Contents

A Note on Decomposition of Tensor Product of Complete Multipartite Graphs into Gregarious Kite

A. Tamil Elakkiya1
1Gobi Arts & Science College, Gobichettipalayam, Erode, Tamil Nadu, India

Abstract

A kite \(K\) is a graph which can be obtained by joining an edge to any vertex of \(K_3\). A kite with edge set \(\{ab,\,bc,\,ca,\,cd\}\) can be denoted as \((a,\,b,\,c;\,cd)\). If every vertex of a kite in the decomposition lies in different partite sets, then we say that a kite decomposition of a multipartite graph is a gregarious kite decomposition. In this manuscript, it is shown that there exists a decomposition of \((K_m \otimes \overline{K}_n) \times (K_r \otimes \overline{K}_s) \) into gregarious kites if and only if
\[
n^2 s^2 m(m-1)r(r-1) \equiv 0 \pmod{8},
\]
where \(\otimes\) and \(\times\) denote the wreath product and tensor product of graphs respectively. We denote a gregarious kite decomposition as \(\it GK\)-decomposition.

Keywords: Decomposition, Kite, Tensor product, Wreath Product

1. Introduction

Many authors have been studied on kite-design and kite decomposition/ kite factorization and their special properties such as gregarious kite decomposition/ gregarious kite factorization. Bermond and Schonheim [1] proved that a kite-design of order \(n\) exists if and only if \(n\equiv 0,\, 1\pmod 8\). Wang and Chang [2,3] considered the problem of a resolvable \((K_{3}+e)\) and \((K_3 + e,\, \lambda)\)-group divisible designs of type \(g^t u^1.\) Wang [4] have proven that the necessary conditions to find a resolvable \((K_{3}+e)\)-group divisible design of type \(g^u\) are also sufficient. Fu et al. [5] have demonstrated that there exists a \(\it GK\)-decomposition of \(K_m(n)\) if and only if \(n \equiv 0,\,1 \pmod 8\) for odd \(m\) or \(n\geq 4\) for even \(m.\) Gionfriddo and Milici [6] considered the uniformly resolvable decompositions of \(K_v\) and \(K_v – I\) into paths and kites. To know more about kite designs, refer [7-11]. Further, the authors A. Tamil Elakkiya and A. Muthusamy [12] have shown that there is a \(\it GK\)-decomposition of \(K_m \times K_n\) if and only if \(mn (m-1) (n-1)\equiv 0 \pmod 8.\) Moreover, the authors A. Tamil Elakkiya and A. Muthusamy [13] considered the existence of a \(\it GK\) factorization of tensor product of complete graphs.

In this way, our main concern is to find a existence of a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\). In this document, it is demonstrated that a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) holds, if and only if, \(n^2 s^2 m(m-1)r(r-1)\equiv 0 \pmod 8.\)

2. Preliminary Results

Definition 1. [12] A partition of a graph \(G\) into a collection of subgraphs \(G_1,\,G_2,\,\ldots,\,G_r\) such that each one of it is distinct mutually by their edges together with \(E(G)= \cup_{i=1}^{r} E(G_i)\) is called a decomposition of \(G;\) We then write \(G\) as \(G=G_1\,\oplus\,G_2\,\oplus\,\ldots \oplus\, G_r,\) where \(\oplus\) denotes edge-disjoint sum of subgraphs. For an integer \(s,\) \(s G\) denotes \(s\) copies of \(G.\)

Definition 2. [12] A kite \(K\) is a graph which can be obtained by joining an edge to any vertex of \(K_3\), for instance see Figure 1. A kite with edge set \(\{ab,\,bc,\,ca,\,cd\}\) can be denoted as \((a,\,b,\,c;\,cd)\).

Figure 1

If every vertex of a kite in the decomposition lies in different partite sets, then we say that a kite decomposition of a multipartite graph is a gregarious kite decomposition. Here we may consider a Gregarious Kite decomposition as a \(\it GK\)-decomposition for short.

