On the Covering Radius of Codes over Z p k

: In this correspondence, we investigate the covering radius of various types of repetition codes over Z p k ( k ≥ 2 ) with respect to the Lee distance. We determine the exact covering radius of the various repetition codes, which have been constructed using the zero divisors and units in Z p k . We also derive the lower and upper bounds on the covering radius of block repetition codes over Z p k .


Introduction
Codes over finite fields have been studied since the inception of coding theory. Due to the rich algebraic structure of rings, the codes over rings gained popularity during the seventies [1][2][3]. In 1994, Hammons et al. [4] obtained the well known non-linear codes as a Gray image of the codes over Z 4 . After that, working on the codes over rings gained greater attention. What started with the ring Z 4 , was later generalized to the rings Z 2 s , Z 2 + uZ 2 , Z 4 + uZ 4 , F p + uF p etc [5][6][7][8]. Covering Radius is a widely discussed parameter for the codes with respect to the Hamming weight [9]. A lot of other weights such as Lee weight [4], Homogenous weight [10] and Euclidean weight have been introduced and used in the literature for convenience.
The Covering Radius for the codes with respect to the Lee distance was first investigated for the ring Z 4 by Aoki [11]. Later, working on the Covering Radius of codes the with respect to the Lee distance gained interest [6,12,13]. We are particulary interested to find the Covering Radius for Repetition Codes, Since the Covering Radius of the Repetition Codes simplifies the process of finding the Covering Radius for many existing codes. For eg., it helps to find the Covering Radius of the well known Simplex and Macdonald Codes, as the generator matrix of Simplex and Macdonald Codes has lot of similarities with the generator matrix of the Repetition Codes. For the Quaternary case, it was discussed in [6].
This motivated us to work on the Covering Radius of Repetition Codes over the ring Z p k . The problem of generalising the results for Z p k starts with defining a proper Lee weight for Z p k and then the extended Gray map defined here is not surjective. Also the zero divisors of different orders are obtained here, which will not be in the case of Z 4 .
In this correspondence, we have investigated the covering radius of the codes over Z p k (k ≥ 2) with respect to the Lee distance in relation to the codes obtained by the Gray map. In Section 2, we have given some basic preliminaries. We have given several upper and lower bounds on covering radius, including Z p k analogue of sphere covering bound, packing bound and Delsarte bound in Section 3. In the next Section, the covering radii of some repetition codes have been discussed, namely repetition codes using the zero divisors in Z p k of different orders and the repetition codes from the units in Z p k . We have ended the section with the upper and lower bounds on the covering radius of the block repetition codes BR (p k −1)n p k and BR (p k −2)n p k . Here we have determined the exact value of ). Finally, we have concluded the paper with the future work that can be proceeded with.

Preliminaries
A linear code C of length n is an additive subgroup of Z n p k . If C is not an additive subgroup of Z n p k , then C is simply called a code of length n. Thus, every linear code C is a Z p k submodule of Z n p k . An element in C is known as a codeword of C. A matrix G is said to be a generator matrix of C if C is the row span of G over Z p k . Two codes are said to be permutation equivalent if one is obtained from another by permuting the coordinates.
The Hamming weight w H (x) of a vector x ∈ Z n p k is the number of non-zero coordinates in x. The Lee weight of x ∈ Z p k in the sense of [14] is given by Note that this weight coincides with the classical Lee weight when p = 2, and is different when p > 2. The Lee weight of x ∈ Z n p k is the sum of the Lee weight of its coordinates. The Hamming (Lee) distance d H (x, y) (d L (x, y)) between two vectors x and y is w H (x − y) (w L (x − y)). The minimum Hamming (Lee) distance is the smallest Hamming (Lee) weight among all non-zero codewords of C. A code of length n, size M, minimum Hamming distance d H , minimum Lee distance d L over Z p k is a (n, M, d H , d L ) code.
The dual code C ⊥ of C is defined as C ⊥ = {x ∈ Z n p k | x 1 · y 1 + x 2 · y 2 + · · · + x n · y n = 0 (mod p k ) for all y ∈ C}. As various distances are possible for the codes over Z p k , we have given a definition of the covering radius for a general distance. Let d be the general distance out of the various possible distances such as Hamming and Lee. The Covering radius r d (C) of a code C over Z p k with respect to the general distance is given by To alleviate notation, we have written r L (resp. r H ) for r d L (resp. r d H ).
In [14], a distance preserving Gray map φ L : by the coordinate wise extention of the Gray map. Let C be a code of length n with M codewords and minimum Lee distance d over Z p k . Then by the above Gray map, the image φ(C) is a code of length np k−1 with M codewords and minimum Hamming distance d. We have summed up the idea below without proof. Proposition 1. [11] If C is a linear code over Z p k of length n, size M and minimum Lee distance d, then the Gray image φ(C) is a code over Z p of length np k−1 , size M and minimum Hamming distance d and also r L (C) ≤ r H (φ(C)).
Note that since the Gray map is injective but not surjective in general, the covering radius of a code C for the Lee metric is at most that of φ(C) for the Hamming metric, but could be different.

