Professional Documents
Culture Documents
______________________________________________________________________________________
Abstract—We investigate linear codes over the ring Z4 +uZ4 + codes over Z4 remain a special topic of interest in the field of
vZ4 + wZ4 + uvZ4 + uwZ4 + vwZ4 + uvwZ4 , with conditions algebraic coding theory because of their relation to lattices,
u2 = u, v 2 = v, w2 = w, uv = vu, uw = wu and vw = wv. designs, cryptography and their many applications”1 [20].
We first analyze the structure of the ring and then define linear
codes over this ring. The Lee weight and the Gray map for Recently, several new families of rings, namely the non-
these codes are defined and MacWilliams relations for complete, chain Frobenius rings, have been studied in connection with
symmetrized, and Lee weight enumerators are derived. The coding theory. These rings have rich mathematical theory,
Singleton bound as well as maximum distance separable codes in particular algebraic structures. Yildiz and Karadeniz [20]
are also considered. Furthermore, cyclic and quasi-cyclic codes derived algebraic structures related to linear codes over the
are discussed, and as an application some new linear codes
over Z4 with the highest known minimum Lee distance are ring Z4 + uZ4 , with u2 = 0. They [20] also constructed
also obtained. several good formally self-dual codes over Z4 from the codes
over Z4 + uZ4 . Bandi and Bhaintwal ( [2], [3]) considered
Index Terms—Linear codes, MacWilliams relations, Maxi-
mum distance separable codes, Cyclic codes, Quasi-cyclic codes, codes over the ring Z4 + vZ4 , with v 2 = v, and Z4 + wZ4 ,
optimal codes. with w2 = 1, 2w, respectively and derived several algebraic
structures including the MacWilliams relation with respect
to Rosenbloom-Tsfasman metric over the ring Z4 + vZ4 and
I. I NTRODUCTION the properties as well as a construction method of self-dual
We, as a human being, cannot not communicate to each codes over the ring Z4 + wZ4 . Moreover, Dian, Detiena,
other. The transmission of information is the heart of com- Suprijanto, and Barra [15] have also obtained some results
munication. Although reliable communication has been an on linear codes over the ring Z2m + vZ2m , with v 2 = v.
unavoidable problem with the human life, it is still a kind Recently, Li, Guo, Zhu, and Kai [14] generalized the ring
of mystery for a long time. It was in 1948, when Claude E. considered by Bandi and Bhaintwal [2] by adding two new
Shannon [16] showed that, if it is given a noisy communi- terms uZ4 and uvZ4 , with the conditions u2 = u, v 2 = v,
cation channel, there is a number called the capacity of the and uv = vu, and derived some properties corresponding to
channel such that reliable communication can be achieved at the linear codes over the ring Z4 + uZ4 + vZ4 + uvZ4 .
any rate below the channel capacity (see Section 13 in [16]). In this paper, we further generalized the ring considered
In other word, the existence of good codes is guarantied, by Li, Guo, Zhu, and Kai [14] to the ring Z4 + uZ4 + vZ4 +
theoretically. This seminal paper [16] has marked the birth wZ4 + uvZ4 + uwZ4 + vwZ4 + uvwZ4 with the conditions
of information theory and coding theory. u2 = u, v 2 = v, w2 = w, uv = vu, uw = wu and
Unfortunately, the proof of Shannon on the existence of vw = wv. We study linear codes over this ring and derive
good codes is not constructive: he proved only the existence some corresponding properties. The paper is organized as
of such codes but did not construct the codes itself. One follows. In Section 2, we study main properties of the ring
main problem in coding theory is to construct good codes Z4 + uZ4 + vZ4 + wZ4 + uvZ4 + uwZ4 + vwZ4 + uvwZ4 .
that satisfy the Shannon’s noisy-channel coding theorem. We then define linear codes, Lee weight, and also a Gray
Codes over finite rings have become an active research map for the linear codes over Z4 + uZ4 + vZ4 + wZ4 +
area in classical coding theory over the recent decades. In uvZ4 + uwZ4 + vwZ4 + uvwZ4 . A Singleton bound as
particular, after the appearance of the work of Hammons, well as maximum distance separable codes are slightly
Kumar, Calderbank, Sloane, and Solé [12], a lot of research considered. In Section 3, several types of weight enumerators
went towards studying (linear) codes over Z4 . Although ”the are defined and related MacWilliams relations are derived.
