, ps −1 , a1, ps −1 , . . . , an−1, ps −1 .

123
On cyclic self-dual codes 511

We have that
⎛ ⎛ s −1
⎞ s −1 s −1
⎞
p p p
ϕ ⎝u ⎝ an−1, j u j ⎠ , a0, j u j , . . . , an−2, j u j ⎠
j=0 j=0 j=0
 
= an−1, ps −1 , a0,0 , a1,0 , . . . , an−2, ps −1 .

This gives that constacyclic shift by u in Rs ( pr , u)n corresponds to a cyclic shift in

F Npr . Let m be the order of pr modulo n, and I a complete set of pr −cyclotomic coset
representatives modulo n, then n| p mr − 1. Hence, the Galois field F pmr contains a
primitive nth root of unity ξn . Let cl pr (i, n) be the pr −cyclotomic coset modulo n con-
taining i, and m i the size of this coset, i.e., cl pr (i, n) = {i, pr i, p 2r i, . . . , p (m i −1)r i}.
We have that m i = |cl pr (i, n)| and m i divides m.
We know that the Discrete Fourier Transform (DFT) is an important tool to deal with
a code. Simple-root cyclic codes and quasi-cyclic codes over finite fields were studied
in the DFT Domain Characterization in [3] and [5] respectively. In the following, we
use the transform approach to consider repeated-root cyclic codes.
Definition 2.1 Let a = (a0,0 , a1,0 , . . . , an−1,0 , a0,1 , a1,1 , . . . , an−1,1 , . . . , a0, ps −1 ,
n−1 p s −1
a1, ps −1 , . . . , an−1, ps −1 ) ∈ F Npr , with a(x) = i=0 j=0 ai, j x
i+ jn ∈ F r [x] the
p
corresponding polynomial. The Discrete Fourier Transform of a(x) is the vector

(A0 , A1 , . . . , An−1 ) ∈ Rs ( pr , u)n

 n−1 p s −1 
with Ah = a(u n ξnh ) = i=0 j=0 ai, j u
n i+ j ξ hi , for 0 ≤ h ≤ n − 1, where
n

nn ≡ 1(mod p ). Define the Mattson–Solomon polynomial of a to be A(z) =
s
n−1
h=0 An−h z (here An = A0 ).
h

The following lemma shows that a vector of F Npr can be recovered from its Discrete
Fourier Transform.
Lemma 2.2 (Inversion Formula) Let a ∈ F Npr with A(z) its Mattson–Solomon poly-
nomial as defined above. Then

  
 1

a=ϕ 1, u −n , u −2n , . . . , u −(n−1)n ∗ A(1), A(ξn ), . . . , A(ξnn−1 )
n

where ∗ denotes componentwise multiplication.

Proof Let 0 ≤ t ≤ n − 1. Then
⎛ s −1
⎞

n−1 
n−1 n−1 p

A(ξnt ) = Ah ξn−ht = ⎝ ai, j u n i+ j ξnhi ⎠ ξn−ht
h=0 h=0 i=0 j=0


n−1  p s −1 
n−1 s −1
p
n  i+ j n t
= ai, j u ξnh(i−t) = (nu ) at, j u j
i=0 j=0 h=0 j=0

123
512 X. Kai, S. Zhu

n−1 h(i−t)
The last equation comes from the fact that h=0 ξn = 0 unless i ≡ t (mod n).
The result follows.
f
Let A = {(A0 , A1 , . . . , An−1 ) ∈ Rs ( pr , u)n : Ah ∈ Rs ( p m i r , u), A pr h = Ah },
where subscripts are calculated modulo n. We make A a ring via componentwise
addition and multiplication. It is easy to verify that A ∼
= ⊕i∈I Rs ( p m i r , u).

Theorem 2.3 Let N = p s n, where ( p, n) = 1. The map γ : F pr [x]/x N − 1 →

⊕i∈I Rs ( p m i r , u) defined by γ (a(x)) = (Ai )i∈I is a ring isomorphism. In particu-
lar, if C is an ideal of F pr [x]/x N − 1, then C ∼ = ⊕i∈I Ci , where Ci is the ideal

{a(u n ξni ) : a(x) ∈ C} ⊆ Rs ( p m i r , u) and I is a complete set of pr -cyclotomic coset
representatives modulo n.

Proof Define the map γ : F pr [x]/x N − 1 → A, where γ (a(x)) = (A0 , A1 , . . . ,

An−1 ). Easily check that γ (a(x) + b(x)) = γ (a(x)) + γ (a(x)). Let a(x), b(x) be
polynomials over F pr of degree less than N , then there exist q(x), r (x) ∈ F pr [x]
such that a(x)b(x) = q(x)(x N − 1) + r (x), where deg(r (x)) < N . So we have
  
a(u n ξni )b(u n ξni ) = r (u n ξni ), which means γ (a(x)b(x)) = γ (a(x)) ∗ γ (b(x)), where
∗ denotes the componentwise product. If γ (a(x)) = 0, then from the inversion for-
p s −1
mula we have j=0 at, j u j = 0 for any 0 ≤ t ≤ n − 1. This gives a(x) = 0, and
s
hence γ is an injection. Also, |A| = i∈I pr m i 2 = pr N , which means that γ is a
bijection. Thus, γ is an isomorphism.

