Journal of Algebra and Related Topics

Vol. 6, No 2, (2018), pp 63-78

BASIS OF A MULTICYCLIC CODE AS AN IDEAL IN

Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i

R. M. LALASOA, R. ANDRIAMIFIDISOA∗ , AND T. J. RABEHERIMANANA

Abstract. First, we apply the method presented by Zahra Sep-

asdar in the two-dimensional case to construct a basis of a three
dimensional cyclic code. We then generalize this construction to a
general s-dimensional cyclic code.

1. Introduction
Multicyclic codes are cyclic codes of dimension s, or s-D cyclic codes,
where s > 2 is an integer. Two-dimensional cyclic codes have been in-
tensively studied ([2, 3, 7, 8, 9, 10, 11]). There are much less results
about general s-D cyclic codes, where s > 3 . Because of their rich
mathematical structure, as in [6], involving Algebraic Geometry or in
[5, 9, 10, 11], using group algebra and Galois Theory or ideals in a
polynomial quotient ring, multicyclic codes are of great importance.

A fundamental problem in coding theory is the construction of a

generator matrix, which allows to find parameters of the code and to
encode messages. The representation of a 2-D cyclic code as an ideal
in a polynomial quotient ring makes the construction of a generator
matrix possible, since it can be deducted from a basis of the ideal
([9, 10, 11]).

MSC(2010): primary: 13F20, 16D25; Secondary: 94B60

Keywords: quotient ring, ideal, ideal basis, multicyclic code, polynomial division algorithm
(by many divisors).
Received: 1 August 2018, Accepted: 3 December 2018.
∗ Corresponding author .
63
64 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

Sepasdar, in [9, 11] presented a method which allowed to construct

an ideal basis of a 2-D cyclic code, which is represented as an ideal
in a two-variables polynomial quotient ring. Her method is based on
an “elimination principle”, and the fact that the coefficients of two-
variate polynomials in the ideals are also polynomials in one variable,
which belongs to principal ideals, and therefore, already have a “gen-
erator polynomial”. From these generator polynomials, one then can
construct a basis of the code, as a vector space.

In the present paper, we apply Sepasdar’s method first to the 3-D

case and then to the general s-D case (s > 2). It is organized as follows :

In Section 2, we give a brief description of multicyclic codes, as ideals

in a polynomial quotient ring. Then we describe the structure of these
polynomial quotient rings. Finally, we describe an auxiliary polyno-
mial quotient ring which will allow us to apply Sepasdar’s method to
higher-dimensional cases

In Section 3, we first describe 2-D cyclic codes while applying what

we saw in Section 2 to the 2-D case. We also present Sepasdar’s result.

In Section 4, we state and prove our main Theorem 4.1. This give the
construction of a basis of a 3-D code, as an ideal in a three-variables
polynomial quotient ring. We prove it by using Sepasdar’s method
with ideals in two variables, a modification of the method and more
calculation to the 3-D case.

In the last Section 5, we state and prove Theorem 5.1, which is the
generalization of Theorem 4.1 to the s-D case, by induction.

2. Multicyclic codes
Throughout this paper, Fq denotes the Galois Field with q elements
(where q is a power of a prime number). Let s > 2 be an inte-
ger, X1 , . . . , Xs distinct letters (or variables) and ρ1 , . . . , ρs > 1 in-
tegers. Let Fq [X1 , . . . , Xs ] the ring of the polynomials in the variables
X1 , . . . , Xs with coefficients in Fq . An element of this ring is of the
form
X
d(X1 , . . . , Xs ) = dα1 ,...,αs X1α1 · · · Xsαs (2.1)
(α1 ,...,αs )∈Ns
 BASIS OF A MULTICYCLIC CODE 65

where dα1 ,...,αs ∈ Fq , the sum being finite.

A multicyclic code, or more precisely, an s-dimensional cyclic code

(s-D multicyclic code), is an ideal I in the quotient ring (and also an
F-algebra)
R = Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i. (2.2)
For σ = 1, . . . , s, let xσ be the residue class of Xσ modulo the ideal
hX1ρ1 − 1, . . . , Xsρs − 1i :
xσ = Xσ + hX1ρ1 − 1, . . . , Xsρs − 1i. (2.3)
We then have
xρσσ = 1. (2.4)
We will need some supplementary notations :

