Malaya J. Mat. 2(3)(2014)

Size: px

Start display at page:

Download "Malaya J. Mat. 2(3)(2014)"

Buddy York
5 years ago
Views:

Malaya J Mat (3)(04) 80-87 On k-step Hamiltonian Bipartite and Tripartite Graphs Gee-Choon Lau a,, Sin-Min Lee b, Karl Schaffer c and Siu-Ming Tong d a Faculty of Comp & Mathematical Sciences,

1 Malaya J Mat (3)(04) On k-step Hamiltonian Bipartite and Tripartite Graphs Gee-Choon Lau a,, Sin-Min Lee b, Karl Schaffer c and Siu-Ming Tong d a Faculty of Comp & Mathematical Sciences, Universiti Teknologi MARA, Shah Alam, Malaysia b 3480 Hollyhock St, Union City, CA 94587, USA c Dept of Mathematics, De Anza College, Cupertino, CA 9504, USA d Dept of Computer Science, Northwestern Polytechnic University, Fremont, CA 94539, USA Abstract For integer k, a (p, q)-graph G = (V, E) is said to admit an AL(k)-traversal if there exist a sequence of vertices (v, v,, v p ) such that for each i =,,, p, the distance between v i and v i+ is k We call a graph k-step Hamiltonian (or admits a k-step Hamiltonian tour) if it admits an AL(k)-traversal and d(v, v p ) = k In this paper we consider k-step Hamiltonicity of bipartite and tripartite graphs As an application, we found that a -step Hamiltonian tour of a graph could sometimes induce a super-edge-magic labeling of the graph Keywords: Hamiltonian tour, -step Hamiltonian tour, bipartite & tripartite graphs, NP-complete problem, super-edge-magic labeling 00 MSC: 05C78, 05C5 c 0 MJM All rights reserved Introduction In 856, Kirkman wrote a paper [3] in which he considered graphs with a cycle which passes through every vertex exactly once The dodecahedron (see Figure ) is a graph with such property that Hamilton played cycle games Hence, such a graph is said to be Hamiltonian The Hamiltonicity of a graph is the problem of determining for a given graph whether it contains a path/cycle that visits every vertex exactly once Figure : Dodecahedron There is no simple characterization on Hamiltonian graphs though they are related to the traveling salesman problem So there are potential practical applications In general we know very little about Hamiltonian graphs though their properties have been widely studied A good reference for recent developments and open problems is [9] In this paper we consider simple graphs with no loops For integer k, a (p, q)-graph G = (V, E) is said to admit an AL(k)-traversal if there exist a sequence (v, v,, v p ) such that for each i =,,, p, the Corresponding author address: geeclau@yahoocom (GC Lau)

