Hamilton-connectivity of Interconnection Networks Modeled by a Product of Graphs

Open access

Abstract

The product graph Gm *Gp of two given graphs Gm and Gp, defined by J.C. Bermond et al.[J Combin Theory, Series B 36(1984) 32-48] in the context of the so-called (Δ,D)-problem, is one interesting model in the design of large reliable networks. This work deals with sufficient conditions that guarantee these product graphs to be hamiltonian-connected. Moreover, we state product graphs for which provide panconnectivity of interconnection networks modeled by a product of graphs with faulty elements.

Abstract

The product graph Gm *Gp of two given graphs Gm and Gp, defined by J.C. Bermond et al.[J Combin Theory, Series B 36(1984) 32-48] in the context of the so-called (Δ,D)-problem, is one interesting model in the design of large reliable networks. This work deals with sufficient conditions that guarantee these product graphs to be hamiltonian-connected. Moreover, we state product graphs for which provide panconnectivity of interconnection networks modeled by a product of graphs with faulty elements.

1 Introduction

Linear arrays and rings are two of the most important computational structures in interconnection networks. So, embedding of linear arrays and rings into a faulty interconnected network is one of the important issues in parallel processing. An interconnection network is often modeled as a graph, in which vertices and edges correspond to nodes and communication links, respectively. Thus, the embedding problem can be modeled as finding fault-free paths and cycles in the graph with some faulty vertices and/or edges. In the embedding problem, if the longest path or cycle is required the problem is closely related to well-known hamiltonian problems in graph theory. In the rest of this paper, we will use standard terminology in graphs(see ref.[2]). It is very difficult to determine that a graph is hamiltonian or not. Readers may refer to [4,5,6].

2 Definitions and Notation

We follow [2] for graph-theoretical terminology and notation not defined here. A graph G = (V,E) always means a simple graph(without loops and multiple edges), where V = V(G) is the vertex set and E = E (G) is the edge set. The degree of a vertex v is denoted by d(v) = dG(v), whereas d = δ(G) and Δ = Δ(G) stand for the minimum degree and the maximum degree of G, respectively. A path is a sequence of adjacent vertices, denoted by x1x2 · · · xl, in which all the vertices x1,x2, · · · ,xl are distinct. Let path P = x1,x2, · · · ,x and P[xi,xj] denotes the sequence of adjacent vertices from xi to x j on path P. That is, P[xi,xj] = xixi+1 · · ·xj-1xj. For two vertices u,v ϵ V(G), a path joining u and v is called a uv-path. A path P in graph G is call a Hamiltonian path if V(P) = V(G). A cycle is a path such that the first vertex is the same as the last one. A cycle is also denoted by x1x2 · · · xlx1. Length of a cycle is the number of edges in it. An -cycle is a cycle of length l. Let c(G) = max{l : -cycle of G}. A cycle of G is called Hamiltonian cycle if its length is |V(G)|. A graph G is Hamiltonian if G contains a Hamiltonian cycle. A graph G is Hamiltonian-connected if any two vertices u,v 2 V(G) exists a Hamiltonian uv-path.

Two edges xy,uv 2 E (G) are called independent if {x,y}∩{u,v} = / 0. A matching is a set of edges that are pairwise independent. A perfect matching between two disjoint graphs G1, G2 with the same order n is a matching consisting of n edges such that each of them has one end vertex in G1 and the other one in G2. See [7, 8, 9, 10, 11, 12, 13, 14].

The construction of new graphs from two given ones is not unusual at all. Basically, the method consists of joining together several copies of one graph according to the structure of another one, the latter being usually called the main graph of the construction. In this regard, Chartrand and Harary introduced in [3] the concept of permutation graph as follows. For a graph G and a permutation π of V(G), the permutation graph Gπ is defined by taking two disjoint copies of G and adding a perfect matching joining each vertex v in the first copy to π(v) in the second. Examples of these graphs include hypercubes, prisms, and some generalized Petersen graphs. The product graph Gm * Gp of two given graphs Gm and Gp, defined in [1] by Bermond et al. in the following way.

Definition 1. Let Gm = (V(Gm),E(Gm)) and Gp = (V(Gp),E(Gp)) be two graphs. Let us give an arbitrary orientation to the edges of Gm, in such a way that an arc from vertex x to vertex y is denoted by exy. For each arc exy, let πexy be a permutation of V(Gp). Then the product graph Gm *Gp has V(Gm)×V(Gp) as vertex set, with two vertices (x,x'), (y,y') being adjacent if either

x=yandxyE(Gp)

or

xyE(Gm)andy=πexy(x).