• Let Z /ρσ Z = {0, 1, . . . , ρσ − 1} be the set of the residue classes of

the integers modulo ρσ . for σ = 1, . . . , s. Now, construct the abelian
groups
Gs = Z /ρ1 Z × . . . × Z /ρs Z,
(2.5)
Gs−1 = Z /ρ1 Z × . . . × Z /ρs−1 Z .
• Let Fq [x1 , . . . , xs ] be the set of polynomials in the variables x1 , . . . , xs .
Using (2.4), we have
X
Fq [x1 , . . . , xs ] ={ rα1 ,...,αs xα1 1 · · · xαs s | rα1 ,...,αs ∈ Fq
(α1 ,...,αs )∈Gs (2.6)
for (α1 , . . . , αs ) ∈ Gs }.
• In the same manner, we define
X αs−1
Fq [x1 , . . . , xs−1 ] ={ rα1 ,...,αs−1 xα1 1 · · · xs−1 | rα1 ,...,αs−1 ∈ Fq
(α1 ,...,αs−1 )∈Gs−1

for (α1 , . . . , αs−1 ) ∈ Gs−1 }.

(2.7)
Note that the set Fq [x1 , . . . , xs ] (resp. Fq [x1 , . . . , Xs−1 ]) is finite and
has cardinality q ρ1 ···ρs (resp. q ρ1 ···ρs−1 ) since it can be identified with
the set of mappings from Gs (resp. Gs−1 ) to Fq .
• Let S be the quotient ring
ρ
S = Fq [X1 , . . . , Xs−1 ]/hX1ρ1 − 1, . . . , Xs−1
s−1
− 1i. (2.8)
The following proposition gives another representation of the ring R :
66 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

Proposition 2.1. With the preceding notations, the rings

R = Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i and Fq [x1 , . . . , xs ] are iso-
morphic.

Proof. Let ϕs be the ring homomorphism defined by

ϕs : Fq [X1 , . . . , Xs ] −→ Fq [x1 , . . . , xs ]
(2.9)
Xi 7−→ xi

for i = 1, . . . , s. Let d(X1 , . . . , Xs ) ∈ Fq [X1 , . . . , Xr ] as in (2.1). Using

(2.4), we then have
X
ϕs (d(X1 , . . . , Xs )) = dα1 ,...,αs xα1 1 mod ρ1 · · · xαs s mod ρs
(α1 ,...,αs )∈Ns

where “ασ mod ρσ ” designs the remainder of ασ by the euclidean di-

vision of ασ by ρσ , Moreover, ϕs is surjective and its kernel is the ideal
hX1ρ1 − 1, . . . , Xsρs − 1i. Therefore, by the first isomorphism theorem
for rings ([4]), there is an isomorphism ϕ̄ which makes the following
diagram commutative :
ϕs
Fq [X1 , . . . , Xs ] Fq [x1 , . . . , xs ]
ϕ¯s
πs

Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i

where πs is the canonical surjection. This proves the proposition.



The following proposition gives another representation of the ring S:

Proposition 2.2. With the preceding notations the rings

ρs−1
S = Fq [X1 , . . . , Xs−1 ]/hX1ρ1 − 1, . . . , Xs−1 − 1i and Fq [x1 , . . . , xs−1 ] are
isomorphic.

Proof. We define the homomorphism ψs by

ψs : Fq [X1 , . . . , Xs−1 ] −→ Fq [y1 , . . . , ys−1 ]
ρ (2.10)
Xi 7−→ yi = Xi + hX1ρ1 − 1, . . . , Xs−1
s−1
− 1i

(It is not the same as ϕs−1 defined by (2.9). Here, yi is the residue
ρs−1
class of Xi modulo the ideal hX1ρ1 − 1, . . . , Xs−1 − 1i). Using the same
arguments as for the mapping ϕs in (2.9), there exists an isomorphism
 BASIS OF A MULTICYCLIC CODE 67

ψ̄s which makes the following diagram commutative :