Definition 3. [12] The tensor product \(G \times H\) and the wreath product \(G \otimes H\) of two graphs \(G\) and \(H\) are defined as follows: \(V(G \times H)\)=\(V(G \otimes H)\) = \(\{(u,\,v) \mid u \in V(G),\, v\in V(H)\}.\) \(E(G\times H)\) = \(\{(u,\,v)\,(x,\,y)\mid u\,x\in E(G)\) and \(v\,y\in E(H)\}\) and \(E(G \otimes H)\) = \(\{(u,\,v)\,(x,\,y)\mid u = x\) and \(v\,y\in E(H)\) or \(u\,x\in E(G)\}.\)

As tensor product has commutative and distributive property over an edge-disjoint sum of subgraphs, if decomposition of \(G\) = \(G_1\,\oplus\,G_2\,\oplus\,\ldots \oplus\, G_r\) holds, we then write as \(G \times H\) = \((G_1\times H)\oplus (G_2\times H)\oplus \ldots \oplus (G_r\times H).\)

Definition 4. [12] A graph \(G\) having partite sets \(V_1,\,V_2,\,\ldots, V_m\) with \(\mid V_i\mid = n,\) \(1\leq i \leq n\) and \(E(G)\) ={\(uv\) \(\mid\) \(u\in V_i\) and \(v \in V_j,\,\forall\, i \neq j\)} is called complete \(m\)-partite graph and is denoted by \(K_m(n).\) Note that \(K_m(n)\) is same as the \(K_m \otimes \overline K_n.\) where \(\overline K_n\) is the complement of a complete graph on \(n\) vertices.

In order to prove our key result, we require the following:

Theorem 1. [5] There exists a \(\it GK\)-decomposition of \(K_m(n)\) if and only if \(m\equiv 0,\,1 \pmod 8,\) for all odd \(n\) (or) \(m \geq 4,\) for even \(n.\)

Remark 1. [5] Let \(K\) is a kite. Then there exists a \(\it GK\)-decomposition of \(K\otimes \overline K_s,\) for all \(s.\)

Lemma 1. [12] For all \(n,\) \(n \neq 2\), there exists a \(\it GK\)-decomposition of \(K \times K_{n},\) where \(K\) is a kite.

Theorem 2. [12] There exists a \(\it GK\)-decomposition of \(K_m \times K_{n}\) if and only if \(mn (m-1) (n-1)\equiv 0 \pmod 8.\)

3. Existence of \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\)

Lemma 2. A \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) exists, if \(m \equiv 0,\,1 \pmod 8,\) \(n,\,s \equiv 1 \pmod 2\) and for all \(r,\) \(r \neq 2.\)

Proof. By Theorem 1, we have a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n).\) Then \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) = \((K \oplus\, K \oplus\,\ldots\,\oplus K)\times (K_r \otimes \overline K_s)\) = \(K \times (K_r \otimes \overline K_s) \oplus\, K \times (K_r \otimes \overline K_s)\oplus\, \ldots \oplus\, K\times (K_r \otimes \overline K_s).\) Now \(K\times (K_r \otimes \overline K_s) \cong (K\times K_r) \otimes \overline K_s.\) As we have a kite decomposition of \(K\times K_r\) from Lemma 1, we then write as \((K\times K_r) \otimes \overline K_s\) = \((K \oplus\, K \oplus\,\ldots\,\oplus\, K) \otimes \overline K_s\) = \((K\otimes\overline K_s) \oplus\, (K\otimes\overline K_s) \oplus\, \ldots \oplus\, (K\otimes\overline K_s).\) A \(\it GK\)-decomposition of \(K\otimes\overline K_s\) can be obtained from Remark 1. Thus all together provides a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s).\) ◻

Lemma 3. A \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) exists, if \(n \equiv 0 \pmod 2\) and for all \(m,\,r,\,s,\) \(\{m,\,r \}\neq 2.\)

Proof. Case 1: Let \(m = 3,\) \(n = 2b,\,b\geq 0\) and for all \(r,\,s,\, (r\neq 2).\) We write \((K_3 \otimes \overline K_n) \times (K_r \otimes \overline K_s) \cong [(K_3 \otimes \overline K_n) \times K_r] \otimes \overline K_s\) = \([(K_3\oplus\, K_3\oplus\, \ldots\oplus\, K_3)\times K_r] \otimes \overline K_s\) = \([(K_3\times K_r)\oplus\,(K_3\times K_r)\oplus\,\ldots \oplus\,(K_3\times K_r)] \otimes \overline K_s\) = \([(K_3\times K_r)\otimes \overline K_s]\oplus\, [(K_3\times K_r)\otimes \overline K_s]\oplus\, \ldots \oplus\, [(K_3\times K_r)\otimes \overline K_s] .\) Theorem 2 gives a \(\it GK\)-decomposition for \(K_3 \times K_r\). Now, we write as \((K_3\times K_r)\otimes \overline K_s\)= \((K \oplus\, K \oplus\,\ldots\,\oplus K) \otimes \overline K_s\) = \((K\otimes\overline K_s) \oplus\, (K\otimes\overline K_s) \oplus \,\ldots \oplus\, (K\otimes\overline K_s).\) A \(\it GK\)-decomposition exists for \(K\otimes\overline K_s\) according to Remark 1. As a consequence of it, a \(\it GK\)-decomposition of \((K_3 \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) has been derived.