results were generalized to many different types of rings, the Finally, in Section 4, cyclic and quasi-cyclic codes over
Z4 + uZ4 + vZ4 + wZ4 + uvZ4 + uwZ4 + vwZ4 + uvwZ4
Manuscript received June 6, 2020; revised November 8, 2020. This are investigated. As an application, several examples of cyclic
research was supported by Institut Teknologi Bandung (ITB) and the
Ministry of Research, Technology and Higher Education (Kementerian Riset codes over Z4 with the highest known minimum Lee distance
dan Teknologi/ Badan Riset dan Inovasi Nasional (Kemenristek/ BRIN)), are obtained.
Republic of Indonesia. Throughout this paper, we follow standard definitions for
Bustomi is a lecturer in the Department of Mathematics, Faculty of
Sciences and Technology, Universitas Airlangga, Campus C Jl. Mulyorejo undefined terms as used in many coding theory books (e.g.
Surabaya, 60115 INDONESIA. (email: bustomi@fst.unair.ac.id) [13]).
A.P. Santika is an assistant professor in the Combinatorial Mathematics
Research Group, Faculty of Mathematics and Natural Sciences, Institut
Teknologi Bandung, Jl. Ganesha 10, Bandung, 40132, INDONESIA. (email: II. S TRUCTURES OF LINEAR CODES OVER R
aditps@math.itb.ac.id)
D. Suprijanto is an associate professor in the Combinatorial Mathe-
Throughout this paper, R denotes the ring Z4 + uZ4 +
matics Research Group, Faculty of Mathematics and Natural Sciences, vZ4 + wZ4 + uvZ4 + uwZ4 + vwZ4 + uvwZ4 with u2 = u,
Institut Teknologi Bandung, Jl. Ganesha 10, Bandung, 40132, INDONESIA.
∗ Corresponding author, email: djoko@math.itb.ac.id 1 [20], pp. 25.
v 2 = v, w2 = w, uv = vu, uw = wu dan vw = wv. We can easily see that φ is isomorphic. Finally, the Gray
The ring R can also be regarded as the quotient ring Z4 , map in R is defined as an extension of the map φ on Rn as
namely Z4 [u, v, w]/hu2 − u, v 2 − v, w2 − wi. This ring is
commutative and has identity. Φ : Rn −→ Z8n 4
The element η ∈ R is called idempotent if η 2 = η. The (c0 , c1 , . . . , cn−1 ) 7−→ (r1,0 , r1,1 , . . . , r1,n−1 ,
elements x and y of R is called orthogonal if xy = 0.
. . . , r8,0 , r8,1 , . . . , r8,n−1 ),
A. Structures of the ring R and a Gray map 8
P
First, we consider a decomposition of R then define a Gray where ci ∈ R and rji ∈ Z4 satisfying ci = rji ηj .
j=1
map from the ring R to Z84 . The Lee weight on Z4 , denoted by wL , is defined as
Consider the idempotent elements of R below
η1 = 1 − u − v − w + uv + uw + vw − uvw 0, x = 0,
= (1 − u)(1 − v)(1 − w), wL (x) := 2, x = 2,
1, x = 1 or 3.
η2 = u − uv − uw + uvw = u(1 − v)(1 − w),
η3 = v − uv − vw + uvw = (1 − u)v(1 − w), From the map φ : r 7−→ (rP 1 , r2 , . . . , r8 ), we define the Lee
8
η4 = w − uw − vw + uvw = (1 − u)(1 − v)w, weight on R as wL (r) = i=1 wL (ri ). The Lee weight of
η5 = uv − uvw = uv(1 − w), a vector c = (c0 , c1 , . . . , cn−1 ) ∈ Rn is defined to be a
η6 = uw − uvw = u(1 − v)w, rational sum of the Lee weight of its components, that is
Pn−1
wL (c) = i=0 wL (ci ). We also defined the Lee distance
η7 = vw − uvw = (1 − u)vw, between c and d ∈ Rn , as dL (c, d) = wL (c − d).