The product graph Gm * Gp can be viewed as formed by |V(Gm)| disjoint copies of Gp, each arc exy, indicating that some perfect matching between the copies Gpx,Gpy( respectively generated by the vertices x and y of Gm) is added. So the graph Gm is usually called the main graph and Gp is called the pattern graph of the product graph Gm * Gp . Moreover, every edge of Gm * Gp that belongs to any of the |E(Gm)| perfect matchings between copies of Gp is an cross edge of Gm * Gp.

Observe that if we choose πexy(x') = x' for any arc exy then Gm * Gp = Gm□Gp. Furthermore, if Gm is K2 we have K2 * G = G* , a permutation graph. Hence, Gm * Gp can be considered as a generalized permutation graph.

Definition 2. A graph G is called f -fault hamiltonian (resp. f -fault hamiltonian-connected) if there exist a hamiltonian cycle (resp. if each pair of vertices are joined by a hamiltonian path) in G -F for any set F of faulty elements with |F| ≤ f.

If a graph G is f -fault hamiltonian (resp. f -fault hamiltonian-connected), then it is necessary that f ≤ δ(G)-2 (resp. f ≤ δ(G)-3), where δ(G) is the minimum degree of G.

Definition 3. A graph G is call f -fault q-panconnected if each pair of fault-free vertices are joined by a path in G -F of every length from q to |V(G-F)|-1 inclusive for any set F of fault elements with |F| ≤ f.

When we are to construct a path from s to t, s and t are called a source and a sink, respectively, and both of them are called terminals. When we find a path/cycle, sometimes we regard some fault-free vertices and edges as faulty elements. They are called virtual faults.

3 Main Results

For the sake of convenience, we write m(p) for the order of Gm( Gp, respectively.) Suppose V(Gm) = {x1,x2, · · · ,xm} and V(Gp) = {y1,y2, · · · ,yp}.

Theorem 1

If Gm is hamiltonian, and Gp is hamiltonian-connected, then Gm * Gp is also hamiltonian-connected.

Proof. Choose two vertices in u,v ϵ V(Gm * Gp). We have the following cases:

Case 1. u,v belong to the same subgraph, say, GPx1. Since Gm is hamiltonian, suppose hamiltonian-cycle C = x1x2 · · ·xm. Since GPx1is hamiltonian-connected, u,v are connected by a hamiltonian-path P1 in GPx1, suppose P1 = (x1,y1)(x1,y2) · · · (x1,yp), where (x1,y1) = u, (x1,yp) = v. Since x1xm ϵ E(Gm), suppose (x1,y2)(xm,y2) ϵ E (Gm * Gp). Then u,v are connected by a hamiltonian-path P' in Gm * Gp by choosing the endvertex, say, (x2,y1) of cross edge (x1,y1)(x2,y1) and have a hamiltonian-cycle C2 passing (x2,y1) in Gpx2.Suppose the last vertex is (x2,yp) in C2, and choosing the endvertex, say, (x3,yp) of cross edge (x3,yp)(x2,yp) and have a hamiltonian-cycle C3 passing (x3,yp) in Gpx3.Suppose the last vertex is (x3,yl) in C3, and so on, choosing the endvertex, say, (xm,yj) (1 ≤ j ≤ p ) of cross edge (xm-1,y1)(xm,yj) and choosing the endvertex (xm,y2) of cross edge (x1,y2)(xm,y2), have a hamiltonian-path Pm passing (xm,yj) and (xm,y2) in Gpxm.Let P' = u(x2,y1)∪C2[(x2,y1),(x2,yp)](x2,yp)(x3,yp)∪C3[(x3,yp),(x3,yl)]∪· · ·∪(xm-1,y1)(xm,yj)∪ Pm[(xm,yj),(xm,y2)](xm,y2)(x1,y2)∪P1[(x1,y2),v]. Then we obtain the desired result.

Case 2. u,v belong to two subgraphs, suppose GPx1, GPxi(1 ≤ i ≤ m). and uGPx1and vGPxi.Suppose u = (x1,y1), v = (xi,yj) (1 ≤ j ≤ p) . Since Gm is hamiltonian, suppose hamiltonian-cycle C = x1x2 · · ·xmx1. Since GPx1is hamiltonian-connected, u belongs to a hamiltonian-cycle C1 in GPx1, suppose C1 = (x1,y1)(x1,y2) · · · (x1,yp)(x1,y1), where (x1,y1) = u.