Covering Radius of Codes
We have discussed several bounds on covering radius of codes in this section, including the Z p k analogue of the packing bound, the sphere covering bound and the Delsarte bound. The following bound is called the packing bound, which is similar to the bound given for Z 4 in [11].
The proof of the following Proposition 2 and 3, is similar to but distinct from the case of Z 4 [11]. Note that the covering radius of a code C for the Lee metric is at most that of φ(C) for the Hamming metric, but could be different.

Proposition 2.
For any code C of length n over Z p k .
This bound is known as the Sphere Covering Bound.
Proof. Because the map φ is an isometry the image of a ball of radius r centered in x is a ball of radius r centered in φ(x) of the same cardinality. The result follows by the usual sphere covering argument.
Let C be a code over Z p k and let s( denotes the number of codewords of the Lee weight i in C ⊥ . Then we have the Delsarte bound Proposition 3. For any C ⊆ Z n p k , r L (C) ≤ s(C ⊥ ).
Proof. As noted above the covering radius r L (C) of a code C for the Lee metric is at most that of φ(C) for the Hamming metric. Denote by r H (φ(C)) that latter quantity. Since φ is an isometry, it can be shown by using the duality of complete weight enumerators and specialization of variables, so that the Hamming weight enumerators of φ(C) and φ(C ⊥ ) are MacWilliams duals of each other, a fact already noted in [4] for k = 2. This implies that the number of Gray weights of C ⊥ , that is the number of weights of φ(C ⊥ ) equals the number of Hamming weights of φ(C) ⊥ . By the Delsarte bound in the Hamming metric we conclude that r H (φ(C)) ≤ s(C ⊥ ). The result follows upon writing r L (C) ≤ r H (φ(C)).
If C is a code of length n over a group (G, +), then the covering radius of the code C is defined by, r(C) = max x∈G n d(x, C) = max x∈G n {min c∈C d(x, c)} = max x∈G n {min c∈C wt(x − c)}. Hence the following result , which is a generalization of codes over finite rings from finite field by G.D Cohen et al. [9]. Theorem 2. Let C be the Cartesian Product of two Codes C 1 and C 2 , then the covering radius of C is r(C) = r(C 1 ) + r(C 2 ) with respect to all distances.
Proof. The proof is the same as the proof of the Theorem 5, which is the more general.
For each i, 1 ≤ i ≤ k − 1 the number of zero divisors of order p i is p i−1 (p − 1) which are given by C p i is a code generated by the generator matrix G p i = n p k−i p k−i · · · p k−i . Hence C p i is an (n, p i , n, np k−i ) code. As we sum up the above ideas we get Proof. Let x ∈ Z n p k and let w i be the number of i coordinates in x for 0 ≤ i ≤ p k − 1. Then Consider, d L (x, 0 0 0) = (w 1 + w p k −1 ) + 2(w 2 + w p k −2 ) + · · · + (p k−1 − 1) We know that the minimum is always less than the average. So we get, It shows that, r L (C p i ) ≤ np k−2 (p − 1). For the reverse inequality, let Hence, r L (C p i ) = np k−2 (p − 1).

Unit Repetition Codes
Let u be a unit in Z p k . Then code C u generated by the matrix G u = [ n uu · · · u] is called unit repetition code. Clearly C u is a (n, p k , n, n) code over Z p k . Theorem 6. Let u be a unit in Z p k , then r L (C u ) = np k−2 (p − 1) Proof. Let x ∈ Z n p k and let w i be the number of i coordinates in x for 0 ≤ i ≤ p k − 1. Consider, d L (x, 0 0 0) = w 0 (0) + (w 1 + w p k −1 ) + · · · + (p k−1 − 1)(w p k −p k−1 +1 + w p k−1 −1 ) We know that the minimum is always less than the average. So we get, Thus, r L (C u ) ≤ np k−2 (p − 1).
Proof. The proof is the same as the proof of the Theorem 7.

Conclusions
We have discussed some well known bounds such as the sphere covering bound, the Delsarte bound and the packing bound with respect to the Lee distance for the codes over Z p k . We have determined the exact value of the covering radius of the zero divisor (unit) repetition codes. We have obtained the lower and upper bounds on the covering radius of the block repetition codes over Z p k . The results obtained in this article is definitely helpful, if we are able to obtain the similarities between the generator matrix of existing codes over Z p k with the generator matrix of the repetition codes over Z p k , then we will able to apply all the existing results on the covering radius. And also it would be an interesting task to discuss the covering radius for the more generalized ring Z n . We can also obtain the weight enumeration of these codes in Lee distance and compare it with the Hamming distance.