From Theorem 2.3, we see that an ideal C of F pr [x]/x N − 1 is isomorphic to

a direct sum ⊕i∈I Ci , where Ci is the ideal of Rs ( p m i r , u). This direct sum will be
called the decomposition of C. In the following, we consider the relationship between
the decomposition of a cyclic code over F pr and its generator polynomial. First, we
give some results about the ring Rs (q, u) with q a power p [6].
s
Theorem 2.4 The ring Rs (q, u) = Fq [u]/u p − 1 is a chain ring with maximal
ideal u − 1 and residue field G F(q).

Theorem 2.5 There are ( p s + 1) distinct cyclic codes of length p s over Fq , and they
are precisely the ideals (u −1)i  of the ring Rs (q, u), for i ∈ {0, 1, . . . , p s }. Self-dual
cyclic codes of length p s over Fq exist if and only if p = 2. When p = 2, there is only
s−1
one self-dual cyclic code (u − 1)2  over Fq of length 2s .

Remark 2.6 We know that the number of distinct cyclic codes over F pr of length
N = p s n is at most ( p s + 1)|I | . On the other hand, from Theorems 2.3 and 2.5, we
deduce that the number of distinct cyclic codes over F pr of length N = p s n is pre-
cisely ( p s + 1)|I | . Therefore, for distinct sequences (ki )i∈I with 0 ≤ ki ≤ p s , cyclic
codes over F pr of length N = p s n are different each other with generator polynomials
i∈I [ f i (x)] .
ki

Lemma 2.7 Let f i (x) be the minimal polynomial of ξni in F pr [x]. Let m i be the size
of the pr −cyclotomic coset modulo n containing i, and n  a positive integer such that
nn  ≡ 1 (mod p s ). Then

123
On cyclic self-dual codes 513

 j
(i) f i (u n ξn ) is a unit in Rs ( pr m i , u) if j ∈ cl pr (i, n);
 
(ii) f i (u n ξni ) ∈ u − 1 but f i (u n ξni ) ∈
/ (u − 1)2 .

Proof (i) Since f i (x) = l∈cl pr (i,n) (x − ξnl ), it follows that

 
 j
    
j j j
f i (u n ξn ) = u n ξn − ξnl = (u n − 1)ξn + (ξn − ξnl ) .
l∈cl pr (i,n) l∈cl pr (i,n)

j
If j ∈ cl pr (i, n), then ξn − ξnl is a unit. Note that

  
(u n − 1)ξn = (u − 1)(u n −1 + u n −2 + · · · + 1)ξn ,
j j

 j  j
and so (u n − 1)ξn is noninvertible. Hence, f i (u n ξn ) is a unit.
 
(ii) As x n −1 = j∈I f j (x), we have j∈I f j (u n ξni ) = (u n ξni )n −1 = u−1. From
 
(i) we know that f j (u n ξni ) is a unit in Rs ( pr m i , u) for j = i. Hence f i (u n ξni ) =

q(u)(u − 1), where q(u) is a unit in Rs ( pr m i , u). This gives f i (u n ξni ) ∈ u − 1.

Suppose that f i (u n ξni ) ∈ (u − 1)2 , then there exists g(u) ∈ Fq r [u] such that

f i (u n ξni ) = g(u)(u − 1)2 . Hence q(u)(u − 1) = g(u)(u − 1)2 . This implies
u − 1 ∈ (u − 1)2 , which means u − 1 ⊆ (u − 1)2 . This is a contradiction.

Therefore, f i (u n ξni ) ∈
/ (u − 1)2 .

Theorem 2.8 Let x n − 1 = i∈I f i (x) be the unique factorization of x n − 1 into a

product of monic irreducible polynomials in F pr [x]. If C is a cyclic code over F pr of
length N = p s n with generator polynomial i∈I [ f i (x)]ki , where 0 ≤ ki ≤ p s , then
for each i ∈ I , C corresponds to the ideal (u − 1)ki  of Rs ( pr m i , u) under the map γ .

Proof Let the decomposition of C be ⊕i∈I Ci , where Ci is the ideal {a(u n ξni ) :
a(x) ∈ C} of Rs ( p m i r , u) . If a(x) = r (x)[ f i (x)]ki where r (x) is prime to f i (x) and
  
0 ≤ ki ≤ p s , then, by Lemma 2.7, a(u n ξni ) = r (u n ξni )[ f i (u n ξni )]ki ∈ (u −1)ki , but
/ (u−1) i . This means if a(x) ∈ C then a(x) = h(x) i∈I [ f i (x)]ki ∈ (u−1)ki ,
∈ k +1

but ∈/ (u − 1)ki −1 , for some h(x) ∈ F pr [x]. Thus, for each i ∈ I , C corresponds to
the ideal (u − 1)ki  of Rs ( pr m i , u) under the map γ .

Now, we consider indecomposable cyclic codes over F pr [x]. In [13], such codes
were described by the group algebra approach. An indecomposable cyclic code C
s
over F pr of length N = p s n has generator polynomial f i (x)l j∈I, j=i f j (x) p , for
some i ∈ I and 0 ≤ l ≤ p s . According to Theorem 2.8, C is isomorphic to the
ideal (u − 1)l  of Rs ( pr m i , u). We take an element Ai ∈ Rs ( p m i r , u). Consider the
element

A = (0, . . . , 0, Ai , 0, . . . , 0) ∈ ⊕i∈I Rs ( p m i r , u).

