Partial domination of maximal outerplanar graphs
Peter Borg1 and Pawaton Kaemawichanurat2
1Department of Mathematics
Faculty of Science
University of Malta
Malta
2Theoretical and Computational Science Center
and Department of Mathematics, Faculty of Science,
King Mongkut’s University of Technology Thonburi
Bangkok, Thailand
Email: 1[email protected], 2[email protected]
Research supported in part by Skill Development Grant 2018, King Mongkut’s University of Technology Thonburi.
Abstract
Several domination results have been obtained for maximal outerplanar graphs (mops). The classical domination problem is to minimize the size of a set S of vertices of an n-vertex graph G such that G−N[S], the graph obtained by deleting the closed neighborhood of S, is null. A classical result of Chvátal is that the minimum size is at most n/3 if G is a mop. Here we consider a modification by allowing G−N[S] to have isolated vertices and isolated edges only. Let ι1(G) denote the size of a smallest set S for which this is achieved. We show that if G is a mop on n≥5 vertices, then ι1(G)≤n/5. We also show that if n2 is the number of vertices of degree 2, then ι1(G)≤6n+n2 if n2≤3n, and ι1(G)≤3n−n2 otherwise. We show that these bounds are best possible.
Keywords: Partial domination; Isolation number; Maximal outerplanar graph.
AMS subject classification: 05C10, 05C35, 05C69.
1 Introduction
Let G be a simple graph with vertex set V(G) and edge set E(G). We denote the degree of v in G by dG(v). The open neighborhood NG(v) of a vertex v of G is the set of neighbors of v, that is, NG(v)={u∈V(G):uv∈E(G)}. The closed neighborhood of v is NG[v]=NG(v)∪{v}. The neighborhood of a subset S of V(G) is the set NG(S)=∪v∈SNG(v). The closed neighborhood of a subset S of V(G) is the set NG[S]=NG(S)∪S. For S⊆V(G), the subgraph of G induced by S is denoted by G[S]. The subgraph obtained from G by deleting all vertices in S (and all edges incident to vertices in S) is denoted by G−S.
A subset S of V(G) is a dominating set of G if each vertex in V(G)∖S is adjacent to at least one vertex in S. The domination number of G is the size of a smallest dominating set of G and is denoted by γ(G). Given a graph H, a subset S of V(G) is an H-isolating set of G if G−NG[S] does not contain a copy of H. Obviously, S is a dominating set if H=K1. The H-isolation number of G is the size of a smallest H-isolating set of G and is denoted by ι(G,H). If H=K1,k+1 for some k≥0, then we may abbreviate ι(G,H) to ιk(G). The study of isolating sets is an appealing and natural direction in domination theory that was introduced recently by Caro and Hansberg [1].
A triangulated disc is a simple planar graph whose interior faces are triangles. A maximal outerplanar graph is a triangulated disc whose exterior face (the unbounded face) contains all vertices. Hence, a maximal outerplanar graph can be embedded in the plane such that all vertices lie on the boundary of the exterior face and all interior faces are triangles. We abbreviate the term maximal outerplanar graph to mop. O’Rourke [14] pointed out that every mop has a unique Hamiltonian cycle. Thus, the Hamiltonian cycle of a mop is the boundary of the mop. This paper’s notation and terminology on mops follows that of [10]; in particular, an edge belonging to the Hamiltonian cycle of a mop is called a Hamiltonian edge, while any other edge of the mop is called a diagonal. A fan of order n≥3, denoted Fn, is the mop obtained from a path Pn−1 by adding a new vertex v and joining it to every vertex of the path. We say that v is the center of Fn.
Domination in mops has been extensively studied since 1975. In the classical paper [3], Chvátal essentially proved that the domination number of an n-vertex mop is at most n/3. Other proofs were obtained by Fish [7] and by Matheson and Tarjan [12]. Caro and Hansberg [1] proved that the K1,1-isolation number of a mop of order n≥4 is at most n/4.
Theorem 1
*If G is a mop of order n, then the following hold:
(a) [3, 7, 12] If n≥3, then γ(G)≤3n.
(b) [1] If n≥4, then ι0(G)≤4n.*
For results on other types of domination in mops, we refer the reader to [1, 2, 4, 5, 6, 8, 9, 15].
Consider any n-vertex mop G. Theorem 1(a) tells us that there exists some S⊆V(G) such that ∣S∣≤n/3 and G−NG[S] is null. Theorem 1(b) tells us that there exists some S⊆V(G) such that ∣S∣≤n/4 and G−NG[S] has isolated vertices only. In this paper, we establish sharp upper bounds on the size of a smallest subset S of V(G) such that G−NG[S] has isolated vertices and isolated edges only, that is, the edges of G−NG[S] form a matching (that is, they are pairwise disjoint). The bounds are detailed in the following two theorems and the proofs are given in Section 2.