ψs
Fq [X1 , . . . , Xs−1 ] Fq [y1 , . . . , ys−1 ]
ψ¯s
πs−1

ρ
Fq [X1 , . . . , Xs−1 ]/hX1ρ1 − 1, . . . , Xs−1
s−1
− 1i,
(2.11)
where πs−1 is the canonical projection. Now define the homomorphism
θ by
θ : Fq [y1 , . . . , ys−1 ] −→ Fq [x1 , . . . , xs−1 ]
(2.12)
yi 7−→ xi .
for i = 1, . . . , s. The mapping θ is obviously surjective and since its
domain and codomain have the same cardinality (see the remark next to
(2.7)), it follows that it is bijective, hence a ring isomorphism. Going
back to the commutative diagram (2.11), we have that the mapping
θ ◦ ψ̄s is an isomorphism between S and Fq [x1 , . . . , xs ]. 
From the propositions 2.1 and 2.2, we then can make the following
identifications:
R = Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i = F[x1 , . . . , xs ]
and
ρ
S = Fq [X1 , . . . , Xs−1 ]/hX1ρ1 − 1, . . . , Xs−1
s−1
− 1i = F[x1 , . . . , xs−1 ].
(2.13)
We directly deduce the following corollary :
Corollary 2.3. Using the notations in (2.13), we have
R = Fq [x1 , . . . , xs−1 ][xs ] = S[xs ]. (2.14)
Remark 2.4. For d(X1 , . . . , Xs ) ∈ Fq [X1 , . . . , Xs ], by the division al-
gorithm ([1, 5]) of d(X1 , . . . , Xs ) by Xσρσ − 1, for σ = 1, . . . s, we can
write
s
X
d(X1 , . . . , Xs ) = qσ (X1 , . . . , Xs )(Xσρσ − 1) + r(X1 , . . . , Xs ), (2.15)
σ=1

with qσ , r ∈ Fq [X1 , . . . , Xs ] and r being of the form

X
r(X1 , . . . , Xs ) = rα1 ,...,αs X1α1 · · · Xsαs (2.16)
(α1 ,...,αs )

where ασ 6 ρσ − 1 for σ = 1, . . . s. Therefore, a representative of the

residue class of d(X1 , . . . , Xs ) modulo the ideal hX1ρ1 − 1, . . . , Xsρs − 1i
is r(X1 , . . . , Xs ). Using this, we can find the result of Proposition 2.1 :
68 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

the class of d(X1 , . . . , Xs ) is the same of that of r(X1 , . . . , Xs ), which

is X
r(x1 , . . . , xs ) = rα1 ,...,αs xα1 1 · · · xαs s ,
(α1 ,...,αs )∈Gs
and immediately deduce that
X
R={ rα1 ,...,αs xα1 1 · · · xαs s | rα1 ,...,αs ∈ Fq } = Fq [x1 , . . . , xs ].
(α1 ,...,αs )∈Gs

Similar considerations also allow to show that S = Fq [x1 , . . . , xs−1 ].

Using Corollary 2.14, an element f (x1 , . . . , xs ) ∈ R can be uniquely
written under the form
ρs −1
X
f (x1 , . . . , xs ) = fi (x1 , . . . , xs−1 )xis , (2.17)
i=0

where fi (x1 , . . . , xs−1 ) ∈ S for i = 0, . . . , ρs − 1.

3. Two dimensional cyclic codes
Let l, n > 1 be integers, X, Y two letters or variables, Fq [X, Y ] the
ring of polynomials in X, Y with coefficients in Fq and R the quotient
ring
R = Fq [X, Y ]/hX l − 1, Y m − 1i.
and S the quotient ring
S = Fq [X]/hX l − 1i.
According to (2.3), let
x = X + hX l − 1, Y m − 1i
y = Y + hX l − 1, Y m − 1i
the residue classes of X and Y modulo the ideal hX l − 1, Y m − 1i. A
two dimensional (2-D) cyclic code is an Ideal I in R. Using (2.13), we
can write
X
R = Fq [x, y] = { dα,β xα y β | dα,β ∈ Fq },
(α,β)∈G2

where G2 = Z /l Z × Z /m Z and
X
S = Fq [x] = { cα xα | cα ∈ Fq }.
α∈Z /l Z

Here is Sepasdar’s result :

Result 3.1 (Sepasdar, [9, 11]). An ideal in Fq [X, Y ]/hX l − 1, Y m − 1i,
i.e. a two-dimensional cyclic code has a finite basis.
 BASIS OF A MULTICYCLIC CODE 69

If I is an ideal in R = Fq [X, Y ]/hX l − 1, Y m − 1i, this result states

that there exist elements p1 (x, y), . . . , pk (x, y) ∈ R (where k > 1 is an
integer) such that an element c(x, y) ∈ I may be written as
k
X
c(x, y) = ui (x, y)pi (x, y)
i=1

with ui (x, y) ∈ R for i = 1, . . . , k.