123
514 X. Kai, S. Zhu

Let ρ denote the inverse map of γ , and a(x) = ρ(A), then a(x) is an element in
F pr [x]/x N − 1 such that

 j Ai , if j = i
a(u n ξn ) = .
0, otherwise.

Recall that the Frobenius map of F pr is defined by a f = a p . We extend the Frobe-

nius automorphism f to Rs ( pr , u) by setting u f = u, and define the trace map from
−1 f i
Rs ( pr , u) to Rs ( p, u) to be T rr (z) = ri=0 z . Let A(z) be the Mattson–Solomon
polynomial of a(x), then for 0 ≤ t ≤ n − 1, we have


n−1  i −1
m i −1
m
f l −it· plr fl
Ah ξn−ht = Ai ξn−th = Ai (ξn−it ) f
l
A(ξnt ) = A i ξn =
h=0 h∈cl pr (i,n) l=0 l=0

i −1
m
(Ai ξn−it ) f = T rm i (Ai ξn−it ).
l
=
l=0

According to Inversion formula,


  

a(x) = ϕ 1, u −n , u −2n , . . . , u −(n−1)n
1

∗ T rm i (Ai ), T rm i (Ai ξn−i ), . . . , T rm i (Ai ξn−(n−1)t ) .
n
s −1
Let Ai = a0 + a1 u + · · · + a ps −1 u p , then

T rm i (Ai ξn−it ) = T rm i (a0 ξn−it ) + T rm i (a1 ξn−it )u + · · · + T rm i (a ps −1 ξn−it )u p

s −1
.

Thus we have proved the following theorem, which gives the trace description for an
indecomposable cyclic code C over F pr of length N = p s n.
Theorem 2.9 Let m be the order of pr modulo n, and m i the size of the pr -cyclotomic
coset modulo n containing i. Let ξn be a primitive nth root of unity in F pr m . If C is an
indecomposable cyclic code over F pr of length N = p s n, then C is equivalent to the
set {(v (a0 ) , v (a1 ) , . . . , v (a ps −1 ) ) : a j ∈ F pr m i }, where

1

v (a j ) = T rm i (a j ), T rm i (a j ξni ), . . . , T rm i (a j ξn(n−1)i ) .
n
From [9, Theorem 5.25] we know that the set

 
Cm i = T rm i (a j ), T rm i (a j ξni ), . . . , T rm i (a j ξn(n−1)i ) : a j ∈ F pr m i

−j
is the [n, m i ] minimal cyclic code of length n over F pr m i with nonzeros {ξn | j ∈
cl pr (i, n)}. Hence we see that the Hamming distance of the indecomposable cyclic

123
On cyclic self-dual codes 515

code defined as in Theorem 2.9 is equal to d((u − 1)l ) · d(cm i ), where (u − 1)l  is
an ideal of Rs ( pr m i , u). The Hamming weight of (u − 1)l  has been determined by
Dinh in [6].

3 Cyclic self-dual codes

Recall that if f (x) is a polynomial of degree h, the reciprocal polynomial of f (x) is

f ∗ (x) = x h f (x −1 ), so that the roots of f ∗ (x) are the reciprocals of the corresponding
roots of f (x). Let C be a cyclic code over F pr of length N with generator polynomial
g(x) and check polynomial h(x) = (x N −1)/g(x), then the dual code C ⊥ is cyclic and
has generator polynomial g ⊥ (x) = h ∗ (x). In the following, let i  be the representative
of the cyclotomic coset containing n − i for each i ∈ I .

Theorem 3.1 Let C be a cyclic code over F pr of length N = p s n with generator

polynomial i∈I [ f i (x)]ki , where f i (x) s are monic irreducible divisors of x n − 1 in
F pr [x] and 0 ≤ ki ≤ p s . Then C is self-dual if and only if ki + ki  = p s for each
i ∈ I.

Proof Let ai denote the constant of f i (x) for each i ∈ I , then ai  s are leading coeffi-
cients of f i∗ (x) s. Since i∈I f i (x) = x n − 1, we have i∈I ai = −1. So
   
ai−1 f i∗ (x) = ai−1 · f i∗ (x) = −x n f i (x −1 )
i∈I i∈I i∈I i∈I
= −x n (x −n − 1) = x n − 1.

Thus, (ai−1 f i∗ (x)) s are monic irreducible divisors of x n − 1 in F pr [x].

If C is self-dual, then
   
 
⊥ ∗ p s −ki −1 ∗ p s −ki
C = [ f i (x)] = [ai f i (x)]
i∈I i∈I
   
 
p s −ki
= [ f i  (x)] = [ f i  (x)] ki 
.
i  ∈I i  ∈I

From Remark 2.6 we have ki  = p s − ki for each i ∈ I .

On the other hand, if ki + ki  = p s for each i ∈ I , then
   
 
⊥
[ f i∗ (x)] p −ki
s
C = = [ f i  (x)] ki 

i∈I i∈I
 

= [ f i  (x)] ki 
= C,
i  ∈I

i.e., C is self-dual.

123
516 X. Kai, S. Zhu

It is well known that {0} is a pr −cyclotomic coset modulo n, hence, if C is a cyclic

code over F pr of length N = p s n, then by Theorem 3.1 it must be 2k0 = p s . This
forces p = 2.