Case 2: A \(\it GK\)-decomposition for \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) have been described as follows:

Let \(m \geq 4,\) \(n = 2b,\,b\geq 0\) and for all \(r,\,s,\, (r\neq 2).\) One can obtain a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n)\) from Theorem 1. Then \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) = \((K \oplus K \oplus\,\ldots\,\oplus K)\times (K_r \otimes \overline K_s)\) = \(K\times (K_r \otimes \overline K_s) \oplus K\times (K_r \otimes \overline K_s)\oplus \ldots \oplus K\times (K_r \otimes \overline K_s).\) Now \(K\times (K_r \otimes \overline K_s) \cong (K\times K_r) \otimes \overline K_s.\) A kite decomposition exits for \(K\times K_r\) by using Lemma 1. Moreover, we then write, \((K\times K_r) \otimes \overline K_s\) = \((K \oplus K \oplus\,\ldots\,\oplus K) \otimes \overline K_s\) = \((K\otimes\overline K_s) \oplus (K\otimes\overline K_s) \oplus \ldots \oplus (K\otimes\overline K_s).\) According to Remark 1, a \(\it GK\)-decomposition of \(K\otimes\overline K_s\) has been proved.

Consequently, Cases \(1\) and \(2\) lead a \(\it GK\)-decomposition for \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) when \(n \equiv 0 \pmod 2\) and for all \(m,\,r,\,s,\) \(\{m,\,r \}\neq 2.\) ◻

Lemma 4. A \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) exists, if \(n \equiv 1 \pmod 2,\) \(s \equiv 0 \pmod 2,\) and for all \(m,r,\) \(\{m,\,r\} \neq 2.\)

Proof. Case 1: Let \(m = 3,\) \(n = 2b+1,\, b\geq 0,\) \(s= 2c,\, c\geq 0\) and for all \(r,\, (r\neq 2).\) Let us consider \((K_3 \otimes \overline K_n) \times (K_r \otimes \overline K_s) \cong [(K_3 \otimes \overline K_n) \times K_r] \otimes \overline K_s\) = \([(K_3\oplus\, K_3\oplus\, \ldots\oplus\, K_3)\times K_r] \otimes \overline K_s\) = \([(K_3\times K_r)\oplus\,(K_3\times K_r)\oplus\,\ldots \oplus\,(K_3\times K_r)] \otimes \overline K_s\) = \([(K_3\times K_r)\otimes \overline K_s]\oplus\, [(K_3\times K_r)\otimes \overline K_s]\oplus\, \ldots \oplus\, [(K_3\times K_r)\otimes \overline K_s] .\) By Theorem 2, a \(\it GK\)-decomposition exists for \(K_3 \times K_r\). Now \((K_3\times K_r)\otimes \overline K_s\) = \((K \oplus\, K \oplus\,\ldots\,\oplus\, K) \otimes \overline K_s\) = \((K\otimes\overline K_s) \oplus\, (K\otimes\overline K_s) \oplus\, \ldots \oplus\, (K\otimes\overline K_s).\) Now, a \(\it GK\)-decomposition of \(K\otimes\overline K_s\) follows from Remark 1. Thus, the above construction provides a \(\it GK\)-decomposition of \((K_3 \otimes \overline K_n) \times (K_r \otimes \overline K_s)\).