η8 = uvw. We also have another kind of weight and distance called
The above eight elements are also pairwise orthogonal, since a Hamming weight and a Hamming distance, and they are:
8
P wH (r) = |{j : rj 6= 0, 0 ≤ j ≤ n − 1}| and dH (r, s) =
ηi ηj = 0 for i 6= j, and satisfy ηi = 1. Hence, by Chinese wH (r − s), for all r, s ∈ Rn , respectively.
i=1
Remainder Theorem, we have
R = Rη1 ⊕ Rη2 ⊕ Rη3 ⊕ Rη4 ⊕ Rη5 ⊕ Rη6 ⊕ Rη7 ⊕ Rη8
B. Linear Codes over R
= Z4 η1 ⊕ Z4 η2 ⊕ Z4 η3 ⊕ Z4 η4 ⊕ Z4 η5 ⊕ Z4 η6 ⊕ Z4 η7
⊕ Z4 η8 . A nonempty subset C ⊆ Rn is called a linear code over
R if C is a submodule of R. To define a dual of the code
Moreover, for any r = a + bu + cv + dw + euv + f uw + C, let us first define the Euclidean inner product on Rn . Let
gvw + huvw ∈ R with a, b, c, d, e, f, g, h ∈ Z4 , we have x = (x0 , x1 , . . . , xn−1 ) and y = (y0 , y1 , . . . , yn−1 ) be two
8
X vectors in Rn . The Euclidean inner product of x and y is
r=r ηi defined as
i=1 Xn
= rη1 + rη2 + rη3 + rη4 + rη5 + rη6 + rη7 + rη8 hx, yi := xj yj ,
j=1
= aη1 + (a + b)η2 + (a + c)η3 + (a + d)η4
+ (a + b + c + e)η5 + (a + b + d + f )η6 where the operations are performed in the ring R.
+ (a + c + d + g)η7 Dual of the code C ⊆ Rn is the code
+ (a + b + c + d + e + f + g + h)η8
C ⊥ = {x ∈ Rn : hx, yi = 0, for all y ∈ C}.
= r1 η1 + r2 η2 + r3 η3 + r4 η4 + r5 η5
+ r6 η6 + r7 η7 + r8 η8 Clearly, C ⊥ is also linear if C is linear over R. Since R is
a Frobenius ring, we also have |C| · |C ⊥ | = 48n [22].
with
Denote r = (r(0) , r(1) , . . . , r(n−1) ) ∈ Rn , and r(i) =
r1 =a ri1 η1 + ri2 η2 + · · · + ri8 η8 , for 0 ≤ i ≤ n − 1. Then r
r2 =a+b can be uniquely expressed as
r3 =a+c
r = r1 η1 + r2 η2 + · · · + r8 η8 ,
r4 =a+d
r5 =a+b+c+e where rj = (r0j , r1j , . . . , rn−1,j ) ∈ Zn4 , for 1 ≤ j ≤ 8. By
r6 =a+b+d+f using this expression, the inner product of any two vectors
x, y ∈ Rn can be written as
r7 =a+c+d+g
r8 = a + b + c + d + e + f + g + h, x · y = (x1 · y1 )η1 + (x2 · y2 )η2 + · · · + (x8 · y8 )η8 ,
and hence r1 , r2 , r3 , r4 , r5 , r6 , r7 , r8 ∈ Z4 . It is clear that the
where x = x1 η1 + x2 η2 + · · · + x8 η8 , xj =
expression r = r1 η1 + r2 η2 + r3 η3 + r4 η4 + r5 η5 + r6 η6 +
(x0j , x1j , . . . , xn−1,j ) ∈ Zn4 , and y = y1 η1 + y2 η2 + · · · +
r7 η7 + r8 η8 is unique. Define the map φ from R to Z84 by n
y 8 , yj = (y0j , y1j , . . . , yn−1,j ) ∈ Z4 , and xj · yj =
P8 ηn−1
r 7−→ (r1 , r2 , r3 , r4 , r5 , r6 , r7 , r8 ). k=0 xkj ykj , for 1 ≤ j ≤ n.
