arXiv:1202.1870v1 [math.CO] 9 Feb 2012 ie,isegsme nya hi omned) lnrgra planar A ends). common their at a only called meet edges its (i.e., of G Let E deset edge set Let Introduction 1 Classifications Subject Mathematics Keywords ecbs nti ae,b prtoso rps h thickn the graphs, on operations by K paper, comple on this graphs, In The complete percubes. few. are are obtained graphs been of have invari imp thickness thicknesses has topological about also other results and with number, 1963 comparing ing in But Tutte design. W.T. VLSI by to defined was which graph, a Abstract ugah nowihtegraph the which into subgraphs ipegah.The graphs. simple ∗ E upre yteNtoa aua cec onaino hn un China of Foundation Science Natural National the by Supported rp ssi obe to said is graph A The ( ( n = G G G  V G G ,te esythat say we then ), and ) 1 G P ( ) 1 union eagahwt etxset vertex with graph a be G m 1 ∪ and ,  ∨ ln graph plane E E ,m n, { H ( v v ( G } G 1 G = ) h thickness The h hcns fcreinproduct cartesian of thickness The fgraphs of G 1 = 2 hcns;creinpout opeegah path. graph; complete product; cartesian thickness; 2 ) ) ≥ 2 esbrpso graph a of subgraphs be ∪ v . ∪ · · · ∪ V 2 r obtained. are 2 E , ( G ( atsa product cartesian Suppose . G or ) ini nvriy ini 002 P.R.China 300072, Tianjin University, Tianjin × 2 G E .The ). planar G v 1 V ( 1 and G v ( sthe is θ H 2 k ( G = ) ∈ n hs deset edge whose and ) eateto Mathematics of Department -al [email protected] E-mail: G fagraph a of ) fi a edano h ln ota otoegscross edges two no that so plane the on drawn be can it if intersection G E 2 1 V vertex-amalgamation E ( G G , stegraph the is H ( ( G G and ) a edcmoe.I satplgclivratof invariant topological a is It decomposed. be can 2 n deset edge and ) and ) G , . . . , fgraphs of a Yang Yan G u If . G G 1 E 05C10. k 1 = ( stemnmmnme fpaa spanning planar of number minimum the is G r pnigsbrpsof subgraphs spanning are G G ∩ G u 1 i G = ) 2 ∪ } and ∩ 2 G . G E E of E 2 1 ( ( G G ( ihvre set vertex with H of ∪ G G ,adgah nti ae r all are paper this in graphs and ),  G 1 j G stegraph the is H = ) 2 and 1 = ) and and s ftecreinproduct cartesian the of ess ebpriegah n hy- and graphs bipartite te ∅ ns ..gnsadcross- and genus e.g. ants, G , e rn o 11126167 No. Grant der ytpso rpswhose graphs of types ly ( { G hdani hswyis way this in drawn ph 2 i G ( K u 1 6= sdfie analogously. defined is 2 1 ∩ n v , tvertex at ,i j i, j,  G V G ratapplications ortant 1  2 ( P )( G = H u m G 1 2 1 = ) if , ∗ { v , hs vertex whose ∪ v 2 } V v , ) denoted , 2 avertex (a ( | G k , . . . , u 2 1 and ) u 2 ) ∈ , 2 then {G1, G2,...,Gk} is a decomposition of G. Furthermore, if G1, G2,...,Gk are all , then {G1, G2,...,Gk} is a planar decomposition of G. The minimum number of planar spanning subgraphs into which a graph G can be decomposed is called the thickness of G, denoted by θ(G). The thickness of a graph first defined by W.T. Tutte[6] in 1963, as a topological invariant of a graph, it is an important research object in topological , and it also has important applications to VLSI design[1]. But the thickness problems are very difficult, the results about them are few. The only types of graphs whose thicknesses have been obtained are complete graphs[2], complete bipartite graphs[3] and hypercubes[4]. The reader is referred to [5] for more background and results about the thickness problems.

In this paper, the thickness of the cartesian product KnPm, n,m ≥ 2 is obtained, in which Kn is the with n vertices and Pm is the path with m vertices.

2 The lower bound for the thickness of KnPm, n, m ≥ 2 From Euler’s polyhedron formula, a planar graph G has at most 3|V |− 6 edges, further- more, one can get the lower bound for the thickness of a graph as follows.

m Lemma 2.1[3] If G = (V, E) is a graph with |V | = n and |E| = m, then θ(G) ≥ 3n−6 .