Then u,v are connected by a hamiltonian-path P' in Gm * Gp by choosing the endvertex, say, (x2,y1) and (x2,y2) of cross edges (x1,y1)(x2,y1) and (x1,y2)(x2,y2). Then there exist a hamiltonian-path P2 connecting (x2,y1) and (x2,y2) in GPx2 since GPx2is Hamilton-connected. Taking adjacent vertices (x2,yt) and (x2,yt+1) in path P2, then P2 = P21 (x2,yt)(x2,yt+1)[P22, where P21 = P2[(x2,y1), (x2,yt)] and P22 = P2[(x2,yt+1), (x2,y2)]. Choosing the endvertices , say, (x3,yt) and (x3,yt+1) of cross edges (x3,yt)(x2,yt) and (x3,yt+1)(x2,yt+1). Then there exist a hamiltonian-path P3 connecting (x3,yt) and (x3,yt+1) in Gpx3since Gpx3is Hamilton-connected. Taking adjacent vertices (x3,yj) and (x3,yj+1) in path P3, then P3 = P31 (x3,yj)(x3,yj+1) ∪ P32, where P31 = P3[(x3,yt), (x3,yj)] and P32 = P3[(x3,yj+1), (x3,yt+1)], and so on. Until we take adjacent vertices (xi-2,ym) and (xi-2,ym+1) in path Pi-2, and the endvertices, say, (xi-1,ym) and (xi-1,ym+1) of cross edges (xi-1,ym )(xi-2,ym) and (xi-1,ym+1)(xi-2,ym+1). Then there exist a hamiltonian-path Pi-1 connecting (xi-1,yμ) and (xi-1,yμ+1) in Gpxi1 since Gpxi1 is Hamilton-connected. Let path P1=u(x2,y1)P21(x2,yt)(x3,yt)P31P(i2)1(xi2,yμ)(xi1,yμ)Pi1(xi1,yμ+1)(xi2,yμ+1)P(i2)2P32(x3,yt+1)(x2,yt+1)P22(x2,y2)(x1,y2)C1[(x1,y2),(x1,yp)].

Choosing the endvertex, say, (xm,yp) of cross edge (x1,yp)(xm,yp). Then there exist a hamiltonian-cycle Cm passing (xm,yp) in Gpxmsince Gpxmis Hamilton-connected. Suppose the last vertex is (xm,y1) in Cm, taking the endvertex, say, (xm-1,yp) of cross edge (xm,y1)(xm-1,yp). Then there exist a hamiltonian-cycle Cm-1 passing (xm-1,yp) in Gpxm1 since Gpxm1 is Hamilton-connected, and so on. Until we take the last vertex, say, (xi+1,y1) in Ci+1, taking the endvertex, say, (xi,yp) of cross edge (xi+1,y1)(xi,yp).

If (xi,yp) = (xi,yj). Then there exist a hamiltonian-cycle Ci passing (xi,yp) in Gpxisince Gpxiis Hamilton-connected, taking the adjacent vertex of (xi,yj), say, (xi,yj-1) in cycle Ci and taking the endvertex, say, (xi+1,yj-1) of cross edge (xi+1,yj-1)(xi,yj-1) (if (xi+1,yj-1) = (xi+1,yp), then taking the other adjacent vertex of (xi,yj) such that (xi+1,yj-1) ≠ (xi+1,yp)). Then there exist a hamiltonian-path Pi+1 connecting (xi+1,yp) and (xi+1,yj-1) in Gpxi+1since Gpxi+1is Hamilton-connected. Let P2=(x1,yp)(xm,yp)Cm[(xm,yp),(xm,y1)](xm,y1)(xm1,yp)Cm1[(xm1,yp),(xm1,y1)]Ci+1[(xi+1,yp),(xi+1,y1)](xi+1,y1)(xi,yp)Pi[(xi,yp),v]

If (xi,yp) ≠ (xi,yj). Then there exist a hamiltonian-path Pi connecting (xi,yp) and (xi,yj) in Gpxisince Gpxiis Hamilton-connected. Let P2=(x1,yp)(xm,yp)Cm[(xm,yp),(xm,y1)](xm,y1)(xm1,yp)Cm1[(xm1,yp),(xm1,y1)]Ci+1[(xi+1,yp),(xi+1,y1)](xi+1,y1)(xi,yp)Pi[(xi,yp),v]

Let P=P1P2Then we obtain the desired result. 2

Corollary 2. If Gm and Gp are hamiltonian-connected, then Gm * Gp is also hamiltonian-connected.