Theorem 2
If G is a mop of order n≥5, then
[TABLE]
Theorem 3
If G is a mop of order n≥5 with n2 vertices of degree 2, then
[TABLE]
The bounds in Theorems 2 and 3 are sharp. For example, let F51,F52,…,F5t be t≥2 vertex disjoint fans of order 5. From each F5i, we choose a vertex of degree 2 and its neighbor of degree 3. Then, we join these 2t vertices by edges to form a mop. The resulting graph Gt is a mop of order n=5t with n2=t vertices of degree 2. An example of the graph Gt with t=4 is illustrated in Figure 1.
Figure 1 : The mop G4, which satisfies ι1(G4)=n/5=20/5=4and ι1(G4)=(n+n2)/6=(20+4)/6=4.
Clearly, for a K1,2-isolating set S of Gt, we have ∣S∩V(F5i)∣≥1 because Gt−NGt[S] does not contain K1,2. Thus, ι1(Gt)≥t. However, Gk has a K1,2-isolating set containing the vertex of degree 4 of each F5i. Consequently, ι1(Gt)=t=(5t)/5=n/5 and ι1(Gt)=t=(5t+t)/6=(n+n2)/6.
Figure 2 : The mop H4, which satisfies ι1(H4)=(n−n2)/3=(20−8)/3=4.
To see the sharpness of the bound ι1(G)≤3n−n2, we first let F51,F52,…,F5t be t≥2 fans of order 5. We choose the two vertices of degree 3 from each F5i. Then we join these 2t vertices by edges to form a mop. The resulting graph Ht is a mop of order n=5t with n2=2t vertices of degree 2. An example of the graph Ht with t=4 is illustrated in Figure 2. Clearly, for a K1,2-isolating set S of Ht, we have ∣S∩V(F5i)∣≥1 because Ht−NHt[S] does not contain K1,2. Thus, ι1(Ht)≥t. However, Ht has a K1,2-isolating set containing the vertex of degree 4 of each F5i. Consequently, ι1(Ht)=t=(5t−2t)/3=(n−n2)/3.
We conclude this section by comparing the bound n/5 with (n+n2)/6 and (n−n2)/3. We see that n/5<(n+n2)/6 when n/5<n2. Moreover, n/5<(n−n2)/3 when n2<2n/5. Thus, the bound in Theorem 2 is better than that in Theorem 3 when n/5<n2<2n/5. We will show that the bound n/5 is sharp when n/5<n2<2n/5. Let F1, F2 and F3 be vertex-disjoint fans of order 5 such that V(F1)={x0,x1,…,x4},V(F2)={y0,y1,…,y4} and V(F3)={z0,z1,…,z4}, where x0,y0,z0 are the centers of the fans and x1,x4,y1,y4,z1,z4 are the vertices of degree 2 of the fans. Let A15 be the graph obtained by taking the union of F1,F2,F3 and a mop that has {z1,z2,x2,x3,y2,y3} as its vertex set and has z1z2, x2x3 and y2y3 among its Hamiltonian edges. Clearly, A15 is a mop of order 15 with 5 vertices of degree 2. This graph is illustrated in Figure 3. Then, we let A151,A152,…,A15t be t vertex-disjoint copies of A15. The graph Bt is constructed by taking the union of A151,A152,…,A15t and a mop that has {x31,y31,x32,y32,…,x3t,y3t} as its vertex set and has x31y31,x32y32,…,x3ty3t among its Hamiltonian edges.
Clearly, Bt is a mop of order n=15t with n2=5t vertices of degree 2. Thus, n/5<n2=5n/15<2n/5. Clearly, each K1,2-isolating set of Bt needs at least one vertex from each fan Fji.
Thus, ιk(Bt)≥3t. Moreover, {x0i:1≤i≤t}∪{y0i:1≤i≤t}∪{z0i:1≤i≤t} is a K1,2-isolating set of B. Thus, ι1(Bt)=3t=n/5.
x_{3}$$x_{2}$$x_{0}$$x_{4}$$x_{1}$$y_{3}$$y_{2}$$y_{1}$$y_{0}$$y_{4}$$z_{1}$$z_{2}$$z_{3}$$z_{4}$$z_{0}Figure 3 : The mop A15.
2 Proofs of the upper bounds
In this section, we prove Theorems 2 and 3. We apply results of O’Rourke [13] in computational geometry that were used in a new proof by Lemańska, Zuazua and Zylinski [10] of an upper bound by Dorfling, Hattingh and Jonck [5] on the size of a total dominating set (a set S of vertices such that each vertex of the graph is adjacent to a vertex in S) of a mop. Before stating these results, we make a related straightforward observation that we will also use.
Given three mops G, G1 and G2, we say that a diagonal d of G partitions G into G1 and G2 if G is the union of G1 and G2, V(G1)∩V(G2)=d and E(G1)∩E(G2)={d}.
Lemma 1
If d is a diagonal of a mop G, then d partitions G into two mops G1 and G2.