4. Ideal basis of a three-dimensional cyclic code

Now, applying what we saw in Section 2 for s = 3, A 3-D cyclic code
is then an Ideal I in R = Fq [X l , Y m , Z n ]/hX l − 1, Y m − 1, Z n − 1i. By
(2.8),
S = Fq [X, Y ]/(X l − 1, Y m − 1).
According to (2.3), let
x = X + hX l − 1, Y m − 1, Z n − 1i,
y = Y + hX l − 1, Y m − 1, Z n − 1i,
z = Z + hX l − 1, Y m − 1, Z n − 1i
the residue classes X, Y and Z modulo the ideal hX l −1, Y m −1, Z n −1i.
Using (2.13), we can write
X
R = Fq [x, y, z] = { dα,β,γ xα y β z γ | dα,β,γ ∈ Fq },
(α,β,γ)∈G3
X
S = F[x, y] = { cα,β xα y β | cα,β ∈ Fq },
(α,β)∈G2

where G3 = Z /l Z × Z /m Z × Z /n Z and G2 = Z /l Z × Z /m Z. By
(2.17), an element f (x, y, z) ∈ R can be uniquely written under the
form
n−1
X
f (x, y, z) = fi (x, y)z i (4.1)
i=0
where fi (x, y) ∈ S for i = 0, . . . , n − 1.
Using the equality z n = 1, we have
zf (x, y, z) = fn−1 (x, y) + f0 (x, y)z + · · · + fn−2 (x, y)z n−1 ,
z 2 f (x, y, z) = fn−2 (x, y) + fn−1 (x, y)z + · · · + fn−3 (x, y)z n−1 ,
.. (4.2)
.
z n−1 f (x, y, z) = f1 (x, y) + f2 (x, y)z + · · · f0 (x, y)z n−1 .

The main task in this section is to prove our main theorem:

70 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

Theorem 4.1. Using the preceding notations, let I be an ideal of

R = Fq [X l , Y m , Z n ]/hX l − 1, Y m − 1, Z n − 1i. For j = 0, . . . , n − 1, let
Ij be the following set
n−1
X
Ij = {gj (x, y) ∈ S | ∃g(x, y, z) ∈ I with g(x,y, z) = gi (x, y)z i
i=j

where gi ∈ S for i = j, . . . , n − 1}.

(4.3)
Then the following hold:
(j) (j)
(1) The Ij ’s are ideals of S, generated by elements p1 , . . . , prj ∈ S,
i.e.
rj
(j) (j)
X
(j)
Ij = hp1 , p2 , . . . , prj i = { p(j)
µ (x, y)qµ (x, y) | qµ ∈ S}.
µ=1
(j) (j)
(2) There exist elements p1 (x, y, z), . . . , prn−1 (x, y, z) ∈ I, such that
n−1
X (j)
p(j)
µ (x, y, z) = piµ (x, y)z i , (4.4)
i=j
(j)
for j = 0, . . . , n − 1, i = j, . . . , n − 1, where piµ (x, y) ∈ S, with
(j) (j)
pjµ (x, y) = pµ (x, y) for µ = 1, . . . , ri .
(j) (j)
(3) The elements p1 , . . . , prj , j = 1, . . . , n − 1 generate I, i.e.
(0) (1) (j) (n−1)
I = hp1 , . . . , p(0) (1) (j)
r1 , p1 , . . . , pr1 , . . . , p1 , . . . , prj , . . . , p1 , . . . , p(n−1)
rn−1 i.

Proof. We see that that all the Ij ’s are non-empty since they contains
the zero polynomial.

(1) Fix an element j ∈ {0, . . . , n − 1}. If g0 (x, y) ∈ Ij , there exists

g(x, y, z) ∈ I such that
n−1
X
g(x, y, z) = gj (x, y)z i .
i=j

First, we have xg0 (x, y), yg0 (x, y) ∈ Ij since I is an ideal of R and
n−1
X n−1
X
i
xg(x, y, z) = xgi (x, y)z , yg(x, y, z) = ygi (x, y)z i
i=j i=j

are elements of I. Second, for g(x, y, z) and g 0 (x, y, z) ∈ I, we have

g(x, y, z) + g 0 (x, y, z) ∈ I since I is an ideal. Thus, with obvious nota-
tions, g0 (x, y) + g00 (x, y) ∈ Ij , for g0 (x, y) and g00 (x, y) ∈ Ij and, Ij is
 BASIS OF A MULTICYCLIC CODE 71

indeed an ideal of S = Fq [X, Y ]/(X l − 1, Y m − 1). By Result 3.1, it has

(j) (j) (j)
a basis {p1 , p2 , . . . , prj } (where rj ∈ N∗ ), as stated in the theorem.