Corollary 3. If Gm and Gp are hamiltonian, then,

  • c(Gm * Gp) ≥ mp — p,

  • There exists a hamiltonian path in Gm * Gp.

Proof. Since Gm is Hamiltonian, suppose Hamiltonian-cycle C' = x1x2 · · · xmx1. Since Gp is Hamiltonian, suppose Hamiltonian-cycle C" = y1y2 · · ·ypy1.

  • We obtain a cycle C as follows: Choosing any vertex, say, (x1,y1), in V(Gpx1). Taking the endvertex, say, (x2,y1), of cross edge (x1,y1)(x2,y1) and have a hamiltonian-cycle C2 passing (x2,y1) in Gpx2. Suppose the last vertex is (x2,yp) in C2. Taking the endvertex, say, (x3,y1), of cross edge (x2,yp)(x3,y1) and have a hamiltonian-cycle C3 passing (x3,y1) in Gpx3The last vertex is (x3,yp) in C3, · · · , and so on, until taking the endvertex, say, (xm,y1), of cross edge (xm,y1)(xm-1,yp) and have a hamiltonian-cycle Cm passing (xm,y1) in GpxmThe last vertex is (xm,yp) in Cm. Taking the endvertex, say, (x1,yj), of cross edge (xm,yp)(x1,yj) and have a hamiltonian-cycle C1 passing (x1,yj) in Gpx1. So the length of ((x1,yj), (x1,y1))-path in Gpx1is at least p2, but if (x1,yj) = (x1,y1), then (i) holds.

  • We obtain a Hamiltonian path as follows: Choosing any vertex, say, (x1,y1), in V(Gpx1), and have a hamiltonian-cycle C1 passing (x1,y1) in Gpx1The last vertex is (x1,yp) in C1. Taking the endvertex, say, (x2,y1), of cross edge (x1,yp)(x2,y1) and have a hamiltonian-cycle C2 passing (x2,y1) in Gpx2The last vertex is (x2,yp) in C2. Taking the endvertex, say, (x3,y1), of cross edge (x2,yp)(x3,y1) and have a hamiltonian-cycle C3 passing (x3,y1) in Gpx3The last vertex is (x3,yp) in C3, · · · , and so on, until taking the endvertex, say, (xm,y1), of cross edge (xm,y1)(xm-1,yp) and have a hamiltonian-cycle Cm passing (xm,y1) in Gpxm. The last vertex is (xm,yp) in Cm. Then we obtain hamiltonian-path ((x1,y1), (xm,yp))-path.

Corollary 4

If there exists hamiltonian path in Gm, Gp is hamiltonian, then there exists a hamiltonian path in Gm * Gp

Fi denote the sets of faulty elements in Gpxi,1imF0 denotes the set of faulty edges in cross edge-set. |Fi| = fi, 0 ≤ i ≤ m. We denote by fvx1,fvx2,,fvxm the number of faulty vertices in Gpx1,Gpx2,,GPxmrespectively, and by fv the number of faulty vertices in Gm * Gp, so that fv=fvx1+fvx2++fvxmNote that the length of a hamiltonian path in Gm * Gp-F is mp — fv-1.

Theorem 5

Let Gm have a hamiltonian-path, Gp is f -fault hamiltonian-connected and f +1-fault hamiltonian, p ≥ 3 f +6, then,

  • for any f ≥ 1, Gm * Gp is f +2-fault hamiltonian, and

  • for f = 0, Gm * Gp with 2 faulty elements has a hamiltonian cycle unless one faulty element is contained in Gpxiand the other faulty element is contained in Gpxj,ij,1i,jm

Proof. (a) Suppose C = x1x2 · · ·xm is hamiltonian-path in Gm. Assuming the number of faulty elements |F| ≤ f +2, we will construct a cycle of length l, l = mp — fv, in Gm * Gp-F.