**Proof. **Let x1x2…xnx1 be the Hamiltonian cycle C of G. We may assume that d=x1xi for some i∈{2,3,…,n}. Let C1 be the cycle x1x2…xix1 of G, and let C2 be the cycle x1xixi+1…xnx1. For each i∈{1,2}, let Gi be the subgraph of G induced by V(Ci). Then, V(G1)∩V(G2)=d and E(G1)∩E(G2)={d}. Each face of Gi is a face of G in the interior of Ci and hence a triangle; thus, Gi is a mop. Since G is a mop and x1xi is a diagonal of G, G has no edge with one vertex in V(C1)\{x1,xi} and the other vertex in V(C2)\{x1,xi}. Thus, E(G)=E(G1)∪E(G2). □
The next lemma was proved by Chvátal in [3] and is restated in [14, Lemma 1.1], and the lemma following it is an extension by O’Rourke [13] that has a central role in our proofs.
Lemma 2
([3])*
If G is a mop of order n≥8, then G has a diagonal d that partitions it into two mops G1 and G2 such that G1 has exactly 4, 5 or 6 Hamiltonian edges of G.*
Lemma 3
([13])*
If G is a mop of order n≥10, then G has a diagonal d that partitions it into two mops G1 and G2 such that G1 has exactly 5, 6, 7 or 8 Hamiltonian edges of G.*
For a graph G and an edge uv of G, the edge contraction of G along uv is the graph obtained from G by deleting u and v (and all incident edges), adding a new vertex x, and making x adjacent to the vertices in NG({u,v})∖{u,v} only. Recall that every mop can be embedded in a plane so that the exterior face contains all vertices. By looking at polygon corners as vertices, we have that a mop is a triangulation of a simple polygon, meaning that its boundary is the polygon and its interior faces are triangles.
Lemma 4
([13])*
If G is a triangulation of a simple polygon P, e is a Hamiltonian edge of G, and G′ is the edge contraction of G along e, then G′ is a triangulation of some simple polygon P′.*
If G is a graph and I⊆V(G) such that uv∈/E(G) for every u,v∈I, then I is called an independent set of G.
For the proof of Theorem 3, we establish the following result about vertices of degree 2.
Lemma 5
If G is a mop of order n≥4, then the set of vertices of G of degree 2 is an independent set of G of size at most 2n.
**Proof. **Let V2 be the set of vertices of G of degree 2. Let x0x1…xn−1x0 be the unique Hamiltonian cycle of G and hence the boundary of the exterior face of G. For each i∈{0,1,…,n−1}, the vertices xi−1modn and xi+1modn are neighbours of xi.
Suppose xi,xj∈V2 such that xixj∈E(G). For each ℓ∈{−1,0,1,2}, let yℓ=xi+ℓmodn. Since n≥4, the vertices y−1,y0,y1,y2 are distinct. Since y0=xi∈V2, we have NG(y0)={y−1,y1}, so xj is y−1 or y1. We may assume that xj=y1. Since xj∈V2, we obtain NG(y1)={y0,y2}. Together with NG(y0)={y−1,y1}, this gives us that the path y−1y0y1y2 lies on an interior face of G, which contradicts the assumption that G is a mop.
Therefore, V2 is an independent set of G. Thus, each edge of G contains at most one vertex in V2. Let H be the set of Hamiltonian edges. For any v∈V2 and e∈H, let χ(v,e)=1 if e contains v, and let χ(v,e)=0 otherwise. By the above, ∑v∈V2χ(v,e)≤1 for each e∈H. If xi∈V2, then the edges containing xi are xi−1modnxi and xixi+1modn, so ∑e∈Hχ(xi,e)=2. We have
[TABLE]
and hence ∣V2∣≤n/2. □
The next lemma lists other facts about vertices of degree 2, which we shall also use.
Lemma 6
*If G is a mop of order n≥3, then the following hold:
(a) Each vertex of G is of degree at least 2.
(b) G has at least 2 vertices of degree 2.
(c) G−v is a mop for any vertex v of G of degree 2.
(d) A graph H is a mop if G=H−w for some w∈V(H) such that dH(w)=2 and NH(w) is a Hamiltonian edge of G.*
**Proof. **(a) This is immediate from the fact that G has a Hamiltonian cycle.
(b) We use induction on n. The result is trivial if n=3. Suppose n≥4. Since G is a mop, G has a diagonal d=xy. By Lemma 1, G is the union of two mops G1 and G2 such that V(G1)∩V(G2)=d and E(G1)∩E(G2)={d}. By the induction hypothesis, for each i∈{1,2}, Gi has two vertices vi,1 and vi,2 such that dGi(vi,1)=dGi(vi,2)=2. By Lemma 5, for each i∈{1,2}, we cannot have dGi(x)=dGi(y)=2, so vi,ji∈/{x,y} for some ji∈{1,2}. Therefore, we have v1,j1=v2,j2, dG(v1,j1)=dG1(v1,j1)=2 and dG(v2,j2)=dG2(v2,j2)=2.