Case 2: Let \(m \geq 4,\) \(n = 2b+1,\, b\geq 0,\) \(s= 2c,\, c\geq 0\) and for all \(r,\, (r\neq 2).\) By Theorem 1, there is a \(\it GK\)-decomposition for \((K_r \otimes \overline K_s)\). Then \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) = \((K_m \otimes \overline K_n)\times (K \oplus K \oplus\,\ldots\,\oplus K)\) = \([(K_m \otimes \overline K_n)\times K] \oplus [(K_m \otimes \overline K_n)\times K]\oplus \ldots \oplus [(K_m \otimes \overline K_n)\times K].\)

Now \([(K_m \otimes \overline K_n)\times K] \cong K\times (K_m \otimes \overline K_n) \cong (K\times K_m) \otimes \overline K_n.\) By Lemma 1, we have a kite decomposition of \(K\times K_m.\) Then \((K\times K_m) \otimes \overline K_n\) = \((K \oplus K \oplus\,\ldots\,\oplus K) \otimes \overline K_n\) = \((K\otimes\overline K_n) \oplus (K\otimes\overline K_n) \oplus \ldots \oplus (K\otimes\overline K_n).\) A \(\it GK\)-decomposition of \(K\otimes \overline K_n\) follows from Remark 1. Thus, all together gives a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s).\) ◻

Lemma 5. A \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) exists, if \(m \equiv 4,\,5 \pmod 8,\) \(n,\,s \equiv 1 \pmod 2\) and for all \(r,\) \(r\neq 2.\)

Proof. Let us consider \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s) \cong [(K_m \otimes \overline K_n) \times K_r] \otimes \overline K_s.\) We write as \((K_m \otimes \overline K_n) \times K_r \cong K_r\times(K_m \otimes \overline K_n) \cong (K_r \times K_m)\otimes \overline K_n.\) By Theorem 2, we have a \(\it GK\)-decomposition of \(K_r \times K_m.\) Then \((K_r \times K_m)\otimes \overline K_n\) = \((K \oplus\, K \oplus\,\ldots\,\oplus\, K) \otimes \overline K_n\) = \((K\otimes\overline K_n) \oplus\, (K\otimes\overline K_n) \oplus\, \ldots \oplus\, (K\otimes\overline K_n).\) By Remark 1, we can obtain a \(\it GK\)-decomposition of \(K\otimes \overline K_n.\) Further, \([(K_m \otimes \overline K_n) \times K_r] \otimes \overline K_s\) = \((K \oplus \, K \oplus\,\ldots\,\oplus \, K) \otimes \overline K_s\) = \((K\otimes\overline K_s) \oplus \, (K\otimes\overline K_s) \oplus\, \ldots \oplus\, (K\otimes\overline K_s).\) A \(\it GK\)-decomposition of \(K\otimes \overline K_s\) follows from Remark 1. Thus, all together gives a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s).\) ◻

Lemma 6. A \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) exists, if \(m \equiv 2,\,3 \pmod 4,\) \(m\neq 2,\) \(n,\,s \equiv 1 \pmod 2\) and \(r \equiv 0,\,1 \pmod 8.\)

Proof. A \(\it GK\)-decomposition of \(K_r \otimes \overline K_s\) can be obtained from Theorem 1. We then write, \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) = \((K_m \otimes \overline K_n)\times (K \oplus K \oplus\,\ldots\,\oplus K)\) = \([(K_m \otimes \overline K_n)\times K] \oplus [(K_m \otimes \overline K_n)\times K]\oplus \ldots \oplus [(K_m \otimes \overline K_n)\times K].\) Now, we write as \([(K_m \otimes \overline K_n)\times K] \cong K\times (K_m \otimes \overline K_n) \cong (K\times K_m) \otimes \overline K_n.\) A kite decomposition of \(K\times K_m\) has been derived from Lemma 1. Then \((K\times K_m) \otimes \overline K_n\) = \((K \oplus K \oplus\,\ldots\,\oplus K) \otimes \overline K_n\) = \((K\otimes\overline K_n) \oplus (K\otimes\overline K_n) \oplus \ldots \oplus (K\otimes\overline K_n).\) A \(\it GK\)-decomposition of \(K\otimes \overline K_n\) exists according to Remark 1. Consequently, a \(\it GK\)-decomposition exists for \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\). ◻

Lemma 7. A \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) exists, if \(m \equiv 2,\,3 \pmod 4,\) \(m\neq 2,\) \(n,\,s \equiv 1 \pmod 2\) and \(r \equiv 4,\,5 \pmod 8.\)