Corollary 3.2 There exists a cyclic self-dual code over F2r with generator polynomial
s−1
(x n − 1)2 for length N = 2s n (n odd).
s−1
Proof Take ki = ki  = 2s−1 for each i ∈ I , then C =  i∈I [ f i (x)]2  = (x n −
s−1 s−1
1)2  = (x 2 n + 1). From Theorem 3.1 we know that C is self-dual.

Thus, we have the theorem below.

Theorem 3.3 Self-dual cyclic codes over F pr of length N = p s n exist if and if only
p = 2.
s−1 s−1
In F2r [x]/(x N − 1), (x n − 1)2 = x 2 n + 1. It is easy to see that the minimum
s−1 s−1
Hamming distance of C = (x n − 1)2  is 2. The code (x n − 1)2  is called the
trivial self-dual code over F2r .

Theorem 3.4 Nontrivial cyclic self-dual codes over F2r of length N = 2s n (n odd)
exist if and only if there exists a monic irreducible factor f (x) ∈ F2r [x] of x n − 1
such that f (x) and f ∗ (x) are not associate.

Proof Assume that there exists a monic irreducible factor f (x) ∈ F2r [x] of x n − 1
such that f (x) and f ∗ (x) are not associate. Then the constant of f (x) is nonzero, it
follows that deg( f (x)) = deg( f ∗ (x)), and f ∗ (x) is a factor of x n − 1. Therefore,
f (x) f ∗ (x) is also a factor of x n − 1. Write x n − 1 = f (x) f ∗ (x)g(x). Since

1 − x n = (x n − 1)∗ = ( f (x) f ∗ (x)g(x))∗ = f ∗ (x) f (x)g ∗ (x),

f ∗ (x) f (x)g ∗ (x) = − f (x) f ∗ (x)g(x). Note that g ∗ (x) and f (x) f ∗ (x) are coprime,
it follows that g ∗ (x)| g(x). Similarly, g(x)| g ∗ (x). Therefore, g(x) = −g ∗ (x). Con-
sider C =  f (x)h f ∗ (x)2 −h g(x)2 , for some 0 ≤ h ≤ 2s and h = 2s−1 . Now by
s s−1

Theorem 3.1
   
C ⊥ = f ∗ (x)2 −h f (x)h g ∗ (x)2 = f (x)h f ∗ (x)2 −h g(x)2
s s−1 s s−1
= C.

Conversely, assume that there is a nontrivial cyclic self-dual code C, then C has
the form C =  i∈I [ f i (x)]ki , where f i (x) s are monic irreducible divisors of x n − 1
in F2r [x] and 0 ≤ ki ≤ 2s . Suppose, to the contrary, that every irreducible fac-
tor f i (x) of x n − 1 in F2r [x] is an associate of f i∗ (x), then f i (x) = f i  (x) for
all i. It follows that i and i  are in the same cyclotomic coset. As C is self-dual,
Theorem 3.1 implies that for all i, ki + ki  = 2s . Hence, ki = ki  = 2s−1 . This
s−1
gives C =  i∈I [ f i (x)]ki  = (x n − 1)2 , which is the trivial self-dual code, a
contradiction.

123
On cyclic self-dual codes 517

For any 0 ≤ q ≤ n − 1, since n is odd, 2r has an inverse modulo n. Thus there exist
positive integers m such that 2r m q ≡ q (mod n). Let m q be the smallest such inte-
ger, then the set cl2r (q, n) = {q, 2r q, 22r q, . . . , 2(m q −1)r q} is the cyclotomic coset
modulo n containing q, where each 2ri q is reduced modulo n. Easily verify that for
q = n − q, cl2r (q, n) = cl2r (n − q, n) if and only if 2ri ≡ −1 (mod n) for some
integer i. Combining the above discussion with Theorem 3.4, we have the following
theorem.

Theorem 3.5 Nontrivial cyclic self-dual codes over F2r of length N = 2s n (n odd)
exist if and only if 2ri ≡ −1 (mod n) for all positive integers i.

Next, we consider the enumeration of cyclic self-dual codes over F2r of length
N = 2s n (n odd).
Now we use the terminology introduced in [10]. A cyclotomic coset is called sym-
metric if n − q ∈ cl2r (q, n) and asymmetric otherwise. The asymmetric cosets come
in pairs cl2r (q, n), cl2r (q  , n) where q  denotes the representative of the cyclotomic
coset containing n − q. Let δ(n) denote the number of such pairs. If cl2r (q, n) is
symmetric, then f q∗ (x) = ε1 f q (x) for some ε1 ∈ F2r , while if cl2r (q, n), cl2r (q  , n)
are an asymmetric pair, then f q∗ (x) = ε2 f n−q (x) for some ε2 ∈ F2r . Let C be a cyclic
self-dual codes over F2r of length N = 2s n (n odd). In the decomposition of C, the
s−1
ideal corresponding to each symmetric coset must be (u − 1)2 , while the ideals
corresponding to each asymmetric pair cl2r (q, n), cl2r (q  , n) can be taken (u − 1)l 
and (u − 1)2 −l  respectively, where l is any number in the range 0 ≤ l ≤ 2s . Thus
s

we obtain the following enumeration result.