(c) Let v be a vertex of G of degree 2. We may label the vertices x1,x2,…,xn so that x1x2…xnx1 is the Hamiltonian cycle C of G and xn=v. Since dG(xn)=2, NG(xn)={x1,xn−1}. The face F having xn−1xn and xnx1 on its boundary must also have xn−1x1 on its boundary (as all interior faces are triangles), meaning that xn−1x1∈E(G). Thus, x1x2…,xn−1x1 is a Hamiltonian cycle of G−v. Also, every interior face of G other than F is a face of G−v (and a triangle). Therefore, G−v is a mop.
(d) Let x1x2…xnx1 be the Hamiltonian cycle C of G. We may assume that NH(w)={x1,x2}. Let C′=x1wx2…xnx1. Then, C′ is a Hamiltonian cycle of H. Let ρ be a plane drawing of G such that all vertices of G lie on the boundary C and all interior faces are triangles. Extend ρ to a plane drawing ρ′ of H by putting w and the edges wx1,wx2 on the exterior of ρ. Then, C′ is the boundary of ρ′. Also, the faces of ρ′ are the faces of ρ together with the face bounded by wx1, wx2 and x1x2, so all faces of ρ′ are triangles. Thus, H is a mop. □
Lemmas 5 and 6(b) tell us that the number n2 of vertices of degree 2 (of a mop) satisfies
[TABLE]
We show in passing that both bounds are sharp. For the upper bound, we start with any mop M with Hamiltonian cycle x1x2…xpx1, add p new vertices y1,y2,…,yp, and add the edges y1x1,y1x2, y2x2,y2x3, …,ypxp,ypx1, and hence the resulting graph G is a mop (by Lemma 6(d)), the vertices of G of degree 2 are y1,y2,…,yp, and p=2∣V(G)∣. For the lower bound, we start with a cycle C=x1x2…xnx1, set p=⌈n/2⌉, and take G with V(G)=V(C) and E(G)=E(C)∪{x2xn,xnx3,x3xn−1,xn−1x4,…,xp−1xn−p+3,xn−p+3xp,xpxn−p+2}, and hence the vertices of G of degree 2 are x1 and xp+1.
The next lemma settles Theorem 2 for 5≤n≤9, and hence allows us to use Lemma 3 in the proof of Theorem 2.
Lemma 7
If 5≤n≤9 and G is a mop of order n, then ι1(G)=1≤5n.
**Proof. **By definition of a mop, ι1(G)≥1.
Suppose n=9. Let x1x2…x9x1 be the unique Hamiltonian cycle of G and hence the boundary of the exterior face of G. By Lemma 2, G has a diagonal d such that G is the union of two mops G1 and G2 such that V(G1)∩V(G2)=d, E(G1)∩E(G2)={d}, and G1 has exactly ℓ Hamiltonian edges of G for some ℓ∈{4,5,6}. We may assume that d is the edge x1xℓ+1 and that V(G1)={x1,x2,…,xℓ+1}.
Suppose ℓ=4. Then, V(G2)={x1,x5,x6,x7,x8,x9}. Let (x1,x5,xj) be the triangular face of G2 containing the edge x1x5 (so j∈{6,7,8,9}). Let x′=x5 if j=9, and let x′=x1 otherwise. If 6≤j≤8, then x1,x2,x5,xj,x9∈NG[x′]. If j=9, then x1,x4,x5,x6,x9∈NG[x′]. Clearly, each component of G−NG[x′] contains at most two vertices. Therefore, {x′} is a K1,2-isolating set of G. Hence, ι1(G)=1<n/5.
Suppose ℓ=5. Then, x6x7,x7x8,x8x9,x9x1 are the 4 Hamiltonian edges of G that belong to G2. Thus, this case is identical to the case ℓ=4.
Suppose ℓ=6. Let (x1,xj,x7) be the triangular face of G1 containing the edge x1x7 (so j∈{2,3,4,5,6}). Suppose j=2. Then, x2x7∈E(G). By Lemma 1, x2x7 partitions G into two mops H1 and H2 such that H1 has exactly 4 Hamiltonian edges of G (namely, x7x8,x8x9,x9x1,x1x2), so the result follows as in the case ℓ=4. By symmetry, it suffices to consider the cases j=3 and j=4; that is, the cases j=6 and j=5 are identical to the cases j=2 and j=3, respectively. In both cases, each component of G−NG[xj] contains at most two vertices, so {xj} is a K1,2-isolating set of G.
Now suppose n=8. Let uv be a Hamiltonian edge of G, and let H be the graph obtained by adding a new vertex w to G and adding the edges wu and wv. By Lemma 6(d), H is a mop of order 9, so ι1(H)=1<n/5. Let {x} be a K1,2-isolating set of H. If x=w, then {x} is a K1,2-isolating set of G. If x=w, then {u} is a K1,2-isolating set of G as NH[w]={u,v,w}⊆NH[u]. Therefore, ι1(G)=1<n/5.
For 5≤i≤7, we obtain the result for n=i from the result for n=i+1 in the same way we obtained the result for n=8 from the result for n=9.