As for the graph KnPm, the numbers of vertices and edges are |V | = mn and |E| = mn(n−1) 2 + n(m − 1) respectively. By using Lemma 2.1,

mn(n−1) 2 + n(m − 1) θ(KnPm) ≥ 3mn − 6 ln + 1 2 m = + 6 6(mn − 2) l m p +1 when n = 6p, 6p + 1, 6p + 2, 6p + 3, 6p + 4; = (1) ( p +2 when n = 6p + 5.

n+7 Lemma 2.2[2] The thickness of the complete graph Kn is θ(Kn)= 6 , except that θ(K9)= θ(K10) = 3.  

3 The thickness of KnP2, n ≥ 2 1 2 Let Kn be the complete graph with n vertices labeled by v1, v2, . . . , vn respectively, Kn 1 is a copy of Kn and it’s vertices labeled by u1, u2, . . . , un respectively, by joining the vertices vi and ui by an edge viui, 1 ≤ i ≤ n, one can get the graph KnP2. Figure 1 illustrates K5P2.

By inserting a vertex wi at edge viui, for 1 ≤ i ≤ n, and merging these n 2-valent vertices w1, w2, . . . , wn into one vertex w, one can get a new graph. This graph also can 3 be seen as the vertex-amalgamation of Kn+1 and Kn+1 at w, denote by Kn+1 ∨{w} Kn+1. Figure 2 shows the graph K6 ∨{w} K6.

v1 u1

v5 v2 u2 u5

v4 v3 u3 u4

Figure 1 The graph K5P2

v1 u1

w u5 v5 v2 u2

v4 v3 u3 u4

Figure 2 The graph K6 ∨{w} K6

Lemma 3.1[7] If G is the vertex-amalgamation of G1 and G2, θ(G1)= n1 and θ(G2)= n2, then θ(G) = max{n1,n2}.

From Lemma 3.1, the thickness of Kn+1 ∨{w} Kn+1 is the same as the thickness of Kn+1. Let θ(Kn+1)= t and {G1, G2,...,Gt} is a planar decomposition of Kn+1, then one can get a planar decomposition of Kn+1 ∨{w} Kn+1 as follows,

{G1 ∨{w} G1, G2 ∨{w} G2,...,Gt ∨{w} Gt}, in which Gi ∨{w} Gi, 1 ≤ i ≤ t are plane graphs. In this way, a planar decomposition of

K6 ∨{w} K6 is shown in Figure 3.

v1 u1 v1 u1

w u5 v5 w u5 v5 v2 u2 v2 u2

v4 v4 v3 u3 u4 v3 u3 u4

Figure 3 A planar decomposition of K6 ∨{w} K6

From the construction of Gi∨{w}Gi, if the edge vqw ∈ Gi∨{w}Gi, then uqw ∈ Gi∨{w}Gi, 1 ≤ q ≤ n. For each graph Gi ∨{w} Gi, 1 ≤ i ≤ t, if vqw, uqw ∈ Gi ∨{w} Gi, then we 4 replace them by a new edge vquq, for 1 ≤ q ≤ n, and delete the vertex w. In this way, we can get a new planar decomposition, which is exactly a planar decomposition of KnP2. Figure 4 illustrates a planar decomposition of K5P2 by using this way.

v1 u1 v1 u1

u5 v5 w u5 v5 v2 u2 v2 u2

v4 v4 v3 u3 u4 v3 u3 u4

Figure 4 A planar decomposition of K5P2

From the argument and construction the above, one can get a planar decomposition of  Kn P2 from that of Kn+1 ∨{w} Kn+1, so we have that  θ(Kn P2) ≤ θ(Kn+1 ∨{w} Kn+1)= θ(Kn+1). (2)