Theorem 3.6 Let δ(n) be the number of pairs of asymmetric 2r -cyclotomic cosets
modulo n. Then the number of distinct cyclic self-dual codes over F2r of length N =
2s n (n odd) is (2s + 1)δ(n) .

Recall that the trace T rr (α) over F2 of an element α ∈ F2r is defined as T rr (α) =
r −1 2i
i=0 α . A basis B = {α1 , α2 , . . . , αr } of F2r over F2 is a self-dual basis if

1, if i = j,
T rr (αi α j ) = .
0, if i = j.

Let B = {α1 , α2 , . . . , αr } be a self-dual basis of F2r over F2 , then α ∈ F2r can be

written as α = ri=1 ai αi with ai ∈ F2 . The Lee weight w LB (α) with respect to B of
the element α is the number of i  s with ai = 1. The Lee weight w LB (c) of a vector
c ∈ F2Nr is the sum of the Lee weights of its components. A self-dual code over F2r is
said to be of Type II with respect to the basis B if the Lee weight of every codeword
in C is a multiple of 4 and Type I otherwise. Type II were studied in [1,2,7].

Theorem 3.7 Cyclic self-dual codes over F2r of length N = 2s n (n odd) are Type I.

Proof Let C be a cyclic self-dual code over F2r of length N = 2s n (n odd), and its
decomposition ⊕i∈I Ci , where Ci is the ideal of Rs (2r m i , u). By Theorem 3.1 it must
s−1
be C0 = (u − 1)2 . Let B = {α1 , α2 , . . . , αr } be a self-dual basis of F2r over

123
518 X. Kai, S. Zhu

s−1 s−1
F2 , then αl (u − 1)2 = αl u 2 + αl ∈ C0 for any 1 ≤ l ≤ r . Consider the vector
s−1
v = (αl u 2 + αl , 0, . . . , 0) ∈ ⊕i∈I Ci . From Theorem 2.9 it follows that ρ(v) must
be permutable to the N −tuple

n n
     
(αl , . . . , αl , 0, . . . , 0, αl , . . . , αl , 0, . . . , 0),

whose Lee weight is 2n. Hence, C contains a codeword of Lee weight 2n. Since 2n
is not a multiple of 4, we have that C is type I.

From Theorem 3.7 we see that no binary cyclic self-dual code of even length is
Type II. This gives another proof for [10, Corollary 2].
In [4], a parity check matrix of a repeated-root cyclic code was constructed via the
so-called Hasse derivative. Now, we also provide a construction method for the parity
check matrix of a cyclic self-dual code over F2r of length N = 2s n (n odd), which
is also suitable to a repeated-root cyclic code in general. Let g(x) be a codeword in
some cyclic self-dual code C over F2r of length N = 2s n (n odd), then g(x) can
2s −1 j 2s 2s
be written in the form g(x) = i=0 x g j (x ), where g j (x ) ∈ F2r [x]. So we
u ξn g j (ξni2 ). By Theorems 2.3 and 2.8, C ∼
 2 −1 n  j i j
s s
have g(u n ξni ) = i=0 = ⊕i∈I Ci and

g(u ξn ) ∈ Ci , where Ci is an ideal of Rs ( p , u). From this, we can obtain some
n i r m i

quantity relationship satisfied by g j (x) s and determine the parity check matrix of C.
We give an example to explain the method.

Example 3.8 Let ω be a primitive element in F8 with ω3 + ω + 1 = 0, then

x 7 − 1 = (x + 1)(x + ω)(x + ω2 ) · · · (x + ω6 )

in F8 [x]. According to Theorem 3.1, the code

 
C = (x + 1)(x + ω)(x + ω6 )(x + ω2 )2 (x + ω3 )2

is a self-dual code of length 14 over F8 . Let c(x) = c0 + c1 x + · · · + c13 x 13 ∈ C,

then c(x) can be written as
   
c0 + c2 x 2 + · · · + c12 x 12 + x c1 + c3 x 2 + · · · + c13 x 12 .

By Theorem 2.8, we have c(uω2 ) = 0 and c(uω3 ) = 0. Hence,

c0 + c2 ω4 + · · · + c12 ω24 = 0, c1 ω2 + c3 ω6 + · · · + c13 ω26 = 0,

and

c0 + c2 ω6 + · · · + c12 ω36 = 0, c1 ω3 + c3 ω9 + · · · + c13 ω39 = 0.

123
On cyclic self-dual codes 519

By Theorem 2.8 again, we have c(u) ∈ u − 1, which implies (u − 1)c(u) = 0. This
13 13 13
gives i=0 ci = 0. Similarly, we get i=0 ci ωi = 0 and i=0 ci ω6i = 0. From the
above equations, we obtain the parity check matrix of C is
⎛ ⎞
1 1 1 1 1 1 1 1 1 1 1 1 1 1
⎜1 ω ω2 ω3 ω4 ω5 ω6 1 ω ω2 ω3 ω4 ω5 ω6 ⎟
⎜ ⎟
⎜1 0 ω4 0 ω 0 ω5 0 ω2 0 ω6 0 ω3 0 ⎟
⎜ ⎟
⎜0 ω2 ω6 0 ω3 0 ω4 0 ω 0 ω5 ⎟
⎜ 0 1 0 ⎟.
⎜1 0 ω6 0 ω5 0 ω4 0 ω3 0 ω2 0 ω 0 ⎟
⎜ ⎟
⎝0 ω3 0 ω2 0 ω 0 1 0 ω6 0 ω5 0 ω4 ⎠
1 ω6 ω5 ω4 ω3 ω2 ω 1 ω6 ω5 ω4 ω3 ω2 ω