Case 1. fi ≤ f , 0 ≤ i ≤ m. Taking two adjacent vertices u,vV(Gpx1F1) Since Gp is f -fault hamiltonian-connected, there exists hamiltonian uv-path, say, P1, in Gpx1F1. We first claim that there exist an edge (x,y) on P1 such that all of , (x, x̄ ), ȳ, and (y, ȳ) do not belong to F, where x¯,y¯Gpx2,(x,x¯),(y,y¯)are two cross edges. Since there are pfvx11candidate edges on P1 and at most f +2 faulty elements can “block” the candidates, at most two candidates per one faulty element. By assumption p ≥ 3 f +6, and the claim is proved. The path P12 can be obtained by merging P1 and a hamiltonian-path P2 in Gpx2F2between x̄, ȳ with the edges (x, ), (y, ȳ ), of course the edge (x,y) is discarded. Similarly, there exist an edge (x',y') on P2 such that all of x¯,(x,x¯),y¯and (y,y¯)do not belong to F, where x¯,y¯Gpx3,(x,x¯),(y,y¯)are two cross edges. The path P123 can be obtained by merging P12 and a hamiltonian-path P3 in Gpx3F3between x¯,y¯with the edges (x,x¯),(y,y¯)of course the edge (x',y') is discarded, and so on, at least we obtain the hamiltonian-path P12· · · m in Gm * GP -F. Therefore P12· · · m ∪ uv is hamiltonian-cycle in Gm * GP-F.

Case 2. There exists some i, j such that fi = f +1, fj = 1. Then ft = 0, 0 ≤ t ≤ m, ti, j. Since f ≥ 1.

Subcase 2.1. i = 0. Taking two adjacent vertices u,vV(Gpx1F1)Since Gp is f -fault hamiltonian-connected, there exists hamiltonian uv-path, say, P1, in Gpx1F1Similar to Case 1, it follows that Gm * Gp is f +2-fault hamiltonian.

Subcase 2.2. i = 1. Since Gp is f + 1-fault hamiltonian, Taking hamiltonian cycle in GPx1F1Similar to Case 1, it follows that Gm * Gp is f +2-fault hamiltonian.

Subcase 2.3. i 6= 1, 0. Since Gp is f +1-fault hamiltonian, Taking hamiltonian cycle Ci in GPxiFiSince fi = f +1, fj = 1. Then ft = 0, 0 ≤ t ≤ m, ti, j, then there exist two adjacent edges (x,y), (y,z) on Ci such that all of , (x, ), ȳ, and (y, ȳ) , z0, and (z, z'), and (y,y') do not belong to F, where x¯,y¯Gpxi+1and y,zGpxi1(x, x̄), (y, ȳ), (y,y'), (z, z') are cross edges. Similar to Case 1, Gpxi1has hamiltonian path Pi-1 connecting the vertex y', z' and Gpxi+1has hamiltonian path Pi+1 connecting the vertex ȳ, . We obtain a cycle C(i-1)i(i+1) by merging Ci and hamiltonian path Pi-1,Pi+1 with the edges (x, x̄), (y, ȳ),(z, z'), (y,y'), of course the edges (x,y) and (y, z) are discarded. Taking an edge α;β in Pi-1 and Pi+1, respectively. Similar to Case 1, at least we obtain cycle C12· · · m, then Gm * Gp is f +2-fault hamiltonian.

Case 3. There exists some fi such that fi = f +2. Then fj = 0, 0 ≤ j ≤ m, ji.

Subcase 3.1. i = 0. Taking two adjacent vertices u,u,vV(Gpx1F1)Since Gp is f -fault hamiltonian-connected, there exists hamiltonian uv-path, say, P1, in Gpx1F1. Since p ≥ 3 f +6, then there exist an edge (x,y) on P1 such that all of ȳ, (x, ), ȳ, and (y, ȳ) do not belong to F, where x¯,y¯Gpx2,(x,x¯),(y,y¯)are two cross edges. Similar to Case 1, it follows that Gm * Gp is f +2-fault hamiltonian.

Subcase 3.2. i = 1. Since Gp is f +1-fault hamiltonian, we select an arbitrary faulty element a in Gpxi. Regarding a as a virtual fault-free element. Taking hamiltonian cycle C1 in GPx1F1. If a is a faulty vertex on C1, let x and y be two vertices on C1 next to a, else if C1 passes through the faulty edge a, let x and y be the endvertices of a. The cycle C12 is obtained by merging C1 and a hamiltonian-path in GPx2 joining and ȳ with cross edges (x, ), (y, ȳ), where x¯,y¯V(Gpx2)Similar to Case 1, it follows that Gm * Gp is f +2-fault hamiltonian.