III. W EIGHT ENUMERATORS AND M AC W ILLIAMS B. The Symmetrized Lee weight, Hamming weight and Lee
RELATIONS weight enumerator
In this section we consider several weight enumerators for In the ring Z4 , we know that wL (1) = 1 = wL (3) and
a linear codes C. We also derived the related MacWilliams the symmetrized Lee weight enumerator for codes over Z4
relations. is defined as
SLW EC (X0 , X1 , X2 ) = CW EC (X0 , X1 , X2 , X1 ).
A. The complete weight enumerator and MacWilliams rela-
tion We define the symmetrized Lee weight enumerator of codes
We knew that the number of elements of R is 65536. over R using similar idea as above. For that purpose, we
The complete weight enumerator (CWE) of a linear code first decompose R into Di = {x ∈ R : wL (x) = i}, for
C ⊆ Rn is defined as 0 ≤ i ≤ 16. Then we have
X 65535 |D0 | = |D16 | = 1,
Y naj (c)
CW EC (X0 , X1 , . . . , X65535 ) = Xj , 8
c∈C j=0
|D1 | = |D15 | = 2 = 16,
1
where nai (c) denotes the number of appearances of ai ∈ R
8 8
in the vector c. |D2 | = |D14 | = 22 + = 120,
2 1
Remark 1. Remember that CW EC (X0 , X1 , . . . , X65535 ) is 8 8 7
|D3 | = |D13 | = 23 +2 = 560,
a homogeneous polynomial in 65536 variables with total 3 1 1
degree of each monomial being the length of the code C, 8 8 6 8
n. Since the code C is linear, then C always contains |D4 | = |D12 | = 24 + 22 + = 1820,
4 2 1 2
the vector 0. It implies that the term X0n always appears
8
8 5
8 7
in CW EC (X0 , X1 , . . . , X65535 ). From the complete weight |D5 | = |D11 | = 25 + 23 +2 = 4368,
5 3 1 1 2
enumerator we may obtain a lot of information related to the
8 8 4 8 6
code, such as the size of the code: |D6 | = |D10 | = 26 + 24 + 22
6 4 1 2 2
X
CW EC (1, 1, . . . , 1) = 1 = |C|. 8
+ = 8008,
c∈C 3
7 8 5 8 3 3 8 5
|D7 | = |D9 | = 2 +2 +2
7 5 1 3 2
As the ring R is a Frobenius ring, the MacWilliams
relation for the complete weight enumerator holds (see [22]). 8 7
+2 = 11440,
To find the exact relation we define the following character 1 3
on R. 8 8 2 8 4
|D8 | = 28 + 26 + 24
Let I be a non-zero ideal in R. Define χ : I → C∗ by 8 6 1 4 2
χ(a + bu + cv + dw + euv + f uw + gvw + huvw) = ih 8 6 8
+ 22 + = 12870.
2 3 4
with C∗ is a unit group in complex number. We know that
χ is a non-trivial character on R. By looking at the elements that have the same Lee weights,
we can define the symmetrized Lee weight enumerator.
Defining the Hadamard transform by Symmetrized Lee weight enumerator (SLWE) of a linear code
X C over R is defined as
fˆ(c) = χ(c · d)f (d),
d∈Rn SLW EC (X0 , X1 , . . . , X16 )
we obtain the following equation = CLW EC (X0 , X1 , . . . , X1 , X2 , . . . , X2 , X3 , . . . , X3 ,
X X | {z } | {z } | {z }
fˆ(c) = |C| f (d). (1) 16 120 560
c∈C d∈C ⊥
X4 , . . . , X4 , X5 , . . . , X5 , X6 , . . . , X6 , X7 , . . . , X7 ,
| {z } | {z } | {z } | {z }
We have the MacWilliams relation for the complete weight 1820 4368 8008 11440
P Proof: For i, j = 0, 1, 2, . . . , 16, we determine Proof: Similar to the proof of Theorem 9 in [14].