2 GC Lau et al / On k-step Hamiltonian 8 distance between v i and v i+ is k We call a graph k-step Hamiltonian (or admits a k-step Hamiltonian tour) if it admits an AL(k)-traversal and d(v, v p ) = k For example, the cubic graph in Figure is -step Hamiltonian and two others admit an AL()-traversal but are not -step Hamiltonian Figure : Example on -step Hamiltonicity There has been much research on Hamiltonicity of bipartite graphs [,, 6, 0,, 4] Clearly, -step Hamiltonian is Hamiltonian In this paper we consider bipartite and tripartite graphs As an application, we found that a -step Hamiltonian tour of a graph could sometimes induce a super-edge-magic labeling of the graph For terms used but not defined, we refer to [3] Definition For a graph G, let D k (G) denote the graph generated from G such that V(D k (G)) = V(G) and E(D k (G)) = {uv d(u, v) = k in G} Lemma A graph G is k-step Hamiltonian or admits an AL(k)-traversal if and only if D k (G) is Hamiltonian or has a Hamiltonian path, respectively Proof It follows directly from Definition Main Results We first give a sufficient condition for a graph to admit no k-step Hamiltonian tour Theorem Suppose G has a clique subgraph K p If V(G\K p ) < p, then G is not k-step Hamiltonian for all k Proof Observe that for any vertices u, v in the clique subgraph K p of G, d(u, v) = Hence, K p induces an empty graph in D k (G) If D k (G) is Hamiltonian, then these p vertices must be not adjacent in a Hamiltonian tour of D k (G) This implies that we need at least p more vertices to form such a Hamiltonian tour Since V(G\K p ) < p, it follows that no Hamiltonian tour exists in D k (G) By Lemma, the theorem follows Theorem The vertex gluing of a graph G and an end-vertex of a path of length n k is not k-step Hamiltonian Proof Let G(P n ) denote the graph such obtained Observe that D k (G(P n )) has a cut-vertex and is not Hamiltonian Theorem 3 If graphs G and H are both k-step Hamiltonian, then so is G H Proof By Lemma, G is k-step Hamiltonian if and only if D k (G) is Hamiltonian We show that D k (G) D k (H) is a subgraph of D k (G H) Then any Hamiltonian cycle in D k (G) D k (H) will also exist in D k (G H) and implies that G H is also k-step Hamiltonian Suppose that edge e = (u, v)(u, w) is an edge in D k (G) D k (H) Then (v, w) must be an edge in D k (H), so the distance between v and w in H is k Let v = v 0, v, v,, v k = w be a length k path from v to w in H Then (u, v), (u, v ), (u, v ),, (u, w) is a length k path from (u, v) to (u, w) in G H, so the distance from (u, v) to (u, w) within G H is no more than k Suppose, however, that the distance from (u, v) to (u, w) is less than k in G H, and let e, e,, e m, be a sequence of edges from (u, v) to (u, w) with m < k All edges in this sequence will either be of the form (z, x)(z, y) where xy is an edge in H, or (x, z)(y, z) where xy is an edge in G Consider the subsequence of edges which are of the first type, (z, x)(z, y) This subsequence must be of the form (z, x 0 )(z, x ), (z, x )(z, x ),, (z n, x n? )(z n, x n ), where x 0 = v, x n = w, and n = m < k Furthermore, x 0 x, x x,, x n? x n must be a sequence of edges in H from v = x 0 to w = x n, which has length n, where n < k This contradicts the fact that the distance from v to w in H is actually k Therefore the distance from

