Skip to Content
AxiomsAxioms
  • Article
  • Open Access

19 November 2026

15 Pages

The Edge Fault-Tolerant Unpaired Two-Disjoint Path Covers of Cayley Graphs Generated by Unicyclic Graphs

and
College of Mathematics and System Sciences, Xinjiang University, Urumqi 830046, China
*
Author to whom correspondence should be addressed.
This article belongs to the Special Issue Advances and Applications in Graph Theory

Abstract

Let G be a graph and let S , T ⊆ V ( G ) be two disjoint vertex subsets with S = T = k . A many-to-many k-disjoint path cover of G is a set of k vertex-disjoint paths, each joining a source in S to a sink in T, whose union covers all vertices of G. The cover is paired if each vertex of S must be joined to a prescribed vertex of T, and unpaired otherwise. Let U G n be an n-dimensional Cayley graph generated by a unicyclic graph with bipartition V 0 and V 1 . Let S ⊆ V 0 and T ⊆ V 1 be two vertex subsets with S = T = 2 . In this paper, we prove that U G n admits an unpaired many-to-many two-disjoint path cover with at most n − 3 faulty edges for n ≥ 3 .

1. Introduction

The topology of an interconnection network is a critical factor in determining the communication efficiency and fault tolerance of parallel and distributed systems [1]. Graph-theoretic models have become fundamental tools for analyzing and comparing network topologies. Typically, an interconnection network is modeled as a graph in which vertices represent processors, memory modules, or switches, and edges represent communication links [2,3,4].
Path embedding is a fundamental problem in network communication and has been extensively studied [5,6,7]. A Hamiltonian path is a spanning path and is a special type of path embedding [8,9]. However, many applications require embedding two or more paths in networks, which gives rise to the many-to-many k-disjoint path cover problem [10,11,12,13,14,15]. Given a graph G, let S = { s 1 , … , s k } and T = { t 1 , … , t k } be two disjoint vertex subsets of equal size k in G. A many-to-many k-disjoint path cover joining S and T is a set of k vertex-disjoint paths whose union covers all vertices of G, with each path connecting one source in S to one sink in T. The cover is paired if each vertex of S must be joined to a designated vertex of T, or unpaired otherwise. A graph G is said to be f-edge fault-tolerant k-disjoint path coverable if, after removing at most f edges, the remaining graph still admits such a cover [16,17,18,19,20].
Among various network topologies, Cayley graphs have received extensive attention owing to their desirable properties, including high symmetry, regularity and vertex-transitivity [2,4,8,21]. Let G be a group with identity element e, and let B be a subset of G \ { e } with B − = B . The Cayley graph C a y ( G , B ) has vertex set G and edge set { u v ∣ u , v ∈ G and v = u s for some s ∈ B } . In this paper, the set { 1 , 2 , 3 , … , n } is denoted by ⟨ n ⟩ , and the symmetric group on ⟨ n ⟩ is denoted by S n . For convenience, the transposition ( i j ) is a permutation that exchanges the elements in positions i and j. The transposition generating graph  G ( B ) is a graph with vertex set ⟨ n ⟩ and edge set { u v ∣ ( u v ) ∈ B } . If G ( B ) is a tree, we call C a y ( S n , B ) a Cayley graph generated by a transposition tree, denoted by Γ n . The properties of Γ n , including connectivity, Hamiltonian laceability under edge faults, and many-to-many two-disjoint path covers, have been extensively studied [22,23,24].
If G ( B ) is a unicyclic graph, we denote it by T c . Then C a y ( S n , B ) is the Cayley graph generated by a unicyclic graph, denoted by U G n . In particular, if T c is a cycle, then C a y ( S n , B ) is called the modified bubble-sort graph, denoted by M B n . The properties of graphs U G n have been extensively studied, such as conditional connectivity [25,26] and disjoint path covers [22,27]. In particular, Qiao et al. [22] investigated the many-to-many two-disjoint path covers of Γ n with edge faults and proved that Γ n is ( n − 4 ) -edge fault-tolerant many-to-many two-disjoint path coverable for n ≥ 4 . As a corollary, they derived that M B n is ( n − 3 ) -edge fault-tolerant many-to-many two-disjoint path coverable. Meanwhile, Ref. [27] shows that U G n admits the paired many-to-many two-disjoint path covers under fault-free conditions.
This paper investigates the edge fault-tolerant unpaired many-to-many two-disjoint path covers of U G n , and shows that for n ≥ 3 , U G n admits such covers in the presence of at most n − 3 faulty edges. Since disjoint path covers serve as a fundamental model for parallel routing, our study on the edge fault-tolerant unpaired two-disjoint path cover of U G n goes beyond a theoretical extension, providing valuable guidance for designing robust and scalable interconnection networks.

2. Preliminaries

For standard definitions not given here, we follow the terminology of Book [28]. For a path P = ⟨ v 1 , v 2 , … , v i , … , v j , … , v n ⟩ in G, we use P [ v i , v j ] to denote the subpath of P from v i to v j . In this paper, we write u ∈ G for u ∈ V ( G ) , and e ∈ G for e ∈ E ( G ) . A path in a graph G that visits every vertex of G exactly once is called a Hamiltonian path. A bipartite graph G is Hamiltonian laceable if it has a Hamiltonian path between any two vertices from distinct partite sets of G.
Proposition 1
([29]). Let G ( B i ) be a transposition graph corresponding to the transposition generating set B i for i = 1 , 2 . Then C a y ( S n , B 1 ) ≅ C a y ( S n , B 2 ) if and only if G ( B 1 ) ≅ G ( B 2 ) .
The following result in [22] indicates the fault tolerance of many-to-many unpaired two-disjoint path covers in Cayley graphs generated by transposition trees  Γ n .
Lemma 1
([22]). Γ n is f-edge fault-tolerant unpaired two-disjoint path coverable with f ≤ n − 4 and n ≥ 4 .
Lemma 2
([30]). Let T be a labeled tree of order n, and let e be an edge of T. The two components of T − e are denoted by T 1 and T 2 , where T 1 = r . Furthermore, let B, B − , B 1 and B 2 be transposition sets on ⟨ n ⟩ satisfying G ( B ) = T , G ( B − ) = T − e , G ( B 1 ) = T 1 and G ( B 2 ) = T 2 . Then, C a y ( S n , B − ) has n r components and each component is isomorphic to C a y ( S r , B 1 ) □ C a y ( S n − r , B 2 ) .
It has been shown that C a y ( S n , B ) is connected if and only if its transposition generating graph G ( B ) is connected [29]. Thus, U G n cannot be disconnected by removing an edge of the cycle in T c , but it can be disconnected by removing a cut edge of T c if T c has one. Let j be a leaf vertex of T c , and let e be the edge incident to j in T c . Then T c has two components T c 1 = G ( B 1 ) and T c 2 = G ( B 2 ) , such that G ( B 1 ) = j and G ( B 2 ) = T c − e . By Lemma 2, we can decompose U G n into n components by removing the edge e, with each component isomorphic to C a y ( S 1 , j ) □ C a y ( S n − 1 , B 2 ) , that is, U G n − 1 .
In this paper, we take a leaf vertex of T c to be n if T c has leaf vertices, and the subgraph of U G n induced by the vertex set { x : x ( n ) = k } is denoted by U G n ( k ) , where x ( n ) means the n-th position of a permutation x. The decomposition of U G n induced by the leaf vertex n yields n copies of U G n − 1 , denoted by U G n ( 1 ) , U G n ( 2 ) , … , U G n ( n ) . Then, for k ∈ ⟨ n − 1 ⟩ , x ′ = x ( k n ) is said to be the outgoing neighbor of x, and x x ′ is called an outgoing edge corresponding to x.
Proposition 2
([31]). There are ( n − 2 ) ! independent edges, denoted by E i , j , between Γ n ( i ) and Γ n ( j ) , where i , j are two distinct integers in ⟨ n ⟩ . Moreover, among all vertices in V ( E i , j ) , there are ( n − 2 ) ! 2 white vertices and ( n − 2 ) ! 2 black vertices in Γ n ( i ) .
Since the transposition generating graph C n of M B n has no leaf vertices and every vertex on C n is incident with exactly two edges, there are 2 ( n − 2 ) ! independent edges between M B n ( i ) and M B n ( j ) by Proposition 2. Moreover, if the transposition generating graph T c of U G n is a unicyclic graph other than a cycle, it contains leaf vertices. This structural difference breaks the bidirectional symmetry, resulting in only half the number of independent edges between U G n ( i ) and U G n ( j ) , which is exactly ( n − 2 ) ! .
Lemma 3
([32]). Let x and y be two distinct vertices in Γ n which comes from different bipartite sets. Suppose that e is an edge in Γ n with e ≠ x y . Then Γ n with n ≥ 4 admits a Hamiltonian path joining x and y passing through the edge e.
Corollary 1.
Let x and y be two distinct vertices in U G n which comes from different bipartite sets. Suppose that e is an edge in U G n with e ≠ x y . Then U G n with n ≥ 3 admits a Hamiltonian path joining x and y passing through the edge e.
Proof. 
For n = 3 , U G 3 ≅ K 3 , 3 . Let e = u v be an edge such that u and x belong to the same partite set, v and y belong to the same partite set, and e ≠ x y . Let the remaining two vertices in U G 3 be w and b, so that the bipartition of U G 3 is { x , u , w } and { y , v , b } . Then a Hamiltonian path P = ⟨ x , b , u , v , w , y ⟩ can be constructed. When n ≥ 4 , for any given edge e ∈ U G n , there exists a transposition i j ∈ E ( T c ) such that e ∈ E ( i j ) ∩ E ( U G n ) . Since the unique cycle in T c has length at least 3, there exists an edge k l ≠ i j on this cycle. Thus, U G n − E ( k l ) ≅ Γ n and the edge e ∈ U G n − E ( k l ) . By Lemma 3, there exists a Hamiltonian path in U G n − E ( k l ) joining x and y passing through the edge e. Because U G n − E ( k l ) is a spanning subgraph of U G n , the Hamiltonian path also exists in U G n . □
Lemma 4
([33]). U G n is ( n − 2 ) -edge fault-tolerant Hamiltonian laceable with n ≥ 3 .
Lemma 5.
For n ≥ 5 and I ⊆ ⟨ n ⟩ , j 1 , j 2 ∈ I with j 1 ≠ j 2 . Set u ∈ U G n ( j 1 ) and v ∈ U G n ( j 2 ) so that they come from different partite sets. Let F be a set of faulty edges in U G n with F ≤ n − 3 . Then U G n ( I ) − F has a Hamiltonian path joining u and v.
Proof. 
Without loss of generality, we can assume that u is a white vertex and v is a black vertex. Let | I | = k and p 1 p 2 ⋯ p k be a permutation on I such that p 1 = j 1 and p k = j 2 . Since ( n − 2 ) ! 2 > n − 3 for n ≥ 5 , for each i ∈ ⟨ k − 1 ⟩ , there exists at least one fault-free edge v i v i ′ ∈ E p i , p i + 1 such that v i ∈ U G n ( p i ) is a black vertex and v i ′ ∈ U G n ( p i + 1 ) is a white vertex. By Lemma 4, there exist a Hamiltonian path P 1 from u to v 1 in U G n ( p 1 ) − F , Hamiltonian paths P i + 1 from v i ′ to v i + 1 in U G n ( p i + 1 ) − F for every i ∈ ⟨ k − 2 ⟩ , and a Hamiltonian path P k from v k − 1 ′ to v in U G n ( p k ) − F . Then a Hamiltonian path is constructed as P = ⟨ u , P 1 , v 1 , v 1 ′ , P 2 , v 2 , … , v i ′ , P i + 1 , v i + 1 , … , v k − 2 ′ , P k − 1 , v k − 1 , v k − 1 ′ , P k , v ⟩ . □
Note that Lemma 5 is derived for n ≥ 5 , as the inequalities used in its proof only hold under this condition. To establish the main result for all n ≥ 3 , we handle the base cases n = 3 and n = 4 separately, which correspond to Lemma 7 and Lemma 9, respectively. For the remaining cases where n ≥ 5 , we proceed with the inductive step using Lemma 5.

