
Chen considered the existence of a Hamiltonian cycle containing a matching and avoiding some edges in an n-cube Qn. In this paper, we considered the existence of a Hamiltonian path and obtained the following result. For n≥4, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤2n−6. Let x and y be two vertices of Qn with different parities satisfying xy∉M. If all vertices in Qn−F have a degree of at least 2, then there exists a Hamiltonian path joining x and y passing through M in Qn−F, with the exception of two cases: (1) there exist two neighbors v and t of x (or y) satisfying dQn−F(v)=2 and xt∈M (or yt∈M); (2) there exists a path xvuy of length 3 satisfying dQn−F(v)=2 and uy∈M or dQn−F(u)=2 and xv∈M.
Citation: Shenyang Zhao, Fan Wang. Hamiltonian paths passing through matchings in hypercubes with faulty edges[J]. AIMS Mathematics, 2024, 9(12): 33692-33711. doi: 10.3934/math.20241608
[1] | Huifeng Zhang, Xirong Xu, Ziming Wang, Qiang Zhang, Yuansheng Yang . ($ 2n-3 $)-fault-tolerant Hamiltonian connectivity of augmented cubes $ AQ_n $. AIMS Mathematics, 2021, 6(4): 3486-3511. doi: 10.3934/math.2021208 |
[2] | Yanling Wang, Shiying Wang . Edge-fault-tolerant strong Menger edge connectivity of bubble-sort graphs. AIMS Mathematics, 2021, 6(12): 13210-13221. doi: 10.3934/math.2021763 |
[3] | Wenke Zhou, Guo Chen, Hongzhi Deng, Jianhua Tu . Enumeration of dissociation sets in grid graphs. AIMS Mathematics, 2024, 9(6): 14899-14912. doi: 10.3934/math.2024721 |
[4] | Yixin Zhang, Yanbo Zhang, Hexuan Zhi . A proof of a conjecture on matching-path connected size Ramsey number. AIMS Mathematics, 2023, 8(4): 8027-8033. doi: 10.3934/math.2023406 |
[5] | Dingjun Lou, Zongrong Qin . The structure of minimally 2-subconnected graphs. AIMS Mathematics, 2022, 7(6): 9871-9883. doi: 10.3934/math.2022550 |
[6] | Xiaohong Chen, Baoyindureng Wu . Gallai's path decomposition conjecture for block graphs. AIMS Mathematics, 2025, 10(1): 1438-1447. doi: 10.3934/math.2025066 |
[7] | Jinyu Zou, Haizhen Ren . Matching preclusion and conditional matching preclusion for hierarchical cubic networks. AIMS Mathematics, 2022, 7(7): 13225-13236. doi: 10.3934/math.2022729 |
[8] | Yuan Zhang, Haiying Wang . Some new results on sum index and difference index. AIMS Mathematics, 2023, 8(11): 26444-26458. doi: 10.3934/math.20231350 |
[9] | G. Nandini, M. Venkatachalam, Raúl M. Falcón . On the r-dynamic coloring of subdivision-edge coronas of a path. AIMS Mathematics, 2020, 5(5): 4546-4562. doi: 10.3934/math.2020292 |
[10] | Yanyi Li, Lily Chen . Injective edge coloring of generalized Petersen graphs. AIMS Mathematics, 2021, 6(8): 7929-7943. doi: 10.3934/math.2021460 |
Chen considered the existence of a Hamiltonian cycle containing a matching and avoiding some edges in an n-cube Qn. In this paper, we considered the existence of a Hamiltonian path and obtained the following result. For n≥4, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤2n−6. Let x and y be two vertices of Qn with different parities satisfying xy∉M. If all vertices in Qn−F have a degree of at least 2, then there exists a Hamiltonian path joining x and y passing through M in Qn−F, with the exception of two cases: (1) there exist two neighbors v and t of x (or y) satisfying dQn−F(v)=2 and xt∈M (or yt∈M); (2) there exists a path xvuy of length 3 satisfying dQn−F(v)=2 and uy∈M or dQn−F(u)=2 and xv∈M.
Define [n] as the set {1,2,…,n}. An n-dimensional hypercube Qn is a graph with vertex set V(Qn)={v:v=v1⋯vn and vi∈{0,1} for every i∈[n]} and edge set E(Qn)={uv∣|D(u,v)|=1}, where D(u,v)={i∈[n]∣ui≠vi}.
One of the most popular and efficient interconnection networks is the hypercube Qn. It is widely recognized for being Hamiltonian for every n≥2 [12]. Havel [13] proved that a Hamiltonian path exists in Qn that connects any two vertices belonging to distinct partite sets. Since then, considerable attention has been given to the research exploring Hamiltonian cycles and paths within hypercubes that exhibit certain additional properties [2,6,7,9,10,11,14,16,19,21,22].
The parity p(u) of a vertex u=u1⋯un in Qn is defined by p(u)=∑ni=1ui(mod 2). Then, there are 2n−1 vertices with parity 0 and 2n−1 vertices with parity 1 in Qn. Observe that Qn is bipartite and the parity form bipartite sets of Qn. Consequently, p(u)≠p(v) if, and only if, d(u,v) is odd. First, let us revisit the following well-established result, initially derived by Havel in [13].
Theorem 1.1. [13] Let x,y∈V(Qn) be such that p(x)≠p(y). Then, there exists a Hamiltonian path joining x and y in Qn.
A forest is deemed linear whenever each of its components is a path. Dvořák [7,8] investigated the problem of Hamiltonian cycles or Hamiltonian paths in hypercubes with given edges and obtained the following results.
Theorem 1.2. [7] For n≥2, let P⊆E(Qn) be such that |P|≤2n−3. Then, there exists a Hamiltonian cycle of Qn containing P if, and only if, the induced subgraph of P is a linear forest.
Theorem 1.3. [8] For n≥2, let P⊆E(Qn) be such that |P|≤2n−4. Let x,y∈V(Qn) be such that p(x)≠p(y). If (1) the induced subgraph of P is a linear forest, and (2) the induced subgraph of P does not contain any path joining x and y, and neither x nor y is incident with more than one edge of P, then there exists a Hamiltonian path joining x and y passing through P in Qn, except the case when n∈{3,4}, d(x,y)=3, and P consists of 2n−4 edges, all of which are of the same dimension and have a pair-wise distance of 2, and each of x, y is incident with one edge of P.
Let P⊆E(Qn) and x,y∈V(Qn) satisfying p(x)≠p(y). We say that {x,y,P} is compatible if {x,y,P} satisfies the conditions (1) and (2) in Theorem 1.3.
A faulty edge implies a forbidden edge, which refers to an edge that cannot be traversed when searching for a Hamiltonian cycle or path structure. In [17], Tsai considered the existence of Hamiltonian paths in hypercubes with faulty edges.
Theorem 1.4. [17] For n≥3, let x,y∈V(Qn) and F⊆E(Qn) be such that p(x)≠p(y) and |F|≤2n−5. If all vertices in Qn−F have a degree of at least 2, then there exists a Hamiltonian path joining x and y in Qn−F.
It is natural to draw the following conclusion from Theorem 1.4.
Corollary 1.5. [17] For n≥3, let F⊆E(Qn) with |F|≤2n−5. If all vertices in Qn−F have a degree of at least 2, then there exists a Hamiltonian cycle in Qn−F.
Chen [5] explored the problem of Hamiltonian cycles containing matchings and avoiding some edges in an n-cube Qn and obtained the following result.
Theorem 1.6. [5] Let n≥3, M⊆E(Qn), and F⊆E(Qn)∖M with 1≤|M|≤2n−4−|F|. If M is a matching and all vertices in Qn−F have a degree of at least 2, then there exists a Hamiltonian cycle containing M in Qn−F.
In [5], Chen pointed out that if |M|=1 or |M|=2, then the upper bound of the number of faulty edges tolerated is sharp.
In this paper, we investigate the existence of a Hamiltonian path in Qn passing through a matching and avoiding faulty edges.
Let M be a matching of Qn (n≥4), and let F be a set of edges in Qn−M such that |M∪F|≤2n−6 and all vertices in Qn−F have a degree of at least 2. Let x,y∈V(Qn) be such that p(x)≠p(y) and xy∉M. We show that the following results hold:
(1) If there exist two neighbors v and t of x (or y) satisfying dQn−F(v)=2 and xt∈M (or yt∈M), then there exists no Hamiltonian path joining x and y passing through M in Qn−F. We denote this deadlock structure by x/y−DS; see Figure 1(1).
(2) If there exists a path xvuy of length 3 satisfying dQn−F(v)=2 and uy∈M or dQn−F(u)=2 and xv∈M, then there exists no Hamiltonian path joining x and y passing through M in Qn−F. We denote this deadlock structure by (x,y)−DS; see Figure 1(2).
(3) If there is neither x/y−DS nor (x,y)−DS in Qn−F, then there exists a Hamiltonian path joining x and y passing through M in Qn−F.
The terminology and notation employed in this paper, yet undefined within its scope, are referred in [1]. The vertex set and edge set of a graph G are denoted by V(G) and E(G), respectively. In a simple graph G, the number of edges incident with a vertex v is referred to as its degree, denoted as dG(v) or simply d(v). The minimum degree of a graph G, denoted by δ(G), is the minimum value of the degrees of all vertices.
Let H and H′ denote two subgraphs of a graph G, and let F be a subset of E(G). We use H+H′ to represent a graph with the vertex set V(H)∪V(H′) and edge set E(H)∪E(H′). We define H+F as the subgraph of G induced by the edge set E(H)∪F. Also, we define G−F as the subgraph of G obtained by removing all edges in F from G. Let S⊆V(G), and we denote by G−S the subgraph of G obtained by removing all vertices in S and all edges incident with vertices in S. When S={s} and F={f}, we simply write G−S, G−F, and H+F as G−s, G−f, and H+f.
The distance of vertices u and v in a graph G, denoted by dG(u,v), is the number of edges in a shortest path joining u and v. The distance dG(u,xy) of a vertex u and an edge xy in a graph G is defined by dG(u,xy):=min{dG(u,x),dG(u,y)}, and the distance dG(uv,xy) of edges uv and xy in a graph G is defined by dG(uv,xy):=min{dG(u,xy),dG(v,xy)}.
