1. Introduction
High-performance computers can be widely used in many fields thanks to the development of high performance computing technology. The topological properties of interconnection networks are very important for high-performance computers. One typically uses an undirected graph to model the topology of a multiprocessor system H, where the processor set of H is represented by and the link set of H is represented by .
Interconnection networks have many important properties, one of which is the connectivity denoted by
. A graph’s connectivity is the minimum number of vertices whose removal makes the graph disconnected or trivial [
1]. Connectivity is an important indicator to measure the fault tolerance and reliability of a network. In a large interconnection network, each vertex has a large number of neighbors. This property has an obvious deficiency, in that it assumes that all the adjacent vertices of a vertex will fail at the same time. However, this situation does not happen frequently in real networks. To address this deficiency, Esfahanian et al. [
2] introduced the concept of restricted connectivity by imposing additionally restricted conditions on a network. Super-connectivity is a special case of restricted connectivity. When determing the super-connectivity of a network, one needs to ensure that each vertex has at least one neighbor. Hence, super-connectivity is a more refined index to judge the fault tolerance of a network.
Let
K be a subset of
.
(or
) denotes a graph obtained by removing all the vertices in
K and edges incident to at least one vertex in
K from
G. If
is disconnected and each component of
has at least two vertices, then
K is called a super vertex cut. Let
S be a subset of
. If
is disconnected and each component of
has at least two vertices, then
S is called a super edge cut. The super-connectivity of
G (or, respectively, the super edge connectivity), denoted by
(or
), is the minimum cardinality of all super vertex cuts (or super edge cuts) in
G, if any exist. Many relevant results have been obtained regarding super-connectivity and super edge connectivity [
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16].
The hypercube
has become one of the most popular interconnection networks, because of its many attractive properties, such as its regularity and symmetry.
is a Cayley graph and hence vertex-transitive and edge-transitive. However, the diameter of
is not optimal. In order to enhance the hypercube, researchers have proposed many variants, such as crossed cubes [
17], locally twisted cubes [
18], and spined cubes [
19]. The
n-dimensional locally twisted cube
was proposed by Yang et al. [
18], whose diameter was only about half that of
. Many research results have been published on the properties of
[
20,
21,
22,
23,
24,
25].
is vertex-transitive if and only if
, and it is edge-transitive if and only if
[
25]. To further enhance the hypercube, inspired by the folded cube [
26], Peng et al. [
27] proposed a new network topology called the folded locally twisted cube
. So far there, no work has been reported on the super-connectivity of
. In this work, we studied the super-connectivity of
and obtained the result that the super-connectivity
is
for
, which is about twice
.
2. Preliminaries
In this paper, we use the terms vertex and node interchangeably. We also use to denote an edge between vertices x and y. For any vertex , the neighboring set of x is denoted by (or for short). Let . The neighboring set of S is defined as (or for short). We define and . We use to represent a binary string of length n, where for is a part of . is the first part of , and is the nth part of . The symbol is used to represent the complement of . As a variant of , has the same number of vertices as . Each vertex of is denoted by a unique binary string of length n. The definition of is given below.
Definition 1 ([
18]).
For , an n-dimensional locally twisted cube, , is defined recursively as follows:(1) is a graph consisting of four nodes labeled with 00, 01, 10, and 11, which are connected by four edges, (00, 01), (00, 10), (01, 11), and (10, 11).
(2) For , is built from two disjointed copies of named and . Let (or, respectively, ) be the graph obtained by prefixing the label of each node of one copy of with 0 (or with 1); each node of is connected to the node of with an edge, where represents modulo 2 addition.
and
are demonstrated in
Figure 1. Each node in
has only one adjacent node in
. The set of edges between
and
is called a perfect matching
M of
. Hence, we can write
. In [
18], Yang et al. also provided a non-recursive definition of
.
Definition 2 ([
18]).
Let and be any two distinct vertices of for . μ and ν are connected if and only if one of the following conditions is satisfied:1. There is an integer such that
(a) ;
(b) ( represents modulo 2 addition);
(c) all the remaining bits of μ and ν are the same.
2. There is an integer such that μ and ν only differ in the kth bit.
Let be any vertex of . By Definition 2, all the n neighbors of are listed as follows:
;
;
;
…
;
.
We call the ith dimensional neighbor of for .
Definition 3 ([
27]).
For any integer , an n-dimensional folded locally twisted cube, denoted by , is a graph constructed based on by adding all complementary edges. Each vertex in is incident to another vertex through a complementary edge, where . We call the added complementary edges
c-links.
has
c-links, and each vertex
is connected to a complementary vertex
by a
c-link. The set of complementary edges between
and
is a perfect matching
C of
. Hence, we can write
or
. Each node
in
(or, respectively,
) has two neighbors,
and
, in
(or
) for
. Compared with
, each vertex in
has one more neighbor. Then, the node degree of
is
and
[
27].
Figure 2 demonstrates
and
, respectively, and
Figure 3 demonstrates
.
3. Super Connectivity of
In this section, we study the super connectivity of for any integer . Since is composed of and the complementary edge set C, we can use some properties of to prove the super-connectivity property of .