3. Fault-Tolerant Unpaired Many-to-Many Two-Disjoint Path Cover of  UG n

Let S = { s 1 , s 2 } and T = { t 1 , t 2 } be two subsets of U G n , where s i is a white vertex and t i is a black vertex for i = 1 , 2 . We will construct an unpaired two-disjoint path cover { P 1 , P 2 } between S and T in U G n . Before proceeding to the main results, recall a result on the unpaired many-to-many two-disjoint path cover of the modified bubble-sort graph as follows.
Lemma 6
([22]). M B n is f-edge fault-tolerant unpaired two-disjoint path coverable with f ≤ n − 3 and n ≥ 3 .
If the transposition generating graph T c of U G n is a cycle, then U G n is the modified bubble-sort graph M B n . Lemma 6 shows that M B n admits an unpaired two-disjoint path cover under n − 3 edge faults. Therefore, in what follows, we consider only unicyclic graphs T c other than cycles. If T c contains a pendant path of length at least 2, let n be its leaf with unique neighbor n − 1 , and let n − 2 be the other neighbor of n − 1 on this path. Otherwise, take an arbitrary leaf n with unique neighbor n − 1 , and let the two neighbors of n − 1 on the cycle be n − 2 and n − 3 .
Lemma 7.
U G 3 is unpaired two-disjoint path coverable.
Proof. 
Let { s 1 , s 2 , s 3 } and { t 1 , t 2 , t 3 } be the bipartition of U G 3 ≅ K 3 , 3 . Let P 1 = s 1 t 1 . Then U G 3 − V ( P 1 ) ≅ K 2 , 2 . Since K 2 , 2 is Hamiltonian laceable, there exists a Hamiltonian path P 2 in U G 3 − V ( P 1 ) joining any two vertices in different partite sets. Hence, P 1 and P 2 are vertex-disjoint and cover V ( U G 3 ) ; thus, { P 1 , P 2 } forms an unpaired two-disjoint path cover of U G 3 . □
In the following, for each transposition ( i j ) ∈ B , let E ( i j ) = { ( x , x ( i j ) ) : x ∈ S n } denote the set of edges of U G n generated by ( i j ) . For distinct j 1 , j 2 ∈ ⟨ n ⟩ , let F c ( j 1 , j 2 ) denote the set of faulty edges between U G n ( j 1 ) and U G n ( j 2 ) , and set f c ( j 1 , j 2 ) = F c ( j 1 , j 2 ) .
Lemma 8.
Let the transposition generating graph of U G 4 be a 3-cycle C 3 with a pendant edge. Let S ⊂ ⟨ 4 ⟩ with | S | = 3 , and F c be the set of faulty cross-edges in U G 4 with | F c | = 1 . Then, for any two distinct vertices x and y in different bipartite sets of U G 4 ( S ) such that at least one of their outgoing neighbors x ′ and y ′ is not in U G 4 ( S ) , there exists a Hamiltonian path P from x to y in U G 4 ( S ) − F c .
Proof. 
Let x be a black vertex and y be a white vertex; assume the outgoing neighbor x ′ is not in U G 4 ( S ) . Without loss of generality, assume that S = { p 1 , p 2 , p 3 } , where { p 1 , p 2 , p 3 } ⊂ ⟨ 4 ⟩ .
Case 1. x and y belong to the same component of U G 4 ( S ) , without loss of generality; assume that x , y ∈ U G 4 ( p 1 ) .
Case 1.1.  f c ( p 1 , p i ) = 0 for i ∈ { 2 , 3 } . If one of the two edges between U G 4 ( p 2 ) and U G 4 ( p 3 ) is faulty, assume that the other fault-free edge connects a white vertex w in U G 4 ( p 2 ) to a black vertex b in U G 4 ( p 3 ) . Let u v ∈ ( E ( 13 ) ∪ E ( 23 ) ) ∩ E ( U G 4 ( p 1 ) ) be an edge such that u ′ is a black vertex in U G 4 ( p 2 ) , and v ′ is a white vertex in U G 4 ( p 3 ) . Since edges u u ′ and v v ′ are outgoing edges between U G 4 ( p 1 ) and U G 4 ( p i ) for i ∈ { 2 , 3 } , u u ′ and v v ′ are fault-free. Given that x ′ is not in U G 4 ( S ) , the edge u v ≠ x y . By Corollary 1, U G 4 ( p 1 ) has a Hamiltonian path H 1 from x to y through u v . By Lemma 4, there exist Hamiltonian paths H 2 from u ′ to w in U G 4 ( p 2 ) and H 3 from b to v ′ in U G 4 ( p 3 ) . Let P = ⟨ x , H 1 [ x , u ] , u , u ′ , H 2 , w , b , H 3 , v ′ , v , H 1 [ v , y ] , y ⟩ . Thus, P is a Hamiltonian path of U G 4 − F c .
Case 1.2.  f c ( p 2 , p i ) = 0 for i ∈ { 1 , 3 } (or f c ( p 3 , p i ) = 0 for i ∈ { 1 , 2 } ). There exist edges u v ∈ E ( 12 ) ∩ E ( U G 4 ( p 1 ) ) and w b ∈ E ( 12 ) ∩ E ( U G 4 ( p 2 ) ) such that u ′ , v ′ ∈ U G 4 ( p 2 ) , w ′ , b ′ ∈ U G 4 ( p 3 ) , and u ′ v ′ ≠ w b . Since f c ( p 2 , p i ) = 0 for i ∈ { 1 , 3 } , the edges u u ′ , v v ′ , w w ′ and b b ′ are fault-free. Given that x ′ is not in U G 4 ( S ) , the edge u v ≠ x y . By Corollary 1, there exist Hamiltonian paths H 1 in U G 4 ( p 1 ) from x to y containing u v and H 2 in U G 4 ( p 2 ) from u ′ to v ′ containing w b . By Lemma 4, there exists a Hamiltonian path H 3 from w ′ to b ′ in U G 4 ( p 3 ) . Let P = ⟨ x , H 1 [ x , u ] , u , u ′ , H 2 [ u ′ , w ] , w , w ′ , H 3 , b ′ , b , H 2 [ b , v ′ ] , v ′ , v , H 1 [ v , y ] , y ⟩ . Then P is the Hamiltonian path of U G 4 − F c .
Case 2. x and y belong to two different components of U G 4 ( S ) , without loss of generality; assume that x ∈ U G 4 ( p 1 ) and y ∈ U G 4 ( p 2 ) .
Case 2.1.  f c ( p 1 , p i ) = 0 for i ∈ { 2 , 3 } (or f c ( p 2 , p i ) = 0 for i ∈ { 1 , 3 } ). Let u v ∈ E ( 12 ) be an edge in U G 4 ( p 1 ) and z = ∗ ∗ p 3 p 1 be a white vertex. Suppose that u ′ , v ′ ∈ U G 4 ( p 3 ) , z ′ ∈ U G 4 ( p 2 ) , since f c ( p 1 , p i ) = 0 for i ∈ { 2 , 3 } , the edges u u ′ , v v ′ , and y y ′ are fault-free. Furthermore, because x ′ is not in U G 4 ( S ) and z ′ ∈ U G 4 ( p 2 ) , it follows that x z ≠ u v . By Corollary 1, there is a Hamiltonian path H 1 in U G 4 ( p 1 ) from x to z through u v ; by Lemma 4, there exist Hamiltonian paths H 2 from z ′ to y in U G 4 ( p 2 ) and H 3 from u ′ to v ′ in U G 4 ( p 3 ) . Let P = ⟨ x , H 1 [ x , u ] , u , u ′ , H 3 , v ′ , v , H 1 [ v , z ] , z , z ′ , H 2 , y ⟩ . Thus, P is the desired path of U G 4 − F c .