The dimension dim(uv) of an edge uv∈E(Qn) is the integer j such that D(u,v)={j}. We denote the set of all j-dimensional edges in Qn by Ej. For every j∈[n] and α∈{0,1}, let Qαn−1,j, with the subscripts j being omitted when the context is clear and the (n−1)-dimensional sub-cube of Qn induced by the vertex sets {u∈V(Qn):uj=α}. Thus, Qn−Ej=Q0n−1+Q1n−1. We say that Qn is split into two (n−1)-cubes Q0n−1 and Q1n−1 at position j. Note that both Q0n−1 and Q1n−1 are isomorphic to Qn−1. Given M⊆E(Qn), let Mα=M∩E(Qαn−1) and Mc=M∩Ej. Every vertex uα∈V(Qαn−1) has in Q1−αn−1 a unique neighbor, denoted by u1−α, and for every edge eα=uαvα∈E(Qαn−1), e1−α denotes the edge u1−αv1−α∈E(Q1−αn−1).
A set of vertex-disjoint paths of a graph G is a spanning k-path if it covers all vertices of G. For a path P in G, we say that P passes through E if E⊆E(P). Similarly, if E⊆⋃ki=1E(Pi), then P1+⋯+Pk passes through E. We use Pxy to denote a path joining vertices x and y. Let Pxy=x⋯u⋯v⋯y, and the sub-path u⋯v of Pxy is denoted by Pxy[u,v].
Next, we present some lemmas.
Lemma 2.1. [18] For n≥2, let e and f be two disjoint edges in Qn. Then, Qn can be split into two (n−1)-cubes such that one contains e and the other contains f.
Lemma 2.2. [7] For n≥2, let x,y∈V(Qn) and e∈E(Qn) be such that p(x)≠p(y) and e≠xy. Then, there exists a Hamiltonian path joining x and y passing through e in Qn.
Theorem 2.3. [23] For n≥2, let F⊆E(Qn) and L⊆E(Qn)∖F be such that |L|+|F|≤n−2. Let x,y∈V(Qn) be such that p(x)≠p(y). If {x,y,L} is compatible, then there exists a Hamiltonian path joining x and y passing through L in Qn−F.
A matching is a special type of a linear forest. Thus, we can easily draw the following corollary.
Corollary 2.4. [23] For n≥2, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤n−2. Let x,y∈V(Qn) be such that p(x)≠p(y) and xy∉M. Then, there exists a Hamiltonian path joining x and y passing through M in Qn−F.
In this section, we present some conclusions about spanning 2-paths. When the path Pvy is an edge vy, we abbreviate Pux+Pvy to Pux+vy for simplicity.
Theorem 3.1. [7] For n≥2, let x,y,u,v∈V(Qn) be pair-wise distinct vertices such that p(x)≠p(y) and p(u)≠p(v). Then, (i) there exists a spanning 2-path Pxy+Puv in Qn; (ii) moreover, in the case when d(u,v)=1, path Puv can be chosen such that Puv=uv, unless n=3, d(x,y)=1, and d(xy,uv)=2.
Theorem 3.2. [3] For n≥4, let x,y,u,v∈V(Qn) be pair-wise distinct vertices such that p(x)=p(y)≠p(u)=p(v). Then, there exists a spanning 2-path Pxy+Puv in Qn.
Theorem 3.3. [15] For n≥2, let x,y,u∈V(Qn) be pair-wise distinct vertices such that p(x)=p(y)≠p(u). Then, there exists a Hamiltonian path joining x and y in Qn−u.
When u=v, the notation Puv denotes the path of a single vertex u. Thus, in Theorem 3.2, when u=v, the conclusion still holds.
Corollary 3.4. [3,15] For n≥4, let x,y,u,v∈V(Qn) be such that x≠y and p(x)=p(y)≠p(u)=p(v). Then, there exists a spanning 2-path Pxy+Puv in Qn.
A set {u1,u2,…,u2k−1,u2k} of distinct vertices of Qn is balanced if the number of odd vertices equals to the number of even vertices.
Lemma 3.5. [4] For n≥4, let {x,y,u,v} be a balanced vertex set in Qn and F⊆E(Qn) with |F|≤1. Then, there exists a spanning 2-path Pxy+Puv in Qn−F.
Theorem 3.6. [4] For n≥4, let F⊆E(Qn) be such that |F|≤2n−7 and the degree of every vertex in Qn−F is at least 3. Assume that {u,x,v,y} is a balanced vertex set in Qn. Then, there exists a spanning 2-path Pux+Pvy in Qn−F.
Lemma 3.7. [20] Let ux and vy be two disjoint edges in Q4 and e∈E(Q4). If {u,v}∩V(e)=∅ and xy≠e, then there exists a spanning 2-path Pux+Pvy passing through e in Q4.
Lemma 3.8. For n≥4, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤n−3. Let ux,vy be two disjoint edges in Qn such that {u,v}∩V(M)=∅ and xy∉M. Then, there exists a spanning 2-path Pux+Pvy passing through M in Qn−F.
Proof. We prove the conclusion by induction on n. When n=4, we have |M∪F|≤1. By Lemma 3.5 when |M|=0 and Lemma 3.7 when |M|=1, there exists a spanning 2-path Pux+Pvy passing through M in Q4−F. Assume that the conclusion holds for n≥4, and we are to show that it holds for n+1. Now, |M∪F|≤(n+1)−3=n−2.
Select j∈[n+1]∖(D(u,x)∪D(v,y)) such that |(M∪F)∩Ej|=0. Split Qn+1 into Q0n and Q1n at position j. Without loss of generality, assume that ux∈E(Q0n). Let Mα=M∩E(Qαn) and Fα=F∩E(Qαn) for α∈{0,1}.
When vy∈E(Q1n), we have |M0∪F0|≤n−2 and |M1∪F1|≤n−2. By Corollary 2.4, there exist Hamiltonian paths Pux and Pvy passing through M0 and M1 in Q0n−F0 and Q1n−F1, respectively. Hence, Pux+Pvy is the desired spanning 2-path. So, we only need to consider the case that vy∈E(Q0n).
Case 1. |M0∪F0|≤n−3.
By induction hypothesis, there exists a spanning 2-path P0ux+P0vy passing through M0 in Q0n−F0. Choose an edge r0w0∈E(P0ux+P0vy)∖M0 satisfying {r1,w1}∩V(M1)=∅. Since |E(P0ux+P0vy)∖M0|−4|M1|=|E(P0ux+P0vy)|−(|M0|+4|M1|)≥(2n−2)−4(n−2)=2n−4n+6>1 for n≥4, such an edge r0w0 exists. Without loss of generality, assume that r0w0∈E(P0ux). Since M1∪{r1w1} is a matching and |(M1∪{r1w1})∪F1|≤n−1<2n−4 for n≥4, by Theorem 1.6 there exists a Hamiltonian cycle C1 containing M1∪{r1w1} in Q1n−F1. Let Pux=P0ux+C1+{r0r1,w0w1}−{r0w0,r1w1} and Pvy=P0vy. Hence, Pux+Pvy is a spanning 2-path passing through M in Qn+1−F.
Case 2. |M0∪F0|=n−2. Now, |M1|=|F1|=0.
When |F0|≥1, choose an edge r0w0∈F0. Since |M0∪(F0∖{r0w0})|=n−3, by induction hypothesis there exists a spanning 2-path P0ux+P0vy passing through M0 in Q0n−(F0∖{r0w0}). If r0w0∉E(P0ux+P0vy), then choose an edge s0t0∈E(P0ux+P0vy)∖M0. If r0w0∈E(P0ux+P0vy), then let s0t0=r0w0. In the above two cases, without loss of generality, assume that s0t0∈E(P0ux). In Q1n, by Theorem 1.1, there exists a Hamiltonian path P1s1t1. Let Pux=P0ux+P1s1t1+{s0s1,t0t1}−s0t0 and Pvy=P0vy. Hence, Pux+Pvy is a spanning 2-path passing through M in Qn+1−F.
When |F0|=0, now |M0|=n−2. Since M0∪{ux,vy} is a linear forest with |M0∪{ux,vy}|=n<2n−3 for n≥4, by Theorem 1.2 there exists a Hamiltonian cycle C0 containing M0∪{ux,vy} in Q0n. Note that C0−ux−vy consists of two paths, having endpoints u and x, respectively. We denote them by Pu and Px. Since {u,v}∩V(M)=∅ and xy∉M, we can choose two edges r0w0∈E(Pu)∖M0 and s0t0∈E(Px)∖M0. Without loss of generality, assume that r0 is closer to u than w0 on Pu and s0 is closer to x than t0 on Px. In Q1n, by Theorems 3.1 and 3.2, there exists a spanning 2-path P1r1s1+P1w1t1. Hence, Pux+Pvy=C0+P1r1s1+P1w1t1+{r0r1,w0w1,s0s1,t0t1}−{ux,vy,r0w0,s0t0} is the desired spanning 2-path.
Lemma 3.9. For n≥4, let ux,vy be two disjoint edges in Qn, and let f∈E(Qn) with vy≠f, then there exists a spanning 2-path Pux+vy in Qn−f.
Proof. By Lemma 2.1, we can split Qn into Q0n−1 and Q1n−1 such that ux∈E(Q0n−1) and vy∈E(Q1n−1). Next, we distinguish two cases to consider.
If f∉E(Q1n−1), then by Theorem 1.4 there exists a Hamiltonian path P0ux in Q0n−1−f. Choose an edge s0t0∈E(P0ux) such that f∉{s0s1,t0t1} and d(s1t1,vy)≠2 when n=4. By Theorem 3.1 there exists a spanning 2-path P1s1t1+vy in Q1n−1. Let Pux=P0ux+P1s1t1+{s0s1,t0t1}−s0t0. Hence, Pux+vy is a spanning 2-path in Qn−f.