Subcase 3.3. i 6= 1, 0. Since Gp is f +1-fault hamiltonian, we select an arbitrary faulty element a in Gpxi. Regarding a as a virtual fault-free element. Taking hamiltonian cycle Ci in GPx1FiIf a is a faulty vertex on Ci, let x and y be two vertices on Ci next to a, else if Ci passes through the faulty edge a, let x and y be the endvertices of a. Let z be the vertices on Ci next to y. The cycle C(i-1)i(i+1) is obtained by merging Ci - α and a hamiltonian-path in GPxi+1 joining and ȳ and a hamiltonian-path in GPxi1 joining z' and y' with cross edges (z, z'), (y,y'), (x, ), (y, ȳ), where z,yV(Gpxi1)and x¯,y¯V(Gpxi+1). Similar to Subcase 2.3, it follows that Gm * Gp is f +2-fault hamiltonian.

(b) If f = 0. Since Gp is 0-fault hamiltonian-connected and 1-fault hamiltonian, we have similar proof subcase 3.2 by regarding a faulty element as virtual faulty-free element if the two faulty elements is contained in subgraph Gpxi. This completes the proof.

Corollary 6. Let Gm = K2 , Gp is f -fault hamiltonian-connected and f +1-fault hamiltonian, p ≥ 3 f +6, then,

  • for any f ≥ 1, K2 * Gp is f +2-fault hamiltonian.

  • for f = 0, K2 * Gp with 2 faulty elements has a hamiltonian cycle unless one faulty element is contained in Gpx1and the other faulty element is contained in Gpx2.

Theorem 7

Let Gm have a hamiltonian-path, Gp is f -fault q-panconnected and f +1-fault hamiltonian, p ≥ 2q+ f +2, q ≥ 2 f +5, then,

  • for any f ≥ 2, Gm * Gp is f +1-fault q+m-panconnected,

  • for f = 1, Gm * Gp with 2 faulty elements has a path of every length q +m or more joining s and t unless s and t are contain in the same subgraph Gpxi and their neighbors are the faulty element in Gpxj,ij,

  • for f = 0, Gm * Gp with 1 faulty element has a path of every length q +m or more joining s and t unless s and t are contained in the same subgraph Gpxi and one of their neighbors is the faulty element in Gpxj,ij

Proof.

  • Suppose P =x1x2 · · ·xm is a hamiltonian-path of Gm. Assuming the number of faulty elements |F|≤ f +1, we will construct a path of every length L, q +m ≤ L ≤ mp — fv-1, in Gm * Gp — F joining any pair of vertices s and t.

Case 1. fi ≤ f , 0 ≤ i ≤ m.

Subcase 1.1. When both s, t is contained in Gpx1. There exists a path P1 of length l1 in Gpx1joining s and t for every q1pfvx11. We are to construct a longer path P12 that passes through vertices in Gpx1and Gpx2. We first claim that there exist one edge (x,y), on P1 such that all of , (x, ), ȳ, and (y, ȳ), do not belong to F, where x¯,y¯Gpx2,(x,x¯),(y,y¯)are two cross edges. Since there are l1 candidate edges on P1 and at most f +1 faulty elements can "block" the candidates, at most two candidates per one faulty element. By assumption q ≥ 2 f +5, and the claim is proved. The path P12 can be obtained by merging P1 and a path P2 in Gpx2F2between , ȳ with the edges (x, ), (y, ȳ), of course the edge (x,y) is discarded. Let l2 be the length of P2, the length l12 of P12 can be anything in the range 2q+1121+2+12pfvx1fvx21. Since p2q+f+2,fvxif, we have 2q+1pfvx11. So there exist path connecting the vertex s and t such that the length of path can be anything in the range [q,2pfvx1fvx21]. Similarly, we claim that there exist one edge (u,v), on P2 such that all of ū , (u, ū), , and (v, ), do not belong to F, where u¯,v¯Gpx3,(u,u¯),(v,v¯)are two cross edges. The path P123 can be obtained by merging P12 and a path P3 in Gpx3F3between ū, with the edges (u, ū), (v, ), of course the edge (u,v) is discarded. Let l3 be the length of p3, the length l123 of P123 can be anything in the range 3q+212312+3+13pfvx1fvx2fvx31Since p2q+f+2,fvxifwe have 2q+1pfvx11and 3q+22pfvx1fvx21and so on, at least the path P12· · · m, and q ≤ l12· · ·m ≤ mp — fv-1.