χ(rs) for r ∈ Di . By definition, we have We also have the MacWilliams relation with respect to the
s∈Dj
Hamming weight enumerator.
SLW EC ⊥ (X0 , X1 , . . . , X16 ) Theorem III.4. Let C be a linear code of length n over R.
= CW EC ⊥ (X0 , X1 , . . . , X1 , X2 , . . . , X2 , X3 , . . . , X3 , Then
| {z } | {z } | {z }
16 120 560 1
HamC ⊥ (X, Y ) = HamC (X + 65536Y, X − Y ).
X4 , . . . , X4 , X5 , . . . , X5 , X6 , . . . , X6 , X7 , . . . , X7 , |C|
| {z } | {z } | {z } | {z }
1820 4368 8008 11440
Proof: Similar to the proof of Theorem 10 in [14].
X8 , . . . , X8 , X9 , . . . , X9 , X10 , . . . , X10 , X11 , . . . , X11 ,
| {z } | {z } | {z } | {z } Next, we consider the other weight enumerator with
12870 11440 8008 4368
respect to the Lee weight, called Lee weight enumerator.
X12 , . . . , X12 , X13 , . . . , X13 , X14 , . . . , X14 , For a linear code C, define Ai as a number of elements
| {z } | {z } | {z }
1820 560 120 of C having Lee weight i. The sequence A0 , A1 , . . . , A16n
X15 , . . . , X15 , X16 ) is called weight distribution in C with respect to the Lee
weight. The Lee weight enumerator for C is defined by
| {z }
16
16 X 16 X 16n
1 X X
= CW EC χ(a1 s)Xj , χ(a2 s)Xj ,
X X
|C| LeeC (X, Y ) = X 16n−wL (c) Y wL (c) = Ai X 16n−i Y i
j=0 s∈Dj j=0 s∈Dj c∈C i=0
16 X
...,
X
χ(a65536 s)Xj . Then we have the following property.
j=0 s∈Dj Theorem III.5. Let C be a linear code of length n over R.
Since for aj , ak ∈ Dj we have Then
16 X 16 X
X X LeeC (X, Y ) = SLW EC (X 16 , X 15 Y, X 14 Y 2 , X 13 Y 3 ,
χ(aj s)Xj = χ(ak s)Xj ,
j=0 s∈Dj j=0 s∈Dj X 12 Y 4 , X 11 Y 5 , X 10 Y 6 , X 9 Y 7 , X 8 Y 8 , X 7 Y 9 ,
then X 6 Y 10 , X 5 Y 11 , X 4 Y 12 , X 3 Y 13 , X 2 Y 14 , XY 15 , Y 16 ).
16
SLW EC ⊥ (X0 , X1 , . . . , X16 ) Proof: Denote wL (c) =
P
ini (c). Then we have
16 X i=0
1 X X
= SLW EC χ(ai s)Xj , . . . , 16 16 16
|C| X X X
ai ∈D0
j=0 s∈Dj 16n−wL (c) = 16ni (c)− ini (c) = (16−i)ni (c).
16 X
X X i=0 i=0 i=0
χ(ai s)Xj .
By definition, we obtain
ai ∈D16 j=0 s∈Dj
X
By direct calculation, we obtain LeeC (X, Y ) = X 16n−wL (c) Y wL (c)
16 X c∈C
X X 16 16
χ(ai s)Xj = Bk X P
(16−i)ni
P
ini
ai ∈Dk j=0 s∈Dj = X i=0 Y i=0
c∈C
for k = 1, 2, . . . , 16. Hence, we have 16
XY
= X 16−i Y i
SLW EC ⊥ (X0 , X1 , . . . , X16 ) c∈C i=0
1 = SLW EC (X 16 , X 15 Y, . . . , Y 16 ).
= SLW EC (B0 , B1 , . . . , B16 ).
|C|
Another weight enumerator of a linear code C, called a The following result gives us a MacWilliams relation with
Hamming weight enumerator, respect to the Lee weight enumerator.