If f∈E(Q1n−1), then by Theorem 1.4 there exists a Hamiltonian path P1vy in Q1n−1−f. Choose two neighbors r1,w1 of v,y on P1vy. If r0w0≠ux, then by Lemma 2.2 there exists a Hamiltonian path P0r0w0 passing through ux in Q0n−1. Let Pux=P0r0w0+P1vy[r1,w1]+{r0r1,w0w1}−ux. Hence, Pux+vy is the desired spanning 2-path. If r0w0=ux, then choose an edge s1t1∈E(P1vy) such that {s0,t0}∩{r0,w0}=∅ and d(s0t0,r0w0)≠2 when n=4. Since |E(P1vy)|−4−1>1 for n≥4, such an edge s1t1 exists. In Q0n−1, by Theorem 3.1 there exists a spanning 2-path P0s0t0+ux. Let Pux=P0s0t0+P1vy[r1,w1]+{uu1,xx1,s0s1,t0t1}−s1t1. Hence, Pux+vy is the desired spanning 2-path.
Lemma 3.10. [20] For n≥5, let uv and e be two disjoint edges in Qn, and let x,y∈V(Qn)∖{u,v} be such that p(x)≠p(y) and xy≠e. Then, there exists a spanning 2-path Pxy+uv passing through e in Qn.
Lemma 3.11. For n≥4, let uv and e be two disjoint edges in Qn, and let x,y∈V(Qn)∖{u,v} be such that p(x)≠p(y) and xy≠e. Then, there exists a spanning 2-path Pxy+uv passing through e in Qn.
Proof. When n≥5, the conclusion holds by Lemma 3.10. So, we only need to consider the case that n=4. By Lemma 2.1, we can split Q4 into Q03 and Q13 such that e∈E(Q03) and uv∈E(Q13).
Case 1. x,y∈V(Q03).
By Lemma 2.2, there exists a Hamiltonian path P0xy passing through e in Q03. Choose an edge s0t0∈E(P0xy)∖{e} such that {s1,t1}∩{u,v}=∅ and d(uv,s1t1)≠2. Since |E(P0xy)∖{e}|−4−1=1, such an edge s0t0 exists. In Q13, by Theorem 3.1 there exists a spanning 2-path P1s1t1+uv. Let Pxy=P0xy+P1s1t1+{s0s1,t0t1}−s0t0. Hence, Pxy+uv is a spanning 2-path passing through e in Q4.
Case 2. x∈V(Q03), y∈V(Q13) (or x∈V(Q13), y∈V(Q03)).
Since 22>2, we can choose a vertex t0∈V(Q03) such that p(t0)≠p(x), t1∉{u,v}, and d(t1y,uv)≠2. In Q03, by Lemma 2.2 there exists a Hamiltonian path P0xt0 passing through e. In Q13, by Theorem 3.1 there exists a spanning 2-path P1t1y+uv. Let Pxy=P0xt0+P1t1y+t0t1. Hence, Pxy+uv is the desired spanning 2-path.
Case 3. x,y∈V(Q13).
By Lemma 2.2 there exists a Hamiltonian path P1xy passing through uv in Q13. Without loss of generality, assume that u is closer to x than v on P0xy. Choose two neighbors r1,w1 of u,v on P1xy[x,u] and P1xy[y,v], respectively. If r0w0≠e, then by Lemma 2.2 there exists a Hamiltonian path P0r0w0 passing through e in Q03. Let Pxy=P1xy[x,r1]+P1xy[y,w1]+P0r0w0+{r0r1,w0w1}. Hence, Pxy+uv is the desired spanning 2-path. If r0w0=e, then choose an edge s1t1∈E(P1xy)∖{uv} such that {s0,t0}∩{r0,w0}=∅ and d(s0t0,r0w0)≠2. Since |E(P1xy)∖{uv}|−4−1=1, such an edge s1t1 exists. In Q03, by Theorem 3.1 there exists a spanning 2-path P0s0t0+e. Let Pxy=P1xy[x,r1]+P1xy[y,w1]+P0s0t0+e+{r0r1,w0w1,s0s1,t0t1}−s1t1. Hence, Pxy+uv is the desired spanning 2-path.
Lemma 3.12. For n≥4, let swrys be a cycle of length four in Qn. Let M be a matching of Qn, and let F be a set of edges in Qn−M such that {s,w,r,y}∩V(M)=∅, ry∉F, and |M∪F|≤n−3. Then, there exists a spanning 2-path Psw+ry passing through M in Qn−F.
Proof. Since swrys is a cycle of length four, then dim(sw)=dim(ry) and dim(rw)=dim(sy). We prove the conclusion by induction on n. When n=4, we have |M∪F|≤1. Thus, the conclusion holds by Theorem 3.1 when |M∪F|=0, Lemma 3.9 when |F|=1, and Lemma 3.11 when |M|=1. Assume that the conclusion holds for n≥4, and we are to show that it holds for n+1. Now, |M∪F|≤(n+1)−3=n−2.
Since n+1>n−2+2, we can select j∈[n+1]∖(D(s,w)∪D(r,w)) such that |(M∪F)∩Ej|=0. Then, |{sw,ry,rw,sy}∩Ej|=0. Split Qn+1 into Q0n and Q1n at position j. Without loss of generality, assume that sw,ry∈E(Q0n).
Case 1. |M0∪F0|≤n−3.
By induction hypothesis, there exists a spanning 2-path P0sw+ry passing through M0 in Q0n−F0. Choose an edge u0v0∈E(P0sw)∖M0 such that {u1,v1}∩V(M1)=∅. Since |E(P0sw∖M0)|−4|M1|=|E(P0sw)|−|M0|−4|M1|≥(2n−2−1)−4(n−2)=2n−4n+5≥1 for n≥4, such an edge u0v0 exists. Since |(M1∪{u1v1})∪F1|≤n−1<2n−4 for n≥4 and M1∪{u1v1} is a matching, by Theorem 1.6 there exists a Hamiltonian cycle C1 containing M1∪{u1v1} in Q1n−F1. Let Psw=P0sw+C1+{u0u1,v0v1}−{u0v0,u1v1}. Hence, P0sw+ry is a spanning 2-path passing through M in Qn+1−F.
Case 2. |M0∪F0|=n−2. Now, |M1∪F1|=0.
When |M0|≥1, choose an edge u0v0∈M0. Since |(M0∖{u0v0})∪F0|=n−3, by induction hypothesis there exists a spanning 2-path P0sw+ry passing through M0∖{u0v0} in Q0n−F0. If u0v0∈E(P0sw), then choose an edge x0t0∈E(P0sw)∖M0. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1x1t1. Let Psw=P0sw+P1x1t1+{x0x1,t0t1}−x0t0. Hence, P0sw+ry is the desired spanning 2-path. If u0v0∉E(P0sw), then without loss of generality assume that u0 is closer to s than v0 on P0sw. Since {s,w}∩V(M0)=∅, we can choose clockwise neighbors x0,t0 of u0,v0 on P0sw. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1x1t1. Let Psw=P0sw+P1x1t1+{u0v0,x0x1,t0t1}−{u0x0,v0t0}. Hence, P0sw+ry is the desired spanning 2-path.
When |M0|=0, we have |F0|=n−2. Choose an edge u0v0∈F0. Since |F0∖{u0v0}|=n−3, by induction hypothesis there exists a spanning 2-path P0sw+ry in Q0n−(F0∖{u0v0}). If u0v0∉E(P0sw), then choose an edge x0t0∈E(P0sw). If u0v0∈E(P0sw), then let u0v0=x0t0. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1x1t1. Let Psw=P0sw+P1x1t1+{x0x1,t0t1}−x0t0. Hence, P0sw+ry is the desired spanning 2-path.
Lemma 4.1. For n≥4, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤2n−6. Let x,y∈V(Qn) be such that p(x)≠p(y) and xy∉M. If δ(Qn−F)=2 and if there is neither x/y−DS nor (x,y)−DS in Qn−F, then there exists a Hamiltonian path joining x and y passing through M in Qn−F.
Proof. We prove the conclusion by induction on n. When n=4, the conclusion holds by Corollary 2.4. Assume that the conclusion holds for n≥4, and we are to show that it holds for n+1. Now, |M∪F|≤2(n+1)−6=2n−4.
When |M|=0, the conclusion holds by Theorem 1.4. So, we only need to consider the case that |M|≥1. Now, |F|≤2n−5. Thus, there is exactly one vertex, denoted by v, of degree 2 in Qn+1−F, and all the other vertices of Qn+1−F are of degree of at least 4. If there is another vertex of degree 2 or 3 in Qn+1−F, then |F|≥(n+1−2)+(n+1−3)−1=2n−4. A contradiction occurs.
Let Fv={e∈F | e is incident with v}. Note that |F|≥|Fv|=(n+1)−2=n−1. Since |M|≤2n−4−|F|≤n−3<|Fv|, there exists a position j such that Fv∩Ej≠∅ and M∩Ej=∅. Split Qn+1 into Q0n and Q1n at position j. We may assume v∈V(Q0n), and denote v by v0. Let Mα=M∩E(Qαn) and Fα=F∩E(Qαn) for α∈{0,1}, and let Mc=M∩Ej and Fc=F∩Ej. Hence, dQ0n−F0(v0)=2, all the other vertices of Q0n−F0 are of degree of at least 3, and δ(Q1n−F1)≥3. Note that |F0|≥n−2, |M0∪F0|≤2n−5, and |M1∪F1|≤n−3.
We claim that Q0n does not contain x/y−DS and (x,y)−DS. If not, then Qn+1 contains x/y−DS or (x,y)−DS. A contradiction occurs.
Case 1. |M0∪F0|≤2n−6.
Sub-case 1.1 x,y∈V(Q0n). Note that |Fc|≤2n−4−(n−2)−1=n−3.