Case 2.2.  f c ( p 3 , p i ) = 0 for i ∈ { 1 , 2 } . Let u be a white vertex in U G 4 ( p 1 ) and v be a black vertex in U G 4 ( p 2 ) , such that u ′ , v ′ ∈ U G 4 ( p 3 ) . Since f c ( p 3 , p i ) = 0 for i ∈ { 1 , 2 } , the edges u u ′ and v v ′ are fault-free. By Lemma 4, there exist Hamiltonian paths H 1 from x to u in U G 4 ( p 1 ) , H 2 from u ′ to v ′ in U G 4 ( p 3 ) , and H 3 from v to y in U G 4 ( p 2 ) . Let P = ⟨ x , H 1 , u , u ′ , H 2 , v ′ , v , H 3 , y ⟩ . Then P is the desired path of U G 4 − F c . □
Since U G n is a Cayley graph generated by the symmetric group, its underlying symmetric group structure inherently guarantees that U G n is vertex-transitive. Specifically, this vertex-transitive property allows us to utilize graph automorphisms to fix a target vertex, say s 1 , in U G n ( n ) . On this basis, we only need to discuss the positional distribution of s 2 , t 1 , and t 2 . This significantly reduces the exhaustive case analysis in the subsequent path constructions.
Lemma 9.
U G 4 is many-to-many unpaired 1-edge fault-tolerant two-disjoint path coverable.
Proof. 
Up to isomorphism, U G 4 has just two types: one whose transposition generating graph is the 4-cycle C 4 , the other whose transposition generating graph is the 3-cycle C 3 with a pendant edge, as shown in Figure 1b,c. For the former, that is, M B 4 , the result follows from Lemma 7.
Let edge e be the unique faulty edge of U G 4 . We now consider U G 4 whose transposition generating graph is a 3-cycle C 3 with a pendant edge. Its edges fall into four types: E ( 12 ) , E ( 23 ) , E ( 13 ) and E ( 34 ) . If e ∈ { E ( 12 ) , E ( 23 ) , E ( 13 ) } , then without loss of generality, assume that e ∈ E ( 12 ) . By Lemma 1, U G 4 − E ( 12 ) is fault-free and isomorphic to Γ 4 . Therefore, U G 4 − e is unpaired two-disjoint path coverable. If e ∈ E ( 34 ) , we decompose U G 4 into four copies of U G 3 along the leaf vertex 4, denoted by U G 4 ( p 1 ) , U G 4 ( p 2 ) , U G 4 ( p 3 ) , and U G 4 ( p 4 ) . Each copy is isomorphic to the complete bipartite graph K 3 , 3 and contains no faulty edges, where p 1 p 2 p 3 p 4 is a permutation on ⟨ 4 ⟩ . There are two independent edges between any two different copies. Since U G 4 is vertex transitive, we may fix s 1 ∈ U G 4 ( p 1 ) , yielding four subcases as follows.
Case 1.  s 1 , s 2 , t 1 , t 2 ∈ U G 4 ( p 1 ) . In this case, by Lemma 7, since U G 4 ( p 1 ) is fault-free, there exists an unpaired two-disjoint path cover { P 1 , P 2 } . Without loss of generality, assume P 1 joins s 1 to t 1 and P 2 joins s 2 to t 2 , where | V ( P 1 ) | = 2 . Let x be the neighbor of s 2 on P 2 . Since there is only one faulty cross edge, at least one of s 1 t 1 and s 2 x has fault-free outgoing edges; assume it is s 2 x . And let S = { p 2 , p 3 , p 4 } , then s 2 , x ∉ U G 4 ( S ) and s 2 ′ , x ′ ∈ U G 4 ( S ) . By Lemma 8, there exists a Hamiltonian path P from s 2 ′ to x in U G 4 ( S ) . Let P 1 = ⟨ s 1 , t 1 ⟩ and P 2 ′ = ⟨ s 2 , s 2 ′ , P , x , P 2 [ x , t 2 ] , t 2 ⟩ . Thus, { P 1 , P 2 ′ } forms a many-to-many unpaired two-disjoint path cover of U G 4 − e .
Case 2.  s 1 , s 2 , t 1 , t 2 lie in two copies of U G 3 , and denote U G 4 ( p 1 ) and U G 4 ( p 2 ) .
Case 2.1.  s 1 , s 2 , t 1 ∈ U G 4 ( p 1 ) , t 2 ∈ U G 4 ( p 2 ) . Since only one outgoing edge is faulty, there exists at least one black vertex x ≠ t 1 in U G 4 ( p 1 ) such that x x ′ is fault-free. By Lemma 7, there is an unpaired two-disjoint path cover { P 1 , P 2 } with P 1 joining s 1 to t 1 and P 2 joining s 2 to x. And let S = { p 2 , p 3 , p 4 } , then x ∉ U G n ( S ) and x ′ , t 2 ∈ U G 4 ( S ) . By Lemma 8, there exists a Hamiltonian path P from x ′ to t 2 in U G 4 ( S ) . Let P 1 = ⟨ s 1 , t 1 ⟩ and P 2 ′ = ⟨ s 2 , P 2 , x , x ′ , P , t 2 ⟩ . Thus, { P 1 , P 2 ′ } forms the desired two-disjoint path cover of U G 4 − e .
Case 2.2.  s 1 , t 1 ∈ U G 4 ( p 1 ) , s 2 , t 2 ∈ U G 4 ( p 2 ) . Then at least one of U G 4 ( p 1 ) and U G 4 ( p 2 ) has no faulty edges to either U G 4 ( p 3 ) or U G 4 ( p 4 ) ; assume it is U G 4 ( p 1 ) . Without loss of generality, assume that the fault-free edge between U G 4 ( p 3 ) and U G 4 ( p 4 ) connects a white vertex w ∈ U G 4 ( p 3 ) to a black vertex b ∈ U G 4 ( p 4 ) . Then in U G 4 ( p 1 ) , there is an edge u v ∈ E ( 13 ) ∪ E ( 23 ) such that u ′ ∈ U G 4 ( p 3 ) is black and v ′ ∈ U G 4 ( p 4 ) is white. Since U G 4 ( p 1 ) has no faulty edges to either U G 4 ( p 3 ) or U G 4 ( p 4 ) , the edges u u ′ ∈ E ( p 1 , p 3 ) , v v ′ ∈ E ( p 1 , p 4 ) are fault-free. By Corollary 1, there is a Hamiltonian path P 1 in U G 4 ( p 1 ) from s 1 to t 1 through u v . By Lemma 4, there exist Hamiltonian paths P 2 from s 2 to t 2 in U G 4 ( p 2 ) , H 1 from u ′ to w in U G 4 ( p 3 ) , and H 2 from v ′ to b in U G 4 ( p 4 ) . Let P 1 ′ = ⟨ s 1 , P 1 [ s 1 , u ] , u , u ′ , H 1 , w , b , H 2 , v ′ , v , P 1 [ v , t 1 ] , t 1 ⟩ and P 2 = ⟨ s 2 , t 2 ⟩ . Then, { P 1 ′ , P 2 } forms the desired two-disjoint path cover of U G 4 − e .
Case 2.3.  s 1 , s 2 ∈ U G 4 ( p 1 ) , t 1 , t 2 ∈ U G 4 ( p 2 ) .
Case 2.3.1.  ∑ i = 2 4 f c ( p 1 , p i ) = 1 . Relabel the four copies as U G 4 ( j 1 ) , U G 4 ( j 2 ) , U G 4 ( j 3 ) , U G 4 ( j 4 ) such that p 1 = j 1 , p 2 = j 4 , and there is no faulty cross edge between U G 4 ( j k ) and U G 4 ( j k + 1 ) for k ∈ ⟨ 3 ⟩ . For each k ∈ ⟨ 3 ⟩ , there exist fault-free edges u k u k ′ and v k v k ′ between U G 4 ( j k ) and U G 4 ( j k + 1 ) with u k , v k ∈ U G 4 ( j k ) black and u k ′ , v k ′ ∈ U G 4 ( j k + 1 ) white. By Lemma 7, there exist unpaired two-disjoint path covers { L i , R i } in U G 4 ( j i ) for i = 1 , 2 , 3 , 4 such that L 1 joins s 1 to u 1 , R 1 joins s 2 to v 1 ; for k = 2 , 3 , L k joins u k − 1 ′ to u k and R k joins v k − 1 ′ to v k ; and L 4 joins u 3 ′ to t 1 , R 4 joins v 3 ′ to t 2 . Let P 1 = ⟨ s 1 , L 1 , u 1 , u 1 ′ , … , u 3 ′ , L 4 , t 1 ⟩ and P 2 = ⟨ s 2 , R 1 , v 1 , v 1 ′ , … , v 3 ′ , R 4 , t 2 ⟩ . Then, { P 1 , P 2 } forms the desired two-disjoint path cover of U G 4 − e . If ∑ i ∈ { 1 , 3 , 4 } f c ( p 2 , p i ) = 1 , we can also use j 1 j 2 j 3 j 4 to relabel the four copies such that f c ( j k , j k + 1 ) = 0 for k ∈ ⟨ 3 ⟩ , then the desired path cover can be constructed.