HamC (X, Y )
X
X n−wH (c) Y wH (c) , Theorem III.6. Let C be a linear code of length n over R.
c∈C
Then
Theorem III.3. Let C be a linear code of length n over R. Proof: By Theorem III.2 dan Theorem III.5, we obtain
Then
LeeC ⊥ (X, Y ) = SLW EC ⊥ (X 16 , X 15 Y, . . . , Y 16 )
HamC (X, Y ) = SLW EC (X, Y, Y, . . . , Y ). 1
| {z } = SLW EC (E0 , E1 , E2 , . . . , E16 )
16 |C|
IV. C YCLIC AND QUASI - CYCLIC C ODES there exist polynomials ft (x), gt (x), pt (x) ∈ Z4 [x] such that
Now, let us look at an important class of linear codes, Ct = hft (x) + 2pt (x), 2gt (x)i, then
namely cyclic codes. We mainly consider the structural
* 8 8 8
+
X X X
properties of cyclic codes over the ring R. C= ηt ft (x) + 2 ηt pt (x), 2 ηt gt (x) .
The notion of cyclic codes is standard for codes over all t=1 t=1 t=1
rings. A cyclic shift on Rn is a permutation T such that Furthermore, if n is odd, then
* 8 8
+
T (c0 , c1 , c2 , . . . , cn−1 ) = (cn−1 , c0 , c1 , . . . , cn−2 ). X X
C= ηt ft (x) + 2 ηt gt (x) .
A linear code C over R is called a cyclic code if C is t=1 t=1
invariant under the cyclic shift T, namely T (C) = C. We use * 8 8
the usual ideas of identifying vectors in Rn and polynomials
X X
Proof: Let D = ηt ft (x) + 2 ηt pt (x),
in the residue class ring R[x]/hxn − 1i as follows: + t=1 t=1
8
X
c = (c0 , c1 , c2 , . . . , cn−1 ) ←→ 2 ηt gt (x) . It is obvious that D ⊆ C. Let c(x) ∈ C.
t=1
c(x) = c0 + c1 x + c2 x2 + · · · + cn−1 xn−1 + hxn − 1i. Because C = ⊕8t=1 ηt Ct and Ct = hft (x) + 2pt (x), 2gt (x)i,
We can see that T (c) is identified by x · c(x) ∈ R[x]/hxn − then there exist ut (x), vt (x) ∈ Z2 [x] such that
1i. This implies that cyclic codes over R are identified by 8
X
ideals in the residue class ring R[x]/hxn −1i. So, we have to c(x) = ηt ((ft (x) + 2pt (x))ut (x) + 2gt (x)vt (x))
understand the structure of the residue class ring R[x]/hxn − t=1
1i in order to understand cyclic codes over the ring R. X8 8
X
The first theorem below is a straightforward generalization = ηt (ft (x) + 2pt (x))ut (x) + ηt 2gt (x)vt (x)
of Theorem 13 proven by Li, Guo, Zhu, and Kai [14]. t=1 t=1
X8 8
X
Theorem IV.1. Let C = C1 η1 ⊕ C2 η2 ⊕ · · · ⊕ C8 η8 . Then C = ηt ut (x) ηt (ft (x) + 2pt (x))
is a cyclic code over R if and only if one of following three t=1 t=1
conditions is satisfied: 8
X X8
(1) For t ∈ {1, 2, . . . , 8}, Ct is a cyclic code over Z4 . + ηt vt (x) 2ηt gt (x).
t=1 t=1
(2) For t ∈ {1, 2, . . . , 8}, Ct⊥ is a cyclic code over Z4 .
(3) C ⊥ is a cyclic code over R. So we have C ⊆ D and hence C = D.
8
By using Theorem IV.2 and the similar technique as in
Proof: Let c =
X
ηt ct ∈ C, and write ct = proof of Theorem IV.3, we obtain generator polynomials for
t=1
the dual of cyclic codes as given in the theorem below.