4 Some nontrivial cyclic self-dual codes over F4 , F8 , F16

4.1 Consider nontrivial cyclic self-dual codes over F4 of length ≤ 22. Let ω be the
primitive element in F4 , where ω2 + ω + 1 = 0.
• N = 6 In F4 [x], x 3 − 1 = (x + 1)(x + ω)(x + ω2 ). There are two equivalent
nontrivial cyclic self-dual codes [6, 3, 3] with generator polynomials

(x + 1)(x + ω)2 , (x + 1)(x + ω2 )2 .

The Hamming weight enumerator is

W (C) = x 6 + 6x 3 y 3 + 27x 2 y 4 + 18x y 5 + 12y 6 .

• N = 12 (1) There are two equivalent nontrivial cyclic self-dual codes

[3, 6, 12] with generator polynomials

(x + 1)2 (x + ω)4 , (x + 1)2 (x + ω2 )4 .

The Hamming weight enumerator is

W (C) = x 12 +12x 9 y 3 +54x 8 y 4 +36x 7 y 5 +60x 6 y 6 +324x 5 y 7

+ 945x 4 y 8 +1116x 3 y 9 +972x 2 y 10 +432x y 11 +144y 12.

(2) There are two equivalent nontrivial cyclic self-dual codes [4, 6, 12]
with generator polynomials

(x + 1)2 (x + ω)3 (x + ω2 ), (x + 1)2 (x + ω)(x + ω2 )3 .

The Hamming weight enumerator is

W (C) = x 12 + 45x 8 y 4 + 168x 6 y 6 + 288x 5 y 7 + 1035x 4 y 8

+ 960x 3 y 9 + 1080x 2 y 10 + 288x y 11 + 231y 12 .

123
520 X. Kai, S. Zhu

• N = 14 In F4 [x], x 7 − 1 = (x + 1)(x 3 + x 2 + 1)(x 3 + x + 1). There are two

equivalent nontrivial cyclic self-dual codes [4, 7, 14] with generator
polynomials

(x + 1)(x 3 + x 2 + 1)2 , (x + 1)(x 3 + x + 1)2 .

The Hamming weight enumerator is

W (C) = x 14 + 42x 10 y 4 + 231x 8 y 6 + 2205x 6 y 8

+ 7686x 4 y 10 + 5544x 2 y 12 + 675y 14 .

• N = 18 In F4 [x], x 9 − 1 = (x + 1)(x + ω)(x + ω2 )(x 3 + ω)(x 3 + ω2 ).

(1) There are two equivalent nontrivial cyclic self-dual codes [4, 9, 18]
with generator polynomials

(x + 1)(x 3 + ω)(x 3 + ω2 )(x + ω)2 ,

(x + 1)(x 3 + ω)(x 3 + ω2 )(x + ω2 )2 .

The Hamming weight enumerator is

W (C) = x 18 + 108x 14 y 4 + 504x 12 y 6 + 2646x 10 y 8

+ 384x 9 y 9 + 11016x 8 y 10 + 13824x 7 y 11
+ 47628x 6 y 12 + 48384x 5 y 13 + 68040x 4 y 14
+ 32256x 3 y 15 + 28593x 2 y 16 + 3456x y 17 + 5304y 18 .

(2) There are two equivalent nontrivial cyclic self-dual codes [3, 9, 18]
with generator polynomials

(x + 1)(x + ω)(x + ω2 )(x 3 + ω)2 ,

(x + 1)(x + ω)(x + ω2 )(x 3 + ω2 )2 .

The Hamming weight enumerator is

W (C) = x 18 + 18x 15 y 3 + 81x 14 y 4 + 54x 13 y 5 + 144x 12 y 6

+ 972x 11 y 7 + 2835x 10 y 8 + 3564x 9 y 9 + 5832x 8 y 10
+ 16362x 7 y 11 + 38907x 6 y 12 + 56862x 5 y 13
+ 60264x 4 y 14 + 43416x 3 y 15
+ 23328x 2 y 16 + 7776x y 17 + 1728y 18 .

123
On cyclic self-dual codes 521

(3) There are four equivalent nontrivial cyclic self-dual codes [3, 9, 18]
with generator polynomials

(x + 1)(x + ω)2 (x 3 + ω)2 , (x + 1)(x + ω)2 (x 3 + ω2 )2

(x + 1)(x + ω2 )2 (x 3 + ω)2 , (x + 1)(x + ω2 )2 (x 3 + ω2 )2 .

The Hamming weight enumerator is

W (C) = x 18 + 18x 15 y 3 + 297x 12 y 6 + 162x 11 y 7 + 2241x 10 y 8

+ 4920x 9 y 9 + 7290x 8 y 10 + 21222x 7 y 11
+ 36774x 6 y 12 + 45360x 5 y 13 + 59805x 4 y 14
+ 53478x 3 y 15 + 23976x 2 y 16 + 5400x y 17 + 1200y 18 .