Case 2.3.2.  f c ( p 3 , p 4 ) = 1 . Without loss of generality, assume that the fault-free edge between the two copies connects a white vertex w in U G 4 ( p 3 ) to a black vertex b in U G 4 ( p 4 ) . Let x , y ∈ U G 4 ( p 1 ) and u ∈ U G 4 ( p 3 ) be black vertices, where x ′ ∈ U G 4 ( p 4 ) and y ′ , u ′ ∈ U G 4 ( p 2 ) . Since f c ( p 3 , p 4 ) = 1 , the edges x x ′ , y y ′ , u u ′ are fault-free. By Lemma 7, there is an unpaired two-disjoint path cover { P 1 , P 2 } in U G 4 ( p 1 ) with P 1 joining s 1 to x and P 2 joining s 2 to y, and an unpaired two-disjoint path cover { H 1 , H 2 } in U G 4 ( p 2 ) with H 1 joining u ′ to t 1 and H 2 joining y ′ to t 2 . By Lemma 4, there exist Hamiltonian paths H 3 from x ′ to b in U G 4 ( p 4 ) and H 4 from w to u in U G 4 ( p 3 ) . Let P 1 ′ = ⟨ s 1 , P 1 , x , x ′ , H 3 , b , w , H 4 , u , u ′ , H 1 , t 1 ⟩ and P 2 ′ = ⟨ s 2 , P 2 , y , y ′ , H 2 , t 2 ⟩ . Then, { P 1 ′ , P 2 ′ } forms the desired unpaired two-disjoint path cover of U G 4 − e .
Case 2.4.  s 1 , t 1 , t 2 ∈ U G 4 ( p 1 ) , s 2 ∈ U G 4 ( p 2 ) . This case is similar to Case 2.1, and the detailed proofs are omitted.
Case 3.  s 1 , s 2 , t 1 , t 2 lie in three copies of U G 3 .
Case 3.1.  s 1 , s 2 ∈ U G 4 ( p 1 ) , t 1 ∈ U G 4 ( p 2 ) and t 2 ∈ U G 4 ( p 3 ) .
Case 3.1.1.  ∑ i = 2 4 f c ( p 1 , p i ) = 1 . Since there is only one faulty outgoing edge and U G 4 ( p 1 ) contains three black vertices, there exist at least two black vertices x and y such that x x ′ and y y ′ are fault-free. If x ′ ∈ U G 4 ( p 2 ) and y ′ ∈ U G 4 ( p 3 ) (or x ′ ∈ U G 4 ( p 3 ) and y ′ ∈ U G 4 ( p 2 ) ), then in U G 4 ( p 2 ) , there is an edge u v ∈ E ( 12 ) with u ′ , v ′ ∈ U G 4 ( p 4 ) and u v ≠ x ′ t 1 , such that the edges u u ′ , v v ′ are fault-free. By Lemma 7, there exists an unpaired two-disjoint path cover { P 1 , P 2 } in U G 4 ( p 1 ) with P 1 joining s 1 to x and P 2 joining s 2 to y. By Corollary 1, there is a Hamiltonian path H 1 in U G 4 ( p 2 ) from x ′ to t 1 through u v . By Lemma 4, there exist Hamiltonian paths H 2 from y ′ to t 2 in U G 4 ( p 3 ) and H 3 from u ′ to v ′ in U G 4 ( p 4 ) . Let P 1 ′ = ⟨ s 1 , P 1 , x , x ′ , H 1 [ x ′ , u ] , u , u ′ , H 3 , v ′ , v , H 1 [ v , t 1 ] , t 1 ⟩ and P 2 ′ = ⟨ s 2 , P 2 , y , y ′ , H 2 , t 2 ⟩ . Then, { P 1 , P 2 } forms a many-to-many unpaired two-disjoint path cover of U G 4 − e .
If x ′ ∈ U G 4 ( p 2 ) and y ′ ∈ U G 4 ( p 4 ) , there exists a black vertex b ∈ U G 4 ( p 4 ) such that b ′ ∈ U G 4 ( p 3 ) and b b ′ is fault-free. The subsequent constructions are analogous to the preceding ones and are therefore omitted.
Case 3.1.2.  f c ( p 1 , p i ) = 0 for i = 2 , 3 , 4 . In this case, there exist two black vertices x and y such that x ′ ∈ U G 4 ( p 2 ) , y ′ ∈ U G 4 ( p 3 ) , and the edges x x ′ , y y ′ are fault-free. Let u v ∈ E ( 12 ) be an edge in either U G 4 ( p 2 ) or U G 4 ( p 3 ) , such that u ′ , v ′ ∈ U G 4 ( p 4 ) . There are two such edges, for at least one of which u u ′ and v v ′ are fault-free. Without loss of generality, assume it lies in U G 4 ( p 2 ) . The rest is similar to Case 3.1.1.
Case 3.2.  s 1 , t 1 ∈ U G 4 ( p 1 ) , s 2 ∈ U G 4 ( p 2 ) and t 2 ∈ U G 4 ( p 3 ) . At least one of U G 4 ( p 2 ) , U G 4 ( p 3 ) , and U G 4 ( p 4 ) is connected to the other two copies by fault-free edges. And let S = { p 2 , p 3 , p 4 } , then s 2 , t 2 ∈ U G 4 ( S ) . Since we aim to construct a Hamiltonian path covering U G 4 ( S ) with endpoints s 2 and t 2 , we do not need to consider the distribution of their outgoing neighbors in this case. By following a construction similar to that in Case 2 of Lemma 8, we obtain such a path P. Let P 1 = ⟨ s 1 , t 1 ⟩ and P 2 = ⟨ s 2 , P , t 2 ⟩ . Thus, { P 1 , P 2 } forms the desired two-disjoint path cover of U G 4 − e .
Case 3.3.  s 1 ∈ U G 4 ( p 1 ) , t 1 , t 2 ∈ U G 4 ( p 2 ) and s 2 ∈ U G 4 ( p 3 ) . The detailed proof is omitted since it is similar to Case 3.1.
Case 3.4.  s 1 ∈ U G 4 ( p 1 ) , s 2 , t 2 ∈ U G 4 ( p 2 ) and t 1 ∈ U G 4 ( p 3 ) . The detailed proof is omitted since it is similar to Case 3.2.
Case 4.  s 1 , s 2 , t 1 , t 2 lie in four distinct copies of U G 3 . Assume that s 1 ∈ U G 4 ( p 1 ) , t 1 ∈ U G 4 ( p 2 ) , s 2 ∈ U G 4 ( p 3 ) and t 2 ∈ U G 4 ( p 4 ) . Since there only exists exactly one faulty cross edge, without loss of generality, assume that f c ( p 1 , p 2 ) = 0 and f c ( p 3 , p 4 ) = 0 . There exist two black vertices x ∈ U G 4 ( p 1 ) and y ∈ U G 4 ( p 3 ) such that x ′ ∈ U G 4 ( p 2 ) , y ′ ∈ U G 4 ( p 4 ) , and the edges x x ′ , y y ′ are fault-free. By Lemma 4, there exist Hamiltonian paths H 1 from s 1 to x in U G 4 ( p 1 ) , H 2 from x ′ to t 1 in U G 4 ( p 2 ) , H 3 from s 2 to y in U G 4 ( p 3 ) , and H 4 from y ′ to t 2 in U G 4 ( p 4 ) . Let P 1 = ⟨ s 1 , H 1 , x , x ′ , H 2 , t 1 ⟩ and P 2 = ⟨ s 2 , H 3 , y , y ′ , H 4 , t 2 ⟩ and { P 1 , P 2 } is the desired two-disjoint path cover. □
Figure 1. Illustrations (a) U G 4 whose transposition generating graph is C 4 ; (b) U G 4 whose transposition generating graph is C 3 with a pendant edge. (White/black dots represent even/odd transpositions, respectively; edges between vertices with identical labels are omitted.)
Theorem 1.
U G n is f-edge fault-tolerant unpaired two-disjoint path coverable with f ≤ n − 3 for n ≥ 3 .
Proof. 