3 8 GC Lau et al / On k-step Hamiltonian (u, v) to (u, w) in G H is also k, and so e = (u, v)(u, w) is also an edge of D k (G H); the argument for edges of the form e = (u, v)(w, v) is identical Since all edges and vertices of D k (G) D k (H) are also in D k (G H), D k (G) D k (H) is a subgraph of D k (G H) Since G and H are k-step Hamiltonian, D k (G) D k (H) is Hamiltonian, and so is D k (G H), implying that G H is k-step Hamiltonian A Hamiltonian graph need not be -step Hamiltonian The simplest example is the complete bipartite graph K(, ) that does not admit an AL()-Hamiltonian traversal, and hence cannot be -step Hamiltonian Theorem 4 All bipartite graphs are not k-step Hamiltonian for even k Proof Suppose G = (V, E) is bipartite graph with bipartition (X, Y) If k is even, the vertex in X cannot connect with vertex in Y, vice versa, in D k (G) Thus D k (G) is a disconnected graph with two components X and Y Hence D k (G) cannot have a Hamiltonian path By Lemma, G is not k-step Hamiltonian We now give a necessary and sufficient condition for cycles to admit a k-step Hamiltonian tour Theorem 5 For integers n 3 and k, the cycle C n is k-step Hamiltonian if and only if n k + and gcd(n, k) = Proof If n k, we have either diam(c n ) < k or D k (C n ) is disconnected Hence, C n is not k-step Hamiltonian We may now assume that n k + Without loss of generality, we may assume that a k-step Hamiltonian tour of C n is given by the sequence u, u k+, u k+,, u (n )k+ Note that {, k +, k +, 3k +,, (n )k + } (mod n) is a set of distinct integers if and only if ik + jk + (mod n) for 0 i < j n if and only if (j i)k 0 (mod n) if and only if k/n = r/(j i) for some integer r if and only if gcd(n, k) = Hence, the theorem holds and the k-step Hamiltonian tour of C n is obtained Theorem 6 The cylinder graph C n P m is -step Hamiltonian for odd n 3 and all m 3 Proof Case n = 3 This case is handled separately since the three vertices in any 3-cycle within C 3 P m are distance from each other Figure 3 shows -step Hamiltonian tours for C 3 P and C 3 P 3 Figure 4 shows -step Hamiltonian tours for C 3 P It is based on the -step Hamiltonian tour for C 3 P Here vertices are labeled (a, b) and we may denote edges by listing their vertices: (a, b)(c, d) Then to modify the -step Hamiltonian tour for C 3 P to one for C 3 P, replace edge (, )(, ) by (, 3)(, ), shown as a dotted line, and also remove edge (, )(, ) The cases in which n = 3 and m is odd are handled with similar constructions, based instead on the -step Hamiltonian tour for C 3 P 3 also as shown below The diagram shows the -step Hamiltonian tour for C 3 P +3 which may be modified for the cases C 3 P by adding edge (, )(, ), again shown shown as a dotted line,,,,,,,,,3,3 m = m = 3 3 Figure 3: A -step Hamiltonian tour in C 3 P m, m =, 3 Case n = k + 5 In this case, we consider two subcases Subcase m = j + 3 Figure 5 gives a -step Hamiltonian cycle of this subcase Note that we have partitioned the vertices in a checkerboard pattern so those whose coordinates have even sum are shown by circular vertices, and those whose coordinates have odd sum are shown by square vertices The circular vertices compose one cycle and the square vertices compose another, except that the two cycles cross over and connect through the edges (j +, k + )(j, ) and (j +, k)(j +, )

4 GC Lau et al / On k-step Hamiltonian 83,,,3,4,5,6,7,8,,,, 3,,,3,4,5,6,7,8,,,, C 3 P,,,,,3,3 3,4,4 4,5,5 5,6,7,6,7 6 7,,,, + +,,,, C 3 P +3 Figure 4: A -Hamiltnoian tour in C 3 P m,,3,,4,5,6,7,,,, k k k+ k,,,3,4,5,6,7, k,,, k k k k k k k+ 4, 4, 4,3 4,4 4,5 4,6 4,7 4, 4, k k 4, 4, k k , 5, 5,3 5,4 5,5 5,6 5,7 k k k k+ j, j+, j+, j+, j+, j+, j+, j+, j+, j+, j+, j+, j+, k k k k+ Figure 5: A -step Hamiltonian tour in C k+ P j+ j,