Subcase 1.2. When both s, t is contained in Gpxi,i1. We first claim that there exist two edges (x,y), (u,v) on Pi such that all of , (x, ), ȳ, and (y, ȳ), ū, (u, ū), , and (v, ) do not belong to F, where x¯,y¯Gpxi+1,u¯,v¯Gpxi1and (x, ), (y, ȳ), (u, ū), (v, ) are four cross edges. Since there are l1 candidate edges on P1 and at most f +1 faulty elements can "block" the candidates, at most two candidates per one faulty element. By assumption q ≥ 2 f +5, and the claim is proved. The path P(i-1)i(i+1) can be obtained by merging Pi and two paths Pi-1, Pi+1 in Gpxi1Fi1,Gpxi+1Fi+1respectively, between ū, with the edges (u, ū), (v, ), and , ȳ with the edges (x, ), (y, ȳ), of course the edge (u,v), (x,y) is discarded. Similar to Subcase 1.1, it follows that the path P12· · · m, and q ≤ l12· · · m ≤ mp — fv-1.

Subcase 1.3. When s is in Gpxi, t is in Gpxj,ij. Since G is f -fault hamiltonian-connected, then f ≤ δ(G)-3, where δ(G) is the minimum degree of G, then there exist one adjacent vertex SGpxi of s such that the cross edges sequence T ∉ F, where T=s(xi+1,yi+1),(xi+1,yi+1)(xi+2,yi+2),,(xj1,yj1)(xj,yj) where =(xi+1,yi+1)Gpxi+1,(xi+2,yi+2)Gpxi+2,(xj1,yj1)Gpxj1,(xj,yj)Gpxj There exists a path P1 of length l1 in Gpxj joining t and (xj,yj) for every q1pfvxj1. Hence a path P1 joining s and t can be obtained by merging P1 and the cross edges sequence T with the length 1of path P1 for every integer in the range q+ji+11pfvxj+ji1. Similarly to Subcase 1.2, it follows that the result completes.

Case 2. There exists some fi such that fi = f +1. Then fj = 0, 1 ≤ j ≤ m, ji. Since f ≥ 2. Similar to case 1, it follows that the result completes.

It immediately follows from Case 1, where the assumption f ≥ 2 is never used, that f = 0, 1, Gm * Gp with f +1 faulty elements has a path of every length q +m or more joining s and t unless s and t are contained in the same subgraph Gpxi and one of their neighbors is the faulty element in Gpxj,ij. Thus, the proof of (c) is done. we testify the case (b), note that Gp is 1-fault q-panconnected. This completes the proof.

Communicated by Juan L.G. Guirao

References

  • [1]

    Bermond J. C. , Delorme C. , and Farhi G., Large graphs with given degree and diameter II , J Combin. Theory, Series B 36 (1984) 32-48. 10.1109/TC.1984.1676504.

  • [2]

    BondyJ. A. and Murty U. S. R., Graph Theory with Applications , 5th printing, American Elsevier Publishing Co., Inc., 1976.

  • [3]

    Chartrand G. and Harary F., Planar permutation graphs , Ann Inst H Poincaré Sec B 3 (1967) 433-438. http://eudml.org/doc/76875

  • [4]

    D. Bauer and E. Schmeichel: On a theorem of Häggkvist and Nicoghossian, In Y. Alavi, F.R.K. Chung, R.L. Graham and D. S. Hsu, eds., Graph Theory, Combinatorics, Algorithms, and Applications-Proceedings 2nd China-USA Graph Theory Conference, SIAM(1991)20-25.

  • [5]

    B. Wei, A Generalization of a result of Bauer and Schmeichel, Graphs and Combin. 9 (1993) 383-389. 10.1007/BF02988325

  • [6]

    D. Bauer, H.J. Broersma, and H.J. Veldman, On generalizing a theorem of Jung, Ars Combinatoria 40 (1995) 207-218.

  • [7]

    S. Wang, C. Wang, J. Liu, On extremal multiplicative Zagreb indices of trees with given domination number, Applied Mathematics and Computation, 332 (2018) 338-350. 10.1016/j.amc.2018.03.058.

  • [8]

    Z. Shao, J. Xu, S. Sheikholeslami, S.Wang, The domination complexity and related extremal values of large 3-D torus, Complexity, Volume 2018, Article ID 3041426, 8 pages. 10.1155/2018/3041426.