The result follows from Lemma 6 when the transposition generating graph of U G n is an n-cycle. Thus, it remains to consider other unicyclic transposition generating graphs. For n = 3 , Theorem 1 follows from Lemma 7. Hence, it suffices to prove that f ≤ n − 3 for n ≥ 4 . By Lemma 9, the base case n = 4 holds. For n ≥ 5 and f ≤ n − 4 , assume that U G n − 1 is f-edge fault-tolerant unpaired two-disjoint path coverable. Decompose U G n into n copies of U G n − 1 , say U G n ( 1 ) , … , U G n ( n ) . Let F be the set of faulty edges in U G n with | F | = f = n − 3 , and let F i be the set of faulty edges in U G n ( i ) with | F i | = f i for i ∈ ⟨ n ⟩ . Let F c be the set of faulty edges between different copies with | F c | = f c . Let I = ⟨ n ⟩ \ { n } . Since U G n is vertex-transitive, we may fix s 1 = 123 … n ∈ U G n ( n ) . We consider the locations of { s 2 , t 1 , t 2 } in four cases.
Case 1.  s 1 , s 2 , t 1 , t 2 ∈ U G n ( n ) . The induction hypothesis yields an unpaired two-disjoint path cover { P 1 , P 2 } between S and T; we assume that path P 1 joins s 1 to t 1 and path P 2 joins s 2 to t 2 in U G n ( n ) .
Case 1.1.  f n ≤ n − 4 . According to the number of faulty edges between U G n ( k ) and U G n ( l ) , we consider two subcases.
Case 1.1.1.  f c ≤ n − 4 .
(a). If T c contains trees of length at least 2, then by Lemma 3, removing E ( n − 1 , n − 2 ) decomposes U G n ( n ) into n − 1 components isomorphic to U G n − 2 . The minimum number of edges in E ( n − 1 , n − 2 ) on P 1 ∪ P 2 needed to join all components is at least n − 3 . Since n − 3 > n − 4 , there exists at least one fault-free edge x y ∈ E ( n − 1 , n − 2 ) on P 1 or P 2 such that both x x ′ and y y ′ are fault-free. Moreover, x ′ and y ′ belong to different copies of U G n − 1 . Without loss of generality, assume x y is on P 2 . By Lemma 5, there exists a path P from x ′ to y ′ spanning all vertices of U G n ( I ) . Then construct a new path P 2 ′ = ⟨ s 2 , P 2 [ s 2 , x ] , x , x ′ , P , y ′ , y , P 2 [ y , t 2 ] , t 2 ⟩ . Thus, { P 1 , P 2 ′ } forms a path cover of U G n − F , as shown in Figure 2a.
(b). If T c contains only trees of length 1, then by Lemma 3, removing E ( n − 1 , n − 2 ) and E ( n − 1 , n − 3 ) decomposes U G n ( n ) into n − 1 components isomorphic to Γ n − 2 . At least n − 3 edges from E ( n − 1 , n − 2 ) ∪ E ( n − 1 , n − 3 ) on P 1 ∪ P 2 are required to join these components. Since n − 3 > n − 4 , there exists at least one fault-free edge x y ∈ E ( n − 1 , n − 2 ) ∪ E ( n − 1 , n − 3 ) on P 1 or P 2 such that both x x ′ and y y ′ are fault-free. The rest of the construction is analogous to that in Case 1.1.1(a) and is omitted here.
Case 1.1.2. Since f c = n − 3 , all edges inside each copy are fault-free. There must exist an edge x y on P 1 or P 2 such that x x ′ and y y ′ are fault-free. If no edge of P 1 or P 2 satisfies this requirement, then there are at least ( n − 1 ) ! − 2 2 faulty edges in F c . Since ( n − 1 ) ! − 2 2 > n − 3 for n ≥ 5 , this contradicts | F | = f = n − 3 . Assume x y lies on P 2 . If x ′ and y ′ are in different copies, this case is similar to Case 1.1.1. Otherwise, assume x ′ and y ′ lie in the same copy U G n ( k ) with k ∈ I . Let e = u v ∈ E ( n − 1 , n − 2 ) be an edge of U G n ( k ) with e ≠ x ′ y ′ , where u = ∗ ∗ ⋯ ∗ l m k and v = ∗ ∗ ⋯ ∗ m l k . Since ( n − 3 ) ( n − 2 ) ! 2 − 1 > n − 3 for n ≥ 5 , we can choose an edge e such that both u u ′ and v v ′ are fault-free. By Corollary 1, there exists a Hamiltonian path H in U G n ( k ) from x ′ to y ′ through the edge e. By Lemma 5, there exists a path P from u ′ to v ′ that spans U G n ( I \ { k } ) . Thus, { P 1 , P 2 ′ } forms a path cover of U G n − F , as shown in Figure 2b, where P 2 ′ = ⟨ s 2 , P 2 [ s 2 , x ] , x , x ′ , H [ x ′ , v ] , v , v ′ , P , u ′ , u , H [ u , y ′ ] , y ′ , y , P 2 [ y , t 2 ] , t 2 ⟩ .
Case 1.2.  f n = n − 3 . Let e = x y ∈ F n be a faulty edge. Applying the induction hypothesis to U G n ( n ) − e , we obtain an unpaired two-disjoint path cover { P 1 , P 2 } between S and T. If e ∉ P 1 ∪ P 2 , it is similar to Case 1.1.1. Otherwise, assume e ∈ P 2 . If x ′ and y ′ are in different copies, this case is also similar to Case 1.1.1. If they are in the same copy U G n ( k ) , it is similar to Case 1.1.2.
Figure 2. Illustrations for Case 1 of Theorem 1.
Case 2.  s 1 , s 2 , t 1 , t 2 lie in two copies of U G n − 1 , denoted by U G n ( n ) and U G n ( i ) , where i ∈ I . There are four subcases.
Case 2.1.  s 1 , s 2 , t 1 ∈ U G n ( n ) and t 2 ∈ U G n ( i ) .
Case 2.1.1. Let f n ≤ n − 4 . Let x = ∗ ∗ ⋯ ∗ k n be a black vertex in U G n ( n ) , where k ∉ { i , n } . Since ( n − 2 ) ( n − 2 ) ! 2 − 1 > n − 3 for n ≥ 5 , such that x ≠ t 1 and x x ′ is fault-free. The induction hypothesis yields an unpaired two-disjoint path cover { P 1 , P 2 } between { s 1 , s 2 } and { t 1 , x } in U G n ( n ) , where P 1 joins s 1 to t 1 and P 2 joins s 2 to x. Then, by Lemma 5, there exists a path P from x ′ to t 2 that spans all vertices of U G n ( I ) . Let P 2 ′ = ⟨ s 2 , P 2 , x , x ′ , P , t 2 ⟩ . Then { P 1 , P 2 ′ } forms the desired path cover of U G n − F , as shown in Figure 3a.
Case 2.1.2.  f n = n − 3 . Since there are no faulty edges between any two copies of U G n − 1 , we have f c = 0 . By Lemma 4, there exists a Hamiltonian path P from s 2 to t 1 in U G n ( n ) − F n . Since s 1 lies on P, let x be a neighbor of s 1 on P. If the outgoing neighbor x ′ of x does not belong to U G n ( i ) , then by Lemma 5, there exists a Hamiltonian path H from x ′ to t 2 in U G n ( I ) . Let P 1 = ⟨ s 1 , P [ s 1 , t 1 ] , t 1 ⟩ and P 2 = ⟨ s 2 , P [ s 2 , x ] , x , x ′ , H , t 2 ⟩ . Then { P 1 , P 2 } forms the desired path cover of U G n − F .