5 84 GC Lau et al / On k-step Hamiltonian Subcase m = j Figure 6 gives a -step Hamiltonian cycle of this subcase We have again partitioned the vertices in a checkerboard pattern so those whose coordinates have even sum are shown by circular vertices, and those whose coordinates have odd sum are shown by square vertices The circular vertices compose one cycle and the square vertices compose another, except in this case the two cycles cross over and connect through the edges (j, k)(j, ) and (j, k + )(j, ) Note that the vertices shown in the rightmost column are identical to those at the bottom of the first column,,3,,4,5,6,7,,,, k k k k+,,,3,4,5,6,7, k,,, k k k k k k k+ 4, 4, 4,3 4,4 4,5 4,6 4,7 4, 4, k k 4, 4, k k+ j, j, j, j, k j, k+ j, j, j, j, j, j, j, j, j, j, j, k k k k+ j, j, Figure 6: A -step Hamiltonian tour in C k+ P j Since C n P m is a subgraph of C n C m, the same -step Hamiltonian cycles work for C n C m, when n is odd, and we have Corollary The graph C n C m is -step Hamiltonian for odd n and all m Let D(n) denote the tripartite donut graph shown in Figure 7 with a given vertex labeling k 3k+4 3k+ k 3k 5k 5k+ k+ 3k k+ 5k 5k 4 3 5k+ k 5k k 3k+3 3k+ 5k+3 k+ 3k k+ 3k+5 3k+7 3 k 3k+6 5k+4 3k 4 5k 6 k 5k 3 3k 3k 4 3k+9 (a) n = 3k = 0 mod (6), k 4 even (b) n = 3k = 4 mod (6), k 5 odd Figure 7: A -step Hamiltonian tour in D(n) Theorem 7 The vertex labeling in graph D(n) gives a -step Hamiltonian tour for all n = 3k, k 4 A ring-worm is a unicyclic graph U n (a, a,, a n ) obtained from a cycle C n with V(C n ) = {v, v,, v n } by identifying vertex v i to the center, c i of a star S i having a i + vertices, {c i, u i,, u i,,, u i,ai } The ringworm has n + a + a + + a n vertices and edges, respectively We can arrange the vertices of the ring-worm as in Figure 8

6 GC Lau et al / On k-step Hamiltonian 85 u, u, u,a u u u a3 u n, u n, u n,an c c n c c 3 c n u, u, u,a u n, u n, u n,an Figure 8: Ring worm graph Theorem 8 If n = 3 and a i > 0 (i =,, 3), or n 5 is odd and a i 0 ( i n), the ring worm U n (a, a,, a n ) is -step Hamiltonian Proof The case n = 3 is obvious Without loss of generality, we assume that n 5 and not all a i = 0 Suppose n = s +, s We can label the vertices by consecutive integers as described below to get a -step Hamiltonian tour for the graph: f (c i+ ) = a + a a i + i +, for i = 0,,,, s, f (c i+ ) = f (c s+ ) + a + a a i+ + i + for i = 0,,,, s, f (u i+,j ) = f (c i+ ) + j for i = 0,,,, s, j =,,, a i+, f (u,j ) = f (c s+ ) + j, j =,,, a, f (u i+,j ) = f (c i ) + j for i =,,, s, j =,,, a i+ We next define two families of cubic graphs Let n be a positive integer The Möbius ladder (also known as the Möbius wheel) is the cycle C n, with n additional edges joining diagonally opposite vertices We will denote this graph by M n, and its vertices by v, v,, v n Then the edges are v v, v v 3,, v n v of the cycle, and the n diagonals are v v n+, v v n+,, v n v n Figure 9 shows the Möbius ladder M n for n = 4, drawn in both the circulant form and the ladder form v 6 v v 8 v v 5 v v 7 v v 6 v 3 v 4 v 3 v 5 v 4 v v v 3 v v v 3 v 4 v 4 v 5 v 6 v 5 v 6 v 7 v 8 Figure 9: Möbius ladder for n = 4 Observe that for odd n, M n is not -step Hamiltonian since it is bipartite For even n, M n is tripartite Theorem 9 For m, M 4m is -step Hamiltonian Proof A -step Hamiltonian tour is given by the sequence v, v 3, v 5,, v 4m, v m, v m,, v, v 4m, v 4m,, v m+, v We now consider the cubic turtle shell graph, TS(n), n even, with a given vertex labeling as shown in Figure 0 Theorem 0 The vertex labeling of the graph TS(n) is a -step Hamiltonian tour for all n = k, k 3