• N = 22 In F4 [x],

x 11 − 1 = (x + 1)(x 5 + ωx 4 + x 3 + x 2 + ω2 x + 1)
×(x 5 + ω2 x 4 + x 3 + x 2 + ωx + 1).

There are two equivalent nontrivial cyclic self-dual codes [6, 11, 22] with
generator polynomials

(x + 1)(x 5 + ωx 4 + x 3 + x 2 + ω2 x + 1)2 ,
(x + 1)(x 5 + ω2 x 4 + x 3 + x 2 + ωx + 1)2 .

The Hamming weight enumerator is

W (C) = x 22 + 330x 16 y 6 + 330x 15 y 7 + 330x 14 y 8

+ 660x 13 y 9 + 9405x 12 y 10 + 25476x 11 y 11
+ 81312x 10 y 12 + 214170x 9 y 13 + 361185x 8 y 14
+ 548130x 7 y 15 + 773157x 6 y 16 + 824472x 5 y 17
+ 677655x 4 y 18 + 430320x 3 y 19 + 185328x 2 y 20
+ 51546x y 21 + 10497y 22 .

4.2 Consider nontrivial cyclic self-dual codes over F8 of length < 30. Let ω be a
primitive element in F8 , where ω3 + ω + 1 = 0.

• N = 14 In F8 [x], x 7 −1 = (x +1)(x +ω)(x +ω2 )(x +ω3 )(x +ω4 )(x +ω5 )(x +ω6 ).

123
522 X. Kai, S. Zhu

(1) There are six equivalent nontrivial cyclic self-dual codes [5, 7, 14]
with generator polynomials

(x + 1)(x + ω)2 (x + ω2 )2 (x + ω3 )2 ,
(x + 1)(x + ω)2 (x + ω3 )2 (x + ω5 )2 ,
(x + 1)(x + ω)2 (x + ω4 )2 (x + ω5 )2 ,
(x + 1)(x + ω2 )2 (x + ω3 )2 (x + ω6 )2 ,
(x + 1)(x + ω2 )2 (x + ω4 )2 (x + ω6 )2 ,
(x + 1)(x + ω4 )2 (x + ω5 )2 (x + ω6 )2 .

The Hamming weight enumerator is

W (C) = x 14 + 294x 9 y 5 + 294x 8 y 6 + 434x 7 y 7 + 8575x 6 y 8

+ 30870x 5 y 9 + 155722x 4 y 10 + 331142x 3 y 11
+ 588784x 2 y 12 + 662284x y 13 + 318752y 14 .

(2) There are two equivalent nontrivial cyclic self-dual codes [4, 7, 14]
with generator polynomials

(x + 1)(x + ω)2 (x + ω2 )2 (x + ω4 )2 ,
(x + 1)(x + ω3 )2 (x + ω5 )2 (x + ω6 )2 .

The Hamming weight enumerator is

W (C) = x 14 + 98x 10 y 4 + 931x 8 y 6 + 336x 7 y 7 + 14749x 6 y 8

+ 16464x 5 y 9 + 160622x 4 y 10 + 312816x 3 y 11
+ 656208x 2 y 12 + 595056x y 13 + 339871y 14 .

(3) There are six equivalent nontrivial cyclic self-dual codes [6, 7, 14]
with generator polynomials

(x + 1)(x + ω)(x + ω6 )(x + ω2 )2 (x + ω3 )2 ,

(x + 1)(x + ω)(x + ω6 )(x + ω5 )2 (x + ω4 )2 ,
(x + 1)(x + ω2 )(x + ω5 )(x + ω)2 (x + ω3 )2 ,
(x + 1)(x + ω2 )(x + ω6 )(x + ω)2 (x + ω4 )2 ,
(x + 1)(x + ω3 )(x + ω4 )(x + ω)2 (x + ω5 )2 ,
(x + 1)(x + ω3 )(x + ω4 )(x + ω6 )2 (x + ω2 )2 .

123
On cyclic self-dual codes 523

The Hamming weight enumerator is

W (C) = x 14 + 343x 8 y 6 + 714x 7 y 7 + 9163x 6 y 8

+ 36946x 5 y 9 + 140336x 4 y 10 + 332318x 3 y 11
+ 611324x 2 y 12 + 641606x y 13 + 324401y 14 .

(4) There are six equivalent nontrivial cyclic self-dual codes [6, 7, 14]
with generator polynomials

(x + 1)(x + ω)(x + ω6 )(x + ω2 )2 (x + ω4 )2 ,

(x + 1)(x + ω)(x + ω6 )(x + ω5 )2 (x + ω3 )2 ,
(x + 1)(x + ω2 )(x + ω5 )(x + ω)2 (x + ω4 )2 ,
(x + 1)(x + ω2 )(x + ω5 )(x + ω6 )2 (x + ω3 )2 ,
(x + 1)(x + ω3 )(x + ω4 )(x + ω6 )2 (x + ω5 )2 ,
(x + 1)(x + ω3 )(x + ω4 )(x + ω)2 (x + ω2 )2 .

The Hamming weight enumerator is

W (C) = x 14 + 343x 8 y 6 + 616x 7 y 7 + 9849x 6 y 8

+ 34888x 5 y 9 + 143766x 4 y 10 + 328888x 3 y 11
+ 613382x 2 y 12 + 640920x y 13 + 324499y 14 .