Now assume that x ′ is in U G n ( i ) . Since f c = 0 , there exists an edge e = u v ∈ E ( n − 1 , n − 2 ) in U G n ( i ) with e ≠ x ′ t 2 , such that the edges u u ′ , v v ′ are fault-free and u ′ , v ′ ∉ U G n ( n ) . By Corollary 1, there exists a Hamiltonian path H in U G n ( i ) from x ′ to t 2 through u v . By Lemma 5, there exists a path P * from u ′ to v ′ that spans all vertices of U G n ( I \ { i } ) . Let P 1 = ⟨ s 1 , P [ s 1 , t 1 ] , t 1 ⟩ and P 2 = ⟨ s 2 , P [ s 2 , x ] , x , x ′ , H [ x ′ , u ] , u , u ′ , P * , v ′ , v , H [ v , t 2 ] , t 2 ⟩ . Then { P 1 , P 2 } forms the desired path cover of U G n − F , as shown in Figure 3b.
Figure 3. Illustrations for Case 2.1 of Theorem 1.
Case 2.2.  s 1 , t 1 , t 2 ∈ U G n ( n ) and s 2 ∈ U G n ( i ) . It is similar to case 2.1.
Case 2.3.  s 1 , t 1 ∈ U G n ( n ) and s 2 , t 2 ∈ U G n ( i ) .
Case 2.3.1.  f n = 0 or f i = 0 . We assume w.l.o.g. that f n = 0 . By Lemma 4, there exists a Hamiltonian path P 2 in U G n ( i ) − F i from s 2 to t 2 . Let x y ∈ E ( n − 1 , n − 2 ) in U G n ( n ) with x y ≠ s 1 t 1 , where x = ∗ ∗ ⋯ ∗ l k n and y = ∗ ∗ ⋯ ∗ k l n for l , k ∈ ⟨ n ⟩ \ { i , n } . Since ( n − 3 ) ( n − 2 ) ! 2 − 1 > n − 3 for n ≥ 5 , the edges x x ′ and y y ′ are fault-free. By Corollary 1, there exists a Hamiltonian path H in U G n ( n ) from s 1 to t 1 through x y . By Lemma 5, there exists a path P from x ′ to y ′ that visits all vertices of U G n ( I \ { i } ) . Let P 1 = ⟨ s 1 , H [ s 1 , x ] , x , x ′ , P , y ′ , y , H [ y , t 1 ] , t 1 ⟩ . Then { P 1 , P 2 } forms the desired path cover of U G n − F , as shown in Figure 4a.
Case 2.3.2.  f n ≥ 1 and f i ≥ 1 . By Lemma 4, there exist a Hamiltonian path P 1 from s 1 to t 1 in U G n ( n ) − F n and a Hamiltonian path P 2 from s 2 to t 2 in U G n ( i ) − F i . Since f n ≥ 1 and f i ≥ 1 with f ≤ n − 3 , at least three copies of U G n − 1 are fault-free, denoted by U G n ( j 1 ) , U G n ( j 2 ) , and U G n ( j 3 ) . Let J = { j 1 , j 2 , j 3 } with j 1 , j 2 , j 3 ∉ { i , n } . Moreover, there are at most n − 5 faults between different copies, thus | F c | = f c ≤ n − 5 . Since 3 ( n − 2 ) ! − 2 2 > n − 3 for n ≥ 5 , there exists a black vertex x = ∗ ∗ ⋯ ∗ j n on P 1 with x ∉ { s 1 , t 1 } such that x x ′ fault-free, where j ∈ J . And if x lies on P 1 , let x 1 and x 2 be two neighbors of x on P 1 . Since the types of any two consecutive edges are different, at least one of x 1 and x 2 does not belong to U G n ( i ) . Denote this vertex by y with y ≠ { s 1 , t 1 } . Since 3 ( n − 2 ) ! − 2 2 − 2 > n − 3 for n ≥ 5 , both x x ′ and y y ′ are fault-free. If y ′ ∉ U G n ( j ) , then by Lemma 5, there exists a path P from x ′ to y ′ spanning all vertices of U G n ( I \ { i } ) . Let P 1 ′ = ⟨ s 1 , P 1 [ s 1 , y ] , y , y ′ , P , x ′ , x , P 1 [ x , t 1 ] , t 1 ⟩ . Then { P 1 ′ , P 2 } forms the desired path cover of U G n − F .
Otherwise, y ′ ∈ U G n ( j ) and all edges in U G n ( j ) are fault-free. Let u v be an edge in U G n ( j ) with u v ≠ x ′ y ′ , where u = ∗ ∗ ⋯ ∗ l k j and v = ∗ ∗ ⋯ ∗ k l j . Since ( n − 3 ) ( n − 4 ) ( n − 3 ) ! 2 − 1 > n − 5 for n ≥ 5 , we can choose an edge u v such that both u u ′ and v v ′ are fault-free. By Corollary 1, there exists a Hamiltonian path H in U G n ( j ) from x ′ to y ′ through the edge u v . By Lemma 5, there exists a path P from u ′ to v ′ spanning all vertices of U G n ( I \ { i , j } ) . Let P 1 ′ = ⟨ s 1 , P 1 [ s 1 , y ] , y , y ′ , H [ y ′ , v ] , v , v ′ , P , u ′ , u , H [ u , x ′ ] , x ′ , x , P 1 [ x , t 1 ] , t 1 ⟩ . Then { P 1 ′ , P 2 } forms the desired path cover of U G n − F , as shown in Figure 4b.
Figure 4. Illustrations for Case 2.3 of Theorem 1.
Case 2.4.  s 1 , s 2 ∈ U G n ( n ) and t 1 , t 2 ∈ U G n ( i ) .
Case 2.4.1.  f i ≤ n − 4 for every i ∈ ⟨ n ⟩ .
(a). f c ≤ n − 4 . Relabel the n copies as j 1 , j 2 , … , j n such that n = j 1 and i = j n . Since ( n − 2 ) ! 2 − 2 ≥ n − 4 for n ≥ 5 , Proposition 2 guarantees a pair of fault-free edges x j k x j k ′ and y j k y j k ′ between U G n ( j k ) and U G n ( j k + 1 ) for k = 1 , 2 , … , n − 1 , where x j k and y j k are two black vertices in U G n ( j k ) . The induction hypothesis yields an unpaired two-disjoint path cover { L 1 , R 1 } in U G n ( j 1 ) − F ( j 1 ) , where L 1 joins s 1 to x j 1 and R 1 joins s 2 to y j 1 , and an unpaired two-disjoint path cover { L k , R k } in U G n ( j k ) − F ( j k ) with L k joining x j k − 1 ′ to x j k and R k joining y j k − 1 ′ to y j k for k = 2 , 3 , … , n − 1 . Moreover, there exists a two-disjoint path cover { L n , R n } in U G n ( j n ) − F ( j n ) , where L n joins x j n − 1 ′ to t 1 and R n joins y j n − 1 ′ to t 2 . Let P 1 = ⟨ s 1 , L 1 , x j 1 , x j 1 ′ , L 2 , x j 2 , … , x j n − 2 ′ , L n − 1 , x j n − 1 , x j n − 1 ′ , L n , t 1 ⟩ and P 2 = ⟨ s 2 , R 1 , y j 1 , y j 1 ′ , R 2 , y j 2 , … , y j n − 2 ′ , R n − 1 , y j n − 1 , y j n − 1 ′ , R n , t 2 ⟩ . Then { P 1 , P 2 } forms the desired path cover of U G n − F , as shown in Figure 5a.
(b). f c = n − 3 . In this case, there are no faulty edges inside the copies, and the induction hypothesis holds in each copy. We relabel the n copies such that n = j 1 and i = j n . After relabeling, the number of faulty edges between U G n ( j k ) and U G n ( j k + 1 ) , denoted by f c ( j k , j k + 1 ) , satisfies f c ( j k , j k + 1 ) ≤ n − 4 for each k ∈ ⟨ n − 1 ⟩ . Since the subsequent construction of the two-disjoint path cover is analogous to that in Case 2.4.1(a), the details are omitted.
Case 2.4.2.  f j = n − 3 for some j ∈ ⟨ n ⟩ , we assume w.l.o.g. that f n = n − 3 . Since f c = 0 , there exists a black vertex x = ∗ ∗ ⋯ ∗ i n of U G n ( n ) , such that x x ′ is fault-free. By Lemma 4, there exists a Hamiltonian path H from s 1 to x. Since s 2 must lie on H, assume that the vertex y is a neighbor of s 2 on H such that the outgoing edge y y ′ is fault-free.