Lemma 1 ([
18]).
For , . Lemma 2 ([
28]).
For any two vertices μ, (), we have . Lemma 3 ([
28]).
Let μ and ν be any two distinct vertices in such that .(1) If and , then the one common neighbor is in , and the other one is in .
(2) If μ, or , then the two common neighbors are in or .
Lemma 4. Let μ and ν be any two distinct vertices in the same for and . If or , then .
Proof. Without loss of generality, we suppose that , , and . Then, is the common neighbor for and . Let and . Next, we consider the neighbors of and in X according to different values of the first part of .
Case 1. .
and
. We list
and
separately in
Table 1.
It is obvious that .
Case 2. .
and
. We list
and
separately in
Table 2.
It is obvious that .
Hence, and have only one common neighbor in and . □
Lemma 5. Let μ be any node in , where and . Then, .
Proof. Let . We consider the different values of the first part of .
Case 1. .
Let
and
. All the neighbors of
and
in
X are listed separately in
Table 3.
It is obvious that .
Case 2. .
Let
and
. All the neighbors of
and
in
X are listed separately in
Table 4.
It is obvious that .
Hence, . □
Lemma 6. Let μ, where . Then .
Proof. Since is constructed from by adding the complementary edge set C, we can study this lemma based on .
Case 1. , are in the same for .
Without loss of generality, we suppose that . According to Lemmas 2 and 3, for , and the two common neighbors are in . According to the definition of , we have , , , and . If and , then and do not have the same neighbors in . Hence, . According to Lemma 4, if or , then and have only one common neighbor in and .
Case 2. and are in a different for .
Without loss of generality, we suppose that and . According to Lemma 2, . Based on the definition of , we have and . According to Lemma 5, and . Hence, we cannot find a vertex , where and have two common neighbors, nor can we find a vertex , where and have two common neighbors. Then, u and v cannot have three or four common neighbors in . Hence, . □
Lemma 7 ([
28]).
If μ and ν are two vertices of and , where , then . Lemma 8. If μ and ν are two vertices of and , where , then .
Proof. According to the position of and , we consider two cases.
Case 1. and are in the same for .
Without loss of generality, we assume that
. According to Lemma 7,
. We have
and
. If
, then
. Otherwise, if
or
, then we let
. All the possible values of
and
are listed in
Table 5.
It is obvious that ; then, we reach a contradiction, and all these values of and are impossible. Hence, .
Case 2. and are in a different for .
Without loss of generality, we assume that
and
. Since
,
should be
or
. If
, let
. Otherwise, If
, let
. Let
. All the possible values of
K are listed in
Table 6.
Since when and , when , and do not have common neighbors. Hence, . □
Lemma 9. Let μ be any node in for any integer . Then, is connected.
Proof. We use mathematical induction on n to prove this lemma. According to Lemma 1, we know that this lemma obviously holds when . Suppose that this lemma holds for . Let be any node in . Without loss of generality, we suppose that . Then, by the induction hypothesis, is connected. Since , according to Lemma 1, is connected. Since each node in is connected to a node in , is connected to . Then, is connected. Hence, this lemma holds. □
Since is the minimum cardinality of all super vertex cuts in , to obtain the upper bound of , we just need to find a super vertex cut F. Then, we have . If we can prove that is connected after removing vertices, then we have the lower bound . With these two results, we can obtain . In the following, we will present two important lemmas to prove the upper bound and lower bound of .
Lemma 10. for any integer .
Proof. Consider an edge . Let . Then, is disconnected, and the edge is one component of . According to Lemma 8, . Let . To prove that is a super vertex cut, we need to show that each vertex has at least one neighbor. According to Lemma 6, and . Since and for , has at least one neighbor in K. Hence, is a super vertex cut and for . □
Lemma 11. for .
Proof. Suppose that F is a super vertex cut of . Then, is disconnected, and each vertex in has at least one neighbor. To prove , we will show that is connected when . Let for , , and . Without loss of generality, we suppose that . Then, .
Case 1. is connected.
Let be any node in . We have . If , then is connected to . Since is connected, then is connected, which means that is connected. Otherwise, since each vertex in has at least one neighbor, there must be a vertex such that . We have . If , then can be connected to through vertex , and is connected. Otherwise, we have , , and . Let . According to Lemma 8, . Since , we can find at least one vertex such that and are connected to through . Hence, is connected.
Case 2. is disconnected.
According to Lemma 1, we have . Since is disconnected, then and . There should be an isolated vertex in and . According to Lemma 9, is connected. For any vertex in where , based on Lemma 8, and do not have common neighbors. Then, there exists a neighbor of in such that . Hence, is connected to through . For any vertex in where , there must exist a neighbor in . Let . According to Lemma 8, . Since , we can find at least vertices in Y connected to . Since there exist two neighbors in for each vertex in Y and when , we can find a vertex in Y such that and are connected to through . Hence, is connected.
Thus, is connected when and for any integer . □
According to Lemmas 10 and 11, we obtain the following result:
Theorem 1. for .