(ct,0 , ct,1 , . . . , ct,n−1 ) ∈ Ct , for 1 ≤ t ≤ 8. Since C is a
Theorem IV.4. Let C = hf (x) + 2p(x), 2g(x)i be a cyclic
cyclic code, we also have
code over Z4 . Then
8 8 8
!
X X X * 8 8
ηt ct,n−1 , ηt ct,0 , . . . , ηt ct,n−2 ∈ C. X X
C⊥ = ηt gbt (x)∗ + 2 ηt xdeg(bgt (x))−deg(ut (x)) ut (x)∗ ,
t=1 t=1 t=1
t=1 t=1
So, (ct,n−1 , ct,0 , . . . , ct,n−2 ) ∈ Ct , for 1 ≤ t ≤ 8, and hence, 8
X
+
Ct is cyclic for 1 ≤ t ≤ 8. The reverse also holds, so the 2 fbt (x)∗ .
first condition is proven. t=1
If Ct is cyclic over Z4 , then Ct⊥ is also cyclic ( [21], Now, let us turn to the special class of cyclic codes called
Proposition 7.9). From condition (1), C ⊥ is a cyclic code a quasi-cyclic codes.
over R, so C is a cyclic code over R. Let σ be a cyclic shift operator over Zn4 . For any positive
We start to observe the generator polynomials of cyclic integer s, let σs be the quasi-shift defined by
code and its dual over R. For that purpose, we need the
following theorem proven by Li, Guo, Zhu, and Kai [14]. σs a(1) | a(2) | · · · | a(s)
Theorem IV.2 ( [14], Theorem 15). Let C = hf (x) +
= σ a(1) | σ a(2) | · · · | σ a(s) ,
2p(x), 2g(x)i be a cyclic code over Z4 . Then
with a(1) , a(2) , . . . , a(s) ∈ Zn4 and ”|” is a vector concate-
C ⊥ = hb
g (x)∗ + 2xdeg(bg(x))−deg(u(x)) u(x)∗ , 2fb(x)∗ i
n n nation. A quaternary quasi-cyclic code C of index s and
length ns is a subset of (Zn4 )s such that σs (C) = C. If
x −1 x −1
with fb(x) := , gb(x) := , and 8
f(x) g(x) X
1 R = ⊕8t=1 ηt Rt , we can write any r ∈ R as r = ηt rt
f (x)∗ := xdeg(f (x)) f . t=1
x with rt ∈ Rt , for 1 ≤ t ≤ 8. We define the mapping
The following two theorems provide generator polynomi- 3 n
als of cyclic code and its dual over R. Φ: Rn −→ Z24
×n−1
i=0 ri 7−→ ×8t=1 ×n−1
i=0 rt,i
Theorem IV.3. Let C = C1 η1 ⊕ C2 η2 ⊕ · · · ⊕ C8 η8 be a
cyclic code of length n over R. If for every t ∈ {1, 2, . . . , 8}, with ri = ×8t=1 rt,i for i = 0, 1, . . . , n − 1 and rt,i ∈ Rt .
Then we have a similar theorem of Theorem 17 in [14]. Then we have C = ⊕8t=1 Ct ηt is a cyclic code over R.
Parameters of Φ(C) is [56, 432 224 , 2].
Theorem IV.5. Let C = C1 η1 ⊕ C2 η2 ⊕ · · · ⊕ C8 η8 be a
If we choose another set of Ci with
cyclic code of length n over R. Then Φ(C) is a quasi-cyclic
code of index 8 and length 8n over Z4 . Ci = h(x3 + x + 1)(x3 + x2 + 1) + 2i, i = 1, 2, . . . , 8,
Proof: Let ×n−1 8
i=0 ci ∈ C. Let ci = ×t=1 ct,i for i = then we have C = ⊕8t=1 Ct ηt is also a cyclic code over R.