Proof. We write \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s) \cong [(K_m \otimes \overline K_n) \times K_r] \otimes \overline K_s.\) Now \((K_m \otimes \overline K_n) \times K_r \cong K_r\times(K_m \otimes \overline K_n) \cong (K_r \times K_m)\otimes \overline K_n.\) By Theorem 2, we have a \(\it GK\)-decomposition of \(K_r \times K_m.\) Then \((K_r \times K_m)\otimes \overline K_n\) = \((K \oplus K \oplus\,\ldots\,\oplus K) \otimes \overline K_n\) = \((K\otimes\overline K_n) \oplus (K\otimes\overline K_n) \oplus \ldots \oplus (K\otimes\overline K_n).\) A \(\it GK\)-decomposition of \(K\otimes \overline K_n\) can be obtained from Remark 1. Further, \([(K_m \otimes \overline K_n) \times K_r] \otimes \overline K_s\) = \((K \oplus K \oplus\,\ldots\,\oplus K) \otimes \overline K_s\) = \((K\otimes\overline K_s) \oplus (K\otimes\overline K_s) \oplus \ldots \oplus (K\otimes\overline K_s)\). A \(\it GK\)-decomposition of \(K\otimes \overline K_s\) follows from Remark 1. Thus the above describes a \(\it GK\)-decomposition for \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s).\) ◻

4. Main Result

Theorem 3. There exists a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) if and only if \(n^2 s^2 m(m-1) r(r-1)\equiv 0 \pmod 8.\)

Proof. Necessity: The number of edges of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) is \(\frac{n^2 s^2 m(m-1) r(r-1)}{2}\) and number of edges of a kite \(K\) is \(4\). Therefore, the edge divisibility condition for a graph \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s)\) is \(\frac{n^2 s^2 m(m-1) r(r-1)}{8}\).

Sufficiency: It has been derived from Lemmas 2 – 7. ◻

5. Conclusion

In Section 3, we give a complete solution for the existence of a \(\it GK\)-decomposition of \((K_m \otimes \overline K_n) \times (K_r \otimes \overline K_s).\) In future, one can find the existence of a \(\it GK\)-factorization of Tensor product complete multipartite graph.

Funding Information

No funds, grants, or other support were received during preparation of this manuscript.

Conflict of Interest

The author have no conflict of interest.

References:

  1. Bermond, J.C. and Schönheim, J., 1977. G-decomposition of Kn, where G has four vertices or less. Discrete Mathematics, 19(2), pp.113-120.
  2. Wang, H. and Chang, Y., 2006. Kite-group Divisible Designs of Type g t u 1. Graphs & Combinatorics, 22(4), 545-571.
  3. Wang, H. and Chang, Y., 2008. (K-3+ e, lambda)-group divisible designs of type g (t) u (1). Ars Combinatoria, 89, pp.63-88.
  4. Wang, L., 2010. On the Existence of Resolvable (K+ e)-Group Divisible Designs. Graphs & Combinatorics, 26(6), pp.879-889.
  5. Fu, C.M., Hsu, Y.F., Lo, S.W. and Huang, W.C., 2013. Some gregarious kite decompositions of complete equipartite graphs. Discrete Mathematics, 313(5), pp.726-732.
  6. Gionfriddo, M. and Milici, S., 2013. On the existence of uniformly resolvable decompositions of Kv and Kv- I into paths and kites. Discrete Mathematics, 313(23), pp.2830-2834.
  7. Colbourn, C.J., Ling, A. C. and Quattrocchi, G., 2005. Embedding path designs into kite systems. Discrete Mathematics, 297(1-3), pp.38-48.

  8. Gionfriddo, L. and Lindner, C. C., 2005. Nesting kite and \(4\)-cycle systems. The Australasian Journal of Combinatorics, 33, pp.247-254.
  9. Kucukcifci, S. and Lindner, C.C., 2002. The Metamorphosis of Lambda-fold Block Designs with Block Size Four into Lambda-fold Kite Systems. Journal of Combinatorial Mathematics and Combinatorial Computing, 40, pp.241-252.
  10. Faro, G.L. and Tripodi, A., 2006. The Doyen–Wilson theorem for kite systems. Discrete mathematics, 306(21), pp.2695-2701.
  11. G. Ragusa, G., 2010. Complete simultaneous metamorphosis of \(\lambda\)-fold kite systems. Journal of Combinatorial Mathematics and Combinatorial Computing, 73, pp.159-180.
  12. Elakkiya, A.T. and Muthusamy, A., 2016. Gregarious kite decomposition of tensor product of complete graphs. Electronic Notes in Discrete Mathematics, 53, pp.83-96.
  13. Elakkiya, A. T. and Muthusamy, A., 2020. \(GK\)-factorization of tensor product of complete graphs. Discussiones Mathematicae Graph Theory, 40, pp.7-24.