□
We now prove Theorems 2 and 3. Recall the statement of Theorem 2.
Theorem 2. If G is a mop of order n≥5, then ι1(G)≤5n.
**Proof. **We use induction on n. Lemma 7 establishes the base case n=5 (in this case, G is a fan F5, and its center is a K1,2-isolating set of G) and also the case 6≤n≤9. Suppose n≥10. We assume that if G′ is a mop of order n′ with 5≤n′<n, then ι1(G′)≤n′/5.
Let x1x2…xnx1 be the unique Hamiltonian cycle C of G and hence the boundary of the exterior face of G. By Lemma 3, G has a diagonal d such that G is the union of two mops G1 and G2 such that V(G1)∩V(G2)=d, E(G1)∩E(G2)={d}, and G1 has exactly ℓ Hamiltonian edges of G for some ℓ∈{5,6,7,8}. We may assume that d is the edge x1xℓ+1 and that V(G1)={x1,x2,…,xℓ+1}. Note that the edges x1x2,x2x3,…,xℓxℓ+1 are the ℓ Hamiltonian edges of G that belong to G1. Let (x1,xj,xℓ+1) be the triangular face of G1 containing the edge x1xℓ+1 (so j∈{2,3,…,ℓ}).
Claim 1
If ℓ=5, then ι1(G)≤5n.
**Proof. **Let G′ be the graph obtained from G by deleting the vertices x2,x3,x4,x5 and contracting the edge x1x6 to form a new vertex y (see Figure 4). Thus, G′ is obtained from G2 by contracting the edge x1x6. By Lemma 4, G′ is a mop.
x_{2}$$x_{3}$$x_{4}$$x_{5}$$x_{1}$$x_{6}$$x_{n}$$x_{7}$$x_{p}$$x_{q}$$x_{r}$$\dots$$\dots$$x_{n}$$x_{6}$$y$$x_{p}$$x_{r}$$x_{q}$$\dots$$\dots$$\RightarrowFigure 4 : The edge contraction of G2.
Let n′ be the order n−5≥5 of G′. By the induction hypothesis, ι1(G′)≤n′/5=n/5−1. Let S′ be a smallest K1,2-isolating set of G′. Then, ∣S′∣=ι1(G′)≤n/5−1. We have that {x1,x6} is a {K1,2}-isolating set of G1 and that NG′(y)⊆NG(x1)∪NG(x6). Thus, if y∈S′, then (S′∖{y})∪{x1,x6} is a {K1,2}-isolating set of G, and hence ι1(G)≤(∣S′∣−1)+2≤n/5. Suppose y∈/S′. Observe that ∣V(G1−NG1[xj])∣≤2. Thus, G1−NG1[xj] does not contain a copy of K1,2. Moreover, S′∪{xj} is a K1,2-isolating set of G as xj is adjacent to both x1 and x6. Therefore, ι1(G)≤∣S′∣+1≤n/5. (□)
Claim 2
If ℓ=6, then ι1(G)≤5n.
**Proof. **Let G′=G2 and n′=∣V(G2)∣. We have V(G′)=V(G)∖{x2,x3,x4,x5,x6}, so n′=n−5≥5. By the induction hypothesis, ι1(G′)≤n′/5=n/5−1. Let S′ be a smallest K1,2-isolating set of G′. Then, ∣S′∣=ι1(G′)≤n/5−1. Suppose j=2. Then, x2x7 is a diagonal of G that partitions G into two mops H1 and H2 with H1=G1−x1. Since H1 has exactly 5 Hamiltonian edges of G (namely, x2x3,x3x4,x4x5,x5x6,x6x7), we have ι1(G)≤5n by Claim 1. Now suppose j≥3. By symmetry, it suffices to consider the cases j=3 and j=4 (see Figure 5); that is, the cases j=6 and j=5 are identical to the cases j=2 and j=3, respectively. In both cases, the order of each component of G1−NG1[xj] is at most 2, so G1−NG1[xj] does not contain a copy of K1,2. Since xj is adjacent to x1 and x7, S′∪{xj} is a K1,2-isolating set of G, and hence ι1(G)≤∣S′∣+1≤n/5. (□)
x_{1}$$x_{2}$$x_{3}$$x_{4}$$x_{5}$$x_{6}$$x_{7}$$x_{1}$$x_{2}$$x_{3}$$x_{4}$$x_{5}$$x_{6}$$x_{7}Figure 5
Claim 3
If ℓ=7, then ι1(G)≤5n.
**Proof. **Let G′=G2 and n′=∣V(G2)∣. We have V(G′)=V(G)∖{x2,x3,x4,x5,x6,x7}, so n′=n−6≥4. Suppose j=2. Then, x2x8 is a diagonal of G that partitions G into two mops H1 and H2 with H1=G1−x1. Since H1 has exactly 6 Hamiltonian edges of G (namely, x2x3,x3x4,…,x7x8), we have ι1(G)≤5n by Claim 2. Similarly, if j=3, then x3x8 is a diagonal which partitions G into two mops H1 and H2 with H1=G1−{x1,x2}, and hence the result follows by Claim 1. Now suppose j≥4. By symmetry, we may assume that j=4 (see Figure 6, left); that is, the cases j=7, j=6 and j=5 are identical to the cases j=2, j=3 and j=4, respectively.