The ideal I0 will be of special interest:

n−1
X
I0 = {g0 (x, y) ∈ S | ∃g(x, y, z) ∈ I with g(x,y, z) = gi (x, y)z i
i=0
where gi ∈ S for i = 0, . . . , n − 1}.
(4.5)
By (4.2), it follows that
z i f (x, y, z) ∈ I0 , (4.6)
which yields that fi (x, y) ∈ I0 for i = 0, . . . , n − 1.
(j)
(2) The assertion results from the fact that pµ (x, y) is an element
of S for j = 0, . . . , n − 1 and µ = 1, . . . , ri and the definition of Ij in
(5.2).

By (4.2) and (4.6), where we replace f by p(j) , we have, by appro-

(j) (j)
priate choices of k, z k pµ (x, y, z) ∈ I, which implies that piµ ∈ I0 for
(0) (0)
i = 0, . . . , n − 1. This latter being generated by p1 , . . . , . . . , pr0 , there
(j)
exist tiµν (x, y) ∈ S such that
r0
X
(j) (j)
piµ (x, y) = p(0)
ν (x, y)tiµν (x, y) (4.7)
ν=1

for j = 0, . . . , n − 1 and µ = 1, . . . , ri . Using (5.3) and (4.7), we then

have
r0
n−1 X
X
(j) (j)
pµ (x, y, z) = p(0) i
ν (x, y)tiµν (x, y)z . (4.8)
i=j ν=1

Now, consider an element f (x, y, z) ∈ I of the form (4.1). Since

(0)
f0 (x, y) ∈ I0 , there exist qµ (x, y) ∈ S, µ = 1, . . . , r0 such that
r0
X
f0 (x, y) = p(0) (0)
µ (x, y)qµ (x, y). (4.9)
µ=1

(0) (0)
Using (4.7) for j = 0 and the fact that p0µ (x, y) = pµ (x, y), from (2),
Theorem 4.1, it follows that
r0 X
X r0
(0)
f0 (x, y) = p(0) (0)
ν (x, y)t0µν (x, y)qµ (x, y). (4.10)
µ=1 ν=1
72 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

Put
r0
X
h1 (x, y, z) = f (x, y, z) − p(0) (0)
µ (x, y, z)qµ (x, y). (4.11)
µ=1

We have
n−1
X r0 X
n−1 X
X r0
i (0)
h1 (x, y, z) = fi (x, y)z − p(0) (0)
ν (x, y)tiµν (x, y)qµ (x, y)z
i

i=0 i=0 µ=1 ν=1

(by (4.8) for j = 0),

n−1
X r0 X
n−1 X
X r0
i (0)
= (f0 (x, y) + fi (x, y)z ) − (p(0) (0) i
ν (x, y)tiµν (x, y)qµ (x, y)z ,
i=1 i=1 µ=0 ν=1
r0 X
X r0 n−1
X
(0)
= (f0 (x, y) − p(0) (0)
ν (x, y)t0µν (x, y)qµ (x, y)) + fi (x, y)z i
µ=1 ν=1 i=1
n−1
X r0 X
X r0
(0) (0)
− ( piµ (x, y)tiµν (x, y)qµ(0) (x, y))z i ,
i=1 µ=1 ν=1
n−1
X n−1 X
X r0
i
= fi (x, y)z − ( p(0) (0)
µ (x, y)qµ (x, y))z
i

i=1 i=1 µ=1

(by (4.10) and (4.7) for j = 0).

(4.12)
(0)
Since f and pµ are elements of I which is an ideal of R, the polynomial
h1 is also an element of I. We remark that
n−1
X (1)
h1 (x, y, z) = hi (x, y)z i
i=1
(1) (1)
where hi ∈ S for i = 1, . . . , n − 1. In other words, h1 (x, y) ∈ I1 .
(1)
Therefore, there exists qµ ∈ S, µ = 1, . . . , r1 such that
r1
X
(1)
h1 (x, y) = p(1) (1)
µ (x, y)qµ (x, y).
µ=1

(1) (1)
Using (4.7), for j = 1, and the fact that p1µ (x, y) = pµ (x, y), from
(2), Theorem 4.1, it follows that
r1 X
X r0
(1) (1)
h1 (x, y) = p(0) (1)
ν (x, y)tiµν (x, y)qµ (x, y). (4.13)
µ=1 ν=1
 BASIS OF A MULTICYCLIC CODE 73