  • [9]

    S. Wang, B. Wei, A note on the independent domination number versus the domina- tion number in bipartite graphs, Czechoslovak Mathematical Journal, 67 (142) (2017), 533-536. 10.21136/CMJ.2017.0068-16.

  • [10]

    S. Ji, S. Wang, On the sharp lower bounds of Zagreb indices of graphs with given number of cut vertices, Journal of Mathematical Analysis and Applications, 458 (1) 2018, 21-29. 10.1016/j.jmaa.2017.09.005.

  • [11]

    Y. Ahmad, U. Ali, M. bilal, S. Zafar, Z. Zahid, Some new standard graphs labeled by 3-total edge product cordial labeling 2 (2017)(1) 61-72. 10.21042/AMNS.2017.1.00005.

  • [12]

    S. Sudhakar, S. Francis, V. Balaji, Odd mean labeling for two star graph, (2) 2017 (1) 195-200, 10.21042/AMNS.2017.1.00016.

  • [13]

    B. Basavanagoud, V.R. Desai, S. Patil, b,α-Connectivity Index of Graphs (2) 2017 (1) 21-30, 10.21042/AMNS.2017.1.00003.

  • [14]

    S. Zhou, L. Xu, Y. Xu, A sufficient condition for the existence of a k-factor excluding a given r-factor (2) 2017 (1) 13-20, 10.21042/AMNS.2017.1.00002.

[1]

Bermond J. C. , Delorme C. , and Farhi G., Large graphs with given degree and diameter II , J Combin. Theory, Series B 36 (1984) 32-48. 10.1109/TC.1984.1676504.

[2]

BondyJ. A. and Murty U. S. R., Graph Theory with Applications , 5th printing, American Elsevier Publishing Co., Inc., 1976.

[3]

Chartrand G. and Harary F., Planar permutation graphs , Ann Inst H Poincaré Sec B 3 (1967) 433-438. http://eudml.org/doc/76875

[4]

D. Bauer and E. Schmeichel: On a theorem of Häggkvist and Nicoghossian, In Y. Alavi, F.R.K. Chung, R.L. Graham and D. S. Hsu, eds., Graph Theory, Combinatorics, Algorithms, and Applications-Proceedings 2nd China-USA Graph Theory Conference, SIAM(1991)20-25.

[5]

B. Wei, A Generalization of a result of Bauer and Schmeichel, Graphs and Combin. 9 (1993) 383-389. 10.1007/BF02988325

[6]

D. Bauer, H.J. Broersma, and H.J. Veldman, On generalizing a theorem of Jung, Ars Combinatoria 40 (1995) 207-218.

[7]

S. Wang, C. Wang, J. Liu, On extremal multiplicative Zagreb indices of trees with given domination number, Applied Mathematics and Computation, 332 (2018) 338-350. 10.1016/j.amc.2018.03.058.

[8]

Z. Shao, J. Xu, S. Sheikholeslami, S.Wang, The domination complexity and related extremal values of large 3-D torus, Complexity, Volume 2018, Article ID 3041426, 8 pages. 10.1155/2018/3041426.

[9]

S. Wang, B. Wei, A note on the independent domination number versus the domina- tion number in bipartite graphs, Czechoslovak Mathematical Journal, 67 (142) (2017), 533-536. 10.21136/CMJ.2017.0068-16.

[10]

S. Ji, S. Wang, On the sharp lower bounds of Zagreb indices of graphs with given number of cut vertices, Journal of Mathematical Analysis and Applications, 458 (1) 2018, 21-29. 10.1016/j.jmaa.2017.09.005.

[11]

Y. Ahmad, U. Ali, M. bilal, S. Zafar, Z. Zahid, Some new standard graphs labeled by 3-total edge product cordial labeling 2 (2017)(1) 61-72. 10.21042/AMNS.2017.1.00005.

[12]

S. Sudhakar, S. Francis, V. Balaji, Odd mean labeling for two star graph, (2) 2017 (1) 195-200, 10.21042/AMNS.2017.1.00016.

[13]

B. Basavanagoud, V.R. Desai, S. Patil, b,α-Connectivity Index of Graphs (2) 2017 (1) 21-30, 10.21042/AMNS.2017.1.00003.

[14]

S. Zhou, L. Xu, Y. Xu, A sufficient condition for the existence of a k-factor excluding a given r-factor (2) 2017 (1) 13-20, 10.21042/AMNS.2017.1.00002.

Journal Information

Metrics

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 56 56 38
PDF Downloads 28 28 21