The order of each component of G1−NG1[x4] is at most 2, so G1−NG1[x4] does not contain a copy of K1,2. Suppose n′=4. Then, n=10 and ∣V(G′)∣=4. Since x4 is adjacent to both x1 and x8, G′−NG[x4] does not contain a copy of K1,2. Thus, {x4} is a K1,2-isolating set of G, and hence ι1(G)=1<n/5. Now suppose n′≥5. By the induction hypothesis, ι1(G′)≤n′/5<n/5−1. Let S′ be a smallest K1,2-isolating set of G′. Then, since S′∪{x4} is a K1,2-isolating set of G, we have ι1(G)≤∣S′∣+1=ι1(G′)+1<n/5. (□)
x_{1}$$x_{2}$$x_{3}$$x_{4}$$x_{5}$$x_{6}$$x_{7}$$x_{8}$$x_{1}$$x_{2}$$x_{3}$$x_{4}$$x_{5}$$x_{6}$$x_{7}$$x_{8}$$x_{9}Figure 6
By Claims 1, 2 and 3, we may assume that ℓ=8. Similarly to the above, the cases j=2, j=3 and j=4 follow by Claims 3, 2 and 1, respectively. Thus, we may assume that j≥5. Also similarly to the above, by symmetry, it suffices to consider j=5 (see Figure 6, right).
The order of each component of G1−NG1[x5] is at most 2, so G1−NG1[x5] does not contain a copy of K1,2. Let G′=G2 and n′=∣V(G2)∣. Then, n′=n−7. Suppose n′≤4. Then, n≤11 and ∣V(G′)∣≤4. Since x5 is adjacent to both x1 and x9, G′−NG[x5] does not contain a copy of K1,2. Thus, {x5} is a K1,2-isolating set of G, and hence ι1(G)=1<n/5. Now suppose n′≥5. By the induction hypothesis, ι1(G′)≤n′/5<n/5−1. Let S′ be a smallest K1,2-isolating set of G′. Then, since S′∪{x5} is a K1,2-isolating set of G, we have ι1(G)≤∣S′∣+1=ι1(G′)+1<n/5. □
Recall the statement of Theorem 3.
Theorem 3. If G is a mop of order n≥5 with n2 vertices of degree 2, then
[TABLE]
**Proof. **We first consider n2>3n. Let V2 be the set of vertices of G of degree 2. Let G′=G−V2 and n′=∣V(G′)∣. By Lemma 5, V2 is an independent set of G, and n2≤n/2. Since n′ is an integer satisfying n′=n−n2≥n/2≥5/2, we have n′≥3. Since G′ is obtained from G by deleting the vertices in V2 one by one, G′ is a mop by Lemma 6(c). Let S be a smallest dominating set of G′. By Theorem 1(a), ∣S∣≤n′/3. Since V2 is an independent set of G, S is a K1,2-isolating set of G. Thus, ι1(G)≤∣S∣≤(n−n2)/3.
We now prove the bound for n2≤3n, using induction on n. We actually prove it for n≥4. If n=4, then we trivially have ι1(G)=1, so ι1(G)≤(n+n2)/6 by Lemma 6(b). By Theorem 2 and Lemma 6(b), we have ι1(G)≤1≤(n+n2)/6 if 5≤n≤9, and we have ι1(G)≤2≤(n+n2)/6 if n=10. We now consider n≥11 and assume that ι1(G′)≤(n′+n2′)/6 for any mop G′ such that ∣V(G′)∣=n′≥4, G′ has n2′ vertices of degree 2, and n′+n2′<n+n2.
Let x1x2…xnx1 be the unique Hamiltonian cycle C of G and hence the boundary of the exterior face of G. By Lemma 3, G has a diagonal d such that G is the union of two mops G1 and G2 such that V(G1)∩V(G2)=d, E(G1)∩E(G2)={d}, and G1 has exactly ℓ Hamiltonian edges of G for some ℓ∈{5,6,7,8}. We may assume that d is the edge x1xℓ+1 and that V(G1)={x1,x2,…,xℓ+1}. Note that the edges x1x2,x2x3,…,xℓxℓ+1 are the ℓ Hamiltonian edges of G that belong to G1. Let (x1,xj,xℓ+1) be the triangular face of G1 containing the edge x1xℓ+1 (so j∈{2,3,…,ℓ}).
Claim 4
If ℓ=5, then ι1(G)≤6n+n2.
**Proof. **By Lemma 6(a), dG2(v)≥2 for each v∈V(G2). Since x1 and x6 are adjacent in G2, Lemma 5 tells us that their degrees cannot be both 2. We may assume that dG2(x1)≥3. This yields dG2(x1)+dG2(x6)≥5.
x_{1}$$x_{6}$$x_{2}$$x_{3}$$x_{4}$$x_{5}$$x_{n}$$x_{7}Figure 7
Claim 4.1
If dG2(x1)+dG2(x6)=5, then ι1(G)≤6n+n2.