By induction hypothesis, there exists a Hamiltonian path P0xy passing through M0 in Q0n−F0. Choose an edge s0t0∈E(P0xy)∖M0 such that {s0s1,t0t1}∩Fc=∅ and s1t1∉M1. Since |E(P0xy)∖M0|−|M1|−2|Fc|=|E(P0xy)|−(|M0|+|M1|+|Fc|)−|Fc|≥2n−1−(n−2)−(n−3)=2n−2n+4>1 for n≥4, such an edge s0t0 exists. Since |M1∪F1|≤n−3<n−2, by Corollary 2.4 there exists a Hamiltonian path P1s1t1 passing through M1 in Q1n−F1. Hence, Pxy=P0xy+P1s1t1+{s0s1,t0t1}−s0t0 is a Hamiltonian path joining x and y passing through M in Qn+1−F.
Sub-case 1.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Choose a vertex z0 in Q0n such that p(x)≠p(z0), z0∉V(M0), z0z1∉Fc, dQ0n−F0(z0,v0)≠1, and z1y∉M1. Since 2(n+1)−2−(|M0|+|Fc|)−n−1≥2n−1−(n−2)−n−1=2n−1−2n+1≥1 for n≥4, such a vertex z0 exists. Note that Q0n does not contain x−DS, and there is only one vertex v0 of degree 2 in Q0n−F0. If Q0n contains z0−DS, then z0∈V(M0) and d(z0,v0)=1. If Q0n contains (x,z0)−DS, then z0∈V(M0) or d(z0,v0)=1. The above two cases both contradict with the choice of z0. Thus, Q0n does not contain x/z0−DS and (x,z0)−DS. By induction hypothesis, there exists a Hamiltonian path P0xz0 passing through M0 in Q0n−F0. By Corollary 2.4, there exists a Hamiltonian path P1z1y passing through M1 in Q1n−F1. Hence, Pxy=P0xz0+P1z1y+z0z1 is a Hamiltonian path joining x and y passing through M in Qn+1−F.
Sub-case 1.3 x,y∈V(Q1n).
By Corollary 2.4, there exists a Hamiltonian path P1xy passing through M1 in Q1n−F1. Choose an edge s1t1∈E(P1xy)∖M1 such that {s0s1,t0t1}∩Fc=∅ and {s0,t0}∩V(M0)=∅. Since |E(P1xy)∖M1|−2|Fc|−4|M0|≥(2n−1)−2×2−4(n−4)=2n−4n+11>1 for n≥4, such an edge s1t1 exists. Since {s0,t0}∩V(M0)=∅, we have that Q0n does not contain s0/t0−DS and (s0,t0)−DS. Hence, by induction hypothesis there exists a Hamiltonian path P0s0t0 passing through M0 in Q0n−F0. Hence, Pxy=P1xy+P0s0t0+{s0s1,t0t1}−s1t1 is the desired Hamiltonian path.
Case 2. |M0∪F0|=2n−5. Now, |Fc|=1 and |Mc|=|M1|=|F1|=0. Note that |F0|≥n−2, 1≤|M0|≤n−3.
Sub-case 2.1 x,y∈V(Q0n).
If |F0|≥n−1, then let r0w0∈F0∖Fv. Note that dQ0n−(F0∖{r0w0})(v0)=2 and all the other vertices of Q0n−(F0∖{r0w0}) are of degree of at least 3. Since |M0∪(F0∖{r0w0})|=2n−6, by induction hypothesis there exists a Hamiltonian path P0xy passing through M0 in Q0n−(F0∖{r0w0}). If r0w0∉E(P0xy), then choose an edge s0t0∈E(P0xy)∖M0 satisfying v0∉{s0,t0}. If r0w0∈E(P0xy), then let s0t0=r0w0. Since r0w0∉Fv, we also have v0∉{s0,t0}. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1s1t1. Hence, Pxy=P0xy+P1s1t1+{s0s1,t0t1}−s0t0 is the desired Hamiltonian path.
If |F0|=n−2, then F0=Fv∖{v0v1}. Denote the two edges incident with v0 in Q0n−F0 by u0v0 and v0s0.
When v0∉{x,y}, first we claim that {x,y,M0∪{u0v0,v0s0}} is compatible. If v0∈V(M0), then let v0s0∈M0. Now, u0v0∉M0. There are two possibilities for u0 up to isomorphism; see Figure 2(1)(2). If v0∉V(M0), then there are three possibilities for {u0,s0} up to isomorphism; see Figure 2(3)-(5). Thus, M0∪{u0v0,v0s0}} is a linear forest. Next, we will verify that {x,y,M0∪{u0v0,v0s0}} satisfies the condition (2) in Theorem 1.3.
For the case (1) in Figure 2, if x or y is an internal vertex, then x=v0 or y=v0. A contradiction occurs. If x,y are endpoints of a path, then since p(x)≠p(y), we have xy∈M0. A contradiction occurs. For the case (2) in Figure 2, if x or y is an internal vertex, then since v0∉{x,y}, we have that x or y is the vertex u0. Now, x (or y) is incident with a matching edge, and dQ0n−F0(v0)=2. Thus, Q0n contains x/y−DS. A contradiction occurs. If x,y are endpoints of a path, then since xy∉M0, we have that x,y are the two endpoints of the path of length 3 in Figure 2(2). Note that dQ0n−F0(v0)=2, thus, Q0n contains (x,y)−DS. A contradiction occurs. For the case (3) in Figure 2, if x or y is an internal vertex, then since v0∉{x,y}, we have that x (or y) is s0 or u0. Now, Q0n contains x/y−DS. A contradiction occurs. If x,y are endpoints of a path, then xy∈M0. A contradiction occurs. For the case (4) in Figure 2, if x or y is an internal vertex, then since v0∉{x,y}, we have that x (or y) is the vertex s0 in Figure 2(4). Now, Q0n contains x/y−DS. A contradiction occurs. If x,y are endpoints of a path, then since xy∉M0, we have that x,y are endpoints of the path of length 3 in Figure 2(4). Thus, Q0n contains (x,y)−DS. A contradiction occurs. For the case (5) in Figure 2, since v0∉{x,y}, p(x)≠p(y), and xy∉M0, we have that {x,y,M0∪{u0v0,v0s0}} also satisfies the condition (2). In conclusion, {x,y,M0∪{u0v0,v0s0}} satisfies the condition (2). The claim is proved.
Since |M0∪{u0v0,v0s0}|≤n−1<2n−4 for n≥4, by Theorem 1.3 there exists a Hamiltonian path P0xy passing through M0∪{u0v0,v0s0} in Q0n. Note that all faulty edges incident with v0 do not lie on the path P0xy. Thus, we have E(P0xy)∩F0=∅.
When v0∈{x,y}, we may assume v0=x. If u0v0∈M0 or v0s0∈M0, then without loss of generality assume that u0v0∈M0, and now v0s0∉M0. Thus, M0∪{u0v0}=M0 and {x,y,M0∪{u0v0}} are compatible. If u0v0∉M0 and v0s0∉M0, then v0∉V(M0). Since xy≠u0v0 or xy≠v0s0, without loss of generality, assume that xy≠u0v0. Thus, M0∪{u0v0} is a linear forest with d(v0)=1 in it. Since xy∉M0, xy≠u0v0, and p(x)≠p(y), we have that {x,y,M0∪{u0v0}} is compatible. Since |M0∪{u0v0}|≤n−2<2n−4 for n≥4, by Theorem 1.3 there exists a Hamiltonian path P0xy passing through M0∪{u0v0} in Q0n. Note that v0=x and u0v0 lies on the path P0xy, and we also have E(P0xy)∩F0=∅.
In the above two cases, choose an edge s0t0∈E(P0xy)∖M0 satisfying v0∉{s0,t0}. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1s1t1. Hence, Pxy=P0xy+P1s1t1+{s0s1,t0t1}−s0t0 is the desired Hamiltonian path.
Sub-case 2.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Since |M0∪F0|=2n−5<2n−4, |M0|≥1, and δ(Q0n−F0)≥2, by Theorem 1.6 there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. Choose a neighbor s0 of x on C0 such that xs0∉M0 and s0≠v0. We claim that the vertex s0 exists. If s0 does not exist, then in one direction along the cycle C0, x is adjacent to v0, and in the other direction, x is incident with a matching edge. Thus, Q0n contains x−DS. A contradiction occurs. Since p(s1)≠p(y), by Theorem 1.1 there exists a Hamiltonian path P1s1y in Q1n. Hence, Pxy=C0+P1s1y+s0s1−xs0 is the desired Hamiltonian path.
Sub-case 2.3. x,y∈V(Q1n).
Since |M0∪F0|=2n−5<2n−4, |M0|≥1, and δ(Q0n−F0)≥2, by Theorem 1.6 there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. Choose an edge s0t0∈E(C0)∖M0 such that v0∉{s0,t0} and s1t1≠xy. Since |E(C0)∖M0|−2−1≥2n−|M0|−3≥2n−(n−3)−3=2n−n>1 for n≥4, such an edge s0t0 exists. By Lemma 2.2, there exists a Hamiltonian path P1xy passing through s1t1 in Q1n. Hence, Pxy=P1xy+C0+{s0s1,t0t1}−{s0t0,s1t1} is the desired Hamiltonian path.
Lemma 4.2. For n≥4, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤2n−6. Let x,y∈V(Qn) be such that p(x)≠p(y) and xy∉M. If δ(Qn−F)=3, then there exists a Hamiltonian path joining x and y passing through M in Qn−F.
Proof. We prove the conclusion by induction on n. When n=4, the conclusion holds by Corollary 2.4. Assume that the conclusion holds for n≥4, and we are to show that it holds for n+1. Now, |M∪F|≤2(n+1)−6=2n−4.
By Theorem 1.4, we only need to consider the case that |M|≥1. Now, |F|≤2n−5. Thus, there are at most two vertices of degree 3 in Qn+1−F. If there are three vertices of degree 3 in Qn+1−F, then |F|≥3(n+1−3)−2=3n−8>2n−5 for n≥4. A contradiction occurs.
Case A. There are two vertices of degree 3, denoted by v and v′, in Qn+1−F.