Put
r1
X (1)
h2 (x, y, z) = h1 (x, y, z) − p(1)
µ (x, y, z)q1µ (x, y). (4.14)
µ=1

We have
n−1
X r1 X
X r0 X
n−1
(1) (1)
h2 (x, y, z) = h1 (x, y, z)z i − p(0)
ν (x, y)tiµν (x, y)q1µ (x, y)z
i

i=1 µ=1 ν=1 i=1

(by (4.8) for j = 1),

n−1
X r1 X
X r0
(1) (1) (1)
= (h1 (x, y)z + h1i (x, y)z i − (p(0)
ν (x, y)t1µν (x, y)q1µ (x, y))z
i=2 µ=1 ν=1
n−1
X r1 X
X r0
(1) (1)
− ( p(0)
ν (x, y)tiµν (x, y)q1µ (x, y))z
i

i=1 µ=1 ν=1

n−1
X r1 X
n−1 X
X r0
(1) (1)
= hi (x, y)z i − p(0) (1)
ν (x, y)tiµν (x, y)qµ (x, y)z
i

i=2 i=1 µ=1 ν=1

n−1
X r1
n−1 X
X (0) (1)
= h1i (x, y)z i − piµ (x, y)q1µ (x, y))z i
i=2 i=1 µ=1

(by (4.13) and (4.7) for j = 1).

(4.15)
(1)
(3) Since h1 (x, y, z) and pµ (x, y, z) are elements of I which is an ideal
of R, the polynomial h2 (x, y, z) is also in I and can be written in the
form
n−1
X (2)
h2 (x, y, z) = hi (x, y)z i
i=2
(2)
where hi (x, y) ∈ S. In other words, h2 (x, y, z) ∈ I2 . Therefore there
(2)
exist qµ (x, y) ∈ S, µ = 1, . . . , r2 such that
r2
X
(2)
h2 (x, y) = p(2) (2)
µ (x, y)qµ (x, y). (4.16)
µ=1

(2) (2)
Using (4.7), for j = 2, and the fact that p2µ (x, y) = pµ (x, y), from
(2), Theorem 4.1, it follows that
r2 X
X r0
(2) (2)
h2 (x, y) = p(0) (2)
ν (x, y)tiµν (x, y)qµ (x, y). (4.17)
µ=1 ν=1
74 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

Put
r2
X
h3 (x, y, z) = h2 (x, y, z) − p(2) (2)
µ (x, y, z)qµ (x, y). (4.18)
µ=1

We then have
n−1
X r2 X
X r0
n−1 X
(2) (2)
= hi (x, y, z)z i − p(0) (2)
ν (x, y)tiµν (x, y)qµ (x, y)z
i

i=2 µ=1 i=2 ν=1

(by (4.8) for j = 2),

n−1
X r2 X
X r0
(2) (2) (2)
= (h2 (x, y)y)z 2 + h2i (x, y)z i − (p(0)
ν (x, y)t2µν qµ (x, y))z
2

i=3 µ=1 ν=1

r2 X
X r0
n−1 X
(2) (2)
− ( p(0)
ν (x, y)tiµν (x, y)q1µ (x, y))z
i

µ=1 i=3 ν=1

n−1
X r2 X
n−1 X
X r0
(2) (2)
= hi (x, y)z i − p(0) (2)
ν (x, y)tiµν (x, y)qµ (x, y)z
i

i=3 i=3 µ=1 ν=1

n−1
X r2
n−1 X
X
(2) (2)
= hi (x, y)z i − piµ (x, y)qµ(2) (x, y)z i
i=3 i=3 µ=1

(by (4.17) and (4.7) for j = 2).

(4.19)
(2)
Since h2 (x, y, z) and pµ (x, y, z) are elements of I which is an ideal of
R, the polynomial h3 (x, y, z) is also in I and can be written in the form
n−1
X (3)
h3 (x, y, z) = hi (x, y)z i , (4.20)
i=3

(3)
with hi (x, y) ∈ S.