**Proof. **Suppose dG2(x1)+dG2(x6)=5. Then, dG2(x1)=3 and dG2(x6)=2. We have x1xn,x6x7∈E(C)∩E(G2). Since x1x6,x6x7∈E(G2) and dG2(x6)=2, NG2(x6)={x1,x7}. Thus, since G2 is a mop, the face having x1x6 and x6x7 on its boundary must also have x1x7 on its boundary (as all interior faces are triangles), that is, x1x7∈E(G2) (see Figure 7). Together with x1x6,x1xn∈E(G2) and dG2(x1)=3, this gives us NG2(x1)={x6,x7,xn}. Thus, since G2 is a mop, the face having x1x7 and x1xn on its boundary must also have x7xn on its boundary, that is, x7xn∈E(G2) (see Figure 7). Let G′=G−{x1,x2,…,x6}. Then, G′=(G2−x6)−x1. Since dG2(x6)=2, G2−x6 is a mop by Lemma 6(c). Since dG2−x6(x1)=2, G′ is a mop by Lemma 6(c). Let n′=∣V(G′)∣ and n2′=∣{v∈V(G′):dG′(v)=2}∣. We have n′=n−6≥5.
By Lemma 5, at most one of x7 and xn has degree 2 in G′. By Lemma 5, at most one of x1 and x6 has degree 2 in G1, and hence, by Lemma 6(b), dG1(xh)=2 for some h∈{2,3,4,5}. Since xh∈V(G1)\V(G2), dG(xh)=dG1(xh). Therefore, n2′≤n2, and hence n′+n2′≤n+n2−6. By the induction hypothesis, ι1(G′)≤(n′+n2′)/6≤(n+n2)/6−1. Let S′ be a smallest K1,2-isolating set of G′. Clearly, ∣V(G1−NG1[xj])∣≤2, so G1−NG1[xj] does not contain a copy of K1,2. Since xj is adjacent to both x1 and x6, it follows that S′∪{xj} is a K1,2-isolating set of G. Thus, we have ι1(G)≤∣S′∣+1=ι1(G′)+1≤(n+n2)/6. (□)
By Claim 4.1, we may assume that dG2(x1)+dG2(x6)≥6. Let G′ be the graph obtained from G by deleting the vertices x2,x3,x4,x5 and contracting the edge x1x6 to form a new vertex y. Then, G′ is obtained from G2 by contracting x1x6. Thus, G′ is a mop by Lemma 4. Let n′=∣V(G′)∣ and n2′=∣{v∈V(G′):dG′(v)=2}∣. We have n′=n−5≥6.
Suppose that every vertex that has degree 2 in G′ also has degree 2 in G. As in the proof of Claim 4.1, dG1(xh)=2 for some h∈{2,3,4,5}, so n2′≤n2−1. Thus, n′+n2′≤n+n2−6. By the induction hypothesis, ι1(G′)≤(n′+n2′)/6≤(n+n2)/6−1. Let S′ be a smallest K1,2-isolating set of G′. Then, ∣S′∣≤(n+n2)/6−1.
We can continue as in the proof of Claim 1 to obtain ι1(G)≤∣S′∣+1≤(n+n2)/6.
Now suppose that G′ has a vertex z such that dG′(z)=2=dG(z). Then z=y or x1,x6∈NG2(z).
Suppose z=y. As noted in the proof of Claim 4.1, x1xn,x6x7∈E(G2). Thus, x7,xn∈NG′(y). Since dG′(y)=2, NG′(y)={x7,xn}. Since dG2(x1)+dG2(x6)≥6, we obtain NG2(x1)={x6,x7,xn} and NG2(x6)={x1,x7,xn}. Since NG2(x1)={x6,x7,xn}, x1x7 is a diagonal of G2. By Lemma 1, we obtain x6xn∈/E(G2), which contradicts NG2(x6)={x1,x7,xn}.