(5) There are six equivalent nontrivial cyclic self-dual codes [4, 7, 14]
with generator polynomials

(x + 1)(x + ω)(x + ω2 )(x + ω5 )(x + ω6 )(x + ω j )2 ( j ∈ {3, 4}),

(x + 1)(x + ω)(x + ω3 )(x + ω4 )(x + ω6 )(x + ω j )2 ( j ∈ {2, 5}),
(x + 1)(x + ω2 )(x + ω3 )(x + ω4 )(x + ω5 )(x + ω j )2 ( j ∈ {1, 6}).

The Hamming weight enumerator is

W (C) = x 14 + 147x 10 y 4 + 1470x 8 y 6 + 112x 7 y 7 + 12887x 6 y 8

+ 21168x 5 y 9 + 149940x 4 y 10 + 317520x 3 y 11
+ 674485x 2 y 12 + 571536x y 13 + 347886y 14 .

4.3 Consider nontrivial cyclic self-dual codes over F16 of length ≤ 10. Let ω be the
primitive element in F16 , where ω4 + ω + 1 = 0.
• N = 6 In F16 [x], x 3 − 1 = (x + 1)(x + ω5 )(x + ω10 ). There are two equivalent
nontrivial cyclic self-dual codes [6, 3, 3] with generator polynomials

(x + 1)(x + ω5 )2 , (x + 1)(x + ω10 )2 .

123
524 X. Kai, S. Zhu

The Hamming weight enumerator is

W (C) = x 6 + 30x 3 y 3 + 135x 2 y 4 + 1170x y 5 + 2760y 6 .

• N = 10 In F16 [x], x 5 − 1 = (x + 1)(x + ω3 )(x + ω6 )(x + ω9 )(x + ω12 ).

(1) There are four equivalent nontrivial cyclic self-dual codes [4, 5, 10]
with generator polynomials

(x + 1)(x + ω3 )2 (x + ω6 )(x + ω9 ),
(x + 1)(x + ω12 )2 (x + ω6 )(x + ω9 ),
(x + 1)(x + ω6 )2 (x + ω3 )(x + ω12 ),
(x + 1)(x + ω9 )2 (x + ω3 )(x + ω12 ).

The Hamming weight enumerator is

W (C) = x 10 + 150x 6 y 4 + 30x 5 y 5 + 3150x 4 y 6 + 14700x 3 y 7

+118725x 2 y 8 + 360150x y 9 + 551670y 10 .

(2) There are four equivalent nontrivial cyclic self-dual codes [4, 5, 10]
with generator polynomials

(x + 1)(x + ω3 )2 (x + ω6 )2 , (x + 1)(x + ω3 )2 (x + ω9 )2 ,
(x + 1)(x + ω12 )2 (x + ω6 )2 , (x + 1)(x + ω12 )2 (x + ω9 )2 .

The Hamming weight enumerator is

W (C) = x 10 + 150x 6 y 4 + 360x 5 y 5 + 1500x 4 y 6 + 18000x 3 y 7

+115425x 2 y 8 + 361800x y 9 + 551340y 10 .

References

1. Betsumiya, K., Gulliver, T.A., Harada, M., Mumemasa, A.: On Type II codes over F4 . IEEE Trans.
Inform. Theory 47, 2242–2248 (2001)
2. Betsumiya, K., Harada, M., Mumemasa, A.: Type II codes over F2r . In: Proc. 14th AAECC (Lecture
Notes in Computer Science), Springer, Berlin, Germany (2001)
3. Blahut, R.E.: Theory and Practice of Error Control Codes. Addison Wesley, Reading, MA (1983)
4. Castagnoli, G., Massey, J.L., Schoeller, P.A., Vov Seemann, N: On repeated-root cyclic codes. IEEE
Trans. Inform. Theory 37, 337–342 (1991)
5. Dey, B.K., Raian, B.S.: DFT domain characterization of quasi-cyclic Codes. AAECC 13, 453–
474 (2003)
6. Dinh, H.Q.: On the linear ordering of some classes of negacyclic and cyclic codes and their distance
distributions. Finite Fields Their Appl. 14, 22–40 (2008)
7. Gaborit, P., Pless, V., Solé, P., Atkin, O.: Type II codes over F4 . Finite Fields Their Appl. 8, 171–
183 (2002)
8. Nedeloaia, C.S.: Weight distributions of cyclic self-dual codes. IEEE Trans. Inform. Theory 49, 1582–
1591 (2003)

123
On cyclic self-dual codes 525

9. Pless, V.S., Huffman, W.C., Brualdi, B.A.: An introduction to algebraic codes, Chap. 1. In: Handbook
of Coding Theory, pp. 3–139. Elsevier, Amsterdam (1998)
10. Sloane, N.J.A., Thompson, J.G.: Cyclic self-dual codes. IEEE Trans. Inform. Theory 5, 364–366 (1983)
11. Tang, L.Z., Soh, C.B., Gunawan, E.: A note on the q−ary image of a q m −ary repeated-root cyclic
codes. IEEE Trans. Inform. Theory 43, 732–737 (1997)
12. Van Lint, J.H.: Repeated-root cyclic codes. IEEE Trans. Inform. Theory 37, 343–345 (1991)
13. Zimmermann, K.H.: On generalizations of repeated-root cyclic codes. IEEE Trans. Inform.
Theory 42, 641–649 (1996)

123