If vv′∉F, then |F|≥2(n+1−3)=2n−4. A contradiction occurs. Thus, vv′∈F. Now, |F|=2n−5, |M|=1, and all the other vertices (except v and v′) of Qn+1−F are of degree of at least 4. Let D(v,v′)={j}. Split Qn+1 into Q0n and Q1n at position j. We may assume v∈V(Q0n), and denote v by v0. Then, v′=v1, δ(Q0n−F0)=3, and δ(Q1n−F1)=3. Note that |F0|=|F1|=n−3 and |M|=1. Let M={e}.
Case 1. e∉Ej. Without loss of generality, assume that e∈E(Q0n).
Sub-case 1.1 x,y∈V(Q0n).
Since |M0∪F0|=n−2, by Corollary 2.4 there exists a Hamiltonian path P0xy passing through M0 in Q0n−F0. Choose an edge s0t0∈E(P0xy)∖M0 such that v0∉{s0,t0}. Since |E(P0xy)∖M0|−2=2n−4>1 for n≥4, such an edge s0t0 exists. Since |F1|=n−3<2n−5 for n≥4, by Theorem 1.4 there exists a Hamiltonian path P1s1t1 in Q1n−F1. Hence, Pxy=P0xy+P1s1t1+{s0s1,t0t1}−s0t0 is a Hamiltonian path joining x and y passing through M in Qn+1−F.
Sub-case 1.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Since 2n−1>2, we can choose a vertex z0 in Q0n such that p(x)≠p(z0), z0≠v0, and xz0≠e. By Corollary 2.4, there exists a Hamiltonian path P0xz0 passing through M0 in Q0n−F0. By Theorem 1.4, there exists a Hamiltonian path P1z1y in Q1n−F1. Hence, Pxy=P0xz0+P1z1y+z0z1 is the desired Hamiltonian path.
Sub-case 1.3. x,y∈V(Q1n).
By Theorem 1.4, there exists a Hamiltonian path P1xy in Q1n−F1. Choose an edge s1t1∈E(P1xy) such that v0∉{s0,t0} and s0t0≠e. Since |E(P1xy)|−3>1 for n≥4, such an edge s1t1 exists. By Corollary 2.4, there exists a Hamiltonian path P0s0t0 passing through M0 in Q0n−F0. Hence, Pxy=P1xy+P0s0t0+{s0s1,t0t1}−s1t1 is the desired Hamiltonian path.
Case 2. e∈Ej. Let e=r0r1. Note that r0≠v0.
Sub-case 2.1 x,y∈V(Q0n) (or x,y∈V(Q1n)).
Since n>2, we can choose a neighbor w0 of r0 in Q0n such that w0≠v0 and r0w0≠xy. Since |{r0w0}∪F0|=n−2, by Corollary 2.4 there exists a Hamiltonian path P0xy passing through r0w0 in Q0n−F0. By Theorem 1.4, there exists a Hamiltonian path P1r1w1 in Q1n−F1. Hence, Pxy=P0xy+P1r1w1+{r0r1,w0w1}−r0w0 is the desired Hamiltonian path.
Sub-case 2.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
When p(x)≠p(r0), by Theorem 1.4 there exist Hamiltonian paths P0xr0 and P1r1y in Q0n−F0 and Q1n−F1, respectively. Hence, Pxy=P0xr0+P1r1y+r0r1 is the desired Hamiltonian path.
When p(x)=p(r0), since xy≠r0r1, by symmetry we may assume y≠r1. Since n>1, we can choose a neighbor w1 of r1 in Q1n satisfying w1≠v1. Since n>2, we can choose a neighbor s1 of y in Q1n satisfying s1∉{v1,w1}. Since |{r0w0}∪F0|=n−2, by Corollary 2.4 there exists a Hamiltonian path P0xs0 passing through r0w0 in Q0n−F0. Note that {r1,w1,s1,y} is a balanced vertex set. Since |F1|=n−3≤2n−7 for n≥4 and δ(Q1n−F1)≥3, by Theorem 3.6 there exists a spanning 2-path P1r1w1+P1s1y in Q1n−F1. Hence, Pxy=P0xs0+P1r1w1+P1s1y+{s0s1,r0r1,w0w1}−r0w0 is the desired Hamiltonian path.
Case B. There is exactly a vertex, denoted by v, of degree 3 in Qn+1−F.
Now, all the other vertices in Qn−F are of degree of at least 4. Let Fv={e∈F | e is incident with v}. Note that |F|≥|Fv|=(n+1)−3=n−2 and |M|≤2n−4−|F|≤n−2.
Case 1. |M|=n−2. Now, |F|=|Fv|=n−2.
Denote the three edges incident with v in Qn+1−F by uv, vr, and vs. If v∈V(M), then let uv∈M.
When v∉{x,y}, first we claim that {x,y,M∪{uv,vs}} is compatible or {x,y,M∪{uv,vr}} is compatible. Note that M∪{uv,vs} and M∪{uv,vr} are both linear forest. If {x,y,M∪{uv,vs}} is not compatible, then {x,y,M∪{uv,vs}} does not satisfy the condition (2) in Theorem 1.3. Now there are three possibilities for {M∪{uv,vs}} up to isomorphism; see Figure 3(1)-(3). Next, we will verify that {x,y,M∪{uv,vr}} satisfies the condition (2) in Theorem 1.3. Note that there are two cases for r, uncovered by M or covered by M. We denote the two cases by r′ and r″. If r∉V(M), then let r=r′; if r∈V(M), then let r=r″; see Figure 3.
If x or y is an internal vertex of the induced subgraph of {M∪{uv,vs}}, then for the case (1)(2) in Figure 3, x=s or y=s; for the case (3) in Figure 3, x=s or u, or y=s or u. Note that x and y cannot be s and u simultaneously because p(x)≠p(y). For the above three cases, without loss of generality, assume that x=s. Since p(x)≠p(y), we have y∉{u,r}. Note that y≠v and xy∉M. Thus, {x,y,M∪{uv,vr}} satisfies the condition (2) in Theorem 1.3. If x,y are endpoints of a path in the induced subgraph of {M∪{uv,vs}}, then since p(x)≠p(y) and xy∉M, we have that {M∪{uv,vs}} must be the case (1) or (2) in Figure 3 and x,y exactly are the two endpoints of the path of length 3. Note that the two internal vertices of this path are v and s, and r∉{x,y,v,s}. Hence, {x,y,M∪{uv,vr}} satisfies the condition (2) in Theorem 1.3. The claim is proved. Without loss of generality, assume that {x,y,M∪{uv,vs}} is compatible.
Since |M∪{uv,vs}|≤n<2(n+1)−4 for n≥4, by Theorem 1.3 there exists a Hamiltonian path Pxy passing through M∪{uv,vs} in Qn+1. Note that all faulty edges incident with v do not lie on the path Pxy. Thus, we have E(Pxy)∩F=∅.
When v∈{x,y}, we may assume v=x. If v∈V(M), then now uv∈M and vr,vs∉M. Thus, M∪{uv}=M and {x,y,M} is compatible. If v∉V(M), then since xy≠uv or xy≠vr or xy≠vs, without loss of generality, assume that xy≠uv. So, {x,y,M∪{uv}} is compatible. Since |M∪{uv}|≤n−1<2(n+1)−4 for n≥4, by Theorem 1.3 there exists a Hamiltonian path Pxy passing through M in Qn+1−F.
Case 2. |M|≤n−3.
Since |M|≤n−3<n−2=|Fv|, there exists a position j∈[n+1] such that |Fv∩Ej|=1 and |M∩Ej|=0. Split Qn+1 into Q0n and Q1n at position j. Assume v∈V(Q0n), and denote v by v0. Hence, δ(Q0n−F0)=3 and δ(Q1n−F1)≥3. Note that |F0|≥n−3, |M0∪F0|≤2n−5, and |M1∪F1|≤n−2.
Sub-case 2.1. |M0∪F0|≤2n−6. Note that |Fc|≤2n−4−(n−3)−1=n−2.
Sub-case 2.1.1 x,y∈V(Q0n) (or x,y∈V(Q1n))
By induction hypothesis, there exists a Hamiltonian path P0xy passing through M0 in Q0n−F0. Choose an edge s0t0∈E(P0xy)∖M0 such that {s0s1,t0t1}∩Fc=∅ and s1t1∉M1. Since |E(P0xy)∖M0|−2|Fc|−|M1|=|E(P0xy)|−(|M0|+|Fc|+|M1|)−|Fc|≥2n−1−(n−1)−(n−2)=2n−2n+2>1 for n≥4, such an edge s0t0 exists. By Corollary 2.4, there exists a Hamiltonian path P1s1t1 passing through M1 in Q1n−F1. Hence, Pxy=P0xy+P1s1t1+{s0s1,t0t1}−s0t0 is the desired Hamiltonian path.
Sub-case 2.1.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Choose a vertex z0 in Q0n such that p(x)≠p(z0), z0z1∉Fc, and {xz0,z1y}∩M=∅. Note that |M|+|Fc|≤2n−4−(|Fv|−1)=n−1. Since 2n−1−|M|−|Fc|≥2n−1−(n−1)=2n−1−n+1>1 for n≥4, such a vertex z0 exists. By induction hypothesis, there exists a Hamiltonian path P0xz0 passing through M0 in Q0n−F0. By Corollary 2.4, there exists a Hamiltonian path P1z1y passing through M1 in Q1n−F1. Hence, Pxy=P0xz0+P1z1y+z0z1 is the desired Hamiltonian path.
Sub-case 2.2. |M0∪F0|=2n−5. Now, |M0|≥1, Fc={v0v1}, and |M1∪F1|=0.
Sub-case 2.2.1. x,y∈V(Q0n).
Since |M0|=|M|≤n−3, we have |F0|≥n−2. So we can choose an edge r0w0∈F0∖Fv. Thus, v0∉{r0,w0}. Note that dQ0n−(F0∖{r0w0})(v0)=3. So, we have δ(Q0n−(F0∖{r0w0}))=3. Since |M0∪(F0∖{r0w0})|=2n−6, by induction hypothesis there exists a Hamiltonian path P0xy passing through M0 in Q0n−(F0∖{r0w0}). If r0w0∉E(P0xy), then choose an edge s0t0∈E(P0xy)∖M0 satisfying v0∉{s0,t0}. If r0w0∈E(P0xy), then let s0t0=r0w0. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1s1t1. Hence, Pxy=P0xy+P1s1t1+{s0s1,t0t1}−s0t0 is the desired Hamiltonian path.