If y ′ ∉ U G n ( i ) , say y ′ ∈ U G n ( l ) with l ∉ { i , n } , then there exists a white vertex z = ∗ ∗ ⋯ ∗ k i in U G n ( i ) , where k ∉ { l , i , n } , such that z z ′ is fault-free. By the induction hypothesis, there exists an unpaired two-disjoint path cover { P 1 , P 2 } in U G n ( i ) between { z , x ′ } and { t 1 , t 2 } , where P 1 joins z to t 1 and P 2 joins x ′ to t 2 . Since y ′ and z ′ lie in different copies, by Lemma 5, there exists a path P from y ′ to z ′ that spans all vertices of U G n ( I \ { i } ) . Let P 1 ′ = ⟨ s 1 , H [ s 1 , y ] , y , y ′ , P , z ′ , z , P 1 , t 1 ⟩ and P 2 ′ = ⟨ s 2 , H [ s 2 , x ] , x , x ′ , P 2 , t 2 ⟩ . Then { P 1 ′ , P 2 ′ } forms the desired path cover of U G n − F , as shown in Figure 5b.
If y ′ ∈ U G n ( i ) , the induction hypothesis yields an unpaired two-disjoint path cover { P 1 , P 2 } in U G n ( i ) between { y ′ , x ′ } and { t 1 , t 2 } , where P 1 joins y ′ to t 1 and P 2 joins x ′ to t 2 . There exists an edge u v on P 1 ∪ P 2 such that u u ′ and v v ′ are fault-free, which holds since f c = 0 . To construct the path cover of U G n − F , we consider two cases: either u ′ and v ′ are in distinct copies, or both are in the same copy, U G n ( k ) . Since the argument is similar to that in Case 2.3.2, so the detailed proofs are omitted.
Figure 5. Illustrations for Case 2.4 of Theorem 1.
Case 3.  s 1 , s 2 , t 1 , t 2 lie in three copies of U G n − 1 , that is, U G n ( n ) , U G n ( j 1 ) and U G n ( j 2 ) , where j 1 , j 2 ∈ I with j 1 ≠ j 2 .
Case 3.1.  s 1 , s 2 ∈ U G n ( n ) , t 1 ∈ U G n ( j 1 ) and t 2 ∈ U G n ( j 2 ) .
Case 3.1.1.  f n ≤ n − 4 . Since ( n − 2 ) ! 2 > n − 3 for n ≥ 5 , there exists a black vertex x = ∗ ∗ ⋯ ∗ j 1 n in U G n ( n ) such that x x ′ is fault-free. Similarly, since ( n − 3 ) ( n − 2 ) ! 2 > n − 3 for n ≥ 5 , there exists a black vertex y = ∗ ∗ ⋯ ∗ k n in U G n ( n ) where k ∉ { j 1 , j 2 } , such that y y ′ is fault-free. The induction hypothesis yields an unpaired two-disjoint path cover { P 1 , P 2 } in U G n ( n ) between { s 1 , s 2 } and { x , y } , where P 1 joins s 1 to x and P 2 joins s 2 to y. By Lemma 4, there exists a Hamiltonian path H from x ′ to t 1 in U G n ( j 1 ) . By Lemma 5, there exists a path P from y ′ to t 2 that spans all vertices of U G n ( I \ { j 1 } ) . Let P 1 ′ = ⟨ s 1 , P 1 , x , x ′ , H , t 1 ⟩ and P 2 ′ = ⟨ s 2 , P 2 , y , y ′ , P , t 2 ⟩ . Then { P 1 ′ , P 2 ′ } forms the desired path cover of U G n − F .
Case 3.1.2.  f n = n − 3 . There exists a black vertex x = ∗ ∗ ⋯ ∗ j 1 n in U G n ( n ) such that x ′ ∈ U G n ( j 1 ) and x x ′ is fault-free. By Lemma 4, there is a Hamiltonian path H from x to s 2 . Since s 1 lies on H, let y be a neighbor of s 1 on H such that y is closer to s 2 on H. Based on the location of y ′ , we have three subcases.
If y ′ ∈ U G n ( k ) for k ∉ { j 1 , j 2 } , then by Lemma 5, there exists a path P from y ′ to t 2 spanning all vertices of U G n ( I \ { j 1 } ) . By Lemma 4, there is a Hamiltonian path L from x ′ to t 1 . Thus, the path cover { P 1 , P 2 } forms the desired path cover of U G n − F , where P 1 = ⟨ s 1 , H [ s 1 , x ] , x , x ′ , L , t 1 ⟩ and P 2 = ⟨ s 2 , H [ s 2 , y ] , y , y ′ , P , t 2 ⟩ .
If y ′ ∈ U G n ( j 1 ) , choose a black vertex z in U G n ( j 1 ) such that z ≠ t 1 and z z ′ is fault-free, where z ′ ∈ U G n ( k ) for k ∉ { j 1 , j 2 , n } . By the induction hypothesis, there is an unpaired two-disjoint path cover { P 1 , P 2 } in U G n ( j 1 ) between { x ′ , y ′ } and { t 1 , z } , where P 1 joins x ′ to t 1 and P 2 joins y ′ to z. By Lemma 5, there exists a path P from z ′ to t 2 spanning all vertices of U G n ( I \ { j 1 } ) . Let P 1 ′ = ⟨ s 1 , H [ s 1 , x ] , x , x ′ , P 1 , t 1 ⟩ and P 2 ′ = ⟨ s 2 , H [ s 2 , y ] , y , y ′ , P 2 , z , z ′ , P , t 2 ⟩ and { P 1 ′ , P 2 ′ } is the desired unpaired two-disjoint path cover, as shown in Figure 6a.
If y ′ ∈ U G n ( j 2 ) , let e = u v be an edge in U G n ( j 2 ) with u v ≠ y ′ t 2 , where u = ∗ ∗ ⋯ ∗ l k j 2 and v = ∗ ∗ ⋯ ∗ k l j 2 for l , k ∈ I \ { j 1 , j 2 } . By Corollary 1, there exists a Hamiltonian path H 1 from y ′ to t 2 through u v . By Lemma 5, there exists a path P from u ′ to v ′ spanning all vertices of U G n ( I \ { j 1 , j 2 } ) . Moreover, by Lemma 4, there is a Hamiltonian path L from x ′ to t 1 . Let P 1 = ⟨ s 1 , H [ s 1 , x ] , x , x ′ , L , t 1 ⟩ and P 2 = ⟨ s 2 , H [ s 2 , y ] , y , y ′ , H 1 [ y ′ , v ] , v , v ′ , P , u ′ , u , H 1 [ u , t 2 ] , t 2 ⟩ . Then { P 1 , P 2 } forms the unpaired path cover of U G n − F , as shown in Figure 6b.
Figure 6. Illustrations for Case 3.1.2 of Theorem 1.
Case 3.2.  s 1 , t 1 ∈ U G n ( n ) , s 2 ∈ U G n ( j 1 ) and s 2 ∈ U G n ( j 2 ) . By Lemma 4, there exists a Hamiltonian path P 1 from s 1 to t 1 ; by Lemma 5, there exists a path P 2 from s 2 to t 2 spanning all vertices of U G n ( I ) . Thus, { P 1 , P 2 } forms the desired unpaired two-disjoint path cover of U G n − F .
Case 3.3.  s 1 ∈ U G n ( n ) , t 1 , t 2 ∈ U G n ( j 1 ) and s 2 ∈ U G n ( j 2 ) . This case is similar to Case 3.1, so we omit the details of its proof.
Case 3.4.  s 1 ∈ U G n ( n ) , s 2 , t 2 ( or t 1 ) ∈ U G n ( j 1 ) and t 1 ( or t 2 ) ∈ U G n ( j 2 ) . This case is similar to Case 3.2, so we omit the details of its proof.
Case 4.  s 1 , s 2 , t 1 , t 2 lie in four distinct copies of U G n − 1 . Assume that s 1 ∈ U G n ( n ) , t 1 ∈ U G n ( j 1 ) , s 2 ∈ U G n ( j 2 ) , and t 2 ∈ U G n ( j 3 ) , where j 1 , j 2 , j 3 ∈ I are distinct. By Lemma 5, there exists a Hamiltonian path P 1 spanning U G n ( { n , j 1 } ) from s 1 to t 1 , and a path P 2 from s 2 to t 2 passing through all remaining vertices of U G n . Thus, { P 1 , P 2 } forms the desired path cover of U G n − F . □