Applying the preceding methods, we get polynomials

h4 (x, y, z), . . . , hn−2 (x, y, z),
qµ(3) (x, y, z)16µ6r3 , . . . , qµ(n−2) (x, y, z)16µ6rn−2
of S. Finally, put
rn−1
X
hn−1 (x, y, z) = hn−2 (x, y, z) − p(n−2)
µ (x, y, z)qµ(n−2) (x, y). (4.21)
µ=1
 BASIS OF A MULTICYCLIC CODE 75

Then hn−1 (x, y, z) is a polynomial of I of the form hn−1

n−1 (x, y)z
n−1
. In
(n−1) (n−1)
other words, hn−1 (x, y) ∈ In−1 . Therefore, there exist qµ (x, y) ∈ S,
µ = 1, . . . , rn−1 such that

rn−1
(n−1)
X
hn−1 (x, y) = p(n−1)
µ (x, y)qµ(n−1) (x, y), (4.22)
µ=1

which yields

rn−1
X
hn−1 (x, y, z) = ( p(n−1)
µ (x, y)qµ(n−1) (x, y))z n−1
µ=0
rn−1
X
= p(n−1)
µ (x, y, z)qµ(n−1) (x, y).
µ=0

For an arbitrary element f (x, y, z) ∈ I we then have shown the follow-

ing equalities :

r0
X
h1 (x, y, z) = f (x, y, z) − p(0) (0)
µ (x, y, z)qµ (x, y),
µ=1
Xr1
h2 (x, y, z) = h1 (x, y, z) − p(1) (1)
µ (x, y, z)qµ (x, y),
µ=1
Xr1
h3 (x, y, z) = h2 (x, y, z) − p(2) (2)
µ (x, y, z)qµ (x, y),
µ=1
···
rn−2
X
hn−1 (x, y, z) = hn−2 (x, y, z) − p(n−2)
µ (x, y, z)qµ(n−2) (x, y),
µ=1
rn−1
X
= p(n−1)
µ (x, y, z)qµ(n−1) (x, y)
µ=1
76 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

and finally have

r0
X
f (x, y, z) = p(0) (0)
µ (x, y, z)qµ (x, y) + h1 (x, y),
µ=1
r0
X r1
X
= p(0) (0)
µ (x, y, z)qµ (x, y) + p(1) (1)
µ (x, y, z)qµ (x, y) + h2 (x, y),
µ=1 µ=1
= ··· ,
r0 rn−1
X X
= p(0) (0)
µ (x, y, z)qµ (x, y) + ··· + p(n−1)
µ (x, y, z)qµ(n−1) (x, y),
µ=1 µ=1

and conclude that

(0) (1) (i) (n−1)
I = hp1 , . . . , p(0) (1) (i)
r1 , p1 , . . . , pr1 , . . . , p1 , . . . , pri , . . . , p1 , . . . , p(n−1)
rn−1 i.

5. Basis for an s-dimensional cyclic code

Now, we generalize Theorem 4.1 for general s-D multicyclic codes
(s > 2). Let
R = Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i,
as in (2.2) and S the quotient ring
ρ
S = Fq [X1 , . . . , Xs−1 ]/hX1ρ1 − 1, . . . , Xs−1
s−1
− 1i.
In section 2, by (2.13) and (2.14), we know that if xi denotes the residue
class of Xi modulo the ideal hX1ρ1 − 1, . . . , Xsρs − 1i, then
S = Fq [x1 , . . . , xs−1 ], R = Fq [x1 , . . . , xs ] = S[xs ]. (5.1)
Theorem 5.1. Let I be an ideal of the quotient ring
R = Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i. For j = 0, . . . , ρs − 1, let
Ij be the following set
Ij = {gj (x1 , . . . , xs−1 ) ∈ S |∃g(x1 , . . . , xs ) ∈ I with
ρs −1
X
g(x1 , . . . , xs ) = gi (x1 , . . . , xs−1 )xis
i=j

where gi ∈ S for i = j, . . . , ρs − 1}.

(5.2)
 BASIS OF A MULTICYCLIC CODE 77

Then the following hold:

(j) (j)
(1) The Ij ’s are ideals of S, generated by elements p1 , . . . , prj ∈ S,
i.e.
rj
(j) (j)
X
Ij = hp1 , p2 , . . . , p(j)
rj i ={ p(j)
µ (x1 , . . . , xs−1 )qµ (x1 , . . . , xs−1 ) | qµ ∈ S}.
µ=1

(j) (j)
(2) There exist elements p1 (x1 , . . . , xs ), . . . , prρs −1 (x1 , . . . , xs ) ∈ I, such
that
ρs −1
X (j)
(j)
pµ (x1 , . . . , xs ) = piµ (x1 , . . . , xs−1 )xis , (5.3)
i=j
(j)
for j = 0, . . . , n − 1, i = j, . . . , ρs − 1, where piµ (x1 , . . . , xs−1 ) ∈ S, with
(j) (j)
pjµ (x1 , . . . , xs−1 ) = pµ (x1 , . . . , xs−1 ) for µ = 1, . . . , ri .
(j) (j)
(3) The elements p1 , . . . , prj , j = 1, . . . , ρs − 1 generate I, i.e.
(0) (1) (j) (ρ −1) s −1)
I = hp1 , . . . , p(0) (1) (j)
r1 , p1 , . . . , pr1 , . . . , p1 , . . . , prj , . . . , p1
s
, . . . , p(ρ
rρs −1 i.