Sub-case 2.2.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Since |M0∪F0|=2n−5<2n−4, |M0|≥1, and δ(Q0n−F0)≥3, by Theorem 1.6 there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. Choose a neighbor s0 of x on C0 satisfying xs0∉M0. If s0≠v0, then by Theorem 1.1 there exists a Hamiltonian path P1s1y in Q1n. Hence, Pxy=C0+P1s1y+s0s1−xs0 is the desired Hamiltonian path. If s0=v0, then since dQ0n−F0(v0)=3, let u0 be the neighbor of v0 in Q0n satisfying dC0(u0,v0)>1 and u0v0∉F0. Note that there are two neighbors of u0 on C0. One neighbor lies on the path joining v0 and u0 on C0 which contains x. We denote it as w0. The other neighbor is denoted as t0.
If u0t0∉M0, then by Theorem 1.1 there exists a Hamiltonian path P1t1y in Q1n. Hence, Pxy=C0+P1t1y+{u0v0,t0t1}−{xv0,u0t0} is the desired Hamiltonian path. If u0t0∈M0, then u0w0∉M0. Let r0 be the neighbor of t0 on C0 which is not u0. Now, r0t0∉M0 and p(w1)=p(t1)≠p(r1)=p(y). By Corollary 3.4, there exists a spanning 2-path P1t1w1+P1r1y in Q1n. Hence, Pxy=C0+P1t1w1+P1r1y+{u0v0,t0t1,r0r1,w0w1}−{xv0,u0w0,r0t0} is the desired Hamiltonian path.
Sub-case 2.2.3. x,y∈V(Q1n).
Since |M0∪F0|=2n−5<2n−4, |M0|≥1, and δ(Q0n−F0)≥3, by Theorem 1.6 there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. Choose an edge s0t0∈E(C0)∖M0 such that v0∉{s0,t0} and s1t1≠xy. Since |E(C0)∖M0|−3>1 for n≥4, such an edge s0t0 exists. In Q1n, by Theorem 2.2 there exists a Hamiltonian path P1xy passing through s1t1. Hence, Pxy=P1xy+C0+{s0s1,t0t1}−{s0t0,s1t1} is the desired Hamiltonian path.
Lemma 4.3. For n≥4, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤2n−6. Let x,y∈V(Qn) be such that p(x)≠p(y) and xy∉M. If δ(Qn−F)≥4, then there exists a Hamiltonian path joining x and y passing through M in Qn−F.
Proof. We prove the conclusion by induction on n. When n=4, the conclusion is by Corollary 2.4. Assume that the conclusion holds for n≥4, and we are to show that it holds for n+1. Now, |M∪F|≤2(n+1)−6=2n−4. By Theorems 1.3 and 1.4, we only need to consider the case that |F|≥1 and |M|≥1.
Select j∈[n+1] such that |(M∪F)∩Ej| is as small as possible. Since |M∪F|≤2n−4, there exists a position j∈[n+1] such that |(M∪F)∩Ej|≤1. When |(M∪F)∩Ej|=1, let (M∪F)∩Ej={w0w1}, and now there are at least six possibilities of such j. We choose the j such that the edge in (M∪F)∩Ej lies in F if possible. Otherwise, the edge in (M∪F)∩Ej is a matching edge for every j of the above six possibilities. In this case, since x and y are incident with at most two edges in M, we can choose j such that the matching edge in (M∪F)∩Ej is not incident with x and y, i.e., {x,y}∩{w0,w1}=∅. Split Qn+1 into Q0n and Q1n at position j. Without loss of generality, assume that |M0∪F0|≥|M1∪F1|. Now, δ(Q0n−F0)≥3 and δ(Q1n−F1)≥3.
Case 1. |M∩Ej|=0. Now, |M1∪F1|≤n−2.
Sub-case 1.1. |M0∪F0|≤2n−6.
Sub-case 1.1.1. x,y∈V(Q0n) (or x,y∈V(Q1n)).
By induction hypothesis when δ(Q0n−F0)≥4 and Lemma 4.2 when δ(Q0n−F0)=3, there exists a Hamiltonian path P0xy passing through M0 in Q0n−F0. Choose an edge u0v0∈E(P0xy)∖M0 such that u1v1∉M1 and {u0u1,v0v1}∩Fc=∅. Since |E(P0xy)∖M0|−|M1|−2|Fc|=|E(P0xy)|−|M|−2|Fc|≥2n−1−(2n−5)−2=2n−2n+2>1 for n≥4, such an edge u0v0 exists. By Corollary 2.4, there exists a Hamiltonian path P1u1v1 passing through M1 in Q1n−F1. Hence, Pxy=P0xy+P1u1v1+{u0u1,v0v1}−u0v0 is a Hamiltonian path joining x and y passing through M in Qn+1−F.
Sub-case 1.1.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Since 2n−1>3 for n≥4, we can choose a vertex z0∈V(Q0n) such that p(x)≠p(z0), {xz0,z1y}∩M=∅, and z0z1∉Fc. By induction hypothesis when δ(Q0n−F0)≥4 and Lemma 4.2 when δ(Q0n−F0)=3, there exists a Hamiltonian path P0xz0 passing through M0 in Q0n−F0. By Corollary 2.4, there exists a Hamiltonian path P1z1y passing through M1 in Q1n−F1. Hence, Pxy=P0xz0+P1z1y+z0z1 is the desired Hamiltonian path.
Sub-case 1.2. |M0∪F0|=2n−5. Now, |Fc|+|M1∪F1|≤1.
Sub-case 1.2.1. x,y∈V(Q0n).
Sub-case 1.2.1.1. F0=∅. Now, |Fc|+|F1|=1, |M1|=0, |M0|=2n−5<2n−4.
By Theorem 1.3, there exists a Hamiltonian path P0xy passing through M0 in Q0n. Choose an edge u0v0∈E(P0xy)∖M0 such that {u0u1,v0v1}∩Fc=∅. Since |E(P0xy)∖M0|−2|Fc|≥2n−1−(2n−5)−2=2n−2n+2>1 for n≥4, such an edge u0v0 exists. Since |F1|≤1<2n−5, by Theorem 1.4 there exists a Hamiltonian path P1u1v1 in Q1n−F1. Hence, Pxy=P0xy+P1u1v1+{u0u1,v0v1}−u0v0 is the desired Hamiltonian path.
Sub-case 1.2.1.2. F0≠∅. Let s0t0∈F0. Note that δ(Q0n−(F0∖{s0t0}))≥3.
Since |M0∪(F0∖{s0t0})|=2n−6, by induction hypothesis and Lemma 4.2, there exists a Hamiltonian path P0xy passing through M0 in Q0n−(F0∖{s0t0}). Next, we distinguish four cases to consider.
If s0t0∉E(P0xy), then choose an edge u0v0∈E(P0xy)∖M0 such that u1v1∉M1 and {u0u1,v0v1}∩Fc=∅. Since |E(P0xy)∖M0|−|M1|−2|Fc|≥2n−1−(2n−5)−2=2n−2n+2>1 for n≥4, such an edge u0v0 exists. If s0t0∈E(P0xy), s1t1∉M1, and {s0s1,t0t1}∩Fc=∅, then let u0v0=s0t0. In the above two cases, since |M1∪F1|≤1<n−2, by Corollary 2.4 there exists a Hamiltonian path P1u1v1 passing through M1 in Q1n−F1. Hence, Pxy=P0xy+P1u1v1+{u0u1,v0v1}−u0v0 is the desired Hamiltonian path.
If s0t0∈E(P0xy) and s1t1∈M1, then now |Fc∪F1|=0. Choose an edge u0v0∈E(P0xy)∖M0 such that {u0,v0}∩{s0,t0}=∅. Since |E(P0xy)∖M0|−3≥2n−1−(2n−6)−3=2n−2n+2>1 for n≥4, such an edge u0v0 exists. In Q1n, by Theorem 3.1 there exists a spanning 2-path P1u1v1+s1t1. Hence, Pxy=P0xy+P1u1v1+s1t1+{s0s1,t0t1,u0u1,v0v1}−{s0t0,u0v0} is the desired Hamiltonian path.
It remains to consider the case s0t0∈E(P0xy) and {s0s1,t0t1}∩Fc≠∅. Now, |M1∪F1|=0. Without loss of generality, assume that Fc={s0s1} and s0 is closer to x than t0 on P0xy. Let N(s0)=\{v∈V(Q0n) | dQ0n(s0,v)=1, dP0xy(s0,v)>1, and vs0∉F0}. Note that t0∉N(s0). Since dQ0n−F0(s0)≥3, we have |N(s0)|≥2.
If there exists a vertex r0∈N(s0) such that r0∈V(P0xy[x,s0]) and r0≠x, then choose a neighbor w0 of r0 on P0xy such that r0w0∉M0. When w0∈V(P0xy[r0,s0]), by Theorem 1.1 there exists a Hamiltonian path P1t1w1 in Q1n. Hence, Pxy=P0xy+P1t1w1+{s0r0,t0t1,w0w1}−{s0t0,r0w0} is the desired Hamiltonian path. When w0∈V(P0xy[x,r0]), choose an edge u0v0∈E(P0xy[r0,s0])∖M0 satisfying s0∉{u0,v0}. Since dP0xy(r0,s0)≥3, such an edge u0v0 exists. In Q1n, by Theorems 3.1 and 3.2 there exists a spanning 2-path P1w1u1+P1t1v1. Hence, Pxy=P0xy+P1w1u1+P1t1v1+{s0r0,w0w1,u0u1,v0v1,t0t1}−{w0r0,u0v0,s0t0} is the desired Hamiltonian path.