4. Conclusions

This paper studies the unpaired many-to-many two-disjoint path cover problem for Cayley graphs generated by unicyclic graphs under edge faults. It is shown that an n-dimensional Cayley graph U G n generated by a unicyclic graph admits an unpaired two-disjoint path cover in the presence of at most n − 3 faulty edges for n ≥ 3 . In view of the degree of U G n , the result is optimal. We provide an example to show that U G n is not ( n − 2 ) -edge fault-tolerant unpaired two-disjoint path coverable. Let v be any vertex of U G n that is incident with exactly two edges, say s 1 v and s 2 v . If we assume that s 1 and s 2 are the sources of the two-disjoint path covers { P 1 , P 2 } . Then v must be on both P 1 and P 2 , which is a contradiction.
The Cayley graph Γ n is ( n − 4 ) -edge fault-tolerant unpaired many-to-many two-disjoint path coverable [22]. Since Γ n is ( n − 1 ) -regular while U G n is n-regular, our result demonstrates an improvement in the edge fault tolerance from n − 4 to n − 3 . Furthermore, the edge fault-tolerant many-to-many k-disjoint path covers of U G n would be future problems.

Author Contributions

Conceptualization, Y.L. and H.Q.; methodology, Y.L. and H.Q.; writing—original draft preparation, Y.L.; writing—review and editing, H.Q.; supervision, H.Q.; funding acquisition, H.Q. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Natural Science Foundation of Xinjiang, China (No. 2024D01C225), the Xinjiang Tianchi Doctoral Project, China (No. 51052601537), and the National Natural Science Foundation of China (No. 12401467).

Data Availability Statement

The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.

Acknowledgments

The authors would like to express their gratitude to the editor and anonymous reviewers for their valuable comments and constructive suggestions on the original manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Xu, J.-M. Topological Structure and Analysis of Interconnection Networks; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2001. [Google Scholar]
  2. Lakshmivarahan, S.; Jwo, J.-S.; Dhall, S.K. Symmetry in interconnection networks based on Cayley graphs of permutation groups: A survey. Parallel Comput. 1993, 19, 361–407. [Google Scholar] [CrossRef] [Scilit]
  3. Schibell, S.T.; Stafford, R.M. Processor interconnection networks from Cayley graphs. Discret. Appl. Math. 1992, 40, 333–357. [Google Scholar] [CrossRef] [Scilit]
  4. Hsieh, S.-Y.; Hsiao, T.-T. The k-degree Cayley graph and its topolog ical properties. Networks 2006, 47, 26–36. [Google Scholar] [CrossRef] [Scilit]
  5. Liu, M.; Liu, H. Paths and cycles embedding on faulty enhanced hypercube networks. Acta Math. Sci. 2013, 33, 227–246. [Google Scholar] [CrossRef] [Scilit]
  6. Hsieh, S.-Y. Embedding longest fault-free paths onto star graphs with more vertex faults. Theor. Comput. Sci. 2005, 337, 370–378. [Google Scholar] [CrossRef] [Scilit]
  7. Tsai, C.H.; Tan, J.J.M.; Liang, T.; Hsu, L.H. Fault-tolerant hamiltonian laceability of hypercubes. Inf. Process. Lett. 2002, 83, 301–306. [Google Scholar] [CrossRef] [Scilit]
  8. Curran, S.J.; Gallian, J.A. Hamiltonian cycles and paths in Cayley graphs and digraphs—A survey. Discret. Math. 1996, 156, 1–18. [Google Scholar] [CrossRef] [Scilit]
  9. Xue, S.; Deng, Q.; Li, P.; Chen, J. Hamiltonian paths and Hamiltonian cycles passing through prescribed linear forests in star graph with fault-tolerant edges. Discret. Appl. Math. 2023, 334, 68–80. [Google Scholar] [CrossRef] [Scilit]
  10. Park, J.-H.; Kim, H.-C.; Lim, H.-S. Many-to-many disjoint path covers in the presence of faulty elements. IEEE Trans. Comput. 2009, 58, 528–540. [Google Scholar] [CrossRef] [Scilit]
  11. Park, J.-H. Paired many-to-many disjoint path covers in restricted hypercube-like graphs. Theor. Comput. Sci. 2016, 634, 24–34. [Google Scholar] [CrossRef] [Scilit]
  12. Chen, X.-B. Paired many-to-many disjoint path covers of the hypercubes. Inform. Sci. 2013, 236, 218–223. [Google Scholar] [CrossRef] [Scilit]
  13. Zhang, S.; Wang, S. Many-to-many disjoint path covers in k-ary n-cubes. Theor. Comput. Sci. 2013, 491, 103–118. [Google Scholar]
  14. Coleman, A.; Fischberg, G.; Gong, C.; Harrington, J.; Wong, T.W.H. Paired (n − 1)-to-(n − 1) disjoint path covers in bipartite transposition-like graphs. Discret. Appl. Math. 2025, 376, 449–461. [Google Scholar] [CrossRef] [Scilit]
  15. Metsidik, M.; Yan, Q. Degree conditions for disjoint path covers in graphs. Discret. Appl. Math. 2026, 379, 45–55. [Google Scholar] [CrossRef] [Scilit]
  16. Chen, X.-B. Paired many-to-many disjoint path covers of hypercubes with faulty edges. Inform. Process. Lett. 2012, 112, 61–66. [Google Scholar] [CrossRef] [Scilit]
  17. Niu, R.; Xu, M. The unpaired many-to-many k-disjoint paths in bipartite hypercube-like networks. Theor. Comput. Sci. 2022, 911, 26–40. [Google Scholar] [CrossRef] [Scilit]
  18. Sitanayah, L.; Brown, K.N.; Sreenan, C.J. A fault-tolerant relay placement algorithm for ensuring k vertex-disjoint shortest paths in wireless sensor networks. Ad. Hoc Netw. 2014, 23, 145–162. [Google Scholar] [CrossRef] [Scilit]
  19. Li, J.; Wang, G.; Chen, L. Paired 2-disjoint path covers of multi-dimensional torus networks with 2n − 3 faulty edges. Theor. Comput. Sci. 2017, 677, 1–11. [Google Scholar] [CrossRef] [Scilit]
  20. Chen, C.; Li, P.; Liu, X. Paired 2-disjoint path cover of burnt pancake graphs with n − 3 faulty elements. Discret. Appl. Math. 2026, 380, 307–325. [Google Scholar] [CrossRef] [Scilit]
  21. Huang, X.; Li, J.; He, W. Hamiltonian laceability of bubble-sort star graphs with edge-faults. Discret. Appl. Math. 2026, 380, 442–457. [Google Scholar] [CrossRef] [Scilit]
  22. Qiao, H.; Meng, J.; Sabir, E. The edge fault-tolerant two-disjoint path covers of Cayley graphs generated by a transposition tree. Discret. Appl. Math. 2024, 356, 174–181. [Google Scholar] [CrossRef] [Scilit]
  23. Yang, W.; Li, H.; Meng, J. Conditional connectivity of Cayley graphs generated by transposition trees. Inform. Process. Lett. 2012, 110, 1027–1030. [Google Scholar]
  24. Li, H.; Yang, W.; Meng, J. Fault-tolerant Hamiltonian laceability of Cayley graphs generated by transposition trees. Discret. Math. 2012, 312, 3087–3095. [Google Scholar] [CrossRef] [Scilit]
  25. Yu, X.; Huang, X.; Zhang, Z. A kind of conditional connectivity of Cayley graphs generated by unicyclic graphs. Inform. Sci. 2013, 243, 86–94. [Google Scholar] [CrossRef] [Scilit]
  26. Zheng, W.; Zhou, S.; Yang, L. Fault tolerability of Cayley graphs generated by transposition unicyclic graphs with a triangle. Discret. Appl. Math. 2026, 384, 372–385. [Google Scholar] [CrossRef] [Scilit]
  27. Qiao, H.; Sabir, E.; Zhang, M. Many-to-many two-disjoint path covers in Cayley graphs generated by unicyclic graphs. Discret. Appl. Math. 2026, 392, 229–235. [Google Scholar] [CrossRef] [Scilit]
  28. Bondy, J.A.; Murty, U.S.R. Graph Theory; Springer: London, UK, 2008. [Google Scholar]
  29. Godsil, C.; Royle, G. Algebraic Graph Theory; Springer: New York, NY, USA, 2001. [Google Scholar]
  30. Feng, Y.; Xie, Y.; Liu, F.; Xu, S. The extendability of Cayley graphs generated by transposition trees. Mathematics 2022, 10, 1575. [Google Scholar] [CrossRef] [Scilit]
  31. Cheng, E.; Lipták, L. Linearly many faults in cayley graphs generated by transposition trees. Inform. Sci. 2007, 177, 4877–4882. [Google Scholar] [CrossRef] [Scilit]
  32. Xue, S.; Deng, Q.; Li, P. Fault-tolerant strongly Hamiltonian laceability and hyper-Hamiltonian laceability of Cayley graphs generated by transposition trees. Comput. J. 2021, 66, 384–398. [Google Scholar] [CrossRef] [Scilit]
  33. Lin, Y.; Qiao, H.; Sabir, E.; Zhang, M. The edge fault-tolerant Hamilton laceability of Cayley graphs generated by unicyclic graphs. Discret. Appl. Math. 2026; submitted.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.