Abstract
Let G = (V, E) be a connected graph. An edge set S ⊂ E is a 3-restricted edge cut, if G - S is disconnected and every component of G - S has at least three vertices. The 3-restricted edge connectivity λ3 (G) of G is the cardinality of a minimum restricted edge cut of G. A graph G is λ3-connected, if 3-restricted edge cuts exist. A graph G is called λ′-optimal, if λ3 (G) = ξ3 (G), where
Introduction
A network is often modeled by a graph G = (V, E) with the vertices representing nodes such as processors or stations, and the edges representing links between the nodes [1–3]. One fundamental consideration in the design of networks is reliability. An edge cut of a connected graph G is a set of edges whose removal disconnects G. The edge connectivity λ (G) of G is the minimum cardinality of an edge cut S of G. The edge connectivity λ (G) is an important feature determining reliability and fault-tolerance of the network. In the definitions of λ (G), no restrictions are imposed on the components of G - S. To compensate for this shortcoming, it would seem natural to generalize the notion of the classical connectivity by imposing some conditions or restrictions on the components of G - S.
Following this idea, k-restricted edge connectivity were proposed in [4]. An edge set S ⊂ E is said to be a k-restricted edge cut, if G - S is disconnected and every component of G - S has at least k vertices. The k-restricted edge connectivity of G, denoted by λ k (G), is the cardinality of a minimum k-restricted-edge-cut of G. If S is a k-restricted edge cut and |S| = λ k (G), then we call S a λ k -cut. If there are k-restricted edge cuts in G, then G is λ k -connected [5–7].
Thus k-restricted edge connectivity is generalization of the classical edge connectivity and can provide more accurate measures for the reliability and the fault-tolerance of a large-scale parallel processing system, and so has received much attention in recent years (see the survey [8]). We will discuss the 3-restricted edge connectivity of graphs.
Let G = (V, E) be a connected graph, d
G
(v) the degree of a vertex v in G (simply d (v)), and δ (G) the minimum degree of G. Moreover, for S ⊂ V, G [S] is the subgraph induced by S, and G - S denotes the subgraph of G induced by the vertex set of V - S. The girth of G is the shortest circle of G. For X, Y ⊂ V, denote by [X, Y] the set of edges with one end in X and the other in Y. Let the vertex subset U ⊆ V and
For 3-restricted edge connectivity of graphs, there is an important result.
Then G is called λ3-optimal, if λ3 (G) = ξ3 (G). If every minimum 3-restricted edge cut is a set of edges incident to a set of three vertices, then G is said to be super 3-restricted edge connected or super-λ3 for simplicity.
If G is a graph with girth g ≥ 4, then a connected subgraph of G with three vertices is a path P = xyz of length two. Thus, ξ3 (G)= min{d (x) + d (y) + d (z) -4 : xyz is a path of length two in G}.
Define the inverse degree of a graph G with no isolated vertices as
For graph-theoretical terminology and notation not defined here we follow [13]. All graphs considered in this paper are simple, finite and undirected.
Super-λ3 and inverse degree
We start the section with the following useful lemmas.
Hence each vertex of X is incident with at least one edge of [X, Y]. Notice that girth g ≥ 6. Take a path xyz in G [X], we have
We get that λ3 (G) = ξ3 (G). That is |N (x) ∩ Y| + |N (y) ∩ Y| + |N (z) ∩ Y| = | [{x, y, z}, Y] | and | (N (x) ∩ X) ∖ {y} | + | (N (y) ∩ X) ∖ {x, z} | + | (N (z) ∩ X) ∖ {y} | = | [X ∖ {x, y, z}, Y] |. And each vertex of ((N (x) ∩ X) ∖ {y}) ∪ ((N (y) ∩ X) ∖ {x, z}) ∪ ((N (z) ∩ X) ∖ {y}) has only one neighbor in Y. Take z′ ∈ (N (x) ∩ X) ∖ {y} and the path yxz′. Applying the above method to yxz′, we can get y has only one neighbor in Y. Similarly, x and z has only one neighbor in Y. Hence |X| ≥ d (x) + d (y) + d (z) -4 ≥ ξ3 (G).
Subcase 2.1. G [X0] is an independent set. Let x ∈ X0, y, z ∈ X1 and xyz be a path in G [X], then N (x) ⊆ X1. We can get
That is
Each vertex in
Since δ (G) ≥2, girth g ≥ 6, take y′ ∈ N (x) - xz′ ∈ (N (y′) ∩ X1) ∖ {y}. Applying the above method to the path xy′z′, we can get y and z have only one neighbor in Y. Hence |X| ≥ d (x) + d (y) + d (z) -4 ≥ ξ3 (G).
Subcase 2.2. G [X0] has an edge e = xy. It is similar to the above case, we have |X| ≥ ξ3 (G). Similarly, we can obtain |Y| ≥ ξ3 (G).
We will introduce a very useful lemma as follows.
Let a1, a2, ⋯, a
p
, A be positive reals with If, in addition a1, a2, …, a
p
, A are positive integers, and a, b are integers with A = ap + b and 0 ≤ b ≤ p, then If f (x) is continuous and convex on an interval [L, R], and if l, r ∈ [L, R], with l + r = L + R, then f (L) + f (R) ≥ f (l) + f (r).
Hence we get ∑y∈Yd (y) < |Y| (|Y|-1) + λ3. By Lemma 3 (2), we have
Because G has a vertex v of degree δ. Without loss of generality, assume v ∈ X. Then
According to Lemma 3 (2), we can obtain
Because of 1/(|Y|-1) ≥1/(n - 3δ + 3), we obtain
We consider the function
And in conjunction with λ3 ≤ ξ3, we have
There is a contradiction.