Therefore, z=y. Thus, x1,x6∈NG2(z). For each i∈{8,9,…,n−1} with dG′(xi)=2, we have NG′(xi)=NG(xi)={xi−1,xi+1}, so z=xi. Thus, z=x7 or z=xn. By symmetry, we may assume that z=x7. Since x7x8∈E(C)∩E(G2) and x1,x6∈NG2(x7), NG2(x7)={x1,x6,x8}. Thus, since G2 is a mop, the face having x1x7 and x7x8 on its boundary must also have x1x8 on its boundary (as all interior faces are triangles), meaning that x1x8∈E(G2). By Lemma 1, x1x8 partitions G into two mops H1 and H2 such that V(H2)={x1,x8,x9,…,xn}. Let G′=H2, n′=V(H2) and n2′=∣{v∈V(H2):dH2(v)=2}∣. We have n′=n−6≥5. By Lemma 5, for each i∈{1,2}, at most one of x1 and x8 has degree 2 in Hi. By Lemma 6(b), dH1(xh)=2 for some h∈V(H1)\{x1,x8}, and hence dG(xh)=2. Therefore, n2′≤n2, and hence n′+n2′≤n+n2−6. By the induction hypothesis, ι1(G′)≤(n′+n2′)/6≤(n+n2)/6−1. Let S′ be a smallest K1,2-isolating set of G′. Let x′=x6 if j=2, and let x′=x1 otherwise. If 3≤j≤5, then x1,x2,xj,x6,x7,x8∈NH1[x′]. If j=2, then x1,x2,x5,x6,x7∈NH1[x′]. Since x1x6 is a diagonal of G, we have x3,x4∈/NG(x8) by Lemma 1. Therefore, S′∪{x′} is a K1,2-isolating set of G. Thus, we have ι1(G)≤∣S′∣+1=ι1(G′)+1≤(n+n2)/6. (□)
Claim 5
If ℓ=6, then ι1(G)≤6n+n2.
**Proof. **As in the proof of Claim 2, we may assume that j=3 or j=4. In both cases, the order of each component of G1−NG1[xj] is at most 2, so G1−NG1[xj] does not contain a copy of K1,2. Let G′=G2, n′=∣V(G2)∣ and n2′=∣{v∈V(G2):dG2(v)=2}∣. We have V(G′)=V(G)∖{x2,x3,x4,x5,x6}, so n′=n−5≥6.
We have x1xj,xjx7∈E(G1). Since 3≤j≤4, x1xj and xjx7 are diagonals of G. By Lemma 1, x1xj partitions G into two mops G1,1 and G1,2 with V(G1,1)={x1,x2,…,xj}, and xjx7 partitions G into two mops G2,1 and G2,2 with V(G2,1)={xj,xj+1,…,x7}. By Lemma 5, at most one of x1 and xj is of degree 2 in G1,1, and at most one of xj and x7 is of degree 2 in G2,1. Thus, by Lemma 6(b), dG1,1(y)=2 for some y∈V(G1,1)\{x1,xj}, and dG2,1(z)=2 for some z∈V(G2,1)\{xj,x7}. We have dG(y)=dG1,1(y)=2 and dG(z)=dG2,1(z)=2. Also, y,z∈/V(G′). Now dG(v)=dG′(v) for each v∈V(G′)\{x1,x7}, and, by Lemma 6(b), at most one of x1 and x7 is of degree 2 in G′. Therefore, n2′≤n2−1.
We have n′+n2′≤n+n2−6. By the induction hypothesis, ι1(G′)≤(n′+n2′)/6=(n+n2)/6−1. Let S′ be a smallest K1,2-isolating set of G′. Then, ∣S′∣=ι1(G′)≤(n+n2)/6−1. Since xj is adjacent to x1 and x7, S′∪{xj} is a K1,2-isolating set of G, and hence ι1(G)≤∣S′∣+1≤(n+n2)/6. (□)
Claim 6
If ℓ=7, then ι1(G)≤6n+n2.
**Proof. **As in the proof of Claim 3, we may assume that j=4. Let G′=G2 and n′=∣V(G2)∣. We have V(G′)=V(G)∖{x2,x3,x4,x5,x6,x7}, so n′=n−6≥5. By the same reasoning used for Claim 5, there are at least two vertices in V(G)\V(G′) of degree 2 in G, at most one of x1 and x8 is of degree 2 in G′, and dG(v)=dG′(v) for each v∈V(G′)\{x1,x8}. Thus, n2′≤n2−1, and hence n′+n2′≤n+n2−7. By the induction hypothesis, ι1(G′)≤(n′+n2′)/6<(n+n2)/6−1. Let S′ be a smallest K1,2-isolating set of G′. Then, since S′∪{x4} is a K1,2-isolating set of G, we have ι1(G)≤∣S′∣+1=ι1(G′)+1<(n+n2)/6. (□)
By Claims 4, 5 and 6, we may assume that ℓ=8. As in the proof of Theorem 2 for the same case, we may assume that j=5. Let G′=G2 and n′=∣V(G2)∣. We have V(G′)=V(G)∖{x2,x3,x4,x5,x6,x7,x8}, so n′=n−7≥4. By the same reasoning used for Claim 5, there are at least two vertices in V(G)\V(G′) of degree 2 in G, at most one of x1 and x9 is of degree 2 in G′, and dG(v)=dG′(v) for each v∈V(G′)\{x1,x9}. Thus, n2′≤n2−1, and hence n′+n2′≤n+n2−8. By the induction hypothesis (recall that the first bound in the theorem is being proved for n≥4), ι1(G′)≤(n′+n2′)/6<(n+n2)/6−1. Let S′ be a smallest K1,2-isolating set of G′. Then, since S′∪{x5} is a K1,2-isolating set of G, we have ι1(G)≤∣S′∣+1=ι1(G′)+1<(n+n2)/6. □