Otherwise, there exists at least one vertex in N(s0) which lies on P0xy[t0,y]. Denote the one closest to t0 by z0. Note that z0≠y, otherwise, the vertex r0 in the above case exists. A contraction occurs. Since p(x)≠p(y) and p(s0)≠p(z0), the number of odd vertices equals to the number of even vertices on P0xy[x,s0]+P0xy[z0,y]. Since s0 has at least |N(s0)|+1 neighbors which lie on P0xy[x,s0]+P0xy[z0,y], we have |V(P0xy[x,s0]+P0xy[z0,y])|≥6. Thus, |E(P0xy[x,s0]+P0xy[z0,y])|≥4.
Choose a neighbor w0 of z0 on P0xy such that z0w0∉M0. If w0∈V(P0xy[z0,y]), then by Theorem 1.1 there exists a Hamiltonian path P1t1w1 in Q1n. Hence, Pxy=P0xy+P1t1w1+{s0z0,t0t1,w0w1}−{s0t0,z0w0} is the desired Hamiltonian path. If w0∈V(P0xy[s0,z0]), then choose an edge u0v0∈E(P0xy[x,s0]+P0xy[z0,y])∖M0 satisfying s0∉{u0,v0}. Since |E(P0xy[x,s0]+P0xy[z0,y])|≥4, such an edge u0v0 exists. In Q1n, by Theorems 3.1 and 3.2 there exists a spanning 2-path P1w1u1+P1t1v1. Hence, Pxy=P0xy+P1w1u1+P1t1v1+{s0z0,t0t1,w0w1,u0u1,v0v1}−{s0t0,z0w0,u0v0} is the desired Hamiltonian path.
Sub-case 1.2.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Since |M0∪F0|=2n−5<2n−4, by Theorem 1.6 when M0≠∅ and Corollary 1.5 when M0=∅, there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. Choose a neighbor s0 of x on C0 such that xs0∉M0.
Sub-case 1.2.2.1. s0s1∉Fc.
When s1y∉M1, since |M1∪F1|≤1<n−2, by Corollary 2.4 there exists a Hamiltonian path P1s1y passing through M1 in Q1n−F1. Hence, Pxy=C0+P1s1y+s0s1−xs0 is the desired Hamiltonian path. When s1y∈M1, now |Fc∪F1|=0 and |M0|≤2n−6. Choose an edge u0v0∈E(C0)∖M0 such that s0∉{u0,v0} and y∉{u1,v1}. Since |E(C0)∖M0|−2−2≥2n−(2n−6)−4=2n−2n+2>1 for n≥4, such an edge u0v0 exists. In Q1n, by Theorem 3.1 there exists a spanning 2-path P1u1v1+s1y passing through M1. Hence, Pxy=C0+P1u1v1+s1y+{u0u1,v0v1,s0s1}−{xs0,u0v0} is the desired Hamiltonian path.
Sub-case 1.2.2.2. s0s1∈Fc. Now |M1∪F1|=0.
Since dQ0n−F0(s0)≥3, we can choose a neighbor t0 of s0 in Q0n such that dC0(t0,s0)≠1 and s0t0∉F0. Choose a neighbor r0 of t0 on C0 such that t0r0∉M0. If r0 lies on one path joining t0 and s0 on C0 and x lies on the other, then by Theorem 1.1 there exists a Hamiltonian path P1r1y in Q1n. Hence, Pxy=C0+P1r1y+{s0t0,r0r1}−{xs0,t0r0} is the desired Hamiltonian path. If r0 and x lie on the same path joining t0 and s0 on C0, then let u0 be the other neighbor of t0 on C0. Now, t0u0∈M0 and u0≠s0. Let v0 be the other neighbor of u0 on C0. So, u0v0∉M0. Since p(r1)=p(u1)≠p(v1)=p(y) and r1≠u1, by Corollary 3.4 there exists a spanning 2-path P1r1u1+P1v1y in Q1n. Hence, Pxy=C0+P1r1u1+P1v1y+{s0t0,u0u1,v0v1,r0r1}−{xs0,u0v0,t0r0} is the desired Hamiltonian path; see Figure 4.
Sub-case 1.2.3. x,y∈V(Q1n).
Since |M0∪F0|=2n−5<2n−4, by Theorem 1.6 when M0≠∅ and Corollary 1.5 when M0=∅, there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. Choose an edge s0t0∈E(C0)∖M0 such that {s0s1,t0t1}∩Fc=∅, s1t1≠xy, and {s1,t1}∩V(M1)=∅. Since |E(C0)∖M0|−2|Fc|−1−4|M1|=|E(C0)|−|M|−(2|Fc|+3|M1|)−1≥2n−(2n−5)−3−1=2n−2n+1>1 for n≥4, such an edge s0t0 exists. Since |(M1∪{s1t1})∪(F1∖{s1t1})|≤2≤n−2 for n≥4 and M1∪{s1t1} is a matching, by Corollary 2.4 there exists a Hamiltonian path P1xy passing through M1∪{s1t1} in Q1n−(F1∖{s1t1}). Hence, Pxy=C0+P1xy+{s0s1,t0t1}−{s0t0,s1t1} is the desired Hamiltonian path.
Sub-case 1.3. |M0∪F0|=2n−4. Now, |Fc|+|M1∪F1|=0 and |M0|≥1.
Since |M0∪F0|=2n−4 and |M0|≥1, by Theorem 1.6 there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0.
Sub-case 1.3.1. x,y∈V(Q0n).
Choose neighbors s0,t0 of x,y on C0 such that xs0∉M0 and t0y∉M0. If s0 lies on one path joining x and y on C0 and r0 lies on the other, then by Theorem 1.1 there exists a Hamiltonian path P1s1t1 in Q1n. Hence, Pxy=C0+P1s1t1+{s0s1,t0t1}−{xs0,t0y} is the desired Hamiltonian path. If s0 and t0 lie on the same path joining x and y on C0, then choose an edge r0w0∈E(C0)∖M0 on the other path. We may assume that r0 is closer to x than w0 on one path joining x and y on C0. By Theorems 3.1 and 3.2, there exists a spanning 2-path P1r1s1+P1w1t1 in Q1n. Hence, Pxy=C0+P1r1s1+P1w1t1+{s0s1,t0t1,r0r1,w0w1}−{xs0,t0y,r0w0} is the desired Hamiltonian path.
Sub-case 1.3.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Choose a neighbor s0 of x on C0 such that xs0∉M0. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1s1y. Hence, Pxy=C0+P1s1y+s0s1−xs0 is the desired Hamiltonian path.
Sub-case 1.3.3. x,y∈V(Q1n).
Choose an edge s0t0∈E(C0)∖M0 satisfying s0t0≠xy. Since |E(C0)∖M0|−1≥2n−(2n−5)−1=2n−2n+4>1 for n≥4, such an edge s0t0 exists. In Q1n, by Lemma 2.2 there exists a Hamiltonian path P1xy passing through s1t1. Hence, Pxy=C0+P1xy+{s0s1,t0t1}−{s0t0,s1t1} is the desired Hamiltonian path.
Case 2. |M∩Ej|=1. Now, Mc={w0w1}, {w0,w1}∩{x,y}=∅, |Fc|=0, and |M1∪F1|≤⌊2n−4−12⌋=n−3<n−2.
Sub-case 2.1. |M0∪F0|≤2n−6.
Sub-case 2.1.1. x,y∈V(Q0n) (or x,y∈V(Q1n)).
By induction hypothesis and Lemma 4.2, there exists a Hamiltonian path P0xy passing through M0 in Q0n−F0. Choose a neighbor r0 of w0 on P0xy. By Corollary 2.4, there exists a Hamiltonian path P1r1w1 passing through M1 in Q1n−F1. Hence, Pxy=P0xy+P1r1w1+{r0r1,w0w1}−r0w0 is the desired Hamiltonian path.
Sub-case 2.1.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
When p(x)≠p(w0), by induction hypothesis and Lemma 4.2, there exists a Hamiltonian path P0xw0 passing through M0 in Q0n−F0. By Corollary 2.4, there exists a Hamiltonian path P1w1y passing through M1 in Q1n−F1. Hence, Pxy=P0xw0+P1w1y+w0w1 is the desired Hamiltonian path.
When p(x)=p(w0), choose a neighbor s1 of y in Q1n such that s1∉V(M1) and xs0∉M0. Since n>(n−3)+1, such a vertex s1 exists. By induction hypothesis and Lemma 4.2, there exists a Hamiltonian path P0xs0 passing through M0 in Q0n−F0.
If there exists a neighbor r0 of w0 on P0xs0 such that r0≠s0 and r1y∉M1, then since |M1∪F1|≤n−3, by Lemma 3.8 there exists a spanning 2-path P1s1y+P1r1w1 passing through M1 in Q1n−F1. Hence, Pxy=P0xs0+P1s1y+P1r1w1+{s0s1,w0w1,r0r1}−r0w0 is the desired Hamiltonian path; see Figure 5(1).
Otherwise, dP0xs0(s0,w0)=1, and the other neighbor r0 of w0 on P0xs0 satisfies r1y∈M1. Choose an edge u0v0∈E(P0xs0)∖M0 such that u1v1≠s1w1 and {u1,v1}∩V(M1)=∅. Since |E(P0xs0)∖M0|−1−4|M1|=|E(P0xs0)|−(|M0|+|M1|)−1−3|M1|≥(2n−1)−(2n−6)−1−3(n−3)=2n−5n+13>1 for n≥4, such an edge u0v0 exists. Since |(M1∪{u1v1}∖{r1y})∪F1|≤n−3 and s1w1r1ys1 is a cycle of length four, by Lemma 3.12 there exists a spanning 2-path P1s1w1+r1y passing through M1∪{u1v1} in Q1n−(F1∖{u1v1}). Hence, Pxy=P0xs0+P1s1w1+r1y+{s0s1,w0w1,r0r1,u0u1,v0v1}−{w0r0,u0v0,u1v1} is the desired Hamiltonian path; see Figure 5(2).
Sub-case 2.2. |M0∪F0|=2n−5. Now, |M1∪F1|=0 and |F0|≥1.
Sub-case 2.2.1. x,y∈V(Q0n).