Proof. We prove by induction on s: the case s = 2 was treated by

Sepasdar and the case s = 3 in Section 4. Now, suppose the theorem
is true for s − 1, where s > 4. Let I be an s-dimensional ideal in the
quotient ring R = Fq [X1 , . . . , Xs ]/hX1ρ1 − 1, . . . , Xsρs − 1i. By (5.1), we
have R ∼ ρ
= S[xs ]. Using the representation S = Fq [X1 , . . . , Xs−1 ]/hX1 1 −
ρs−1
1, . . . , Xs−1 − 1i, by the induction hypothesis, each of the Ij ’s as in the
(j)
theorem have a basis {pµ (x1 , . . . , xs−1 )}, since they are ideals of S. As
(j)
we have done in Section 4, there are also polynomials pµ (x1 , . . . , xs )
whose set is a basis of I. Thus, the result is true for s and, therefore,
by induction, for all s ∈ N∗ . 
Acknowledgments
The authors would like to thank the referee for careful reading.

References
1. D. Cox D, J. Little, D. O’Shea, Ideals, Varieties and Algorithms, Springer,
Second Edition, 1997.
2. T. Ikai, H. Kosako, Y. Kojima, Two-dimensional cyclic Codes, Electronics and
Communications in Japan, 57-A (1975), 27-35.
3. H. Imai, A theory of two-dimensional cyclic Codes, Information and Control,
34 (1977), 1-21.
4. N. Jacobson, Basic Algebra II (2nd ed.), Dover, New York, 2009.
78 LALASOA , ANDRIAMIFIDISOA, AND RABEHERIMANANA

5. R. M. Lalasoa, R. Andriamifidisoa and T. J. Rabeherimanana, Multicyclic codes

and algebraic dynamical Systems, British Journal of Mathematics & Computer
Science, (2) 21 (2017), 1-22.
6. K. Saints and C. Heegard, Algebraic-geometric codes and multidimensional
cyclic Codes : A unified Theory and Algorithms for decoding using Groebner
Bases, IEEE Trans. IT, (6) 41 (1995), 1733-1751 .
7. S. Sakata, Decoding binary 2-D cyclic codes by the 2-D BerlekampMassey algo-
rithm, IEEE Trans. Inf. Theory IT, (4) 37 (1991), 12001203.
8. S. Sakata, S. On determining the independent Point Set for doubly periodic Ar-
rays and encoding two-dimensional cyclic Codes and their Duals, IEEE Trans.
Inf. Theory IT, (5) 27 (1981), 556565.
9. Z. Sepasdar, Some notes on the characterization of two dimensional skew cyclic
Codes, J. Algebra Relat. Topics, (2) 4 (2016), 1-8.
10. Z. Sepasdar, Characterizations of some two-dimensional cyclic Codes corre-
spond to the Ideals of Fq [x, y]/hxs − 1, y 2k − 1i, Finite Fields Appl., 41 (2016),
97112.
11. Z. Sepasdar, Generator Matrix for two-dimensional cyclic Codes of arbitrary
Length, arXiv:1704.08070v1, [math.AC], 26 Apr 2017.

Rufine Marius Lalasoa

Department of Mathematics and Computer Science, University of Antananarivo,
p.O.Box 906, 101 Antananarivo, Madagascar.
Email: larissamarius.lm@gmail.com

Ramamonjy Andriamifidisoa
Department of Mathematics, University of Antananarivo, p.O.Box 906, 101 An-
tananarivo, Madagascar,
And
Higher Polytechnics Institute of Madagascar (ISPM), Ambatomaro Antsobolo,
101 Antananarivo, Madagascar.
Email: ramamonjy.andriamifidisoa@univ-antananarivo.mg

Toussaint Joseph Rabeherimanana

Department of Mathematics, University of Antananarivo, p.O.Box 906, 101 An-
tananarivo, Madagascar.
Email: rabeherimanana.toussaint@yahoo.fr