Theorem 3.1 The thickness of the cartesian product KnP2(n ≥ 2) is θ(KnP2) = n+8   6 , except that θ(K8 P2) = θ(K9 P2) = 3 and possibly when n = 6p + 4 (n is a nonnegative  integer). Proof When n 6= 8, 9, from (1), (2) and Lemma 2.2, we can get that θ(KnP2) = θ(Kn+1), except possibly when n = 6p + 4 (p is a nonnegative integer). When n = 8, we suppose that θ(K8P2) = 2, and {G1, G2} is a planar decomposition of K8P2 in which G1 and G2 are plane graphs. We insert 2-valent vertex wi at edge viui, for each 1 ≤ i ≤ n, in G1 and G2, then merge these 2-valent vertices into vertices w and w˜ in G1 and G2 respectively. In this way, we get two new plane graphs, denote by G˜1 and ˜ ˜ ˜ G2, and {G1, G2} is a planar decomposition of K9 ∨{w} K9. From (2), θ(K9)= θ(K9 ∨{w} K9) ≤ 2, it is a contradiction to the fact θ(K9) = 3. Hence, θ(K8P2) = θ(K9) = 3. With a similar argument, we can get θ(K9P2)= θ(K10) = 3. Summarizing the above, the theorem is obtained. 

4 The thickness of KnPm, n ≥ 2, m ≥ 3 We use the similar method with that in section 3. Firstly, we insert a 2-valent vertex into each ”path edge” (the edges come from Pm). Secondly, we merge these (m − 1)n 2-valent vertices into m − 1 vertices, each of which joint two adjacent Kn, then we get a new graph G˜. The graph G˜ can be seen as a vertex-amalgamation of m graphs, in which the first and the mth graphs are Kn+1, the others are Kn+2 − e. From Lemma 3.1, one can get that

θ(G˜)= θ(Kn+2 − e) ≤ θ(Kn+2). (3) 5

With a similar argument to that in section 3, we can get a planar decomposition of

KnPm from a planar decomposition of G˜, hence

θ(KnPm) ≤ θ(Kn+2 − e) ≤ θ(Kn+2). (4)

Theorem 4.1 The thickness of the cartesian product KnPm(n ≥ 2,m ≥ 3) is  n+9 θ(Kn Pm)= 6 , except possibly when n = 6p+3, 6p+4 and n = 8 (p is a nonnegative integer).   Proof When n 6= 7, 8, from (1), (4) and Lemma 2.2, we can get that θ(KnPm) =

θ(Kn+2), except possibly when n = 6p + 3, 6p + 4 (p is a nonnegative integer). When n = 7, from (1) and (4), we have 2 ≤ θ(K7Pm) ≤ θ(Kn+2 − e). We can give a planar decomposition of K9 − e as shown in Figure 5, and K9 − e is a non-planar graph, hence θ(K9 − e) = 2, θ(K7Pm) = 2. Summarizing the above, the theorem follows.  8 1 8

12343 567 3 5264

9 7 9

Figure 5 A planar decomposition of K9 − e

5 Concluded Remarks Determining the thickness for a arbitrary graph is a difficult problem. The known results and methods are few. In [7], the authors first considered getting the thickness of graphs by operations on graphs. In this paper, the author use operations on graphs and some

conclusions in [7], obtain the thickness of the cartesian product KnPm.  n+8 For the exceptional cases in theorem 3.1, when n = 6p+4, θ(Kn P2) is either ⌊ 6 ⌋ or n+8  ⌊ 6 ⌋− 1. For the exceptional cases in theorem 4.1, when n = 6p + 3, 6p + 4, θ(Kn Pm) n+9 n+9  is either ⌊ 6 ⌋ or ⌊ 6 ⌋− 1; when n = 8, θ(K8 Pm) is either 2 or 3. What are the exactly numbers for these exceptional cases are still open.

References

[1] A. Aggarwal, M.Klawe, and P.Shor, Multilayer grid embeddings for VLSI, Algorithmica, 6(1991) 129-151 [2] L.W. Beineke and F. Harary, The thickness of the complete graph, Canad. J. Math., 17 (1965) 850-859 [3] L.W. Beineke, F. Harary, and J.W. Moon, On the thickness of the , Proc. Cambridge. Philo. Soc., 60 (1964) 1-5 6

[4] M. Kleinert, Die Dicke des n-dimensionale W¨urfel-Graphen, J. Comb. Theory, 3 (1967) 10-15. [5] E. M¨akinen and T. Poranen. An annotated bibliography on the thickness, outerthickness, and of a graph. Technical report, University of Tampere, 2009. http://www.sis.uta.fi/ tp54752/pub/thickness-bibliography/. [6] W.T. Tutte, The thickness of a graph, Indag. Math., 25 (1963) 567-577 [7] Y. Yang, X.H. Kong. The thickness of amalgamations of graphs. http://arxiv.org/abs/1201.6483.