Let u0v0∈F0. By induction hypothesis and Lemma 4.2, there exists a Hamiltonian path P0xy passing through M0 in Q0n−(F0∖{u0v0}). If u0v0∉E(P0xy), then choose a neighbor r0 of w0 on P0xy. If u0v0∈E(P0xy) and dP0xy(w0,u0v0)=0, then let r0w0=u0v0. In Q1n, by Theorem 1.1 there exists a Hamiltonian path P1r1w1. Hence, Pxy=P0xy+P1r1w1+{r0r1,w0w1}−r0w0 is the desired Hamiltonian path. If u0v0∈E(P0xy) and dP0xy(w0,u0v0)≥1, then since w0∉{x,y}, we can choose a neighbor r0 of w0 on P0xy such that r0∉{u0,v0}. In Q1n, by Theorem 3.1 there exists a spanning 2-path P1u1v1+P1r1w1. Hence, Pxy=P0xy+P1u1v1+P1r1w1+{r0r1,w0w1,u0u1,v0v1}−{r0w0,u0v0} is the desired Hamiltonian path.
Sub-case 2.2.2. x∈V(Q0n), y∈V(Q1n) (or x∈V(Q1n), y∈V(Q0n)).
Since |M0∪F0|=2n−5<2n−4, by Theorem 1.6 when M0≠∅ and Corollary 1.5 when M0=∅, there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. If dC0(w0,x)=1, then by Theorem 1.1 there exists a Hamiltonian path P1w1y in Q1n. Hence, Pxy=C0+P1w1y+w0w1−xw0 is the desired Hamiltonian path. If dC0(w0,x)=2, then choose neighbors s0,r0 of x,w0 on C0, respectively, such that xs0∉M0 and r0≠s0. Now, p(s1)=p(r1)≠p(w1)=p(y) and s1,y,r1,w1 are distinct vertices. If dC0(w0,x)≥3, then choose neighbors s0,r0 of x,w0 on C0, respectively, such that r1≠y and xs0∉M0. Now, p(s1)≠p(y), p(r1)≠p(w1), and s1,y,r1,w1 are distinct vertices. In the above two cases, by Theorem 3.1 there exists a spanning 2-path P1s1y+P1r1w1 in Q1n. Hence, Pxy=C0+P1s1y+P1r1w1+{s0s1,r0r1,w0w1}−{xs0,r0w0} is the desired Hamiltonian path.
Sub-case 2.2.3. x,y∈V(Q1n).
Since |M0∪F0|=2n−5<2n−4, by Theorem 1.6 when M0≠∅ and Corollary 1.5 when M0=∅, there exists a Hamiltonian cycle C0 containing M0 in Q0n−F0. Choose a neighbor r0 of w0 on C0. Since r1w1≠xy, by Lemma 2.2 there exists a Hamiltonian path P1xy passing through r1w1 in Q1n. Hence, Pxy=P1xy+C0+{r0r1,w0w1}−{r0w0,r1w1} is the desired Hamiltonian path.
We investigate the existence of a Hamiltonian path in Qn passing through a matching and avoiding faulty edges. From Lemmas 4.1, 4.2, and 4.3, we can obtain the following conclusion.
Theorem 5.1. For n≥4, let M be a matching of Qn, and let F be a set of edges in Qn−M with |M∪F|≤2n−6. Let x,y∈V(Qn) be such that p(x)≠p(y) and xy∉M. If the degree of every vertex in Qn−F is at least 2, and there is neither x/y−DS nor (x,y)−DS in Qn−F, then there exists a Hamiltonian path joining x and y passing through M in Qn−F.
Shenyang Zhao: Methodology and Writing-original & draft; Fan Wang: Conceptualization and Writing-review & editing. All authors have read and agreed to the published version of the manuscript.
The authors would like to express their gratitude to the anonymous referees whose helpful comments and suggestions have led to a substantial improvement of the paper.
This work is supported by NSFC (grant no. 12061047), and supported by Jiangxi Provincial Natural Science Foundation (nos. 20212BAB201027 and 20192BAB211002).
The authors declare that they have no conflicts of interest in this work.
[1] | J. Bondy, U. Murty, Graph Theory with Applications, London: Macmillan Press, 1976. |
[2] |
R. Caha, V. Koubek, Hamiltonian cycles and paths with a prescribed set of edges in hypercubes and dense sets, J. Graph Theory, 51 (2006), 137–169. https://doi.org/10.1002/jgt.20128 doi: 10.1002/jgt.20128
![]() |
[3] |
R. Caha, V. Koubek, Spanning multi-paths in hypercubes, Discrete Math., 307 (2007), 2053–2066. https://doi.org/10.1016/j.disc.2005.12.050 doi: 10.1016/j.disc.2005.12.050
![]() |
[4] |
X. Chen, Paired many-to-many disjoint path covers of hypercubes with faulty edges, Inform. Process. Lett., 112 (2012), 61–66. https://doi.org/10.1016/j.ipl.2011.10.010 doi: 10.1016/j.ipl.2011.10.010
![]() |
[5] |
X. Chen, Matchings extend to Hamiltonian cycles in hypercubes with faulty edges, Front. Math. China, 14 (2019), 1117–1132. https://doi.org/10.1007/s11464-019-0810-8 doi: 10.1007/s11464-019-0810-8
![]() |
[6] |
D. Dimitrov, T. Dvořák, P. Gregor, R. Škrekovski, Gray codes avoiding matchings, Discrete Math. Theor. Comput. Sci., 11 (2009), 123–148. https://doi.org/10.46298/dmtcs.457 doi: 10.46298/dmtcs.457
![]() |
[7] |
T. Dvořák, Hamiltonian cycles with prescribed edges in hypercubes, SIAM J. Discrete Math., 19 (2005), 135–144. https://doi.org/10.1137/S0895480103432805 doi: 10.1137/S0895480103432805
![]() |
[8] |
T. Dvořák, P. Gregor, Hamiltonian paths with prescribed edges in hypercubes, Discrete Math., 307 (2007), 1982–1998. https://doi.org/10.1016/j.disc.2005.12.045 doi: 10.1016/j.disc.2005.12.045
![]() |
[9] |
J. Fink, Perfect matchings extend to Hamilton cycles in hypercubes, J. Combin. Theory Ser. B, 97 (2007), 1074–1076. https://doi.org/10.1016/j.jctb.2007.02.007 doi: 10.1016/j.jctb.2007.02.007
![]() |
[10] |
J. Fink, Connectivity of matching graph of hypercube, SIAM J. Discrete Math., 23 (2009), 1100–1109. https://doi.org/10.1137/070697288 doi: 10.1137/070697288
![]() |
[11] |
P. Gregor, Perfect matchings extending on subcubes to Hamiltonian cycles of hypercubes, Discrete Math., 309 (2009), 1711–1713. https://doi.org/10.1016/j.disc.2008.02.013 doi: 10.1016/j.disc.2008.02.013
![]() |
[12] | L. Gros, Théorie du Baguenodier, Lyon: Aimé Vingtrinier, 1872. |
[13] | I. Havel, On Hamiltonian circuits and spanning trees of hypercube, Čas. Pěst. Mat., 109 (1984), 135–152. http://dml.cz/dmlcz/108506 |
[14] | G. Kreweras, Matchings and hamiltonian cycles on hypercubes, Bull. Inst. Combin. Appl., 16 (1996), 87–91. |
[15] |
M. Lewinter, W. Widulski, Hyper-Hamilton laceable and caterpillar-spannable product graphs, Comput. Math. Appl., 34 (1997), 99–104. https://doi.org/10.1016/S0898-1221(97)00223-X doi: 10.1016/S0898-1221(97)00223-X
![]() |
[16] |
J. J. Liu, Y. L. Wang, Hamiltonian cycles in hypercubes with faulty edges, Inform. Sci., 256 (2014), 225–233. https://doi.org/10.1016/j.ins.2013.09.012 doi: 10.1016/j.ins.2013.09.012
![]() |
[17] |
C. Tsai, Linear array and ring embeddings in conditional faulty hypercubes, Theor. Comput. Sci., 314 (2004), 431–443. https://doi.org/10.1016/j.tcs.2004.01.035 doi: 10.1016/j.tcs.2004.01.035
![]() |
[18] |
C. Tsai, Y. Lai, Conditional edge-fault-tolerant edge-bipancyclicity of hypercubes, Inform. Sci., 177 (2007), 5590–5597. https://doi.org/10.1016/j.ins.2007.06.013 doi: 10.1016/j.ins.2007.06.013
![]() |
[19] |
W. Wang, X. Chen, A fault-free Hamiltonian cycle passing through prescribed edges in a hypercube with faulty edges, Inform. Process. Lett., 107 (2008), 205–210. https://doi.org/10.1016/j.ipl.2008.02.016 doi: 10.1016/j.ipl.2008.02.016
![]() |
[20] |
F. Wang, H. P. Zhang, Small matchings extend to Hamiltonian cycles in hypercubes, Graphs Combin., 32 (2016), 363–376. https://doi.org/10.1007/s00373-015-1533-6 doi: 10.1007/s00373-015-1533-6
![]() |
[21] |
F. Wang, H. P. Zhang, Hamiltonian laceability in hypercubes with faulty edges, Discrete Appl. Math., 236 (2018), 438–445. https://doi.org/10.1016/j.dam.2017.10.005 doi: 10.1016/j.dam.2017.10.005
![]() |
[22] |
J. M. Xu, Z. Z. Du, M. Xu, Edge-fault-tolerant edge-bipancyclicity of hypercubes, Inform. Process. Lett., 96 (2005), 146–150. https://doi.org/10.1016/j.ipl.2005.06.006 doi: 10.1016/j.ipl.2005.06.006
![]() |
[23] |
Y. Yang, N. Song, Z. Zhao, Hamiltonian Laceability of Hypercubes with Prescribed Linear Forest and/or Faulty Edges, J. Combinat. Math. Combinat. Comput., 120 (2024), 393–398. https://doi.org/10.61091/jcmcc120-36 doi: 10.61091/jcmcc120-36
![]() |