7 86 GC Lau et al / On k-step Hamiltonian 3k+ k+ 3k 3 k+ k+3 3k+3 k+5 k k k k+ (a) n = k = 0 (mod4), k 4 even k+ 3k k 3 k+ k+ 3k+ k+4 3k+4 k k k (b) n = k = (mod4), k 3 odd Figure 0: Graph TS(n), n even 3 An Application For a (p, q)-graph G, a labeling of the vertices and edges of G given by bijections f : V(G) {,,, p} and f + : E(G) {p +, p +,, p + q} is called a super-edge-magic (SEM) labeling if f (u)+ f (v)+ f + (uv) is a constant for every edge uv in E(G) Such a graph is called SEM Theorem 3 ([4, 7]) A graph G is SEM if and only if it admits a bijection f (V(G)) such that { f (u)+ f (v)} consists of q consecutive integers Observe that a -step Hamiltonian labeling for each odd cycle and the ring worm with odd cycle correspond to a vertex labeling that induces an edge labeling f + such that the edge labels form a sequence of consecutive integers However, we are not able to find another -step Hamiltonian labeling that corresponds to a SEM labeling Problem Does there exist infinitely many families of -step Hamiltonian graphs whose labeling corresponds to a SEM labeling? The problem of determining whether a graph is Hamiltonian is NP-complete even for planar graphs In 97, Karp [] proved that finding such a path in a directed or undirected graph is NP-complete Later, Garey and Johnson [8] proved that the directed version restricted to planar graphs is also NP-complete, and the undirected version remains NP-complete even for cubic planar graphs In 980, Akiyama, Nishizeki, and Saito [] showed that the problem is NP-complete even when restricted to bipartite graphs We end this paper with the following conjecture and problem Conjecture The -step Hamiltonian problem for tripartite graphs is NP-complete Problem Study the k-step Hamiltonicity of complete multipartite graph with certain edges deleted References [] TSN Akiyama, T Nishizeki, NP-completeness of the Hamiltonian Cycle Problem for Bipartite Graphs, Journal of Information Processing, 3(980), [] J-C Bermond, Hamiltonian Graphs In Selected Topics in Graph Theory Edited by L W Beineke and R J Wilson Academic, London(978), 7 67 [3] G Chartrand and P Zhang, Introduction to Graph Theory, Walter Rudin Student Series in Advanced Mathematics, McGraw-Hill, 004

8 GC Lau et al / On k-step Hamiltonian 87 [4] Z Chen, On super edge-magic graphs, J of Comb Math Comb Comput, 38 (00), [5] VV Dimakopoulos, L Palios and AS Poulakidas, On the Hamiltonicity of the Cartesian Product, available online: (accessed October 0) [6] MN Ellingham, JD Horton, Non-Hamiltonian 3-connected Cubic Bipartite Graphs, J of Comb Theory B, 34(3)(983), [7] RM Figueroa-Centeno, R Ichishima and FA Muntaner-Batle, The place of super edge-magic labelings among other classes of labelings, Discrete Math, 3(00), [8] MR Garey and DS Johnson, Computers and Intractability: A Guide to the Theory of NP-Completeness, Freeman, New York, 979 [9] R Gould, Advances on the Hamiltonian Problem, A Survey, Graphs and Combinatorics, 9(003), 7 5 [0] D Holton and REL Aldred, Planar Graphs, Regular Graphs, Bipartite Graphs and Hamiltonicity, Australasian J of Combinatorics, 0(999), 3 [] A Itai, CH Papadimitriou and JL Szwarcfiter Hamilton paths in grid graphs SIAM Journal on Computing, (4)(98), [] RM Karp, Reducibility among combinatorial problems, Complexity of Computer Computations, (97) [3] TP Kirkman, On the representation of polyhedra, Phil Trans Royal Soc, 46(856), [4] J Moon and L Moser, On Hamiltonian bipartite graphs, Israel J of Math, (3)(963), Received: October 9, 03; Accepted: February 6, 04 UNIVERSITY PRESS Website:

Super edge-magic labeling of graphs: deficiency and maximality

Electronic Journal of Graph Theory and Applications 5 () (017), 1 0 Super edge-magic labeling of graphs: deficiency and maximality Anak Agung Gede Ngurah a, Rinovia Simanjuntak b a Department of Civil