0, 1, . . . , n − 1 and ct,i ∈ Ct . Since C is a cyclic code, we
Parameters of Φ(C) is [56, 48 248 , 6]. Let us choose
have Ct is cyclic for 1 ≤ t ≤ 8. This means that for every
t ∈ {1, 2, . . . , 8}, we have σ(×n−1 n−1
i=0 rt,i ) ∈ Ct , if ×i=0 rt,i ∈ C1 = C2 = C3 = h(x + 1)(x3 + x + 1)(x3 + x2 + 1)
n−1 8 n−1
Ct . Write Φ(×i=0 ci ) = ×t=1 ×i=0 rt,i . Then
+ 2(x3 + x + 1)i,
σ8 (×8t=1 ×n−1
i=0 rt,i ) = ×8t=1 σ(×n−1
i=0 rt,i ) ∈ Φ(C). C4 = C5 = C6 = C7 = C8 = h(x3 + x2 + 1) + 2i.
So we have Φ(C) is a quasi-cyclic code C of index 8 and
We have C = ⊕8t=1 Ct ηt is also a cyclic code over R.
length 8n over Z4 .
Parameters of Φ(C) is [56, 423 227 , 2]. J
Furthermore, by using the Theorem 18 of [14] below, we
obtain directly the type of Φ(C) as given in Corollary IV.7. Example IV.11. Let n = 9. In Z4 [x], x9 − 1 = (x + 1)(x2 +
x + 1)(x6 + x3 + 1). Choose
Theorem IV.6 ( [14], Theorem 18). Let Ct , t ∈ {1, 2, . . . , 8}
be a cyclic code of length n (n is odd) over Z4 . Write Ci = h(x2 + x + 1)(x6 + x3 + 1) + 2(x6 + x3 + 1)i, i =
Ct = hf1,t (x)+2f2,t (x)i with f1,t (x) and f2,t (x) are monic 1, 2, . . . , 8,
factors of xn − 1 over Z4 and f2,t (x) | f1,t (x). Then the
cardinality of Ct , for 1 ≤ t ≤ 8, is then C = ⊕8t=1 Ct ηt is a cyclic code over R. Parameters of
Φ(C) is [72, 48 216 , 6].
4n−deg(f1,t (x)) 2deg(f1,t (x)−deg(f2,t (x)) . If we choose
The corollary below follows directly. Ci = h(x2 +x+1)(x6 +x3 +1)+2(x2 +x+1)i, i = 1, 2, . . . , 8,
8
We have C = ⊕8t=1 Ct ηt is a cyclic code over R. Parameters
Y
Corollary IV.7. Let Φ(C) = Ct be a linear code of
t=1 of Φ(C) is [72, 48 232 , 3]. J
length 8n (n is odd) over Z4 and Ct is a cyclic code over
Z4 for every t ∈ {1, 2, . . . , 8}. Then the cardinality of Φ(C) Example IV.12. Let n = 15. In Z4 [x], x15 − 1 = (x +
is 1)(x2 + x + 1)(x4 + x + 1)(x4 + x3 + 1)(x4 + x3 + x2 + 1).
P8 P8 Choose
4 t=1 (n−deg(f1,t (x))) 2 t=1 (deg(f1,t (x)−deg(f2,t (x)) .
8
Y Ci = h(x4 + x + 1)(x4 + x3 + 1)(x4 + x3 + x2 + 1)
Now, consider a linear code Φ(C) = Ct of length 8n + 2(x4 + x + 1)(x4 + x3 + 1)i, for 1 ≤ i ≤ 4,
t=1
(n is odd) over Z4 and let dL be the Lee distance of Φ(C). Ci = h(x4 + x + 1)(x4 + x3 + 1)(x4 + x3 + x2 + 1)
Let min1≤t≤8 dL (Ct ) = dL (Cj ), for some j, and let c ∈ Cj + 2(x4 + x3 + 1)(x4 + x3 + x2 + 1)i, for 5 ≤ i ≤ 8.
such that wL (c) = dL (Cj ). Then
We have C = ⊕8t=1 Ct ηt is a cyclic code over R. Parameters
−1
dL (Φ (0, . . . , 0, c, 0, . . . , 0)) = dL (Cj ), of Φ(C) is [120, 424 232 , 10].
and hence dL = min1≤t≤8 dL (Ct ). If we choose